EPA-600/R-97/121
    EPA TRACEABILITY PROTOCOL FOR ASSAY AND
CERTIFICATION OF GASEOUS CALIBRATION STANDARDS

                      September 1997
          U.S. Environmental Protection Agency (MD-47)
             National Exposure Research Laboratory
         Human Exposure and Atmospheric Sciences Division
               Research Triangle Park, NC 27711

-------
                                                EPA-600/R-97/121
    EPA TRACEABBLITY PROTOCOL FOR ASSAY AND
CERTIFICATION OF GASEOUS CALIBRATION STANDARDS

                       September 1997
          U.S. Environmental Protection Agency (MD-47)
              National Exposure Research Laboratory
         Human Exposure and Atmospheric Sciences Division
               Research Triangle Park, NC 27711

-------
                                    ADDENDUM
       Monitoring needs for acid deposition, and source and ambient air quality may exceed
inventories of SRMs and PRMs available for required traceability of standards. To facilitate
support for research and monitoring projects for which standards of traditional NIST or
equivalent traceability are unavailable, the Traceability Protocol, as revised September  1997,
incorporates the use of Research Gas Mixtures (ROMs).  Candidate ROMs will be analyzed
and certified by NIST under their ROM program, and assertions as to stability and quality
will be as specified by NIST. Standards certified against ROMs may be used directly only
for calibration or.audit, and may not be used to certify subordinate standards. Enquiries
about the ROM program should be directed to NIST at the address given below.

              National Institute of Standards and Technology
              Analytical Chemistry Division
              Chemical Science and Technology Laboratory
              B-324 Chemistry
              Gaithersburg, MD.20899

-------
                              TABLE OF CONTENTS
Section
                                                                            Page

          List of Figures	 jx
          List of Tables 	x
          Acknowledgments 	 Xj
          Glossary	  xijj

   1       INTRODUCTION	1-1

   2       EPA TRACEABILITY PROTOCOL FOR ASSAY AND CERTIFICATION OF
          COMPRESSED GAS CALIBRATION STANDARDS 	2-1

          2.1  GENERAL INFORMATION	2-1
              2.1.1     Purpose and Scope of the Protocol	2-1
              2.1.2    Reference Standards	2-1
                      2.1.2.1   Gas Manufacturer's Intermediate Standard	2-4
                      2.1.2.2   Recertification of Reference Standards	   2-5
              2.1.3    Using the Protocol 	2-5
              2.1.4    Certification Documentation	2-5
              2.1.5    Certification Label	2-6
              2.1.6    Assay/Certification of Compressed Gas Calibration
                      Standards	2-6
                      2.1.6.1   Incubation of Newly Prepared Compressed
                              Gas Calibration Standards	2-6
                      2.1.6.2   Stability Test for Reactive Gas Mixtures 	2-6
                      2.1.6.3   Certification Periods for Compressed Gas
                              Calibration Standards	 2-7
                      2.1.6.4   Minimum Cylinder Pressure	2-9
                      2.1.6.5   Assay/Certification of Multicomponent
                              Compressed Gas Calibration Standards	2-9
             2.1.7     Analyzer Calibration	2-10
                     2.1.7.1   Basic Analyzer Calibration Requirements	2-10
                     2.1.7.2   Analyzer Multipoint Calibration	2-10
                     2.1.7.3   Zero and Span Gas Checks	2-13
                     2.1.7.4   Reference Standards for Multipoint Calibrations
                              and Zero and Span Gas Checks 	  2-14
                     2.1.7.5   Uncertainty of the Calibration Curve 	2-14
             2.1.8     Uncertainty of the Estimated Concentration of
                     the Candidate Standard	2-15
             2.1.9    Zero Gas	2-17
             2.1.10   Accuracy Assessment of Commercially Available
                     Standards	2-17

         2.2 PROCEDURE G1: ASSAY AND CERTIFICATION OF A COMPRESSED
             GAS CALIBRATION STANDARD WITHOUT DILUTION	2-18
             2.1.1    Applicability	• • • • 2-18

                                        iii

-------
                         TABLE OF CONTENTS (continued)
Section                                                                      Page
                                            «
              2.2.2    Limitations ....	2-18
              2.2.3    Assay Apparatus 	2-18
              2.2.4    Pollutant Gas Analyzer	2-20
              2.2.5    Analyzer Calibration		2-20
                      2.2:5.1   Multipoint Calibration  	2-20
                      2.2.5.2   Analyzer Range	2-20
                      2.2.5.3   Linearity	2-20
                      2.2.5.4   Zero and Span Gas Checks	2-21
              2.2.6    Assay Gases	2-22
                      2.2.6.1   Candidate Standard	2-22
                      2.2.6.2   Reference Standard	2-22
                      2.2.6.3   Zero Gas	2-23
              2.2.7    Assay Procedure	 2-23
              2.2.8    Stability Test for Newly Prepared Candidate Standards	2-24
              2.2.9    Certification Documentation	2-25
              2.2.10   Recertification Requirements	2-25

         2.3  PROCEDURE G2: ASSAY AND CERTIFICATION OF A COMPRESSED
              GAS CALIBRATION STANDARD USING DILUTION	2-25
              2.3.1    Applicability	2-25
              2.3.2    Limitations	2-25
              2.3.3    Assay Apparatus 	2-26
              2.3.4    Pollutant Gas Analyzer	2-29
              2.3.5    Analyzer Calibration	2-29
                      2.3.5.1   Multipoint Calibration	 2-29
                      2.3.5.2   Analyzer Range 	2-30
                      2.3.5.3   Linearity	2-30
                      2.3.5.4   Zero and Span Gas Checks	2-30
              2.3.6    Selection of Gas Dilution- Flow Rates or Gas
                      Concentration Settings	 2-32
              2.3.7    Flowmeter Type and Flowmeter Calibration 	2-32
              2.3.8    Assay Gases	2-33
                      2.3.8.1   Candidate Standard	2-33
                      2.3.8.2   Reference Standard	2-34
                      2.3.8.3   Zero Gas	2-34
              2.3.9    Assay Procedure  	2-34
              2.3.10   Stability Test for Newly Prepared Standards	2-36
              2.3.11   Certification Documentation	2-37
              2.3.12   Recertification Requirements	2-37

  3      EPA TRACEABILITY PROTOCOL FOR ASSAY AND CERTIFICATION OF
         PERMEATION DEVICE CALIBRATION STANDARDS  	3-1
                                        IV

-------
Section
                          TABLE OF CONTENTS (continued)
                                                                              Page

          3.1  GENERAL INFORMATION	3-1
              3.1.1    Purpose and Scope of the Protocol	3-1
              3.1.2    Reference Standards	3-1
              3.1.3    Selecting a Procedure  	;•	3-2
              3.1.4    Using the Protocol	 3-3
              3.1.5    Certification Documentation	3-3
              3.1.6    Certificatiori Label	3.4
              3.1.7    Assay/Certification of Candidate Permeation Device
                      Calibration Standards	3-4
                      3.1.7.1   Permeation Device Design	3-4
                      3.1.7.2   Precautions for Use and Storage of
                               Permeation Devices	'	3-4
                      3.1.7.3   Equilibration of Newly Prepared Permeation
                               Devices	3-5
                      3.1.7.4   Certification Conditions for Permeation
                               Device Calibration Standards	3-5
              3.1.8    Technical Variances	3-5

          3.2  PROCEDURE P1:  ASSAY AND CERTIFICATION OF PERMEATION
              DEVICE CALIBRATION STANDARDS REFERENCED TO A
              PERMEATION DEVICE REFERENCE STANDARD  	3-6
              3.2.1    Applicability	3-6
              3.2.2    Limitations	3-6
              3.2.3    Assay Apparatus  	3-7
              3.2.4    Pollutant Gas Analyzer	 3-9
              3.2.5    Analyzer Calibration	3-9
                      3.2.5.1   Multipoint Calibration 	3-9
                      3.2.5.2   Analyzer Range 	3-10
                      3.2.5.3   Linearity  	3-10
                      3.2.5.4   Zero and Span Gas Checks	3-10
              3.2.6    Selection of Gas Dilution Flow Rates	3-10
              3.2.7    Flowmeter Type and Flowmeter Calibration	3-10
              3.2.8    Permeation Devices	3-11
                      3.2.8.1   Candidate Standard	3-11
                      3.2.8.2   Reference Standard	3-11
                      3.2.8.3   Zero Gas	3-11
              3.2.9    Assay Procedure  	3-12
              3.2.10   Stability Test for Newly Prepared Permeation
                      Devices	3-14
              3.2.11   Certification Documentation	3-14
              3.2.12   Recertification Requirements .. -.	3-15

-------
                         TABLE OF CONTENTS (continued)
Section    '                                                                 Page

         3.3 PROCEDURE P2: ASSAY AND CERTIFICATION OF PERMEATION
             DEVICE CALIBRATION STANDARDS REFERENCED TO A
             COMPRESSED GAS REFERENCE STANDARD  	3-15
             3.3.1    Applicability	3-15
             3.3.2    Limitations	3-15
             3.3.3    Assay Apparatus	3-15
             3.3.4    Pollutant Gas Analyzer	3-17
             3.3.5    Analyzer Calibration	3-18
                     3.3.5.1   Multipoint Calibration 	3-18
                     3.3.5.2   Analyzer Range 	3-19
                     3.3.5.3   Linearity  	3-19
                     3.3.5.4   Zero and Span Gas Checks	3-19
             3.3.6    Selection of Gas Dilution Flow. Rates	3-19
             3.3.7    Flowmeter Type and Flowmeter Calibration  	3-19
             3.3.8    Candidate Standard	3-20
             3.3.9    Reference Standard	3-20
             3.3.10   Zero Gas	3-21
             3.3.11   Assay Procedure  	3-21
             3.3.12   Equilibrium Test for Newly Prepared Permeation
                     Devices	3-24
             3.3.13   Certification Documentation	3-24
             3.3.14   Recertification Requirements	3-24

         3.4 PROCEDURE P3: ASSAY AND CERTIFICATION OF PERMEATION
             DEVICE CALIBRATION STANDARDS REFERENCED TO A MASS
             REFERENCE STANDARD	3-24
             3.4.1    Applicability	3-24
             3.4.2    Limitations	3-24
             3.4.3    Assay Apparatus  	3-25
                     3.4.3.1   Analytical Balance 	3-25
                     3.4.3.2   Temperature-controlled Chamber	3-25
                     3.4.3.3   Electrostatic Charge Neutralization  	3-27
             3.4.4    Weighing Interval 	3-27
             3.4.5    Assay Procedure  	3-28
             3.4.6    Number of Weighings of the Candidate Standard	3-29
             3.4.7    Calculation of Certified Permeation Rate	3-29
             3.4.8    Uncertainty of Certified Permeation Rate for
                     Candidate Standard	3-31
             3.4.9    Certification Documentation	3-31
             3.4.10   Recertification Requirements	  3-31

  4      REFERENCES	4-1
                                       VI

-------
                     TABLE OF CONTENTS (continued)








Appendix                                                        Page



  A     INSTRUCTIONS FOR CALIBRATION WORKBOOK	A-1



  B     INSTRUCTION FOR PERMEATION RATE WORKBOOK	B-1



  C     CALCULATION OF TOTAL ANALYTICAL UNCERTAINTY	C-1



  D     MATRIX NOTATION 	D-1
                                 VII

-------
                                  LIST OF FIGURES
Number     •                                                                    Page

  2-1      Example regression curve and confidence bands from
           multipoint calibration  	2-12

  2-2      One possible design of the apparatus for the assay of compressed gas
           calibration standards without dilution (Procedure G1)	 2-19

  2-3      One possible design of the apparatus using flow controllers for assay of
           compressed gas calibration standards with dilution (Procedure G2) 	2-27

  2-4      One possible design of the apparatus using a gas dilution system for
           assay of compressed  gas calibration standards with dilution (Procedure
           G2)	2-28

  3-1      One possible design of the apparatus for the assay of permeation device
           calibration standards referenced to a permeation device reference
           standard (Procedure P1) 	3-8

  3-2      One possible design of the apparatus for the assay of permeation device
           calibration standards referenced to a compressed gas reference standard
           (Procedure P2)	3-16

  3-3      Chamber for maintaining permeation tubes at constant
           temperature 	3-26

  3-4      Example of spreadsheet graphic output for calculating permeation rates  	3-30
                                          IX

-------
                           LIST OF TABLES
Number                                                        Page

  2-1     SUMMARY OF COMPRESSED GAS SRMs THAT ARE AVAILABLE
         FROM NIST	2-2

  2-2     SUMMARY OF COMPRESSED GAS PRMs THAT ARE AVAILABLE
         FROM NMi	2-3

  2-3     CERTIFICATION PERIODS FOR COMPRESSED GAS CALIBRATION
         STANDARDS IN ALUMINUM CYLINDERS THAT ARE CERTIFIED
         UNDER THIS PROTOCOL 	2-8

  2-4     SOME LINEARIZING TRANSFORMATIONS FOR MULTIPOINT
         CALIBRATION DATA	2-16

  3-1     NIST SRM PERMEATION DEVICE REFERENCE STANDARDS	3-2

-------
                                ACKNOWLEDGMENTS
       The information in this document has been funded wholly by the U.S. Environmental
Protection Agency (EPA) under Contract No. 68D40091 to the Research Triangle Institute (RTI),
Research Triangle Park, North Carolina.   •

       This traceability protocol was prepared by Robert S. Wright and Michael J. Messner of
RTI under RTI Project No. 91U-6699-208.  It is a revision of an earlier protocol that was
prepared in 1993 by the same authors. The authors thank the EPA work assignment manager,
Beme I. Bennett, for his guidance and technical discussion as the protocol was being written.
The authors also thank the following RTI staff members who provided administrative, technical,
editorial, word*processing, and graphic arts support: Malcolm J. Bertoni,  Michael Y.  Cross,
W. Gary Eaton, Tonda J. Gentry, Beryl C. Pittman, Helen M. Reading, Sue S. Preston, and
Jan L Shirley.

This protocol was reviewed by personnel of EPA, the National Institute of Standards and
Technology (NIST), and the Netherlands Measurement Institute (NMi) as it was being prepared.
The authors wish to thank the following individuals for their technical assistance:

       Pamela M. Chu (NIST)
       Ed W.B. DeLeer (NMi)
       Willliam D. Dorko (NIST)
       David L. Duewer (NIST)
       Gerald D. Mitchell (NIST)
       John T. Schakenbach (EPA)
       Rob Wessel (NMi)

Numerous comments from specialty gas producers provided useful comments and suggestions
for revisions to this protocol. The authors thank the following organizations for their assistance:

       Airgas, Inc.
       Air Liquide America Corporation
       Air Products and Chemicals, Inc.
       BOC Gases
       Matheson Gas Products
       MG Industries
       Praxair, Inc.
       Scott-Marrin, Inc.
       Scott Specialty Gases, Inc.
       Spectra Gases, Inc.
                                          XI

-------
                                    GLOSSARY
ANSI    American National Standards Institute
ASTM   American Society for Testing and Materials
GMIS    Gas Manufacturer's Intermediate Standard
ILAC    International Laboratory Accreditation Program
ISO      Jnternational Standardization Organization
NIST    National Institute of Standards and Technology
Nmi      Netherlands Measurement Institute
NTRM   NIST Traceable Reference Material
NVLAP   National Voluntary Laboratory Accreditation Program
PRM     Primary Reference Material
RGM    Research Gas Mixture
SRM     Standard Reference Material
                                        XII

-------
                                      SECTION 1

                                    INTRODUCTION
        In 1993, the U.S. Environmental Protection Agency (EPA) in Research Triangle Park, North
Carolina, revised its 1987 version of its traceability protocol for the assay and certification of
compressed gas and permeation-device calibration standards.1-2 The protocol allows producers of
gaseous standards, users of gaseous standards, and other analytical laboratories to establish
traceability between their protocol gases and gaseous Standard Reference Materials (SRMs)
produced by the National Institute of Standards and Technology (NIST). Parts 50, 58, 60, and 75
of Title 40 of the Code of Federal Regulations (CFR) require using SRMs or gaseous standards
traceable to SRMs for calibrating and auditing ambient air and stationary source pollutant monitoring
systems.3"6

        As the protocol was being revised, the 1987 version was published in the January 11,1993,
issue of the Federal Register. It was inserted as Appendix H to 40 CFR Part 75, which concerns
the Acid Rain Program.  In April 1996, EPA's Acid Rain Program suggested that the 1987 and 1993
versions of the protocol be consolidated into a single protocol. In discussions at that time,  RTI
suggested that the protocol be revised in a manner that would allow EPA Protocol Gases to be
produced either with the higher accuracy (within ±2 percent) needed by EPA's Acid Rain Division
or with less stringent accuracy requirements that might be acceptable to other users.  In 1997, RTI
solicited additional suggested revisions from specialty gas producers and other interested parties.

        The purpose of this work assignment was to revise the statistical procedures in the  1993
version of the protocol and to consolidate its two published versions in order to make the revised
protocol useful to EPA's Acid Rain Division as well as other users. Improvement of the statistical
procedures had the highest priority.

        The current revision has several significant changes from the 1993 version as listed below:

        1.  Statistical techniques are used to calculate the total uncertainty of the candidate
           standard.  Data from the multipoint calibration and the assays are used in these
           calculations (see Section 2.1.4);

        2.  The uncertainty of reference standards is now included in the calculation of the total
           uncertainty of the candidate standard (see Section 2.1.2);

        3.  Statistical techniques are used to calculate the stability of candidate standards.  The
           intermediate performance standard of 1 percent agreement between assays has  been
           deleted  (see Section 2.1.6.2);

       4.  Statistical spreadsheets were developed to assist in these calculations, but equivalent
           statistical techniques may also be used.  These spreadsheets replace most of the
           manual calculations required in the 1993 version of the protocol (see Appendices A
          • through D).


                                          1-1

-------
5.  Multipoint calibration .data may be fitted to straight-line, quadratic, cubic, or nuartic
    linear regression models ny the spreadsheets, although the use of cubic and quartic
    models is discouraged (see Appendix A);                  -  .    i  -         :
                      i •                     /V                   i
6.  If a quadratic or higher-order model is used to fit the multipoint calibration data, at least
    two reference standards, having different concentrations, must be measured during the
    assays (see Section2*2.7);   .         ,      .,.

7.  The correction for minor calibration drift following the multipoint calibration is 'embedded
    in the spreadsheets' calculation of the uncertainty of a single assay;

8.  Primary Reference Standards from the Netherlands Measurement Institute (NMi) are
    accepted as being equivalent to SRMsfrom NIST (see Section 2.1.2);

9.  The analyst may substitute a low-concentration reference standard in the place of the
    zero gas during assays of the candidate standard (see Section 2.1.9);

10. A summary of EPA's audit results from 1992 to the present is available at an EPA
    website, http://www.epa.gov/ttn (see Section 2.1.10);

11. A strip chart recorder is no longer required as part of the assay apparatus, but a high-
    precision data acquisition system must produce an electronic or paper record of the
    analyzer's response during assays. This record must be maintained for 3 years after
    the standard's certification date (see Section 2.2A).
                                   1-2

-------
                                     SECTION 2

         EPA TRACEABILITY PROTOCOL FOR ASSAY AND CERTIFICATION OF
                    COMPRESSED GAS CALIBRATION STANDARDS
2.1    GENERAL INFORMATION

2.1.1  Purpose and Scope of the Protocol

       This protocol describes two procedures for assaying the concentration of compressed gas
calibration standards and for certifying that the assayed concentrations are traceable to a National
Institute of Standards and Technology (NIST) Standard Reference Material (SRM). This protocol
is mandatory for certifying the compressed gas calibration standards used for the pollutant
monitoring that is required by the regulations of 40 CFR Parts 50, 58, 60 and 75 for the calibration
and audit of ambient air quality analyzers and continuous emission monitors.  This protocol may be
used to assay and certify gas mixtures that have the same components as compressed gas SRMs.'
A multiple-component standard may be assayed and certified under this protocol if compressed gas
SRMs that contain the individual components in the standard exist. This protocol may be used by
specialty gas-producers, standard users,'or other analytical laboratories. The assay procedure may
involve the direct  comparison of the standards to  reference standards without dilution (i.e.,
Procedure G1) or the indirect comparison of the standards to reference standards with dilution (i.e.,
Procedure G2). A candidate standard having a concentration that is lower or higher than that of the
reference standard may be certified under this protocol if both concentrations (or  diluted
concentrations) fall within the well-characterized region of the pollution gas analyzer's calibration
curve. This protocol places no restrictions  on cylinder sizes and the same analytical procedures
must be used in assays of all cylinder sizes.

2.1.2  Reference Standards

       Parts 50, -58, 60, and 75 of the monitoring  regulations require that gaseous pollutant
concentration  standards used for calibration and audit of ambient air quality analyzers and
continuous emission monitors be traceable to either a NIST SRM or a NIST Traceable Reference
Material (NTRM).7 In 1996, NIST and the Netherlands Measurement Institute (NMi) issued a joint
declaration that specific NMi Primary Reference Materials (PRMs)8 can be considered as being
equivalent to the corresponding NIST SRMs. The compressed gas SRMs that are available from
NIST are  listed in Table 2-1. The current SRM-equivalent compressed gas PRMs that are available
from NMi are listed in Table 2-2.  Other gas mixtures  are under study by NIST and NMi and they
may  be added to the Declaration of Equivalence. PRMs produced by other national metrology
organizations will be considered equivalent to NIST SRMs when a declaration of equivalence is
issued jointly by NIST and the national metrology organization.  The generic terms "Primary
Reference Material" and "PRM" are used in this document to refer to any SRM-equivalent standard
that has received such equivalency status.

       The uncertainty of SRMs, NTRMs,  and  PRMs is expressed as a 95-percenfconfidence
interval, which is the one-sigma uncertainty multiplied by a coverage factor almost always equal to


                                         2-1

-------
                                     .ems-
             TABLE£-1. .SUMMARY OF COMPRESSED GAS SRMs
                     THAT ARE AVAILABLE FROM MIST
      Certified, component
  Balance
    gas
 Concentration1* range
       for SRMs
Ambient nonmethane organics-(15
components)
Ambient toxic organics (19
components)
Aromatic organic gase?c
Carbon dioxide
Carbon dioxide
Carbon monoxfde
Carbon monoxide,
Carbon monoxide, propane, and
carbon dioxide

Hydrogen sulfide
Methane
Methane
Methane and propane
Nitric oxide
Oxides of nitrogen
(i.e., nitrogen dioxide plus
nitric acid)
Oxygen
Propane
-'ropane ~
Sulfur dioxide
Nitrogen

Nitrogen

Nitrogen
Air
Nitrogen
Air
Nitrogen
Nitrogen
Nitrogen
Air
Nitrogen
Air
Nitrogen
Air

Nitrogen
Air
Nitrogen
Nitrogen
5ppb

5ppb

0.25to10ppm
345 to 365 ppm
0.5 to 16 percent
10 to 45 ppm
10 ppm to 13 percent
1.6 to 8 percent CO
600 to 3,000 ppm C3H8
0 to 14 percent CO2
5 to 20 ppm
1 to 10 ppm
50 to 100 ppm
4 ppm CH4,1 ppm CgHB
5 to 3,000 ppm
100 ppm

2 to 21 percent
0.25 to 500 ppm
100 ppm to 2 percent
50 to 3,500 ppm
 All SRMs may not be available at all times. Other compressed gas SRMs may be developed in
 the future and could be used as reference standards. Contact NIST for information about SRM
 availability at (301) 975-6776 or http://gases.nist.gov.
 SRM concentrations are by mole.
 Aromatic organic gases are benzene, bromobenzene, chlorobenzene, and toluene.
                                     2-2

-------
                  TABLE 2-2. SUMMARY OF COMPRESSED GAS PRMs
                           THAT ARE AVAILABLE FROM NMi'
     Certified component
Balance gas
Concentration range for PRMsb
Carbon dioxide
Carbon monoxide
Ethanol
Nitric oxide
Oxygen
Propane
Sulfur dioxide
Nitrogen
Nitrogen
Nitrogen '
Nitrogen
Nitrogen
Nitrogen
Nitrogen
100ppmto 15 percent
1 00 ppm to 6 percent
100to259ppm
10 to 4000 ppm
2 to 22 percent
500 to 3000 ppm
100 to 3500 ppm
  Information about PRMs can be obtained from:

       Nederlands Meetinstituut B.V.
       Van Swinden Laboratorium
       Department of Chemistry
       P.O. Box 654
       2600 AR DELFT
       The Netherlands
       Telephone: 31 152691680
       Telefax: 31 15261 2971
       E-mail: SecChemie@NMi.nl
            The NMi office in the United States is:

            NMi USA, Inc.
            36 Gilbert Street South
            Tinton Falls, NJ 07701
            P.O. Box 7758
            Shrewsbury, NJ 07701
            Telephone: (908) 842-8900
            Telefax: (908) 842-0304
            E-mail: NMiUSANJ@aol.com
  SRM-equivalent PRMs from other national metrology organizations may be added in the future.
  Users of this protocol will be advised if such additions occur.

b Within the listed ranges, any concentration is available. PRMs are prepared individually in 5-L
  cylinders according to ISO Standard 6142 (Gas Analysis-Preparation of calibration gas mixtures-
  weighing methods). After preparation, the composition is verified against Dutch Primary Standard
  Gas Mixtures. The stability is normally guaranteed for a period of 2 years.  Uncertainties depend on
  the certified concentration and vary from 0.1 percent (relative) for binary mixtures to 1.0 percent
  (relative) maximum for certain constituents in multicomponent mixtures.

  2.9 This estimate includes allowances for the uncertainties of known sources of systematic error as
  well as the random error of measurement. A value of one-half of the stated uncertainty of these
  reference standards should be used in calculating the total analytical uncertainty of standards that
  are certified under this protocol (see Appendix C).

         The EPA regulations define a "traceable" standard as  one that has been compared and
  certified, either directly or via not more than one intermediate standard, to a primary standard  such
  as a NIST SRM or an NTRM.3-4 Comparison of a candidate standard directly to an SRM, an SRM-
  equivalent PRM, or an NTRM is  preferred and recommended.  However, the use of  a Gas
  Manufacturer's Intermediate  Standard (GMIS)  (see  Subsection 2.1.2.1) in the comparison is
  permitted. A GMIS is an intermediate reference standard that has been compared directly to  an
                                            2-3

-------
SRM, an  SRM-equivalent PRM, or a NTRM according to Procedure G1.  It is an acceptable
reference standard for the assay of candidate standards. However, purchasers of standards that
have been compared to L 3MIS should Delaware that in conformity with the above definition, such
a standard could only bt used directly for calibration or audit. Such a standard.could not be used
as a second-generation  intermediate reference standard to certify other compressed gas calibration
standards.                                     '                             •

      .Accordingly, the reference standard used for assaying and certifying a compressed gas
calibration standard under this protocol must be an SRM, a NTRM, an SRM-equivalent PRM, or a
GMIS. The reference standard must be within its certification period.

       Volume reference standards must be traceable to NIST primary standards by calibration at
a NIST-accredited state weights and measures laboratory or at a calibration laboratory that is
accredited by the National Voluntary  Laboratory  Accreditation  Program (NVLAP) or by the
International Laboratory Accreditation Conference (ILAC).10-" These volume reference standards
are required for assays  using procedure G2 (see Subsection 2.3.7).

2.1.2.1 Gas Manufacturer's Intermediate Standard—
       A  GMIS is a compressed  gas calibration  standard that has been assayed by direct
comparison to an SRM, an SRM-equivalent PRM, or a NTRM, that has been assayed and certified
according to  Procedure  G1, and that also meets the following requirements:

       1.  A candidate  GMIS must be assayed on at least three separate dates that are uniformly
          spaced over at least a 3-month period.  During each of these  assays, the candidate
          GMIS  must  be measured at least three times. All these assays must use the same
          SRM,  SRM-equivalent  PRM, or NTRM as the reference standard  to avoid errors
          associated with the use of different reference standards for different assays.

      2.  For each assay, the analyst must calculate the mean and 95-percent uncertainty for the
          three or more measured concentrations of the candidate GMIS according to the
          statistical procedures described in Appendix A or equivalent statistical techniques.  The
          95-percent  uncertainty  must be  less than or equal to  1.0  percent of the mean
          concentration.

      3.  After the three or more  assays have been completed, the analyst must calculate the
          overall mean estimated concentration and the 95-percent uncertainty for the candidate
          GMIS  using the spreadsheet described  in Appendix C or equivalent statistical
          techniques.

      4.  If  the  95-percent  confidence  limits (i.e., estimated concentration  plus or minus
          uncertainty)  for the assays overlap, the candidate GMIS can be considered to  be stable
          and can be  used as a reference standard for assays of candidate standards. In the
          Appendix C spreadsheet, all cells in the comparisons table will be "true." If the
          confidence limits do not  overlap, the candidate GMIS may be unstable  or there may be
          analytical problems associated with the assays or the reference standards. One or
          more cells in the comparisons table will be "false."  The analyst must  either disqualify
          the candidate GMIS or investigate why the confidence limits do  not overlap.  The
          analyst may discard the data from a questionable assay and then conduct another
          assay. The candidate GMIS can be used as a reference standard if the confidence

                                         2-4

-------
           limits for the remaining assays and the new assay overlap. The candidate GMIS cannot
           be used if it appears to be unstable.
                                     -1 -
       5.  A'GMIS must be recertified every 2 years. Use the spreadsheet described in Appendix
           C or equivalent statistical techniques to compare the confidence limits from a single
           recertification  assay with the confidence limits from the previous assays.  If the
           confidence intervals overlap, the GMIS can be recertified. If the reassayed GMIS fails
           to meet this requirement, it must undergo a full certification as described in Step 1
           above before it can be used again. There is no requirement that the same reference
           standard must be used in the original assays  and the recertification assay,  but this
           practice is desirable if possible.

2.1.2.2 Recertification of Reference Standards—
       Recertification requirements for SRMs and NTRMs  are specified by NIST.  Recertification
requirements for PRMs are specified by NMi.  See Subsection 2.1.2.1 for GMIS recertification
requirements.

2.1.3  Using the Protocol

       The assay/certification protocol described here is designed to minimize both systematic and
random errors in the assay process. Therefore, the protocol should be carried out exactly as it is
described. The assay procedures in this protocol include one or more possible designs for the
assay apparatus.  The analyst is not required to use these designs and may use alternative
components  and configurations that produce equivalent-quality measurements.   Inert materials
(e.g., Teflon®, stainless steel, or glass) and clean, noncontaminating components should be used
in those portions of the apparatus that are in contact with the gas mixtures being assayed.

2.1.4  Certification Documentation

       Each certified compressed gas  calibration  standard must be documented in a written
certification report and this report must contain at least the  following information:

       1.   Cylinder identification number (e.g., stamped cylinder number).

       2.   Certified concentration for the compressed gas calibration standard, in parts per million
           by mole or mole percent. This value should be reported to 3 significant digits. The
           certified concentration is the mean of all  assayed concentrations  for which the
           candidate standard is considered to be stable.

       3.   Balance gas in the gas mixture.

       4.   Cylinder pressure at certification and the statement that the standard should not be
           used when its gas pressure is below 1.0 megapascals (i.e., 150 psig).

       5.   Date of the assay/certification.

       6.   Certification expiration date (i.e., the certification date plus the certification period) (see
           Subsection 2.1.6.3).
                                          2-5

-------
       7.  Identification of the reference standard used in the assay:, NIST SRM number, NIST
           sample number, cylinder identification number and certified concentration for an SRM;
 •   :: ,,,   v cylinder identification number and certified concentration for an SRM-equivalent PRM,
         • ,;,:* NTRM, pr a GMIS.  The certification documentation must identify the reference
         :  standard as being an SRM, an SRM-equivalent PRM, a NTRM, or a GMIS.

       8,  Statement that the assay/certification was performed according to this protocol and that
      •.-.;.  lists the assay procedure (e.g., Procedure G1) used.

       9.  The analytical method that was used in the assay.

      10.  Identification of the specialty gas producer or other laboratory (i.e., name and location)
           where the standard was assayed and certified. This identification must be given in the
           same orjarger font as the other required information in the report.
  *.**.  .             '
      11.  Chronological record of all certifications for the standard.

      12.  If applicable, statement that the certified concentration of specified component has been
           corrected for analytical interferences from other specified components.

      13.  An estimate of the total uncertainty associated with the assay of the candidate standard.
           This estimate must include the uncertainties of the reference standards, the analyzer
           multipoint calibration, and any interference correction.  Use the spreadsheet described
           in Appendix C or equivalent statistical techniques to calculate the total uncertainty.

       This certification documentation  must be given to the purchaser of the standard.  The
specialty gas producer must maintain laboratory records and certification documentation for 3 years
after the standard's certification date. A specialty gas producer or other vendor may redocument
an assayed and certified standard that it has purchased from another specialty gas producer and
that it wishes to sell to a third party.  However, the new certification documentation must clearly
identify the specialty gas producer or other laboratory (i.e., name and location) where the standard
was assayed.

2.1.5   Certification Label

      A label or tag bearing the information described in items 1-6,8, and 10 of Subsection 2.1.4
must be attached to the standard.

2.1.6   Assay/Certification of Compressed Gas Calibration Standards

2.1.6.1 Incubation of Newly Prepared Compressed Gas Calibration Standards—
       Newly prepared compressed gas calibration standards must be incubated at least 4  days
before being assayed and certified.

2.1.6.2 Stability Test for Reactive Gas Mixtures-
       Compressed gas calibration standard? that contain reactive gas mixtures,  including
hydrogen sulfide (H2S), nitric oxide (NO), oxides of nitrogen (NOJ, sulfur dioxide (SO2), and carbon
monoxide (CO), and that have not been previously certified, must be tested for stability as discussed
herein. Conduct an initial assay of the candidate standard and determine a concentration for the

                                          2-6

-------
 standard.  The candidate standard must be measured at least three times during the assay.
 Reassay the standard at least 7 days after the first assay.  Use the spreadsheet described in
 Appendix C or equivalent statistical techniques to compare the 95-percent confidence limits for
 the two assays.  If the confidence limits overlap, the candidate standard can be considered to
 be stable and can be certified.  In the spreadsheet, all cells in the comparisons table will be
 "true."  If the  confidence intervals do not overlap, the candidate standard may be unstable or
 there may be analytical problems associated with the assays or the reference standards.  The
 analyst must  wait an additional 7 days or more and conduct a third  assay.  If the confidence
 interval for the third assay overlaps either of the two previous assays, the candidate standard
 can be certified  using the data from the two  overlapping assays to determine the  certified
 concentration and the total uncertainty.  The analyst must disqualify  the candidate standard if
 none of the three confidence intervals overlap.   The analyst should investigate the cause of the
 lack  of agreement among  the  three  assays  and should  correct any  problems  that  are
 discovered.

 2.1.6.3 Certification Periods for Compressed Gas Calibration Standards—
       The certification of a compressed gas calibration standard is valid for only a specified
 period following  its certification date,  which is the date of its last assay.  In general,  the
 certification period should be no longer than the period for which similar standards have been
 shown to be stable.12"14  The certification periods for various standards are specified in Table
 2-3.  These certification periods are for standards that are contained in aluminum cylinders.  In
 general, the certification period for standards that are contained in nonaluminum cylinders is 6
 months.  However, an exception is made for the following three gas  mixtures: carbon dioxide with
 a  concentration  >0.5 percent; oxygen with a concentration >0.8 percent; and propane with a
 concentration >0.1  percent.  The certification period for standards containing these three gas
 mixtures in nonaluminum cylinders is given in Table 2-3.

       If a standard is to be used after its certification period has ended, it must be recertified in
 accordance with  this protocol.  The recertification assay must be performed using the same
 analytical  procedure (e.g., Procedure G1) as was used for the original assay of the standard. The
 purpose of this assay is to determine whether the standard has remained stable since its original
 certification. The standard must be measured at least three times during the recertification assay.
A second assay is not needed for recertification 'of the standard. There is no requirement that the
same reference standard must be used in the  original and recertification assays, although this
practice is desirable if possible. Record the results of the recertification assay in the laboratory's
records.  Use the spreadsheet described in Appendix C or equivalent statistical techniques to
compare the confidence  limits for the recertification assay with those for the previous assays.
If the confidence limits overlap, the  standard can be recertified.  The second certification period
for the standard is the same as that given in Table 2-3.

       A standard that was certified under this protocol may be recertified by a laboratory other
than the one that performed the original certification. In such a case, the 95-percent confidence
limits for the recertification assay  must overlap the certified  concentration  plus or minus  the
total uncertainty that was given  in  the original  certification documentation.   If the confidence
limits do not overlap, a second  recertification assay must be conducted and the  confidence
limits for the two recertification assays  must overlap before the  standard can be recertified.
The  recertification  documentation  must -list the  information from  the original  certification
documentation plus the corresponding information  from  the  recertification  assays.   Both  the
original and  the  recertification   laboratories  must  be   identified  in  the  recertification
documentation.

                                          2-7

-------
             TABLE 2-3. CERTIFICATION PERIODS FOR COMPRESSED GAS
             n CALIBRATION STANDARDS IN ALUMINUM CYLINDERS THAT
                       ARE CERTIFIED UNDER THIS PROTOCOL
                                                        Applicable
                                                      concentration
Certification
Certif ietf components
Ambient nonmethane organics (15
components)
Ambient toxic organics (19 components)
Aromatic organic gases
Carbon dioxide
Carbon monoxide
Hydrogen sulfide
Methane
Nitric oxide
Nitrous oxide
Oxides of nitrogen
(i.e., sum of nitrogen dioxide
and nitric acid)
Oxygen
Propane
Sulfur dioxide
Sulfur dioxide
Multicomponent mixtures
Mixtures with lower concentrations
Balance gas
Nitrogen
Nitrogen
Nitrogen
Nitrogen or air*
Nitrogen or air
Nitrogen
Nitrogen or air
Oxygen-free
nitrogen"
Air
Air
Nitrogen
Nitrogen or air
Nitrogen or air
Nitrogen or air
—
^—
range
5ppb
5ppb
20.25 ppm
2300 ppm
28 ppm
24 ppm
2! ppm
s4 ppm
2300 ppb
280 ppm
20.8%
21 ppm
40 to 499 ppm
2500 ppm
—
— —
period (months)
24
24
36
36
36
12
36
24
36
24
36
36
24
36
See text
See text
' When used as a balance gas, "air* is defined as a mixture of oxygen and nitrogen where the minimum
 concentration of oxygen is 10 percent and the concentration of nitrogen is greater than 60 percent.
" Oxygen-free nitrogen contains <0.5 ppm of oxygen.
       The spreadsheet described in Appendix C to calculate the total analytical uncertainty of a
candidate standard has provision for data from only three assays.  If more than three assays are
conducted, only the data from the three most recent assays should be used in the spreadsheet.

       The certification periods given in Table 2-3 apply to specific concentration ranges over which
the gas  mixtures have been found to be stable.  These concentration ranges match the
concentration ranges for NIST SRMs. The protocol described here allows the certification of
standards with concentration:  -nat may be lower than those of the corresponding SRMs.  If the
concentration of the standard    ss than the applicable concentration range given in Table 2-3, the
initial certification period for •    -tandard is 6 months.  After this period, the standard must be
recertified before further use    e standard must be measured at least three times during the
recertification assay.  If the ,    dence limits fot the recertification  assay overlap those for the
previous assays^ the standar   •  be recertified for the period shown in Table 2-3. For example,
                                         2-8

-------
a 35-ppm sulfur dioxide in nitrogen standard will have an initial certification period of 6 months.  After
a successful recertification, this standard will have a recertification period of 24 months.

        If the confidence limits from the recertification assay do not overlap those from the original
assays, the analyst must either disqualify the standard for further use  under this protocol or
investigate why there is an apparent difference between the original assays and the recertification
assay.  This difference  may  be  due to an actual instability of the gas mixture, to a reference
standard problem, to an analytical instrumentation problem, or to some other problem. If the analyst
can find a reasonable explanation for the difference and if this cause is not instability, then the
standard can be recertified. The analyst must append a brief report on the investigation to the
recertification documentation and to the laboratory's records.

        A multiple-component standard can be certified for a period equal to that of its most briefly
certifiable component. For example, a standard containing sulfur dioxide, carbon monoxide, and
propane in nitrogen can be certified for 24 months because the shortest certification period  is 24
months.

        A standard may be recertified if the gas pressure remaining in the cylinder is greater than
3.4 megapascals (i.e., 500 psig).

2.1.6.4  Minimum Cylinder Pressure—
        In general,  a compressed gas  calibration standard should not be  used when its gas
pressure is below 1.0 megapascals (i.e., 150 psig). NIST has found that some gas mixtures (e.g.,
nitric oxide in nitrogen) have exhibited a concentration change when the cylinder pressure fell below
this value.

2.1.6.5  Assay/Certification of Multicomponent Compressed Gas Calibration Standards—
       This protocol may be used to assay and  certify a multiple-component standard  if
compressed gas SRMs, SRM-equivalent PRMs, or  NTRMs exist that  contain  the individual
components of the  multiple-component standard.  If any component in the multiple-component
standard interferes with the assay of any other component, the analyst must conduct an interference
study to determine an interference correction equation. This study must be conducted using the
same analyzer or analyzers as will be used to assay the standard. The study must use single-
component  and  multiple-component reference  standards that have  been  assayed  using
interference-free analyzers..  The study must cover the same range of  concentrations for all
components as will exist for the standards being assayed and certified according to this protocol.

Data from the interference study must be evaluated using multiple-variable least-squares regression
analysis. The analyst should consult with a statistician before beginning the study or evaluating its
data.  The regression analysis must produce an interference correction equation and an estimate
of the 95-percent uncertainty associated with the corrected concentrations for the assayed
components. The interference correction equation will be valid for the range of concentrations
covered in the study for which the uncertainty of the corrected concentration is ^1  percent of the
corrected concentration.  The analyst must add the interference correction uncertainty to the total
uncertainty of the standard.  The certification documentation must include a statement that the
certified concentration of a specified component has been corrected for interferences from other
specified components.  An interference study is not needed if the assay analyzer is interference
free.
                                          2-9

-------
2.1.7  Analyzer Calibration

2.1.7.1  Basic Analyzer Calibration Requirements—
       The assay procedures described In this protocol employ a data'reduction technique to
calculate the concentration of a candidate compressed gas standard that corrects for minor analyzer
calibration., variations (i.e., drift).  This technique does not require the absolute accuracy of the
analyzer's calibration curve at the time of the assay. Requirements for the analyzer follow:  (1) it
must have a, well-characterized  calibration  curve for the pollutant of interest (see Subsection
2.1,7.2); (2)tit must have good resolution and low noise; (3) its calibration must be known and must
be reasonably  stable or  recoverable  during  the assay/certification process; and (4) all
measurements of candidate standards must fall within a well-characterized region of its calibration
curve.

2.1.7.2 Analyzer Multipoint Calibration—
       The gas analyzer used for the assay must have had a multipoint calibration within 1 month
prior to the assay date. This calibration is not used directly to interpret analyzer response during
the assay of the candidate compressed gas calibration standard. The data reduction technique
corrects the analyzer response on the assay date for any  minor calibration drift during the period
between the multipoint calibration and the assay date. The corrected analyzer response is then
used with the multipoint calibration data to calculate a  concentration value  for the candidate
standard.

       The analyzer's zero and span controls may be adjusted before the start of the multipoint
calibration. If a zero or span adjustment is made, allow the analyzer to stabilize for at least one hour
before beginning the multipoint  calibration.  The waiting period  is necessary because some
analyzers' calibrations drift for a period of time following a  zero or span control adjustment.

       The multipoint calibration must consist of one or more measurements of the analyzer
responses to at least five different concentrations.  The  use of a zero gas in the calibration is
recommended, but is not required. Record these measurements and the analyzer's zero and span
control settings in  the laboratory's records.   These   calibration  concentrations  should be
approximately evenly spaced over the concentration range. The multipoint calibration is valid only
for the concentration range lying between  the largest and smallest measured concentrations. The
concentrations may be produced by undiluted reference standards or by dilution of reference
standards using a gas dilution system.  See Subsection 2.1.7.4 for reference standard requirements.
If a gas dilution system is used, it  must have a specified accuracy of no worse than 1.6 percent of
the undiluted reference standard concentration.  The accuracy of the gas dilution system must be
checked by the analyst at monthly intervals by comparing diluted reference standards to undiluted
reference standards having approximately the same concentration.

       If the analyzer has multiple concentration ranges, a multipoint calibration should be done
for at1 ranges that will be used later for the assay of candidate standards. A multipoint calibration
tha-  s conducted on one range is not valid for an assay that is conducted on another range.

       Data from tht multipoint calibration must be evaluated using least-squares regression
anavsis.15 This statisi;cal analysis technique will be used to determine the analyzer's calibration
curve and to characterize the uncertainty associated with the'calibration. The concentration values
are the independent (i.e., X) values in the analysis and their units may be parts per million, mole
percent, or any other appropriate units. The analyzer response values are the dependent (i.e., Y)

                                          2-10

-------
values  in the analysis  and their units may be volts, millivolts, percent of scale  or any other
measurable analyzer response units.  The analyzer response values should have a resolution of
less than or equal to 1 percent of the maximum measured analyzer response.

        Because an analyzer's response has a random error component, repeated measurements
of the same reference standard will not produce identical analyzer responses. The analyst may
investigate the analyzer's precision by making  replicate measurements of multiple  reference
standards. Least-squares regression analysis is normally conducted under the assumption that the
precision is the same at all concentrations. However, this assumption may not be true for real-world
analyzers and the analyst may need to use alternate statistical procedures to analyze the multipoint
calibration data.

        Calculate the least-squares regression coefficients of the calibration equation using the
spreadsheets described in Appendix A or using equivalent statistical techniques (e.g., the worksheet
for linear relationships given in Chapter 5 of Reference 15). The spreadsheets allow the multipoint
calibration data to be fitted to straight-line, quadratic, cubic, or quartic linear regression models.
EPA discourages the use of the cubic and quartic models and believes that better fits of the data
can be obtained by performing multipoint calibrations over more limited concentration ranges and
by using straight-line or quadratic models.  Inclusion of cubic  and  quartic models  in the
spreadsheets is for experimental use or for situations in which there is a theoretical basis for the use
of such higher-order models. Analysts should be aware that apparent higher-order calibration
curves may be caused by artifacts such as inaccurate reference standards or leaks in a gas dilution
system.  They should not use higher-order regression models to fit multipoint calibration data that
have inadequate precision and that should be fitted to lower-order regression models. If analysts
suspect that the precision is inadequate, they should make replicate measurements at each different
concentration. Additionally, a multipoint calibration should not change regression model orders from
one month to the next.

       The spreadsheet described in Appendix A will suggest the best regression  model for the
multipoint calibration data, but the analyst should choose the model that best fits the measurement
process on theoretical grounds.

       Plot the values from the multipoint calibration and the regression curve with confidence
bands as shown in Figure 2-1. These plots will provide a graphical representation of the calibration
and will permit a qualitative assessment of the uncertainty associated with the calibration.  Record
the regression coefficients and other statistical results in the laboratory's records.

       However, a quantitative assessment of the calibration's uncertainty is needed to allow the
analyst to determine whether the multipoint calibration data adequately characterizes the "true"
calibration curve  for the analyzer.  The  criterion to be used to evaluate the uncertainty of the
multipoint calibration is the 95-percent uncertainty (i.e., a = 0.05) for a concentration predicted from
the regression line using measured values of the analyzer response.   This 95-percent uncertainty
for the predicted concentration can be calculated using the spreadsheets described in Appendix A
or using equivalent statistical techniques. Record the uncertainty calculations in the laboratory's
records.  A multipoint calibration will be considered  to be well-characterized for all concentrations
that are within the range of the multipoint calibration measurements and for which the magnitude
of the 95-percent confidence limits for the regression-predicted analyzer response are s±1  percent
of the measured response for the largest concentration in the multipoint calibration.  For example,
assume that a multipoint calibration was conducted between 0 and 100 ppm and that the measured

                                           2-11

-------
                       Calibration Points
                   *** Uppor Confidooco Band
                   — Estimated CaBxation Line
                       Lower Confidence Band
                      10     20     30    40    SO    60
                                          Concentration
70
80
90    100
                  Figure 2-1. Example regression curve and confidence
                            bands from multipoint calibration.

responses ranged between 0 and  10 volts.   The calibration  is well-characterized  for  all
concentrations for which the 95-percent confidence limits are  £±0.1 volt, which is equal to ±1
percent of 10 volts. Step 4 of the spreadsheet described In Appendix A allows the analyst to enter
var.ous concentrations and obtain the corresponding regression-predicted analyzer response and
con'idence limits.

       In effect, the .-5-percent uncertainty value is a measure of how well the multipoint calibration
da;, Mt an equation *  nich the analyst assumes is the "true" calibration equation for the analyzer.
Corr.parison of uncertainty values from straight-line and quadratic equations permits the analyst to
select the equation that best represents the calibration data.

       A multipoint calibration may fail to  meet this uncertainty criterion for  several possible
reasons:

       •   inadequate analytical precision;

       •   inaccuracy of the reference standards or the gas dilution system; or

           excessive uncertainty in the analyzer's calibration equation due to incorrect assump-
           tions about the form of the equation.
                                           2-12

-------
        The effect of inadequate analytical precision can be reduced by increasing the number of
 replicate measurements at each calibration concentration or by increasing the number of different
 concentrations used in the multipoint calibration.  Additionally, precision can be improved by using
 an averaged analyzer response, rather than  an instantaneous  analyzer  response, for each
 measurement. Reference standard inaccuracy  is reduced by using the most accurate reference
 standards that are  available.  An inaccurate gas dilution system can be detected by comparing
 measurements of the concentration of a diluted reference standard to  the theoretically equal
 concentration of  another, undiluted reference standard.   It can also be detected by comparing
 measurements of two theoretically equal concentrations obtained by dilution of two  reference
 standards having significantly different concentrations.  An inaccurate gas dilution system must not
 be used for the  multipoint calibration.  The effect of excessive uncertainty in a straight-line
 calibration equation can be eliminated by using a quadratic calibration equation or  by transforming
 the calibration'data mathematically so that they may be fitted to a straight line regression equation.
 See Subsection 2.1.7.5 for a discussion of such mathematical transformations.

        Note that possibly a more restrictive uncertainty criterion applies for the assay of the
 candidate  standard.   The  magnitude  of the  95-percent confidence limits for the estimated
 concentration of the candidate standard must be  s±1 percent of the concentration of the reference
 standard  (see Subsections 2.2.2 and 2.32).  For example, assume that a 70-ppm candidate
 standard is being assayed using a 50-ppm reference standard.  The 95-percent confidence limits
 for the candidate  standard's estimated concentration must be £±0.5 ppm.

 2.1.7.3 Zero and Span Gas Checks—
        On any day after the multipoint calibration that the analyzer will be used for the assay of a
 candidate  standard, its calibration drift must be  measured. This drift is calculated relative to the
 analyzer response during the multipoint calibration. The purpose of the zero and span gas checks
 is to verify that the calibration drift has remained within acceptable  limits since the  multipoint
 calibration. The  criterion that is used to assess the drift is the relative difference between the
 analyzer's current response and the  corresponding value from the multipoint calibration.  The
following equation is used for this calculation:
             Relative Difference = 100
                                      Current Response   Calibration Response
                                     Calibration Response for Reference Standard

       This calculation is performed in Step 6 of the spreadsheet described in Appendix A.

       Note that the relative difference is always calculated relative to the calibration response for
the reference standard, even when the zero gas is being measured. This calculation is performed
for the zero gas measurements and for the reference standard measurements.  If the reference
standard was not measured during the multipoint calibration, use the regression-predicted response
for a concentration equal to that of the reference standard.

       If the relative differences for the zero and span gas checks are each less than or equal to
5.0 percent, the analyzer's current calibration'is considered to be approximately the same as during
the multipoint calibration and the assay may be conducted. The zero and span controls do not have
to be adjusted following the zero and span checks because the data reduction technique used in
this protocol does not depend on the absolute accuracy of the analyzer calibration equation at the
time of the assay.

                                           2-13

-------
        If the relative differences forifie'^zero or span gas checks are greater than 5.0 percent, the
 analyzer is considered to be out ofj&tibration.  A new multipoint calibration may be conducted
 before the'candidate standard is assayed or the analyzer's z&o'and; span controls may be adjusted
 to return the analyzer's response to the original calibration levels! For some analyzers such as
 nondispersive infrared instruments,  daily changes in environmental variables such as barometric
 pressure may shift the calibration. After any adjustment of controls, the analyst should repeat the
 zero and span gas checks and recalculate the relative differences to verify that the analyzer is in
 calibration.

       The zero gas and reference standard measurements that are performed for the assay of the
 candidate standard may also be used for the zero and span gas checks.
                         . i'i                        ,.?•'-'
       Between the time of the multipoint calibration and the time of the zero and span gas checks,
 the analyst may adjust the analyzer's zero and span controls for assays that will hot be certified
 according to this protocol.   However,  these controls must be returned to their settings at the
 multipoint calibration before the zero and span gas checks or assays under this protocol.

 2.1.7.4 Reference Standards for Multipoint Calibrations and Zero and Span Gas Checks—
       The reference standards for the  multipoint calibration and for the span gas checks must be
 diluted or undiluted SRMs, SRM-equivalent PRMs, NTRMs, or GMISs as specified in Subsection
 2.1.2. The reference standard for the span gas check need not be the same as one of those used
 for the multipoint calibration or for the assay of the candidate standard.

       Pure gases may be diluted to prepare gas mixtures for use in multipoint calibrations, but
 such mixtures may not be used as the reference standards for the span gas check or for the assay
 of the candidate standard.  Pure gases  may not be diluted by more than a factor of 100.

       The zero gas must meet the requirements in Subsection 2.1.9.  For some analyzers such
 as gas chromatographs, the analyst may have reason to believe that the zero gas reading may not
 accurately represent the zero-intercept of the calibration equation. The analyst may substitute a
 low-concentration, NIST-traceable  reference standard for the  zero  gas, providing  that  the
concentration of this standard is less than the concentration of the candidate standard.

2.1.7.5 Uncertainty of the Calibration Curve—
       The data reduction technique used in this protocol is based on the assumption that the
analyzer has a well-characterized calibration curve. The accuracy of the certified concentration of
a candidate compressed gas calibration standard is dependent upon this assumption.  The analyst
cannot assume that the analyzer's calibration curve is a straight line between the measured values
for the zero gas and the reference standard. The analyst must calculate the calibration equation
and the uncertainty for its predicted concentrations by statistical analysis of the measurements
obtained during the multipoint calibration.

       The total uncertainty of the certified concentration for a candidate standard is composed of
several  components.  The first component is the  accuracy  associated with the certified
 concentration of the reference standard. This uncertainty is minimized by using an SRM, an SRM-
 equivalent PRM, a NTRM, or a GMIS as the reference standard.  The second component is the
 precision of the measurements of the  reference and candidate standards.  This uncertainty is
 minimized- by making replicate measurements of these standards.  The third component is the
 uncertainty associated with the concentrations that are predicted from  the analyzer's calibration

                                          2-14

-------
curve. This uncertainty concerns whether an assumed calibration equation accurately represents
the "true" calibration curve.

        This third component of uncertainty does not exist if the concentrations of the reference and
candidate standards are equal. The assumed calibration equation and the true calibration curve
will pass through the data for the reference standard regardless of whether they diverge elsewhere
and the equation will be accurate for that single concentration.  However, the uncertainty does exist
if the concentrations of the reference  and  candidate standards differ. The assumed and true
calibration curves may pass through different points for concentrations not equal to that of the
reference standard.  Analytical errors will develop because of this difference.

        The measure of this uncertainty that is most directly useful to the analyst is the 95-percent
uncertainty for a regression-predicted concentration given  one or more measurements of the
candidate standard.  The uncertainty may be  calculated using the spreadsheet described in
Appendix A or using equivalent statistical techniques. Several points should be noted about this
uncertainty value.  First, its magnitude decreases  as  n increases where n is  the number of
measurements in the multipoint calibration.  Second, its magnitude decreases as n' increases,
where n' is the number of measurements of the candidate standard.  Third, its magnitude increases
as the mean measured analyzer response (y') for the candidate standard diverges from the overall
mean measured analyzer response (y) for the multipoint calibration. These points mean that it
becomes easier to satisfy the uncertainty criterion as one increases  the number of measurements
in the multipoint calibration and in the assay of the candidate standard. Additionally, the absolute
uncertainty of the regression predicted concentration is larger at the extremes of the calibrated
concentration range than at the middle of the range.

        For analyzers having an inherently nonlinear, but precise response, the calibration equation
can be calculated using quadratic or higher-order polynomial  regression analysis.  Alternatively, a
nonlinear equation may be linearized with a simple mathematical transformation of the multipoint
calibration data. Examples of some linearizing  transformations are given in Table 2-4, which is
reproduced from Reference 15.  The multipoint calibration  data may need to undergo  several
different transformations  before the optimum transformation is determined. Using appropriately
transformed calibration data, a calibration equation can be calculated with an acceptable 95-percent
uncertainty for the regression-predicted concentration. Subsequently, data obtained from the assay
of the candidate .standard must  be similarly transformed to calculate a concentration for the
candidate standard.

2.1.8    Uncertainty of the Estimated Concentration of the Candidate Standard

        Uncertainty in the concentration estimated for a candidate standard is due to many different
sources, including uncertainty  in the reference standards, uncertainty in the analyzer multipoint
calibration, uncertainty in the zero/span correction factors, and measurement imprecision. Some
of these sources can be assessed using standard statistical techniques, but others cannot be
assessed with the limited data that are produced when implementing this protocol.

        For those cases where the candidate standard is assayed at the same time as the multipoint
calibration, the candidate standard's concentration is determined directly from the calibration curve.
The total uncertainty of the concentration is calculated by using the spreadsheets described in
                                          2-15

-------
           TABLE 2-4. SOME LINEARIZING TRANSFORMATIONS
                 FOR MULTIPOINT CALIBRATION DATA
If the relationship
is of the form:
•-
b
Y = a * -
X

•i
Y

a + bX'
or
4
- - a + bX
Y

Y- X
a + bX
Y ab*
Y = ae bx
aXb

Y = a + bXn,
where n is known
STEP ONE:
Plot the
transformed
calibration data
V Y
TT~ AT~-
1
Y X
.


Y X





X X
Y
log Y X
log Y X
log Y '°9 X

Y X"

STEP TWO:
Fit the
straight line
VT b0.b,XT
Use the normal
procedures for
calculating the
regression line using
the transformed
calibration data.
Calculate the 95-
percent uncertainties
for the predicted
transformed
concentrations and
compare them to the
uncertainty criterion.







STEP THREE:
Convert straight line
constants (bc and b,)
to original
constants:
bo = b, -
i. .
a b




a b




au
D
log a log b
log a b log e
log a b
r
a b

Source: Reference 15.
                               2-16

-------
Appendices A and C or equivalent statistical techniques. It combines the uncertainty of the assay
with the uncertainty of the reference standard using the following equation:
2 +
                  UncertaintyTOTAL = /(Uncertainty^)

       For those cases where the candidate standard is assayed on a date following the multipoint
calibration, the spreadsheet includes the uncertainty associated with the zero gas and reference
standard measurements in the calculation of total uncertainty.

       If an interference-correction equation has been used to obtain a corrected concentration for
the candidate standard, the 95-percent uncertainty for the corrected concentration must be included
in the assessment of the total analytical uncertainty of the candidate standard's concentration using
the following equation:
        UncertaintyTOTAL = /(Uncertainty^Y)2 + (^cer^n^co^^an)2 + (Unc*rtaintyCTANDARD)2
       The analyst may report the total analytical uncertainty of the candidate standard's certified
concentration on the certification documentation or may report this estimate as a percentage that
is relative to the certified concentration using the following equation:
                UncertaintyRaATWE = 100 [ UncertaintyTOTAL / Certified Concentration]  .


2.1.9  Zero Gas

       Zero gas used for zero gas checks or for dilution of any candidate or reference standard
should be clean, dry, zero-grade air or nitrogen containing no detectable concentration of the
pollutant of interest. It should match the balance gas in the candidate standard and the reference
standard, unless it has been demonstrated that the analyzer is insensitive to differences in the
balance gas composition.  The zero gas also should contain no contaminant that causes a
detectable response on the analyzer or that suppresses or enhances the analyzer's response. The
oxygen content of zero  air should be approximately that  of  ambient  air,  unless  it has been
demonstrated that varying the oxygen content does not suppress or  enhance the analyzer's
response. The water vapor concentration in the zero gas should be less than 5 ppm.

       The  analyst may substitute a low-concentration, NIST-traceable reference standard for the
zero gas in zero gas checks and assays if there is reason to believe that the zero gas reading may
not accurately represent the zero-intercept of the calibration equation.

2.1.10 Accuracy Assessment of Commercially Available Standards

       Periodically, the U.S. EPA will assess the accuracy of compressed gas calibration standards
that have been assayed and certified according to this protocol. The accuracy of representative
standards will be assessed by EPA audits. The audit results, identifying the specialty gas producers
or other analytical laboratories that assayed and certified the standards, will be published as public
information.  A summary of EPA's audit results from 1992 to the present is available as a
WordPerfect 6.1 file at EPA's Ambient Monitoring Technology Information Center (Internet Address:
                                           2-17

-------
 http://www.epa.gov/ttn).   This  document can  be found in the Directory of TTNWeb Sites/
 AMTIC/Publications/QRD-NERL Docurg^its.

 22    PROCEDURE G1: ASSAY AND CERTIFICATION OF A COMPRESSED GAS
        CALIBRATION STANDARD WITHOUT DILUTION
                                                •
 2.2.1   Applicability

        This procedure may be  used to assay the concentration of a candidate compressed gas
 calibration standard, based on the concentration of a compressed gas reference standard of the
 same gas mixture. This procedure allows a specialty gas producer, a standard user, or other
 analytical laboratory to certify that the assayed concentration for the candidate standard is traceable
 to  the reference standard.  The procedure employs a pollutant gas  analyzer to compare the
 candidate and* reference standards' concentrations by direct measurement without dilution of either
 gas.

        This procedure may be used for the assay of more than one candidate standard during the
 same assay session. Criteria that apply to the assay of one candidate standard apply to the assay
 of multiple candidate standards.

 2.2.2   Limitations              :

        The concentration of the candidate standard may be greater than or lesser than the
 concentration of the reference standard.  However, both concentrations must lie within the well-
 characterized region  of the multipoint calibration (see Subsection 2.1.7.2).  Additionally, the
 magnitude of the 95-percent confidence limits for the estimated concentration of the  candidate
 standard must s±1 percent of the reference standard concentration. This criterion may be more
 restrictive than the corresponding criterion for the multipoint calibration, but it allows the analyst
 greater flexibility in the selection of a reference  standard for the assay of a particular candidate
 standard. For example, assume that a 70-ppm candidate standard is being assayed using a 50-
 ppm  reference standard' and that the analyzer's calibration was found to be well-characterized
 between 20 and 80 ppm. The 95-percent confidence limits for the candidate standard's estimated
 concentration must be less than or equal to ±0.5 ppm.

       The balance gas must be the same in  both the candidate standard and  the reference
 standard, unless it has been demonstrated that the analyzer's response is insensitive to differences
 in the balance gas composition.

 2.2.3   Assay Apparatus

        Figure 2-2 illustrates one possible design of apparatus for the assay of compressed gas
 calibration standards without dilution.  This apparatus is designed to allow the convenient routing
 of the gas mixtures to the pollutant gas analyzer. The gas mixture to be measured is selected by
 rotation of two three-way valves (i.e., V1 and V2). Pressure regulators and gas flow controllers (i.e.,
 C1  and C2) control the flow rates from the individual cylinders.  The gas flow controllers may be
 needle values, capillary tubes, thermal mass flow controllers, or other flow control devices. The gas
 mixtures are routed to the analyzer through a union tee tube fitting. Gas in excess of the analyzer's
 demand is vented, which helps to ensure that the gas entering the analyzer is at near-ambient
 pressure. Normally, the excess gas is vented to the atmosphere without any obstructions in the
 tubing. However, the excess gas can be routed through an uncalibrated rotameter by rotation of
 a three-way valve (i.e., V3). The rotameter is  used to demonstrate that the total gas flow rate
 exceeds the sample flow rate of the analyzer and that no room air is being drawn in through the vent
. line.

                                         2-18

-------
                                                                                        t
                                                                                              Gas Flow
                                                                                               to Vent
                       Pressure
                       Regulator
Three-way
Valve (V1)
CD
                                                          Pressure
Gas Flow
Controller
  (C1)
                                                 I
                                                                                                        o
Three-way  in-
Valve fV3) -
                                                      Rotameter
11UI
s??\
?]y
) 	
(C2)




                                  Three-way
                                  Valve (V2)
                                                         Gas Flow to
                                                          Analyzer
                Candidate
                Standard
                                 Reference
                                 Standard
                                                     Zero
                                                     Gas
                             Figure 2-2.  One possible design of the apparatus for the assay of compressed
                                        gas calibration standards without dilution (Procedure G1)

-------
        The apparatus may be modified in several ways that will not diminish its performance.  The
two three-way valves could be replaced by solenoid valves or by a single four-way valve with three
input ports and one output port.  Alternatively, a single length of tubing with a gas flow controller
could be connected manually to individual cylinders' pressure regulators in succession. See  also
Subsection 2.1.3.                     -  -.-....

2.2.4   Pollutant Gas Analyzer

        The pollutant gas analyzer must have a well-characterized calibration curve and must be
capable of measuring directly the concentration of both the candidate and the reference standards
without dilution.  See Subsection 2.1.7.1. It must have good resolution, good precision  a stable
response, and low output signal noise. In addition, the analyzer should have good specificity for the
pollutant of interest so that it has no detectable response to any other component or contaminant
that may be contained  in either  the candidate or reference standards.  If any component  in a
multiple-component standard interferes with the assay of any other component,  the analyst must
conduct an interference study to determine an interference correction equation. If the candidate and
reference standards contain dissimilar balance gases (e.g., air versus nitrogen or different  pro-
portions of oxygen in the balance air), it must have been demonstrated that the analyzer's response
is not sensitive  to differences in the balance gas  composition.  This demonstration can be
accomplished by showing that no difference exists in the analyzer's response when measuring a
compressed gas calibration standard  that  has been diluted with identical flow rates of different
balance gases.

        The analyzer should be connected to a  high-precision data acquisition system (e.g., a strip
chart recorder), which must produce an electronic or paper record of the analyzer's response during
the assay. A high-precision digital panel meter,  a digital voltmeter, a data logger or some other data
acquisition system with four-digit resolution can be used to obtain numerical values of the analyzer's
response. More precise  values will be obtained if this system has a data-averaging capability.  The
assay record must be maintained for 3 years after the standard's certification date.

        If the analyzer has not been in continuous operation, turn it on and allow it to stabilize (e.g.,
for at least 12 hours) before beginning the measurements.

2.2.5   Analyzer Calibration

2.2.5.1  Multipoint Calibration-

        See  Subsections 2.1.7.2 and 2.1.7.4.

2.2.5.2  Analyzer Range—
        The  range  of the analyzer must include the concentrations of the zero gas, the candidate
standard and the reference standard. The concentrations of the candidate and reference standards
must fall within the well-characterized region of the analyzer's calibration curve.  In general, the
analyst should use a range that will produce the largest on-scale analyzer response.

2.2.5.3  Linearity—
        The data reduction technique used in this procedure requires that the analyzer have a well-
cr  -acterized, but not  necessarily  linear, calibration curve (see Subsection 2.1.7.5).  High-
er   . ntration-range analyzers of the type that are required for this procedure may not be inherently
lir    but in such cases they usually have a predictable, non-linear calibration  curve that can be
des jnDed by a polynomial  equation or can be mathematically transformed to produce a straight-line
calibration curve that is suitable for use in this procedure.  Any such polynomial equation  or
mathematical transfc -nation should be verified during the multipoint calibration. Caution should be

                                           2-20

-------
exercised in using a transformed calibration curve because zero or span control adjustments to the
analyzer may produce unexpected effects in the transformed calibration curve.

2.2.5.4  Zero and Span Gas Checks-
        See Subsections 2.1.7.3 and 2.1.7.4.  Prior to carrying out the assay of the candidate
standard, use zero and span gases to check for calibration drift in the analyzer since the multipoint
calibration.  Zero gas and span gas checks must be performed on any day after the multipoint
calibration that candidate standards are assayed.  If multiple assays are being performed on the
same analyzer range, the analyst needs to perform only a single set of zero gas and span gas
checks for this range. However, another set must be performed if the range is changed.

        The gas mixtures to be used during the zero and span gas checks need not be the same
as any of the reference standards used for the assay of the candidate standard or for the multipoint
calibration. The reference standard for the span gas check must be traceable to a NIST SRM, a
SRM-equivalent PRM, or an NTRM.   Information concerning  this  standard  (e.g.,  cylinder
identification number, certified concentration) must be recorded in the laboratory's records.

        A source  of clean, dry zero gas is recommended, but not required. The analyst may
substitute a low-concentration, NIST-traceable reference standard for the zero gas if there is reason
to believe that the zero  gas reading may not accurately represent the zero-intercept of the
calibration equation.

        Make three or more discrete measurements of the zero gas and three or more discrete
measurements of the reference standard. "Discrete" means that the analyst must change the gas
mixture  being sampled by the analyzer between measurements. For example, the analyst might
alternate between measurements of the reference standard and measurements of the zero gas.
Record  these measurements in the laboratory's records.

        Next, verify that the analyzer's precision is acceptable. Calculate the mean and standard
deviation of the analyzer's responses to the zero gas. Repeat the calculations for the reference
standard measurements. These calculations are performed in Step 6 of the spreadsheet described
in Appendix A.  Record these calculations in the laboratory's records. The standard error of the
mean for each set of measurements must be less than or equal to 1 .0 percent of the mean response
to the reference standard. That is,
                                             100
where
      s  = standard deviation of the analyzer's response;
      n  = the number of independent measurements of the gas mixture; and
         = the mean analyzer response to the reference standard.
       The value of the standard error of the mean can be made smaller by increasing the number
of measurements. This calculation will enable the analyst to determine how many replicate
measurements are needed during the assay of the candidate standard to obtain acceptable
precision. The analyst may wish to use a data logger or data acquisition system with averaging
capability to obtain more precise measurements.  If the value of the standard error of the mean is
not acceptable, then the analyzer must be repaired or another analyzer must be used for the assay.


                                         2-21

-------
        Next, verify that excessive calibration drift has not occurred since the multipoint calibration.
 For the zero gas measurements, calculate the relative difference (in percent) between the current
 mean analyzer response during the zero gas check and the  corresponding response that is
 predicted from the multipoint calibration regression equation. That is,
                    .   •-,';.  f  '.              •  .r

             Relative Difference   .100 [  ,CUrrent flesP°nse  \Calibration  ResPonse  1  .
                                  [  Calibration  Response for Reference  Standard J


       This calculation is. performed in Step 6 of the spreadsheet described in Appendix A.

       Note that the relative difference is always calculated relative to the calibration response for
 the reference standard, even when the zero gas is being measured. Repeat this calculation for the
 reference standard measurements.  Record these calculations in the laboratory's records. If the
 reference standard was not measured during the multipoint calibration, use the regression-predicted
 response for a concentration equal to that of the reference standard.

       Then, if the relative differences for the zero and span gas checks are less than or equal to
 5.0 percent, the analyzer is considered to be sufficiently in calibration.  The zero and span controls
 need not be adjusted and the assay may be conducted. The data reduction technique does not
 require the absolute accuracy-of the analyzer calibration.  Some minor calibration drift is acceptable
 because the effect of any drift will be corrected during the reduction of the assay data.

       However, if the relative difference for either the zero or the span gas check is greater than
 5.0 percent, then the analyzer is considered to be out of calibration.  A new multipoint calibration
 may be conducted before the candidate standard is assayed or the analyzer's zero and span
 controls may be adjusted to return the analyzer's response to the original calibration levels. For
 some analyzers such as  nondispersive infrared instruments,   daily changes in environmental
 variables such as barometric pressure may shift the calibration. After any adjustment of controls,
 the analyst should repeat the zero and span gas checks and recalculate the relative differences to
 verify that the analyzer is sufficiently in calibration. The analyzer will be considered  to be out of
 calibration if the relative differences remain greater than 5.0 percent.

       The zero gas and reference standard measurements that are performed for the assay of the
candidate standard may also be used for the zero and span gas checks.

2.2.6  Assay Gases

2.2.6.1 Candidate Standard—
       See Subsections 2.1.6 and 2.2.2.

2.2.6.2 Reference Standard—
       See Subsections  2.1.2 and 2.2.2.  The reference standard used for the assay of the
candidate standard must be a NIST SRM, an SRM-equivalent PRM, an NTRM or a GMIS.  This
standard need not be the same as any of the reference standards used for the span gas check or
for the  multipoint  calibration.   Information concerning the reference standard (e.g., cylinder
 identification number, certified concentration, etc.) must be recorded  in the laboratory's records.

       If the multipoint calibration data have been fitted to a linear (i.e., straight-line) regression
 model, then only a single reference standard need be measured during the assay of the candidate
 •andard.  If these data have been fitted to a quadratic or higher-order regression model, then at
  dst two reference s:andards must be measured.  One reference standard is adequate to determine
 ne slope of a linet  equation, but additional reference standards are .needed to determine the


                                          2-22

-------
curvature of quadratic and higher-order polynomial equations. The concentrations of the additional
reference standards should be located at the maximum difference between the polynomial curve
and the corresponding straight line between the zero gas and the highest-concentration reference
standard.

2.2.6.3 Zero Gas—
       See Subsection 2.1.9. A source of clean, dry zero gas is recommended, but not required.
The analyst may substitute a low-concentration, NIST-traceable reference standard for the zero gas
during zero gas checks and assays if there is reason to believe that the zero gas reading may not
accurately represent the zero-intercept of the calibration equation.  Information concerning the zero
gas should be recorded in the laboratory's records.

2.2.7  Assay Procedure

       1.  Verify that the assay apparatus is properly configured, as described in Subsection 2.2.3
           and shown in Figure 2-1. Inspect the analyzer to verify that it appears to be operating
           normally  and that all controls are set to their expected values.  Record these control
           values in the laboratory's records.

       2.  Verify that a multipoint calibration of the analyzer has been performed within 1  month
           prior to the assay date.  (See Subsections 2.1.7.2, 2.1.7.5 and 2.2.4). Additionally,-
           verify that the zero and span gas checks indicate that the analyzer is in calibration (see
           Subsection 2.2.5.4).  Finally, verify that the concentrations of the candidate and
           reference standards fall within the well-characterized region of the analyzer's calibration
           curve (see Subsection 2.2.2).

       3.   Measure and adjust the flow rates of the gas mixtures (i.e., reference standard(s),
           candidate standard, and zero gas) to approximately the same value that will provide
           enough flow for the analyzer and sufficient excess to assure that no ambient air will be
           drawn into the vent line.

       4.   In succession, measure the zero gas, the reference standard(s), and the candidate
           standard(s) using the analyzer.  Use valves V1  and V2 to select each of the gas
           mixtures for measurement.  For each measurement, allow ample time for the analyzer
           to achieve a stable response.  If the response for each measurement is  not stable, the
           precision of the measurements will decline and the candidate standard may  not be
           certifiable under this protocol.  Record the analyzer response for each measurement
           in the laboratory's records, using the same response units (e.g., volts, millivolts, percent
           of scale,  etc.) as was used for the multipoint calibration. At this point, do not convert
           these data into concentration values using the calibration equation. Do not perform any
           necessary mathematical transformation of these data. These steps will be done later.
           Do not make any zero control, span control, or other adjustments to the analyzer during
           these measurements.

           The analyst may assay multiple candidate standards during the  same assay session.
           For example, a single set of  measurements may involve a zero gas,  a reference
           standard  and three candidate standards.  Criteria that  apply  to the  assay of one
           candidate standard apply to the assay  of multiple candidate standards. The analyst
           should  be aware that the effect of any short-term calibration drift will be greater when
           multiple candidate  standards are assayed.  This greater  effect is due to the  longer
           period of time between reference standard measurements.  Unacceptable uncertainties
         '  of the estimated concentrations for the candidate standards  may occur as a result of the
           longer assay session.

                                         2-23

-------
       5,  Conduoicat least two additional sets of measurements, as described in step 4 above.
   .v -,  n   HoweraeMor these subsequent sets of measurements, change the order of the three
           ^measurements (e.g., measure reference standard, zero gas, and candidate standard
           for the second set and measure zero gas, candidate standard, and reference standard
           for the third set). Changing the order that the gas mixtures are measured helps the
           analyst to discover  any effect of that  one  measurement has on  subsequent
           measurements.  The number of sets of measurements will have been determined
           during analysis of .
The spreadsheet  described in Appendix  C or equivalent statistical techniques must be used to
evaluate the stability of the candidate standard and to calculate the overall estimated concentration
and the total uncertainty for the candidate standard.
                                         2-24

-------
        The stability is evaluated by comparing the 95-percent confidence limits (i.e., estimated
 concentration ±95-percent uncertainty) for the candidate standard from the two or more assays.
 If the confidence limits overlap, the candidate standard can be considered to be stable and may be
 certified. In the spreadsheet, all cells, in the comparisons table will be "true." If the confidence limits
 do not overlap, the candidate standard may be unstable or there may be analytical problems
 associated with the assays or the reference standards. One or more cells in the comparisons table
 will be "false."  The analyst must either disqualify the candidate standard or investigate why the
 confidence limits do not overlap.  The analyst may conduct additional assays until stability is
 achieved and add the additional data to the spreadsheet.  Data from a nonoveriapping assay may
 be discarded and the remaining data used to determine the overall estimated concentration and the
 total uncertainty provided the confidence limits overlap. Record these values and any discarded
 data in the laboratory's records.

 2.2.9   Certification Documentation

        See Subsections 2.1.4 and 2.1.5.

 2.2.10  Recertification Requirements

        See Subsections 2.1.6.3 and 2.1.6.4.

 2:3     PROCEDURE G2:  ASSAY AND CERTIFICATION OF A COMPRESSED GAS
        CALIBRATION STANDARD USING DILUTION

 2.3.1    Applicability

       This procedure may be used to assay the concentration of a diluted candidate compressed
 gas calibration standard, based on the concentration of a diluted compressed gas reference
 standard of the same gas mixture. This procedure allows a specialty gas producer, a standard user,
 or other analytical laboratory to certify that the assayed concentration for the candidate standard
 is traceable to the reference standard. The procedure employs a low-concentration-range (i.e.,
'ambient air quality level) pollutant gas analyzer to compare quantitatively diluted gas samples of
 both the candidate and reference standards.

       Dilution of the candidate and reference standards  with zero gas allows greater flexibility in
the range of concentrations of both the candidate and reference standards that can be assayed.
 Additionally, dilution allows the use of a low-concentration-range analyzer, which is more likely to
 have an inherently Jinear response than a high-concentration-range analyzer. However, the dilution
technique introduces additional error into the assay which would  not be present if the standards
were assayed without dilution. This additional error is measured by an accuracy check of the assay
 apparatus which is performed as part of the multipoint calibration.

       This procedure may be used for the assay of multiple candidate standards at the same time.
Criteria that apply to the assay of one candidate standard apply to the assay of multiple candidate
standards.

2.3.2   Limitations

       1.  The concentration of the diluted candidate standard may be greater than or lesser than
           the concentration of the diluted reference standard. However, both concentrations must
           lie within the well-characterized region of the analyzer's multipoint calibration (see
           Subsection 2.1.7.2).
                                          2-25

-------
           Additionally, the magnitude'of the 95-oercent confidence limits for the estimated
           concentration of the candidate standard must be <;±1 percent of the reference standard
           concentration. This criterion may be more restrictive than the corresponding criterion
           for the multipoint calibration, but it allows the analyst greater flexibility in the selection
           of a reference standard for the'assay of a particular candidate standard. For example,
           assume that a 70-ppm candidate standard is being assayed using a 50-ppm reference
           standard and that the analyzer's calibration was found to be well characterized between
           20 and 80  ppm.  The 95-percent-confidence limits for the  candidate  standard's
           estimated concentration must be <&OJ5 ppm.               'r

       2.  An accurate system for.flow measurement and gas dilution is required. :"

       3.  The balance gas in both the candidate and reference standards must be identical,
           unless either a high dilution flow rate ratio (i.e., at least 50 parts zero gas to 1 part
           standard) is used for the assay or it has been demonstrated that the analyzer is
           insensitive to differences in the balance gas.

2.3.3  Assay Apparatus          -                                  '       •

       The components  of the assay  apparatus  can be assembled  in several different
configurations without diminishing  performance. Two possible designs of the assay apparatus are
illustrated in Figures 2-3 and 2-4. The former figure shows a configuration in which  discrete
components (i.e., three-way valves, gas flow controllers, and a mixing chamber) are used to dilute
the reference  and candidate standards.  The latter figure shows a configuration  in which a
commercially available gas dilution system is used to dilute the standards. Both designs share the
important characteristic that the candidate standard is diluted by the same components that dilute
the reference standard.

       In Figure 2-3, either zero gas or a diluted standard can be routed to the analyzer by rotation
of three three-way values (i.e.f V1, V2, and V3).  One gas flow controller (i.e., C1) regulates the flow
rates of the reference and candidate standards.  These flow rates can be measured by a single
flowmeter connected to an outlet port on valve V2 or by a flowmeter built into C1. Another gas flow
controller (i.e., C2) regulates the flow rate of the zero gas. This flow rate can be measured by a
flowmeter connected to  an outlet port on valve V3 or by a flowmeter built into C2. The gas flow
controllers may be needle valves, capillary tubes, thermal mass flow Controllers, or other suitable
devices (see Subsection 2.3.7).  If different flow rates are used for.  £ reference and candidate
standards during the assay (see Subsection 2.3.6), separate gas flow controllers may be used for
the two standards. However, the  same flowmeter must be used to measure the two flow rates to
minimize error in the measurement (see Subsection 2.3.7). Flow  rates should be controlled  and
measured with a relative  uncertainty of 1 percent or less. For large dilutions of the standards, the
reference and candidate standard  flow rates may be quite small. Therefore, the internal volume of
the tubing and components should be kept small to minimize the flushing time when valve V1 is
rotated.

       The  mixing chamber combines the two gas streams and should be designed  to produce
turbulent flow to ensure thorough  mixing of the gas streams.  The diluted gas mixtures are routed
to the analyzer through a union tee tube fitting, which vents excess gas flow. Normally,  the excess
gas is  vented to the atmosphere  without any obstructions in the tubing and the gas entering the
analyzer is at  near-atmospheric pressure.  However, the  excess  gas can be routed through an
uncalibrated rotameter by rotation of  a three-way valve (iie., V4).  The rotameter  is used to
demonstrate that the total gas flow rate exceeds the sample flow rate of the analyzer and that no
room air is being drawn in through the vent line.
                                          2-26

-------
                                                                                                         Gas Flow
                                                                                                         to Vent
                                                                                Flow
                                                                             Measurement
                                                                                Port
                  Pressure
                  Regulator
Three-way
Valve (V1)
53
  Gas Flow
Controller (C1)
                                                                                      Three-way
                                                                                      Valve (V2)
                                                                Gas Flow
                                                              Controller (C2)
                                                                                                                        Rotameter
                                                                                                                        Gas Flow to
                                                                                                                         Analyzer
            Candidate
            Standard
                                           Three-way
                                           Valve (V3)
                           Reference
                           Standard
                                      Flow
                                  Measurement
                                      Port
                             Figure 2-3. One possible design of the apparatus using flow controllers for assay
                                   of compressed gas calibration standards with dilution (Procedure G2)

-------
                                                                                     Gas Flow
                                                                                      to Vent
       Pressure
                                 Three-way
                                 Valve (V1)
Three-way
Valve (V2)
                                                                                                          c 3. g o
                                                                                                          w 3 S. r
                                                                                                           •      •
                                                                                                     Rotameter
                                                                    Gas
                                                                  Dilution
                                                                  System
                  Gas Flow to
                   Analyzer
Candidate
Standard
                Reference
                Standard
                Figure 2-4.  One possible design of the apparatus using a gas dilution system for
                  assay of compressed gas calibration standards with dilution (Procedure G2)

-------
       The apparatus in Figure 2-3 may be modified in several ways that will not diminish its
performance.  For  example, the three-way valves could  be replaced by  solenoid valves.
Alternatively, valve V1 could be replaced by a single length of tubing that is connected manually to
the two standards' pressure regulators in succession (see also Subsection 2.1.3).

       In Figure 2-4, the reference and candidate standards are diluted with a gas dilution system.
This gas dilution system may use capillary tubes, positive-displacement pumps, thermal mass flow
controllers, or other suitable devices to dilute the standards.  If a gas dilution system is used, it must
have a specified accuracy of not greater than 1.0 percent of the undiluted  reference standard
concentration. The analyst must check the accuracy of the gas  dilution system during the multipoint
calibration (see Subsections 2.1.7.2 and 2.3.5.1).

2.3.4  Pollutant Gas Analyzer

       The pollutant gas analyzer must have a well-characterized calibration curve and must have
a range  that is capable  of measuring the diluted concentration of  both the candidate and the
reference standards (see Subsection 2.1.7.1).  It must have  good resolution, good precision, a
stable response, and low output signal noise. In addition, the analyzer should have good specificity
for the pollutant  of interest so that it has no detectable response  to any other component or
contaminant that may be contained in either the candidate or reference standards.  If any com-
ponent in a multiple-component standard interferes with the assay of any other component,  the •
analyst must conduct an interference study to determine an interference correction equation. A
suitable  analyzer with acceptable performance  specifications may  be selected from the list of
EPA-designated  reference and equivalent method analyzers.17  If the candidate and reference
standards contain dissimilar balance gases (e.g., air versus  nitrogen or different proportions of
oxygen in the balance air), either a high dilution flow-rate ratio  (i.e., at least 50 parts zero gas to 1
part standard) should be used or it must have been demonstrated that the analyzer's response is
not sensitive to  differences in  the  balance  gas composition.  This  demonstration  may  be
accomplished by showing that no difference exists in an analyzer's response when measuring a
compressed gas calibration standard that has been diluted with identical flow rates of different
balance gases.

       The analyzer should be connected to a high-precision data acquisition system (e.g., a strip
chart recorder) which must produce an electronic or paper record of the analyzer's response during
the assay. A high-precision digital panel meter, a digital voltmeter, a data logger or some other data
acquisition system with four-digit resolution can be used to obtain numerical values of the analyzer's
response. More precise values will be obtained if this system has a data-averaging capability. The
assay record must be maintained for  3 years after the standard's certification  date.

       If the analyzer has not been in continuous operation, turn it on and allow it to stabilize (e.g.,
for at least 12 hours) before beginning any measurements.

2.3.5.  Analyzer Calibration

2.3.5.1 Multipoint Calibration-
       See Subsections 2.1.7.2 and 2.1.7.4. Following completion of the multipoint calibration, the
accuracy of the assay apparatus must be checked to verify that the error associated with the dilution
is not excessive. This accuracy check involves the measurement of an undiluted or diluted check
standard.  The check standard must be an SRM, an SRM-equivalent PRM, or an NTRM, or a GMIS
as specified in Subsection 2.1.2.  It must have a certified concentration that is different from that of
the reference standard used in the multipoint calibration.  Information concerning this standard (e.g.,
cylinder identification number, certified concentration) must be recorded in the laboratory's records.
                                          2-29

-------
        If an undiluted check standard is used, its concentration must fall in the well-characterized
 region of the calibration curve.  If a diluted check standard is used, the diluted concentration must
 fall in the well-characterized region.

       . Make three or more discrete measurements of the undiluted or diluted check standard.
 "Discrete" means that the analyst must change the gas mixture being sampled by the analyzer
 between measurements.  For example, the analyst might alternate between measurements of the
 check standard and the zero gas.  Record these measurements in the laboratory's records.

        Next the analyst must verify that the dilution error is not excessive. For the check standard
 measurements, calculate the relative difference (in percent) between the mean analyzer response
 and the corresponding response that is predicted from the multipoint calibration regression equation
 and the undiluted or diluted check standard concentration.  That is,
            Relative Difference   100
                                  Mean Analyzer Response   Predicted Response
                                              Predicted Response
       If the relative difference is greater than 1.0 percent, the dilution error is considered to be
excessive. The analyst must investigate why the relative difference is excessive. The problem may
be due to errors in the reference standard and check standard concentrations, errors in assay
apparatus or to some other source.  Assays may not be conducted until the relative difference for
a subsequent accuracy check is less than or equal to 1.0 percent.

2.3.5.2 Analyzer Range—
       The range of the analyzer must include the concentrations of the zero gas and of the diluted
candidate and reference standards (see Subsection 2.3.6).  The concentrations of the diluted
reference and candidate standards must fall within the well-characterized region of the analyzer's
calibration curve.  Because the selection of the dilution ratio or ratios to be used in the assay
provides  great flexibility in the choice of concentrations to be  measured by the analyzer, the
analyzer range should be selected based on optimum accuracy, stability, and linearity.

2.3.5.3 Linearity—
       The data reduction technique used in this procedure requires that the analyzer have a well-
characterized, but not necessarily linear, calibration curve  (see Subsection 2.1.7.5).   Many
lower-concentration analyzers of the type that may be used for this procedure have straight-line
calibration curves. If not, they usually have a predictable nonlinear calibration curve that can be
des:  Ded by a polynomial equation or can be mathematically transformed to produce a straight-line
caliL  tion curve suitable for use in this procedure.  Any such polynomial equation or mathematical
transformation should be verified during the multipoint calibration.  Caution should be exercised in
using a transformed calibration curve because zero or span control adjustments to the analyzer may
produce unexpected effects in the transformed calibration curve.

2.3.5.4 Zero and Span Gas Checks—
       See Subsections 2.1.7.3 and 2.1.7.4. Prior to carrying out the assay of the candidate
standard, use zero and span gases to check for calibration drift in the analyzer since the multipoint
calibration.  Zero gas and span gas checks must be performed on  any day after the multipoint
calibration that candidate standards are assayed.  If multiple assays are being performed on the
same analyzer range, the analyst needs to perform only a single set of zero gas and span gas
check;.  However, another set must be performed if the range is changed.

         ne gas mixtures to be used during the zero and span gas checks need not be the same
as ar    the reference standards used for the assay of the diluted candidate standard or for the


                                          2-30

-------
multipoint calibration. The reference standard for the span gas check must be traceable to a NIST
SRM, an SRM-equivalent PRM, or an NTRM. Information concerning this standard (e.g., cylinder
identification number, certified concentration) must be recorded in the laboratory's records..

       Make three or more discrete measurements of the zero gas and three or more independent
measurements of the diluted reference standard. Record these measurements in the laboratory's
records.              •

        Next, the analyst must verify that the analyzer's precision is acceptable.  Calculate the mean
and standard deviation of the analyzer's response to the zero gas. Repeat these calculations for
the diluted reference standard measurements. These calculations are performed in Step 6 of the
spreadsheet described in Appendix A. Record these calculations in the laboratory's records.  The
standard error of the mean for each set of measurements must be less than or equal to 1 .0 percent
of the mean response to the diluted reference standard. That is,
                                        vfi   100

where

       s      =    standard deviation of the analyzer's response;
       n      =    the number of independent measurements of the gas mixture; and

      RDRS     =    the mean analyzer response to the diluted reference standard.

       The value of the standard error of the mean can be made smaller by increasing the number
of measurements.  This calculation will enable the analyst to determine how many replicate
measurements are needed during the assay of the diluted candidate standard to obtain acceptable
precision.  The analyst may wish to use a data logger or data acquisition system with data averaging
capability to obtain more precise measurements. If the value of the standard error of the mean is
not acceptable, then the analyzer must be repaired or another analyzer must be used for the assay.

       Next  the analyst must verify that excessive calibration drift has not occurred since the
multipoint calibration.  For the zero gas measurements, calculate the relative difference (in percent)
between the mean analyzer response during the zero gas check and the corresponding response
that is predicted from  the multipoint calibration regression equation.  That is,


                 _.„       
-------
        Then, if the relative differences for the zero and spa,    ?cks are less than or equal to 5.0
 percent, the analyzer is considered to be in calibration. Th     : and span controls need not be
 adjusted and the assay may be conducted.  The data reduc     echnique used in this procedure
 does not require the absolute accuracy of ne analyzer's ca     on. Some minor calibration drift
 is acceptable because the drift will be corrected for during th    auction of the assay data.
                                                 •
        However, if the relative difference for either the zero c me span gas checks is greater than
 5.0 percent, then the analyzer is considered to be out of caiicration. A new multipoint calibration
 may be conducted before the candidate standard is assayed or the analyzer's zero and span
 controls may be adjusted to return the analyzer's response to the original calibration levels. For
 some  analyzers such as nondispersive infrared  instruments, daily changes in environmental
 variables such as barometric pressure  may shift the  calibration.  After any adjustment of the
 controls, the analyst should repeat the zero and span gas checks and recalculate the relative
 differences to verify that the analyzer is sufficiently in calibration. The analyzer will be considered
 to be out of calibration if the relative differences remain greater than 5.0 percent.

        The zero gas and diluted reference standard measurements that are performed for the
 assay of the diluted candidate standard may also be used for the zero gas and span gas checks.

 2.3.6   Selection of Gas Dilution Flow  Rates or Gas Concentration Settings

        The flow rates or settings used for the zero gas, reference standard, and candidate standard
 should be selected carefully to provide diluted concentrations for both the candidate and reference
 standards that fall in the well-characterized region of the analyzer's calibration curve. The diluted
 concentration of the candidate standard may be greater than or lesser than the diluted concentration
 of the reference standard. Any assay error due to the dilution process will be reduced if the same
 dilution flow-rate ratio or concentration setting can be used for both the candidate and reference
 standards.  Select the diluted concentrations of the reference and candidate standards, and select
flow rates or concentration settings that will produce the highest analyzer responses within the well-
 characterized region of the analyzer's calibration curve.

        If the same dilution flow-rate ratio or concentration setting cannot be used for both the
 candidate and reference standards, select different ratios or settings for the candidate and reference
 standards to produce concentrations that  are  approximately equal and that fall in  the well-
 characterized region of the analyzer's calibration-curve. 'Select flow rates or settings such that only
 one of the apparatus controls must be adjusted when switching from the reference standard to the
candidate standard, or vice versa.  Where a choice of analyzer ranges is available, higher dilution
 ratios or lower concentration settings will reduce the consumption of the standards.

2.3.7    Flowmeter Type and Flowmeter Calibration

        Figure 2-3 shows flow measurement ports on valves V2 and V3.  In this configuration, a
single  flowmeter can be used to measure both the standard flow rate and the zero gas flow rate.
 Such an approach would reduce measurement errors arising from differences in the calibration of
 multiple flowmeters. Alternatively, the flow rates  can  be measured at the outlet of the dilution
 apparatus, with the rotameter vent temporarily plugged. In either case, a NIST-traceable volumetric
flowmeter such as a wet test meter, a thermal mass flowmeter, or a soap bubble flowmeter can be
 used (see Subsection 2.1.2)  Each flow rate must be measured separately while the other flow
 rates are set to zero. Care must  be exercised to ensure that ^ach measured flow rate remains
 constant when combined with the other flow rate(s) and betweer- the time of measurement and the
 time of the assay. Additionally, care must be taken to ensure that the flowmeter does not cause any
 back  pressure in the gas stream and any resulting charge in the flow rate  through the flow
 controller. '

                                          2-32

-------
       If in-line flowmeters are mounted directly downstream of the flow controllers, they may not
operate at atmospheric pressure because of back pressure from downstream components. Also,
this back  pressure may vary as a function of the total flow rate.  Thus, the flowmeters must
compensate for the variable in-line pressure. Thermal mass flowmeters do not need to be corrected
for pressure effects. Measurements from pressure-sensitive flowmeters such as rotameters or from
volumetric flowmeters such as wet test meters must be carefully corrected for the actual gas
pressure during the flow measurement. An in-line flowmeter must not contaminate or react with the
gas mixture passing through it.

       The flowmeters used should be stable, repeatable, and linear and have good resolution.
If possible, select flow rates or a flowmeter range such that the flow rates to be measured fall in the
upper half of the flowmeter's range. The flowmeters should be carefully calibrated at several flow
rates to prove linearity. The calibration should be accurate to plus or minus 1 percent and must be
referenced to an accurate flow rate or volumetric standard traceable to a NIST primary standard.
Flowmeter calibrations should be checked and recertified periodically, as determined by stability
information such as a chronological control chart of calibration data.

       All volumetric flow-rate measurements must  be corrected or referenced to the  same
temperature and pressure conditions, such as EPA-standard conditions (i.e., 760 millimeters of
mercury (mm Hg), 25 °C) or the ambient temperature and pressure conditions prevailing  in the
laboratory during the assay. Measurements using wettest meters and soap bubble flowmeters also
must be corrected for the saturation of the gas stream with water vapor in the moist interiors of these
flowmeters. The equation to correct the flow rate for temperature, pressure, and humidity effects
is given below:
                          Flow Rate = Volume
                                      Time


where

       PM     =    measured barometric pressure (mm Hg);
       PWV    =    partial pressure of water vapor (mm Hg);
       Ps     =    standard pressure (mm Hg);
       Ts     =    standard temperature (298.2 K); and
       TM     =    measured ambient temperature (273.2 + °C).

       Measurement of reference and candidate standard flow rates with the same flowmeter and
measurement of  both dilution zero gas flow rates  with the same flowmeter tend to  reduce
measurement errors, associated with the use  of multiple flowmeters.  These errors are more
pronounced at higher dilution flow rate ratios. Note that the impact of any flow measurement error
is reduced if the same dilution ratio can be used for both the reference standard and candidate
standard measurements.

2.3.8   Assay Gases

2.3.8.1 Candidate Standard-
       See Subsections 2.1.6, 2.3.2, and 2.3.6.
                                         2-33

-------
2.3.8.2 Reference Standard— jus.;                                     .,.-,
        See Subsections 2.1.2,2. JJ&4 2.3.2, and 2.3.6. The reference standard used for the assay
of the candidate standard must be a NIST SRM, an SRM-equivalent PRM, an NTRM or a GMIS.
This standard need not be the same, as any of the reference standards used for the span gas check
or for the multipoint calibration.  Information concerning the reference standard (e.g., cylinder
identification number, certified concentration, etc.) must be recorded in the laboratory's records.

If the multipoint calibration data have been fitted to a linear (i.e., straight-line) model, then only a
single reference standard need be measured during the assay of the candidate standard. If these
data have been fitted to a quadratic^ higher-order polynomial model, then at least two reference
standards must be measured. One reference standard is adequate'to determine the slope of a
linear equation, but additional reference standards  are needed to determine the curvature of
quadratic or higher-order polynomial equations. The concentrations^ the additional reference
standards should be located at the maximum difference between the  polynomial curve and the
corresponding straight line between the zero gas and the highest-concentration reference standard.

2.3.8.3  Zero Gas—
        See Subsection 2.1.9. Use the same zero gas for dilution of both candidate and reference
gases. The analyst may substitute a low-concentration! NIST-traceable reference standard for the
zero gas in zero gas checks and assays if there is reason  to believe that the zero gas reading may
not accurately represent the zero-intercept of the calibration equation. Information concerning the
zero gas should be recorded in the laboratory's records.

2.3.9  Assay Procedure

        1.    Verify that the assay apparatus is properly configured as shown in Figure 2-3 or
             Figure 2-4 and as described in Subsection 2.3.3.  Inspect the analyzer to verify that
             it appears  to be operating normally and that all controls are set to their expected
             values. Record these control values in the laboratory's records.

       2.    Verify that the flowmeters, if used in the assay apparatus, are properly calibrated
             (see Subsection 2.3.7).

       3.    Verify that a multipoint calibration of the analyzer has been performed within 1 month
             prior to the assay date and that the dilution error is not excessive (see Subsections
             2.1.7.2,2.1.7.5, 2.3.4, and 2.3.5.1).  Additionally, verify that the zero and span gas
             checks indicate that the analyzer is in calibration (see Subsection 2.3.5.4). Finally,
             verify that the concentrations of the diluted reference and candidate standards fall
             within the well-characterized region  of  the  analyzer's calibration  curve  (see
             Subsection 2.3.2).

       4.    Determine  and establish the flow rates or concentration settings of the gas mixtures
             (i.e., reference standard(s), candidate standard, and zero gas) that will be used for
             the assay (see Subsections 2.3.6,2.3.7, and 2.3.5.2). Also check that the total flow
             rate coming from the mixing chamber will provide enough flow for the analyzer and
             sufficient excess to ensure  that no ambient air will be drawn into the vent line.
             Changes in the sample pressure may change the calibration curve. When using the
             same flow rates for both candidate and reference standards,  carefully set the
             delivery pressures of the two stands ^s1 pressure regulators to the same value so
             that there is no change in the flow i^te when  switching from one standard to the
             other.
                                          2-34

-------
       Calculate  the  diluted reference  standards' concentration using the following
       equation:
              Diluted Standard Cone. =
(Undiluted Standard Cone.) (Standard now Rate)
                                     (Standard Flow Rate + Zero Gas Flow Rate)
       Record the measured flow rates and the undiluted and diluted reference standard
       concentrations in the laboratory's records.

5.     In succession, measure the zero gas, the diluted reference standard(s) and the
       diluted candidate standard using the analyzer. For each measurement, adjust the
     •  flow rates, if necessary, to those determined in step 4, and allow ample time for the
       analyzer to achieve a stable reading. If the reading for each measurement is not
       stable, the precision of the measurements will decline and the candidate standard
       might not be certifiable under this protocol.  Record the analyzer response for each
       measurement using the same response units (e.g., volts, millivolts, percent of scale,
       etc.) as was used for the multipoint calibration. At this point, do not convert the data
       into concentration values using  the  calibration equation.  Do not perform any
       mathematical transformations of the data. These steps will be done later. Do not
       make any zero control, span control, or other adjustments to the analyzer during this
       set of measurements.  Record these analyzer responses in the laboratory's records.

       The analyst may assay multiple candidate standards -during the same assay
       session.  For example, a single set of measurements may involve a zero gas, a
       diluted reference standard, and three diluted candidate standards. Criteria that
       apply to  the  assay of one  candidate  standard  apply to the assay  of multiple
       candidate standards. The analyst should be aware that the effect of any short-term
       calibration drift will be greater when multiple candidate standards are assayed. This
       greater effect is due to. the longer period  of time between reference standard
       measurements.  Unacceptable uncertainties of the estimated concentrations for the
       diluted candidate standards may occur as a result of the longer assay session.

6.     Conduct at least two additional sets of measurements, as described in step 5 above.
       However, for these subsequent sets of measurements, change the order of the three
      'measurements (e.g., measure the reference standard,  zero gas, and candidate
       standard for the second set and  measure the zero  gas, candidate standard, and
       reference standard for the third set, etc.). Changing the order that the gas mixtures
       are measured helps the analyst to discover any effect that one measurement has on
       subsequent measurements.  The number of sets of measurements will have been
       determined during the analysis of the multipoint calibration data such that the 95-
       percent  uncertainty for the regression-predicted concentration of the candidate
       standard is <;1 percent of the concentration of the reference standard.

7.     If any one or more of the measurements of a set of measurements is invalid or
       abnormal for  any reason, discard all three measurements and repeat the set of
       measurements.  Such measurements may be discarded  if the analyst can
       demonstrate that the experimental conditions were inappropriate during these
       measurements.  Data cannot be discarded just because they appear to be outliers,
       but may be discarded if they satisfy statistical criteria for testing outliers.15  The
       analyst must record the discarded data and a brief explanation as to why these data
       were discarded in the laboratory's records.

                                  2-35

-------
        8.     The spreadsheet des: ;bed in Appendix A or equivalent statistical techniques must
              be used to calculate.-  estimated concentration and a 95-percent uncertainty for the
              candidate standard ^ased on data from the assay measurements and from the
              multipoint calibration.   The use of both  sets of data in the statistical analysis
              'produces an estimated concentration with smaller uncertainty while correcting for
              any minor calibration drift that may have occurred since the multipoint calibration.
              Record the estimated concentration and the 95-percent uncertainty in the labora-
              tory's records.

              The spreadsheet also calculates the percentage of the uncertainty that is due to the
             -multipoint calibration.  This percentage is needed for the total uncertainty calcula-
              tions when two or more assays fall under the same multipoint calibration.  Record
              this value in the laboratory's records.

              The analyst should investigate any of the measurements that appear to be outliers.
              Such  data may be discarded if the analyst can demonstrate that the experimental
              conditions were inappropriate during these measurements.  Data cannot be dis-
              carded just because they appear to be outliers, but may be discarded if they satisfy'
              statistical criteria for testing outliers. The analyst must record any discarded data as
              well as a brief summary of the investigation in the laboratory's records.

        9.     If the  multipoint calibration data and the assay data underwent any mathematical
              transformations before their statistical analysis, perform the reverse transformations
              for the estimated concentration and the 95-percent uncertainty.  Record the trans-
              formed values in the laboratory's records.

        10.    Finally, the certified undiluted concentration for a candidate standard containing a
              unreactive gas mixture and requiring only a single assay can be calculated from the
              mean concentration of the diluted candidate standard as follows:


                         Certified Undiluted Cone.  =  (Mean Diluted Cone.) (Total Gas Flow  Rate)
                                                         (Standard Flow Rate)

                         where Total Gas Row  Rate = Standard Row Rate + Zero Gas  Flow Rate .
2.3.10 Stability Test for Newly Prepared Standards

       Newly prepared candidate standards that contain reactive gas mixtures must be assayed
on at least two dates that are separated by at least 7 days. See Subsections 2.1.6.1  and 2.1.6.2.
The spreadsheet described in Appendix C or equivalent statistical techniques must be used to
evaluate the stability of the candidate standard and to calculate the overall estimated concentration
and the total uncertainty for the candidate standard.

       The stability is evaluated by comparing the 95-percent confidence limits (i.e., estimated
concentration ±95-percent uncertainty) for the candidate standard from the two or more assay:,
If the confidence limits overlap, the candidate standard can be considered to be stable and may be
certified. In the spreadsheet, all cells in the comparisons table will be "true." If the confidence limit-
do  not overlap, the candidate standard may be unstable or there  may  be analytical problerr
associated with the assays or the reference standards. One or more cells in the comparisons tab'.
will be "false." The analyst must either disqualify the candidate standard or investigate why try.
confidence limits do not overlap.  The analyst may conduct additional  assays until stability -.*


                                          2-36

-------
achieved and add the additional data to the spreadsheet.  Data from a nonoverfapping assay may
be discarded and the remaining data used to determine the overall estimated concentration and the
total uncertainty provided the confidence limits overlap. Finally calculate a certified undiluted
concentration using  the above equation.  Record these values and any discarded  data in the
laboratory's records.
2.3.11 Certification Documentation

       See Subsections 2.1.4 arid 2.1.5.

2.3.12 Recertification Requirements

       See Subsections 2.1.6.3 and 2.1.6.4.
                                         2-37

-------
                                     SECTION 3

          EPA TRACEABILITY PROTOCOL FOR ASSAY AND CERTIFICATION OF
                   PERMEATION DEVICE CALIBRATION STANDARDS
3.1    GENERAL INFORMATION

3.1.1  Purpose and Scope of the Protocol

       This protocol describes three procedures for assaying the permeation rate of a permeation
device calibration standard and for certifying that the assayed permeation rate is traceable to
National Institute of Standards and Technology (NIST) reference standards.  This protocol is
mandatory for certifying the permeation device calibration standards used  for the  pollutant
monitoring that is required by the regulations of the Code of Federal Regulations, Chapter 40, Parts
50 and 583'4 for the calibration and audit of ambient air quality analyzers. This protocol covers the
assay and certification of  sulfur dioxide (SO2) and nitrogen dioxide (NO2) permeation device
calibration standards. This protocol may be used by permeation device producers, standard users,
or other analytical laboratories.  The  assay procedure may involve the  comparison  of these
standards to permeation device .reference standards (i.e., Procedure  P1), to  compressed gas
reference standards (i.e., Procedure P2), or to mass reference standards (i.e., Procedure P3).

3.1.2  Reference Standards

       The permeation device reference standards that may be used under this protocol are NIST
Standard Reference Material (SRM) numbers 1625 and 1626. These SRMs (listed in Table 3.1)
are permeation tubes containing SO2. In the future, NIST may develop additional SRMs, which may
be used as reference standards under this protocol.

       The compressed gas reference standards that may be used under this protocol are NIST
SRMs, Netherlands Measurement Institute Primary Reference Materials (NMi PRMs) that are
equivalent  to SRMs,  NIST traceable  reference  materials  (NTRMs), or gas manufacturer's
intermediate standards (GMISs). These standards are described in Subsection 2.1.2 of this report.

       The uncertainty of SRMs,  NTRMs, and PRMs is expressed as a 95-percent confidence
interval, which is the one-sigma uncertainty multiplied by a coverage factor almost always equal to
2.9 This estimate includes allowances for the uncertainties of known sources of systematic error as
well as the random error of measurement. A value of one-half of the stated uncertainty of these
reference standards should  be used in calculating the total analytical uncertainty of standards that
are certified under this protocol (see Appendix C).

       Mass reference standards that may be used under this protocol must be traceable to NIST
mass standards.18"20 Additionally, they must have an individual tolerance of no more than 0.05 mg.
Examples of mass reference standards that meet these specifications are American National
Standards Institute/American Society for Testing and Materials (ANSI/ASTM) Classes 1, 3, and 4.
The mass reference standards must be recalibrated on a regular basis (e.g., yearly) at a NjST-


                                         3-1

-------
         TABLE 3-1. NIST SRM PERMEATION DEVICE REFERENCE STANDARDS
Nominal concentration (in

NIST
SRM no.
1625
1626

Permeation
device type
Sulfur dioxide
Sulfur dioxide
Device
length
(cm)
10
' 5'
Nominal
permeation rate at
30 °C (ug/min)
3.7
2.1
umol/mol) at various dilution
gas flow
1
1.4
0.8
rates (Umin)
5 10
0.28 0.14
0.16 0.08
accredited State weights and measures laboratory or at a calibration laboratory that is accredited
by the National Voluntary Laboratory Accreditation Program (NVLAP),10-11 which is administered by
NIST, or  by the International Laboratory Accreditation Conference (ILAC).  The recalibration
frequency is to be determined from records of previous recalibrations of these standards.

       Two separate sets of mass reference standards are recommended. Working calibration
standards should be usec tor routine permeation device weighings and should be kept next to the
analytical balance in a protective container. Laboratory primary standards should be handled very
carefully and should be keot in a locked compartment. The working standards should be compared
to the laboratory primary standards every 3 or 6 months to check for mass shifts associated with
handling or contamination  The current masses of the working standards as traced to the laboratory
primary standards shoulc DG recorded in a laboratory notebook and should be used to check the
calibration of the analytical balance.

       Always use smooth, nonmetallic forceps for handling mass reference standards.  The
standards are handled only with these forceps, which are not used for any other purpose.  Mark
these forceps to distinguish them from the forceps that are used to  handle permeation devices.
Handle the standards carefully to avoid damage that may alter their masses.

       The temperature reference standards that may be used under this protocol must be liquid-in-
glass thermometers having scale tolerances and uncertainties that conform to NIST Special
Publication 250-23.21 They must have an uncertainty of no more than 0.05 (°C). The thermometers
must have serial numbers etched or permanently marked into the  glass and a manufacturer's
calibration certificate of traceability to NIST standards.  They must be recalibrated on a regular
schedule (e.g., yearly) according to NIST guidelines22 by the user, a NIST-accredited State weights
and measures laboratory, or an NVLAP- or ILAC-accredited calibration laboratory.

3.1.3  Selecting a Procedure

       Procedures P1 and P2 are applicable to the assay and certification of candidate permeation
device calibration stanaards using an ambient air quality analyzer.  Procedure P1  provides for the
assay to be referenced to a permeation device reference standard.  Procedure P2 provides for the
assay to be referenced to a compressed gas reference standard.
                                          3-2

-------
       Procedure P3 is applicable to the assay and certification of candidate standards using an
analytical balance. This procedure provides for the assay to be referenced to a mass reference
standard.

3.1.4  Using the Protocol

       The assay/certification protocol described here is designed to minimize both systematic and
random errors in the assay process. Therefore, the protocol should be carried out exactly as it is
described.  The assay procedures in this protocol  include one possible design for the assay
apparatus.  The analyst is not required to use this design and may use alternative components and
configurations that produce equivalent-quality measurements. Nonreactive materials (e.g., Teflon*
stainless steel, or glass) and clean, noncontaminating components should be used in those portions
of the apparatus that are in contact with the gas mixtures being assayed.

3.1.5  Certification Documentation

       Each certified permeation device calibration standard must be documented in a written
certification report and this report must contain at least the following information:

       1.  Permeation device identification number;

       2.  The contents of the permeation device;

       3.  Certified permeation rate (in nanograms (ng) per minute);

       4.  The certification temperature (in °C to the nearest 0.1 °);

       5.  The dilution gas (i.e., air or nitrogen) used during the assay (for procedures P1 and P2);

       6.  Date of the assay/certification;

       7.  Identification of the  reference standards used in the assay:  NIST SRM number,  NIST
           sample number, and certified concentration or  permeation rate for an SRM; cylinder
           identification  number and certified concentration for an  SRM-equivalent PRM,  an
           NTRM or GMIS; manufacturer, model  number and serial number for a mass or
           temperature reference standard. The certification documentation must identify the type
           of reference standard used in the assay;

       8.  Statement that the assay/certification was performed according to this protocol and that
           lists the assay procedure (e.g., Procedure  P1) used;

       9.  The analytical method that was used in the assay;

       10.  Identification of the laboratory where the standard was assayed and certified;

       11.  Chronological record of all certifications for the  standard by the certifying laboratory;

       12.  A statement that the standard will retain its certification only as long as 5 percent of the
           original liquid weight or a visible amount of liquid remains in it;

                                          3-3

-------
        13.  Thq environmental exposure conditions (e.g., temperature and moisture) that will
            invalidate the certification; and
                       '                ' U      '• 'I.: :     .   .-;•.-
        14.  A statement of the overall analytical uncertainty estimate associated with the assay of
            the candidate standard. The estimate must include the uncertainty associated with the
            reference standard.

 This certification documentation, must be given to the purchaser of the standard. The permeation
 device producer must maintain laboratory records and certification documentation for 3 years after
 the standard's certification date.: A permeation device producer or other vendor may redocument
 an assayed and certified standard that it has purchased from another permeation device producer
 and that it wishes to sell to a third party.  However, the new certification documentation must clearly
 list the permeation device producer or other laboratory where the standard was assayed.

 3.1.6   Certification Label

        The permeation device calibration standard must be labeled with its identification number.

 3.1.7   Assay/Certification of Candidate Permeation Device Calibration Standards

 3.1.7.1  Permeation Device Design—
        Permeation devices are designed and constructed in various ways, but all devices consist
 of a sealed chamber containing liquified gas and a permeable area through which the gas is allowed
 to permeate. The permeated gas is-swept and diluted with a measured volumetric flow rate of dry
 air or nitrogen to create a quantitative concentration of the pollutant gas.

 3.1.7.2  Precautions for Use and Storage of Permeation Devices—
       The permeation rate of all permeation devices is critically dependent on temperature; a
 permeation device is useful as a concentration  standard only when its temperature is precisely
 controlled and accurately measured, and an accurately metered dilution gas flow rate is provided.

       The  inaccuracy of gaseous pollutant  concentration standards that  are  produced  by
permeation devices may increase due to physical or chemical sorption of the permeated gas in the
permeation system. This sorption will have a larger: effect on the inaccuracy as the concentration
decreases.  Nonreactive materials (e.g., Teflon*,  stainless steel, or glass) and clean, noncon-
taminating components should be used in those portions of  the permeation system that are in
contact with the permeated gas.

       The reproducibility of the certified permeation rate of a permeation device may be adversely
affected by exposure of the device to temperatures greater than the specified operating or storage
temperature range for the device or by exposure to excessive  moisture.  NO2 permeation devices
 must be stored under dry conditions and preferably at a temperature between 20 and 35  °C, or as
otherwise  recommended by the manufacturer.   SO2 permeation devices may be refrigerated for
storage.

        It appears that there is a limited temperature range at which NO2 permeation devices can
 be used as standards.  This temperature range  is conservatively given as 20 to 35 °C.23 Low or
 high temperature storage of NO2 permeation devices is not recommended.
                                          3-4

-------
       NO2 permeation devices should be stored in and used in dry dilution gas.  One study
showed that N02 permeation rates were significantly lower in moderately humid air (i.e., 30
to 40 percent relative humidity) than in dry air on the preceding  day.24 Furthermore, the
permeation rates did not return to the original levels after dry air had passed over the device
for 24 hours. Another study found that N02 concentrations from a permeation device declined
by about one-third as relative humidity levels increased from 0 to 100 percent.25

       Candidate standards being certified under Procedure P3 must be stored under constant
temperature conditions between assays. A storage container for this application is described
in the procedure.

       When stored at a temperature other than the assay temperature, some permeation
devices require an equilibration period at the assay temperature to reach thermal equilibrium
and  a stable,  accurate  permeation rate.  When transferred  from a  different  storage
temperature, thin-walled permeation devices should be maintained at the assay temperature
wjth  a fixed dilution flow rate  for at  least 48 hours before use or before certification.
Temperature changes of > 10 °C may require equilibration periods of up to 15'days for N02
permeation devices to attain a stable  permeation rate.23-26  Upon return to the original
temperature, some devices may not return to the same permeation rate  as  before the
temperature change.   Other types of permeation devices may require longer equilibration
periods.  Observe any manufacturer's recommendations for equilibration and use.

3.1.7.3 Equilibration  of Newly Prepared Permeation Devices—
       A newly prepared permeation device must be equilibrated for at least 48 hours at the
assay temperature before being assayed for the first time. The equilibration  period may be
100 hours  or longer for some permeation devices.26 This period will vary as a function of the
permeating compound, the  material and the thickness of the permeating surface and the
temperature.

3.1.7.4 Certification  Conditions for Permeation Device Calibration Standards—
       A standard will retain its certification only as long as 5 percent of the original liquid
weight or a visible  amount of liquid remains in it.  A standard loses its certification if it is
exposed for prolonged periods of time to excessive moisture or to temperatures greater than
15 °C above its certification temperature.  A  standard that loses its  certification must be
reassayed  before it can be certified for  further  use.

3.1.8  Technical Variances

       Permeation device producers, standard users, and other analytical laboratories may
petition the U.S. EPA for  technical variances to the assay  procedures in this protocol.  A
technical variance allows the use of a specific alternative  assay procedure  for candidate
standards,  which can be certified under this protocol.  The petitioner must send a written
request with a detailed description of the alternative assay procedure and supporting analytical
data to: EPA Traceability Protocol Project, U.S. EPA, Mail Code 47, Research Triangle Park,
NC 27711.  The supporting analytical data  must demonstrate the equivalence  of the
alternative  assay procedure  with the procedures given in this protocol. Technical variances
may also be given for alternative temperature ranges of certifying or storing permeation
devices provided that supporting analytical data are provided with the  written request.
                                         3-5

-------
       Permeation device prod   ?rs, standard users,  and other analytical laboratories may
petr- • JD the U.S. EPA t allow t   assay and certification of permeation devices that contain
gase; or liquified gases, other th"~ S02i;and NOjt 'The petitioner must send a written request
with a detailed description of tn& permeation; device and  supporting analytical data to the
address given above.  The supporting analytical data must" demonstrate that the permeation
rate for the proposed device can be accurately determined, that only the specified compound
is permeating, that the rate is stable over the lifetime of the device,  and that the rate is not
changed by temperature and humidity effects.     T .. n . ^ o           :••  • •    i
              • - :. .              ,-  t       "*"      - • •-           i r  •• *-
3.2    PROCEDURE PI: ASSAY AND CERTIFICATION OF PERMEATION  DEVICE
       CALIBRATION STANDARDS REFERENCED TO A PERMEATION DEVICE
       REFERENCE STANDARD

3.2.1  Applicability

       This procedure may be used to assay the permeation rate of a candidate SO2 or N02
permeation device calibration standard, based on the permeation rate of a permeation device
reference  standard of the  same pollutant compound, and  to certify that the assayed
permeation rate is  traceable to the reference standard.  The  procedure employs  a  low-
concentration range (i.e., ambient  air  quality level) pollutant  gas analyzer to  compare
quantitatively diluted concentrations from the two permeation devices for the assay of the
candidate device. This procedure may be used for the assay of multiple candidate standards
during the same assay session.  Criteria that apply to the  assay of one candidate standard
apply to the assay of multiple  candidate standards.  This procedure  may be used by
permeation device producers, standard users, or other analytical laboratories.

3.2.2  Limitations

       1.  The permeation rate of the candidate standard may be greater than or lesser than
          the  permeation  rate  of the  reference  standard.    However,  the  diluted
          concentrations from both standards must lie within the well-characterized region
          of the analyzer's multipoint calibration (see Subsection 2.1.7.2). Additionally, the
          95-percent uncertainty for the regression-predicted concentration of the diluted
          candidate standard must be ^1.0 percent of  the concentration of the  diluted
          reference standard. This uncertainty is obtained from the statistical analysis of the
          multipoint calibration data using the spreadsheet described in Appendix A or using
          equivalent statistical techniques (e.g., the worksheet for linear relationships given
          in Chapter  5  of  Reference 15).  This criterion  means that the uncertainty
          associated with the multipoint calibration determines the concentration range over
          which a  diluted candidate standard may  be assayed.

       2.  A quantitatively accurate flow  measurement and dilution system is required.

       3.  A source of clean, dry zero gas is required.

       4.  This procedure is designed to assay the permeation rate c-f a candidate standard
          that is mounted in a specially designed assay dilution system; the procedure does
          not  accommodate the certification of a candidate standard that is mounted in its
        •  own self-contained dilution/flow measurement system.


                                         3-6

-------
3.2.3   Assay Apparatus

        Figure 3-1 illustrates the components and configuration of one possible design for the assay
apparatus, including a common dilution system for both the reference and candidate standards.
The configuration is designed to allow convenient routing of zero gas and dHuted concentrations of
the reference and candidate  standards, in turn, to the analyzer for measurement, as selected by
valves V1, V2, and V3. Three gas flow controllers (i.e., C1, C2, and C3) regulate the total dilution
flow rates for both the reference and candidate permeation devices and the purge gas flow rate.
These gas flow controllers may be needle valves, capillary tubes, thermal mass flow controllers, or
other suitable devices. The flow rates must be controlled to within 1.0 percent variation during the
assay.

        The total dilution flow  rate is measured by a single, common flowmeter (i.e., M1). Valve V1
directs a portion of the total dilution flow through one or the other of the two temperature-controlled
permeation device chambers to sweep up the permeated pollutant gas. This sweep flow rate is
monitored by an auxiliary flowmeter for each permeation device (i.e., M2 and M3). These auxiliary
flowmeters  need not be accurately calibrated, since only the total dilution flow measured by
flowmeter M1 is  used in the dilution calculation. Gas flow controllers C1 and C3 can be used to
adjust and balance the flow rates of the two gas streams sweeping through  the permeation device
chambers. The permeation device that is not being analyzed receives a purge gas stream to avoid
the buildup of high pollutant concentrations in the chamber.  This purge gas flow is vented through
valve V2 and is not measured by flowmeter M1.

        The assay apparatus illustrated in Figure 3-1 may be modified by the addition of multiple
candidate standard chambers.  These chambers may be set to different temperatures.

        If it  is necessary to  use different dilution flow rates  for the  candidate and reference
permeation devices (see Subsection 3.2.6), separate flow controllers for the two permeation devices
may be used for the two different flows. However, the same flowmeter should always be used to
measure these two flow rates to minimize systematic flow measurement errors.

        The mixing chamber combines the gas streams and should be designed to provide
turbulence in the  flow to ensure thorough mixing of the two gas streams. The diluted gas mixtures
are routed to the analyzer through a union tee tube fitting, which vents excess gas flow. Normally,
the excess gas is vented to  the atmosphere without any obstructions in the tubing and the gas
entering the analyzer is at near-atmospheric pressure.  However, the excess gas can be routed
through an uncalibrated rotameter by rotation of a three-way valve (i.e., V4). The rotameter is used
to demonstrate that the total gas flow rate exceeds the sample flow rate of the analyzer and that no
room air is being  drawn in through the vent line (also see  Subsection 3.1.4). Check the apparatus
carefully for leaks and correct all leaks before use.

        The mean temperatures of the reference  standard chamber and the candidate standard
chamber must be controlled to within 0.05 °C of the setpoint with a temperature stability of ±0.05 °C.
These temperatures must be measured with a NIST-traceable thermometer  having a measurement
uncertainty of ±0.05 °C or less.
                                          3-7

-------
                                                                 Gas
                                                               Flowmeter
                                                                 (M2)
w
do
                            Dilution Gas
                             Flowmeter
                               (M1)
          Pressure
         Regulator
                       4-way
                       Valve
                        (V1)
                                                      Gas Flow
                                                      Controller
                                                        (C1)
                                                                             Candidate
                                                                             Standard
                                                                             Chamber
                                                                                t
                                                   Excess Gas
                                                   Flow to Vem
                                                              i

                                                       Rotameter
                                                                                        3-way
                                                                                      Valve (V4)
1
                           Purge Gas Flow
Gas Flow
Controller
  (C2)
        Purge Gas
       Flow to Vent
                                              Gas
                                           Flowmeter
                                              (M3)
        Zero
        Gas
                                                                                                            Gas Flow to
                                                                                                             Analyzer
                                                                                    Gas Flow
                                                                                     to Vent
                                                         Reference
                                                         Standard
                                                         Chamber
                                  Gas Flow
                                  Controller
                                    (C3)


Figure 3-1. One possible design of the apparatus for the assay of permeation device calibration
      standards referenced to a permeation device reference standard (Procedure P1).

-------
 3.2.4   Pollutant Gas Analyzer

        See  Subsection 2.3.4.  The pollutant gas  analyzer  must have  a well-characterized
 calibration curve and a range capable of measuring the diluted concentrations of both the candidate
 and the reference standards.  It must have good resolution, good precision, a stable response, and
 low output signal noise.  In addition, the analyzer must have good specificity for the pollutant of
 interest so that it has no detectable response to any contaminant that may be contained in the
 standards. A suitable analyzer with acceptable performance specifications may be selected from
 the list of EPA-designated reference and equivalent method analyzers."

        The analyzer should be connected to a high-precision data acquisition system (e.g., a strip
 chart recorder), which must produce an electronic or paper record for documentation of the analyzer
 responses obtained during the assay.  Additionally, a digital panel meter with four-digit resolution,
 a digital voltmeter, data logger, or other data acquisition system must be used to obtain numerical
 values of the analyzer's response. More precise values will be obtained if these instruments have
 some data  averaging capability.  The assay record must be  maintained  for 3 years after the
 standard's certification date.

        If the analyzer has not been in continuous operation, turn it on and allow it to stabilize (e.g.,
 for at least 12 hours) before beginning any measurements.

 3.2.5   Analyzer Calibration

 3.2.5.1 Multipoint Calibration-
       See Subsections 2.1.7:2 and 2.1.7.4. Following completion of the multipoint calibration, the
 accuracy of the assay apparatus must be checked to verify that the error associated with the dilution
 is not excessive.  This accuracy check involves the measurement of a diluted check standard.  This
 check standard must be a permeation device that is traceable to a NIST SRM.   It must have a
 certified  permeation rate that is different from that  of the reference standard used during the
 multipoint calibration.  Information concerning this standard (e.g., permeation device identification
 number, certified permeation rate) must be  recorded in the laboratory's records.  The diluted
concentration of  the check standard must fall in the weJI-characterized region of the calibration
curve.

       Make three or more  discrete measurements of the diluted check standard.   "Discrete"
means that  the analyst must change  the gas mixture being sampled by the analyzer between
measurements.  For example, the  analyst might alternate between measurements of the diluted
check standard and the zero gas. Record these measurements in the laboratory's records.

       Next the analyst must verify that the dilution error is not excessive.  For the diluted check
standard measurements, calculate the relative difference (in percent) between the mean analyzer
response and the corresponding response that is predicted from the multipoint calibration regression
equation and the diluted check standard concentration. That is,

            _ . ..  r..u        inn Mean Analyzer Response - Predicted Response
            Relative Difference = 100          7	
                                              Predicted  Response

If the relative difference is greater than 1.0 percent, the dilution error is excessive. The analyst must
investigate why the dilution error is excessive. The problem may be due to errors in the reference
                                           3-9

-------
 standard and check standard permeation rates, errors in the assay apparatus or to some other
 source. Assays may not be conducted until the relative difference for a subsequent accuracy check
 is £ 1.0 percent.
                       ^*          '•'. *
 3.2.5.2 Analyzer Range^-
        See Subsection 2.3.5.2.

 3.2.5.3 Linearity—
        See Subsection 2.3.5.3.

 3.2.5.4 Zero and Span Gas Checks—
        See Subsection 2.3.5.4.

 3.2.6   Selection of Gas Dilution Flow Rates

       The dilution flow rates used for the reference and candidate standards should be selected
 carefully to provide diluted concentrations for both standards that fall in the well-characterized region
 of the analyzer's calibration curve.  Potential errors in the assayed permeation rate due to dilution
 flow rate measurement error will be greatly reduced if the same dilution flow rates can be used for
 both standards. If the same dilution flow rates cannot be used for both standards, select different
 dilution flow rates for the candidate and. reference devices to provide approximately equal diluted
 concentrations  that fall in the well-characterized region.  Additionally, the magnitude  of the 95-
 percent confidence limits for the estimated concentration of the diluted candidate standard must be
 £±1 percent of the concentration of the diluted reference standard.

3.2.7  Fiowmeter Type and Flowmeter Calibration

       Flowmeter M1, as shown in Figure 3-1, measures in-line flow rates and does not operate
at atmospheric pressure  because of backpressure from downstream components.  Also, this
backpressure is variable, depending on the total dilution flow rate.  Thus, the type of flowmeter used
must compensate for the variable in-line pressure. Measurements from a pressure-sensitive
flowmeter such as a rotameter or a wet test meter must be carefully corrected for the actual in-line
pressure during the total dilution flow rate measurement.

       Alternatively, the flow rates can be measured at the outlet of the dilution apparatus, with the
excess gas flow vent temporarily plugged. In this case, a volume-type meter such as a wet test
meter  or a soap film  flowmeter can be used, and flow  measurements may be conveniently
referenced to atmospheric pressure. Each flow rate must be  measured independently while the
other flow rate is set to zero. Great care must then be exercised to ensure that each measured flow
rate remains constant between the time of measurement and the time of the assay.

       The flowmeter used should be stable, repeatable, linear, and have good resolution.  The
flowmeter must  not contaminate or react with the gas mixture passing through it. If possible, select
flow rates or a  flowmeter range such that the  measured flow rates fall in the upper  half of the
flowmeter's range. The flowmeter should be carefully calibrated at several flow rates to prove
linearity. The calibration should be accurate to ±1.0 percent, referenced to an accurate flow or
volume standard traceable to a NIST primary  standard (see  Subsection 3.1.2). The flowmeter
calibration should be checked and recertified on a regular schedule (e.g., yearly). The recertification
                                         3-10

-------
 frequency is to be determined from stability information such as a chronological control chart of
 calibration data.

        All volumetric flow rate  measurements must be corrected or referenced to the same
 temperature and pressure conditions, such as EPA standard conditions (25 °C and 760 mm Hg)
 or the ambient temperature and pressure conditions prevailing in the laboratory during the assay.
 Measurements using wet test meters and soap bubble flowmeters also must be corrected for the
 saturation of gas stream with water vapor in the moist interiors of these flowmeters. The equation
 to correct the flow rate for temperature, pressure, and humidity effects is given below:


                         Flow Rate = Volume
                                     Time

 where

     PM   = measured barometric pressure (mm Hg);
    PWV   = partial pressure of water vapor (mm Hg);
     Ps   = standard pressure (mm Hg);
     Ts   = standard temperature (298.2 K);
     TM   = measured ambient temperature (273.2 + °C).

       Measurement of both dilution flow rates with the same flowmeter tends to reduce systematic
 flow measurement error.  Note particularly that flow measurement error is greatly reduced if the
 same dilution flow rates can be used for both the reference and candidate standards.

 3.2.8  Permeation Devices

 3.2.8.1 Candidate Standard—
       See Subsections 3.1.7 and 3.2.2.  Follow the manufacturer's instructions for equilibration
 and for use of the candidate standard and for selecting the temperature at which it is to be assayed
 and certified. The candidate standard should be assayed at the same temperature at which it will
 be subsequently used. The mean operating temperature of the candidate standard chamber must
 be  controlled to within 0.05 °C  of the setpoint with a temperature stability of ±0.05 °C.  This
temperature must be measured with a NIST-traceable thermometer with a measurement uncertainty
 of ±0.05 °C or less (see Subsection 3.1.2).

3.2.8.2 Reference Standard-
       See Subsections 3.1.2, 3.1.7.1,3.1.7.2, and 3.2.2. Follow NIST's instructions for equilibra-
tion and use of the SRM reference standard and for selecting an operating temperature within its
certified range. The mean operating temperature of the reference standard chamber must be
controlled to within 0.05  °C of the setpoint with a temperature  stability of ±0.05 °C.   This
temperature must be measured with a NIST-traceable thermometer with a measurement uncertainty
of ±0.05 °C or less.

3.2.8.3 Zero Gas-
       See Subsection 2.1.8.  Use the same zero gas for dilution of both the candidate and
reference standards.
                                         3-11

-------
3.2.9  Assay Procedure
            '        .     :

       1.  Verify that the asj^y^pparatus is properly configured as shown in Figure 3-1 and
           described in Subsfijcgpjn 3.2.3.  Inspect .the analyzer to verify that it appears to be
           operating normally^f^that all controls are set to their expected values. Record these
           control values in the^gbbratory's records.
       2.  Determine and establish the operating temperatures for the reference and candidate
           standards in their respective temperature-controlled chambers. Install the standards
           and, with zero gas flowing over both standards, allow ample time for the standards to
           equilibrate (see Subsection 3.1.7.2).  Record the temperatures in the laboratory's
           records.                       -.^ ,<

       3.  Verify that flowmeter M1 is properly calibrated (see Subsection 3.2.7).
                                      :    '- *•»
       4.  Verify that a multipoint calibration of the analyzer has been performed within 1 month
           prior to the assay date (see Subsections 2.1.7.2,2.1.7.5, and 2.3.4). Additionally, verify
           that the zero and span gas checks indicate that the analyzer is in calibration (see
           Subsection 2.3.5.4).

       5.  Determine and establish the dilution flow rates and diluted concentrations for the
           reference and candidate standards that will be used for the assay (see Subsections
           3.2.6,3.2.7, and 3.2.5.2). Use an estimated permeation rate for the candidate standard
           in these calculations. Calculate the diluted standard concentrations (in ppm) using the
           following equation:

                   Diluted Standard Cone. = M f M ] [ Permeation Rate 1
                                        I   J[MW][Dilution Row Rate]

           where

           MV  =  Molar volume of the dilution gas (liters/mole);

                =  (0.08206) Tm

          MW  =  Molecular weight of the dilution gas (grams/mole); permeation rate is given in
                   nanograms/minute; and dilution flow rate is given in liters/minute.

           Ensure that the diluted candidate and reference standard concentrations are within the
           well-characterized region of the analyzer's calibration curve (see Subsection 2.3.2).
           Also check that both dilution flow rates will provide enough flow for the analyzer, with
           sufficient excess to ensure that no ambient air will be drawn into the vent line, and
           without increasing the pressure of the sample delivered to the analyzer.  If possible, use
           the same dilution flow rate  for both standards.  Record the measured flow rates in the
           laboratory's records.

       6.   In succession, measure the zero gas, the diluted reference standard and the diluted
           candidate standard using the analyzer. Use valves V1, V2, and V3 to select each of
           the three gas mixtures for measurement. For each measurement, adjust the flow rates,

                                          3-12

-------
    if necessary, to those determined in step 5, and allow ample time for the analyzer to
    achieve a stable reading. If the reading for each measurement is not stable, the
    precision of the measurements will decline and the candidate standard might not be
    certifiable under this protocol. Record the analyzer response for each measurement,
    using the same response units (e.g., volts, millivolts, percent of scale, etc.) as was used
    for the multipoint calibration. At this point, do not convert the data into concentration
    values  using   the  calibration  equation.   Do  not  perform  any  mathematical
    transformations of the data. These steps will be done later. Do not make any zero
    control, span  control,  or other adjustments to the  analyzer  during  this set of
    measurements. Record these analyzer responses in the laboratory's records.

    The analyst may assay multiple candidate standards during the same assay session.
    For example, a single set of measurements may involve a zero gas, a diluted reference
    standard, and three diluted candidate standards. Criteria that apply to the assay of one
    candidate standard apply to the assay of multiple candidate standards. The analyst
    should be aware that the effect of any short-term calibration drift will be greater when
    multiple candidate  standards are assayed.  This greater effect is due to the longer
    period of time between reference standard measurements. Unacceptable uncertainties
    of the estimated concentrations for the diluted candidate standards may occur as a.
    result of the longer assay session.

7.  Conduct at least two additional sets of measurements, as described in step 6 above.
    However, for these subsequent sets of measurements, change the order of the three
    measurements  (e.g., measure the reference standard, zero gas, and candidate
    standard for the second  set and measure the  zero gas,  candidate standard, and
    reference standard for the third set, etc.).  Changing the order that the gas mixtures are
    measured helps the analyst to discover any effect that one measurement has on'
    subsequent measurements.  The number of sets of measurements will have been
    determined during  the analysis of the multipoint calibration data such that the 95-
    percent uncertainty for  the regression-predicted  concentration is zl percent of the
    concentration of the reference standard.

8.  If any one or more of  the measurements  of a set of measurements is invalid or
    abnormal for any reason, discard all three measurements and repeat the set of
    measurements.  Such measurements may be discarded if the analyst can demonstrate
    that the experimental conditions were inappropriate during these measurements.  Data
    cannot be discarded just because they appear to be outliers, but may be discarded if
    they  satisfy statistical criteria for testing outliers.16 The analyst must record any
    discarded  data and a brief explanation about why the data were discarded  in the
    laboratory's records.

9.  The spreadsheet described in Appendix A or equivalent statistical techniques must be
    used to calculate an  estimated concentration and a 95-percent uncertainty for the
    diluted candidate standard based on data frorp the assay measurements and from the
    multipoint calibration. The use of both sets of data in the statistical analysis produces
    an estimated concentration with smaller uncertainty while correcting for any minor cali-
    bration drift that may have occurred since the multipoint calibration.  Record the
  '  estimated concentration and the 95-percent uncertainty in the laboratory's records.
                                  3-13

-------
           The spreadsheet also calculated the percentage of the uncertainty that is,due to the
           multipoint calibration. This percentage is needed for the total uncertainty calculations
           when two or more assays fall under the same multipoint calibration. Record this value
         ... in the laboratory's records.
              •_ L'          '               '
           : ,o v.-     ; -      •• :   :  '•••  r ;-•
           The analyst should investigate any. of the measurements that appear to be outliers.
           Such data may be discarded if the analyst can demonstrate that the experimental
           conditions were inappropriate during these measurements. Data cannot be discarded
           just because they appear to be outliers, but may be discarded if they satisfy statistical
           criteria for testing outliers.  The analyst must record any discarded data as  well  as a
           brief summary of the investigation in the laboratory's records.

        10. If  the multipoint calibration data  and the  assay data underwent any mathematical
           transformations before their statistical analysis, the analyst must perform to reverse
           transformations for the estimated concentration and the 95-percent uncertainty. Record
           the transformed values in the laboratory's records.  --

        11. Finally, calculate the certified permeation rate (in nanograms/minute) and uncertainty
           for the candidate standard  using the equations below:
                                                   MW
Certified Permeation Rate =  1Q3  —  ,     «        ,,    _ .
                                      Cone.          Rate
                                                   MV
I Diluted Standard] [Dilution Flow
ll [Dili
                      Uncertainty of Permeation Rate  . n |=| | <%™£Z\ \DM™£™
3.2.10 Equilibration Test for Newly Prepared Permeation Devices

       A permeation device that has not been previously assayed must be tested for a stable
permeation rate as follows: Reassay the permeation rate at least 24 hours after the first assay and
compare the two assayed concentrations. The spreadsheet described in Appendix C or equivalent
statistical techniques must be used to evaluate the stability of the permeation rate by comparison
of the confidence limits from the two assays. If the confidence intervals overlap, the permeation rate
can be considered to be stable and the candidate standard may be certified for use.  Otherwise,
equilibrate the device longer at the operating temperature and repeat the test, using the second and
third assays as if they were the first and second. This process may be repeated until the rate
stabilizes. Permeation devices that are not stable may not be used for calibration or audit purposes.
Candidate standards that fail the initial and the repeat stability tests are unstable and are disqualified
for farther use under this protocol.

3.2.    Certification Documentation

       See Subsections 3.1.5 and 3.1.6.
                                          3-14

-------
 3.2.12  Recertification Requirements

        See Subsection 3.1.7.3.   •

 3.3     PROCEDURE P2: ASSAY AND CERTIFICATION OF PERMEATION DEVICE
        CALIBRATION STANDARDS REFERENCED TO A COMPRESSED GAS
        REFERENCE STANDARD

 3.3.1   Applicability

        This procedure may be used to assay the permeation rate of a candidate SO2 and NO2
 permeation device calibration standard, based on the concentration of a compressed gas reference
 standard of the same pollutant compound, and to certify that the assayed permeation rate is
 traceable to the reference standard. The procedure employs a  low-concentration range (i.e.,
 ambient air quality level) pollutant gas analyzer to compare quantitatively diluted concentrations
 from the permeation device calibration standard with quantitatively diluted concentrations from the
 compressed gas reference standard.  This procedure may be used for the assay of multiple
 candidate standards during the same assay session.  Criteria that apply to the assay of one
 candidate standard apply to the assay of multiple candidate standards. This procedure may be
 used by permeation device producers, standard users or other analytical laboratories.

 3.3.2   Limitations

        1.  The concentration of the diluted candidate standard may be greater than or less than
           the  concentration of  the  diluted  reference standard.    However, the  diluted
           concentrations from both standards must lie within the well-characterized region of the
           analyzer's calibration curve (see  Subsection 2.1.7.2).  Additionally, the 95-percent
           uncertainty for the regression-predicted concentration of the diluted candidate standard
           must be  sl.O percent of the concentration of the diluted reference standard.  This
           uncertainty is obtained from the statistical analysis of the multipoint calibration data
           using the spreadsheet described in Appendix  A or using equivalent statistical
           techniques (e.g., the worksheet for linear relationships given in Chapter 5 of Reference
           15). This criterion means that the uncertainty associated with the multipoint calibration
           determines the concentration range over which a diluted candidate standard may be
           assayed.

       2.   A quantitatively accurate dilution and flow measurement system is required.

       3.   A source of clean, dry zero gas is required.

       4.   This procedure is designed to assay the permeation rate of a candidate standard that
           is mounted in a specially designed assay dilution system. The procedure does not
           accommodate the certification of a candidate standard that is mounted in its own self-
           contained dilution/flow measurement system,

3.3.3   Assay Apparatus

       Figure 3-2 illustrates the components and configuration of one possible design for the assay
apparatus, including a common dilution system for both the reference and candidate standards.

                                         3-15

-------
                                                          Gas Flowmeter
                                                              (M2)
            Pressure
            Regulator
Dilution Gas
 Flowmeter
   (M1)
3-way
Valve
 (VI)
                                                   Gas Flow
                                                   Controller
                                                     (C1)
                                                                         Candidate
                                                                          Standard
                                                                          Chamber

o>

Gas Flow
Controller
  (C2)
  4-way
Valve (V2)
                             Purge Gas Flow
                                                            Purge Gas
                                                           Flow to Vent
  3-way
Valve (V3)
                Zero
                Gas
            Pressure
            Regulator
                                    Reference
                                    Standard
                                  Flowmeter (M3)
                                                                                                        Excess Gas
                                                                                                        Flow to Vent
                                                                                     das Flow
                                                                                    to Analyzer
                                                           Gas
                                                          Flow to
                                                           Vent
                                                   Gas Flow
                                                 Controller (C3)
                Reference
                Standard
                    Figure 3-2. One possible design of the apparatus for the assay of permeation device calibration
                        ~  standards referenced to a compressed gas reference standard (Procedure P2).

-------
The configuration is designed to allow convenient routing of zero gas and diluted concentrations of
the reference standard and the candidate standard, in turn, to the analyzer for measurement, as
selected by valves V1, V2, and V3. Three gas flow controllers (i.e., C1, C2, and C3) regulate the
total dilution 'flow.rate for the candidate standard, the purge gas flow rate,  and the reference
standard flow rate.  These gas flow controllers may be needle valves, capillary tubes, thermal mass
flow controllers, or other suitable devices.  The flow rates should be controlled to within 1.0 percent
variation during the assay. The dilution flow rates  for the reference and candidate standards is
measured by a single, common flowmeter (i.e., M1). The reference standard and purge gas flow
rates may be measured at the vent port of valve V2 or by flowmeters M2 and M3 that are mounted
in the two gas streams.

        When the candidate standard is being measured, valve V1  directs a  portion of the total
dilution flow through the candidate standard chamber. This sweep flow rate is regulated by gas flow
controller C1 and is measured by gas flowmeter M2.  This flowmeter need not be accurately
calibrated because only the total dilution flow rate, measured by flowmeter M1, is used in the dilution
calculations. When the reference standard is being measured, valve V1 directs the purge gas
through the candidate standard chamber. The purge gas prevents the buildup of high pollutant
concentrations in the chamber. It is vented through valve V2 and is not measured by flowmeter M1.

       The assay apparatus illustrated in Figure 3-2 may be modified by the addition of multiple
candidate standard chambers.  These chambers may be set to different temperatures.

        If it is necessary to use different dilution flow rates for the reference standard and the
candidate standard (see Subsection 3.2.6), separate flow controllers for the two dilution flow rates
may be used. However, the same flowmeter should  be used to measure both dilution flow rates to
help reduce systematic flow measurement errors.

       The mixing chamber combines the gas streams and should be designed to provide
turbulence in the flow to ensure thorough mixing of the two gas streams. The diluted gas mixtures
are routed to the analyzer through a union tee tube fitting, which vents excess gas flow. Normally,
the excess gas is vented to the atmosphere without any obstructions in the tubing and the gas
entering the analyzer is at near-atmospheric pressure.  However, the  excess  gas can be routed
through an uncalibrated rotameter by rotation of a three-way valve (i.e., V4).  The rotameter is used
to demonstrate that the total gas flow rate exceeds the sample flow rate of the analyzer and that no
room air is being drawn in through the vent line (also see Subsection 3.1.4). Check the apparatus
carefully for leaks and correct all leaks before use.

       The mean temperature of the candidate  standard chamber must be controlled to  within
0.05 °C of the setpoint with a temperature  stability of ±0.05 °C.  This  temperature must be
measured with a NIST-traceable thermometer having a measurement uncertainty of not more than
0.05 °C.

3.3.4   Pollutant Gas Analyzer

       See Subsection  2.3.4.   The pollutant gas analyzer must  have a  well-characterized
calibration curve and a range capable of measuring the diluted concentrations of both the candidate
and reference standards.  It must have good resolution, good precision, a stable response, and low
output signal noise. In addition, the analyzer must have good specificity for the pollutant of interest
so that it has no detectable response to any contaminant that may be contained in the standards.

                                          3-17

-------
 A suitable analyzer with acceptable performance specifications may be selected from the list of
 EPA-designated reference and equ   ant method analyzers.17 If the balance gas of the reference
 standard must be different from the .0 gas used for dilution (e.g., air versus nitrogen or different
 proportions of oxygen), either a high dilution ratio (i.e., at least 50 parts zero gas to 1  part standard)
 should be used, or the analyzer must be proven to be not sensitive to differences in the balance gas
 composition. The latter may be demonstrated by showing no difference in an analyzer's response
 when measuring a calibration standard that has been diluted with identical flow rates of the different
 balance gases.       ,,  a-i -.c

        The analyzer should be connected to a high-precision data acquisition system (e.g., a strip
 chart recorder) which must produce an electronic or paper record of the analyzer response during
 the assay.  Additionally, a digital panel meter with four-digit resolution, a digital voltmeter, data
 logger, or other data acquisition system must be used to obtain numerical values of the analyzer's
 response.  More precise values will be obtained if these instruments have some data averaging
 capability.  The assay record must be maintained for 3 years after the standard's certification date.

        If the analyzer has not been in continuous operation, turn it on and allow it to stabilize (e.g.,
 for at least 12 hours) before beginning any measurements.

 3.3.5   Analyzer Calibration

3.3.5.1  Multipoint Calibration-
        See Subsections 2.1.7.2 and 2.1.7.4. Following completion of the.multipoint calibration, the
accuracy of the assay apparatus must be checked to verify that the error associated with the dilution
is not excessive. This accuracy check involves the measurement of an undiluted or diluted check
standard.   The check standard must be a NIST SRM, an SRM-equivalent PRM, an NTRM, or a
GMIS as specified in Subsection 2.1.2. It must have a certified concentration that is different from
that  of the reference standard used in the multipoint calibration.  Information concerning this
standard (e.g., cylinder identification number, certified  concentration)  must be recorded in the
laboratory's records.

        If an undiluted check standard is used, its concentration must fall in the well-characterized
region of the calibration curve. If a diluted check standard is used, the diluted concentration must
fall in the well-characterized region.

        Make three or more discrete  measurements of the undiluted or diluted check standard.
"Discrete"  means that the analyst must change the gas mixture  being sampled by the analyzer
between measurements. For example, the analyst might alternate between measurements of the
check standard and the zero gas.  Record these measurements in the laboratory's records.

        Next the analyst  must verify that the dilution error is not excessive. For the .check standard
measurements, calculate the relative difference (in percent) between the mean analyzer response
and the corresponding response that is predicted from the multipoint calibration regression equation
and the undiluted or diluted check standard concentration. That is,
            Relative Difference    100
Mean Analyzer Response - Predicted Response
            Predicted Response
                                          3-18

-------
 If the relative difference is >1.0 percent, the dilution error is considered to be excessive. The analyst
 must investigate why the relative difference is excessive.  The program may be due to errors in the
 reference standard and check standard concentrations, errors in assay apparatus or to some other
 source. Assays may not be conducted until the relative difference for a subsequent accuracy check
 is £ 1.0 percent.

 3.3.5.2 Analyzer Range—
       See Subsection 2.3.5.2.

 3.3.5.3 Linearity-
       See Subsection 2.3.5.3.

 3.3.5.4 Zero and Span Gas Checks-
       See Subsection 2.3.5.4.

 3.3.6  Selection of Gas Dilution Flow Rates

     .  The dilution flow rates used for the reference standard and the candidate standard should
 be selected carefully to provide diluted concentrations for both standards that fall in the well-.
 characterized region of the analyzer's calibration curve. Potential errors in the assayed permeation
 rate due to dilution flow rate measurement error will be reduced if the same dilution flow rates can
 be used for both the reference and candidate, standards. This should be feasible by appropriate
 selection of the reference standard flow rate.  Select a combination of reference standard flow rate
 and dilution flow rate that produces approximately equal diluted reference standard and candidate
 standard concentrations that fall in the well-characterized region of the analyzer's calibration curve.
 Additionally, the magnitude of the 95-percent confidence limits for the estimated concentration of
 the diluted candidate standard must be s±1  percent of the concentration of the diluted reference
 standard.

 3.3.7  Flowmeter Type and Flowmeter Calibration

       Flowmeters  M1 and M3, shown in Figure 3-2, measure in-line flow rates and do not operate
 at atmospheric pressure because of backpressure from downstream components.  Also, this
 backpressure is variable, depending on the total dilution and reference standard flow rates. Thus,
the flowmeters must compensate for the variable in-line pressure. Thermal mass flowmeters do not
 need to be corrected for pressure effects. Measurements from pressure-sensitive flowmeters such
as rotameters or wet test meters must be carefully corrected for the actual in-line pressure during
the flow rate measurements.

       Alternatively, the flow rates can be measured at the outlet of the dilution apparatus, with the
excess gas flow vent temporarily plugged. In this case, a volume-type meter such as a wet test
meter  or a soap film flowmeter can be used, and flow  measurements may be conveniently
 referenced to atmospheric pressure.  Each flow rate must be measured independently while the
other flow rate is set to zero. Great care must be exercised to ensure that each measured flow rate
 remains constant when combined with the other flow rate and between  the time of measurement
and the time of the assay.

       The flowmeters used should'be stable, repeatable, linear, and have good resolution. The
flowmeters must not contaminate or react with the gas mixture passing through them.  If possible,

                                          3-19

-------
 select flpwrates or r«pwmeter ranges such that the measured flow rates fall in the upper half of the
 flowmeters' ranges. The flowmeters should be carefully calibrated at several flow rates to prove
 linearity. Thacalibratjpr\of the zero gas flowmeter should be accurate to ±1.0 percent, referenced
 to an accurate flow or volume standard traceable to a NIST primary standard (see Subsection
 3.1.2).  This flowmeter calibration should be checked and recertified on a regular schedule (e.g.,
 yearly).  The recertifjcation frequency is to be determined from stability information such as a
 chronological control chart of calibration data.

        It is desirable to measure both dilution flow rates with the same flowmeter (i.e., M1).  This
 practice reduces measurement errors associated with the use of multiple flowmeters. Note that the
 impact  of any flow measurement error is reduced if the same dilution ratio can be used for both
 candidate standard and reference standard measurements.

        All volumetric  flow-rate measurements must be corrected or referenced to the same
 temperature and pressure conditions, such as EPA-standard conditions (i.e., 760 millimeters of
 mercury (mm Hg) and 25 °C) or the ambient temperature and pressure conditions prevailing in the
 laboratory during the assay. Measurements using wet test meters and soap bubble flowmeters also
 must be corrected for the saturation of the gas stream with water vapor in the moist interiors of these
 flowmeters.  The equation to correct the flow rate for temperature, pressure, and humidity effects
 is given below:
                         .Flow Rate  =  Volume
                                      Time

where

     PM  = measured barometric pressure (mm Hg);
    PWV  = partial pressure of water vapor (mm Hg);
     Ps  = s:andard pressure (mm Hg);
     Ts  = : landard temperature (298.2 K); and
     TM  = measured ambient temperature (273.2 + °C).

3.3.8  Candidate Standard

       See Subsections 3.1.7 and 3.2.2.  Follow the manufacturer's instructions for equilibration
and for use of the candidate standard and for selecting the temperature at which it is to be assayed
and certified. The candidate standard should be assayed at the same temperature at which it will
be subsequently used. The mean operating temperature of the candidate standard chamber must
be controlled to within 0.05 °C  of the  setpoint with a temperature stability of ±0.05  °C.  This
temperature must be measured with a NIST-traceable thermometer with a measurement uncertainty
±0.05 °C or less (see Subsection 3.1.2).

3.3.9  Reference Standard

       See Subsections 2.1.2,2.1.6.4,2.3.2,2.3.6, and 3.1.2.
                                         3-20

-------
3.3.10 Zero Gas

       See Subsection 2.1.8.  If possible, the zero gas should be the same as the balance gas of
the reference standard.
                    /
3.3.11 Assay Procedure

       1.  Verify that the assay apparatus is properly configured as shown in Figure 3-2 and
           described in Subsection 3.3.3.  Inspect the analyzer to verify that it appears to be
           operating normally and that all controls are set to their expected values. Record these
           control values in the laboratory's records.

       2.  Determine and establish the operating temperature for the candidate standard in its
           temperature-controlled chamber.  Install the candidate standard, start the purge gas
           flow, and allow ample time for the device to equilibrate  (see Subsection 3.1.7.2).
           Record the temperature in the laboratory's records.

       3.  Verify that the flowmeters are property calibrated (see Subsection 3.3.7).

       4.  Verify that a multipoint calibration of the analyzer has been performed within 1  month
           prior to the assay date (see Subsections 2.1.7.2,2.1.7.5, and 2.3.4). Additionally, verify
           that the zero and  span gas checks indicate that the  analyzer is in calibration (see
           Subsection 2.3.5.4).

       5.  Determine and establish the reference standard flow rate and the dilution flow rates and
           diluted concentrations for the reference standard and the candidate standard that will
           be used  for the assay (see Subsections 2.3.5.2 and 3.3.6).  Ensure that the diluted
           reference standard and diluted candidate standards concentrations are within the well-
           characterized region of the analyzer's calibration curve (see Subsection 2.3.2). Also
           check, that both dilution  flow rates will provide enough flow for the analyzer, with
           sufficient excess to ensure  that no ambient air will be drawn into the vent line.  If
           possible, use the  same  dilution flow rate for both the reference standard and the
           candidate standard. Also adjust the flow  rate of the portion of the dilution flow that
           passes over the candidate standard (i.e., flow controller C3), and adjust the purge flow
           rate (i.e., flow controller C2) to approximately the same value.


                               Stendard = [(Undiluted Standard Cone.) (Standard Flow Rate)
                                 Cone.    i (Standard Flow Rate  * Zero Gas Flow Rate)

           Calculate the diluted reference standard concentration using the following equation:
           Calculate the diluted candidate standard concentration (in ppm) using the following
           equation:

                      ' rvi ^ j o.  A A  o      ho~3l [ MV 1 f Permeation Rate ]
                       Diluted Standard  Cone. =  IU   ——   ———-—-—
                                             L    J  MW   D ution Row Rate
                                          3-21

-------
    where

    MV   = Molar volume of the dilution gas (liters/mole);
             'urr
          = (0.08206) Tm

    MW   = Molecular weight of the dilution gas (grams/mole); permeation rate is given in
           nanograms/minute; and dilution flow rate is given in liters/minute.
                               i  -.p     v  .     :
    Use an estimated permeation rate for the candidate standard in this calculation.

    Record the measured flow'rates and the undiluted and diluted reference standard
    concentrations in the laboratory's records.

6.  In succession, measure the zero gas, the diluted reference standard and the diluted
    candidate standard using the analyzer. Use valves V1, V2, and V3 to select each of
    the three gas mixtures for measurement For each measurement, adjust the flow rates,
    if necessary, to those determined  in step 5, and allow ample time for the analyzer to
    achieve a stable reading.  If the  reading for each measurement is not stable, the
    precision of the measurements will decline and the candidate standard might not be
    certifiable under this protocol. Record the analyzer response for each measurement,
    using the same response units (e.g., volts, millivolts, percent of scale, etc.) as was used
    for the multipoint calibration. At this point, do not convert the data into concentration
    values using  the  calibration  equation.   Do  not  perform any  mathematical
    transformations Of the data.  These steps will be done later.  Do not make any zero
    control, span control, or other adjustments  to the analyzer during  this set  of
    measurements. Record these analyzer responses in the laboratory's records.

    The analyst may assay multiple candidate standards during the same assay session.
    For example, a single set of measurements may involve a zero gas, a diluted reference
    standard, and three diluted candidate standards. Criteria that apply to the assay of one
    candidate standard apply to the assay of multiple candidate standards.  The analyst
    should be aware that the effect of any short-term calibration drift will be greater when
    multiple candidate standards are assayed.  This greater effect is due to the longer
    period of time between reference standard measurements. Unacceptable uncertainties
    of the estimated concentrations for the diluted candidate standards may occur as a
    result  of the longer assay session.

7.  Conduct at least two additional sets of measurements, as described in step  6 above.
    However, for these subsequent sets of measurements, change the order of the three
    measurements (e.g.,  measure the reference standard, zero gas,  and candidate
    standard  for the  second set and measure the zero gas, candidate standard, and
    reference standard for the third set, etc.). Changing the order that the gas mixtures are
    measured helps  the analyst to discover any effect that one measurement has on
    subsequent measurements.  The number of sets of measurements ,vill have been
    determined during the analysis of the multipoint calibration data SL -  that the 95-
    percent uncertainty of the regression-predicted concentration of the candidate standard
    z 1 percent of the concentration of the reference standard.
                                  3-22

-------
8.  If any one or more of the measurements of a set of measurements is invalid or
    abnormal for any reason, discard all three  measurements and repeat the set of
    measurements. Such measurements may be discarded if the analyst can demonstrate
    that the experimental conditions were inappropriate during these measurements. Data
    cannot be discarded just because they appear to be outliers, but may be discarded if
    they satisfy statistical criteria for  testing outliers.16  The analyst must record any
    discarded data and a brief explanation about why the data were discarded in the
    laboratory's records.

9.  The spreadsheet described in Appendix A or equivalent statistical techniques must be
    used to calculate an estimated  concentration and a 95-percent uncertainty for the
    diluted candidate standard based on data from the assay measurements and from the
    multipoint calibration. The use of both sets of data in the statistical analysis produces
    an estimated concentration with smaller uncertainty  while correcting for any minor
    calibration drift that may have occurred since the multipoint calibration.  Record the
    estimated concentration and the 95-percent uncertainty in the laboratory's records.

    The spreadsheet also calculated the percentage of the uncertainty that is due to the
    multipoint calibration.  This percentage is needed for the total uncertainty calculations
    when two or more assays fall under the same multipoint calibration.  Record this value
    in the laboratory's records.

    The analyst should investigate any of the measurements that appear to be outliers.
    Such data may be discarded if the analyst can demonstrate that the experimental
    conditions were inappropriate during these measurements. Data cannot be discarded
    just because they appear to be outliers, but may be discarded if they satisfy statistical
    criteria for testing outliers.  The analyst must  record any discarded as well as a brief
    summary of the investigation in the laboratory's records.

10. If the multipoint  calibration data and the assay data underwent any mathematical
    transformations before their statistical analysis, the analyst must perform to reverse
    transformations for the estimated concentration and the 95-percent uncertainty. Record
    the transformed values in the laboratory's, records.

11. Finally, calculate the certified permeation rate  (in nanograms/minute) and uncertainty
    for the candidate standard using the equations below:
                Certified Permeation  Rate =
                                  [Diluted Standard
                                  [     Cone.
           Dilution Flow
               Rate
Uncertainty of Permeation Rate = '10
                                                MW
                                                MV
95-Percent
 Uncertainty
Dilution Flow
    Rate
                                   3-23

-------
 3.3.12 Equilibration Test for Newly Prepared Permeation Devices

       A permeation device that has not been previously assayed must be tested for a stable
 permeation rate as follows: Reassay the permeation rate at least 24 hours after the first assay and
 compare the two assayed concentrations. The spreadsheet described in Appendix C or equivalent
 statistical techniques must be used to evaluate the stability of the permeation rate by comparison
 of the confidence limits from the two assays. If the confidence intervals overlap, the permeation rate
 can be considered to be stable and the candidate standard may be certified for use.  Otherwise,
 equilibrate the device longer at the operating temperature and repeat the test, using the second and
 third assays as if  they were the first and second. This process may be repeated until the rate
 stabilizes. Permeation devices that are not stable may not be used for calibration or audit purposes.
 Candidate standards that fail the initial and the repeat stability tests are unstable and are disqualified
 for further use under this protocol.

 3.3.13 Certification Documentation

       See Subsections 3.1.5 and 3.1.6.

 3.3.14 Recertification Requirements

       See Subsection 3.1.7.3.
3.4    PROCEDURE P3: ASSAY AND CERTIFICATION OF PERMEATION DEVICE
       CALIBRATION STANDARDS REFERENCED TO A MASS REFERENCE STANDARD

3.4.1   Applicability

       This procedure may be used to assay the permeation rate of a candidate SO2 or NO2
permeation device calibration standard based on mass reference standards, and to certify that the
assayed  permeation rate is traceable to the reference standard. The procedure employs an
analytical balance to measure the weight loss in the candidate standard. It may be used for the
assay of  multiple candidate standards during the same assay session. Criteria that apply to the
assay of one candidate standard apply to the assay of multiple candidate standards.  This
procedure may be used by permeation device producers, standard users, or other analytical
laboratories.

3.4.2   Limitations

       1. This procedure is intended only for the assay of candidate standards containing SO2
          or NO2.  These liquid compounds must be anhydrous grade (minimum purity 99.99
          percent) or phosphorous pentoxide-dried commercial purity grade (minimum purity 99.5
          percent).

       2. An accurate analytical balance with a NIST-traceable calibration is required to weigh
          the candidate standard.

       3. A temperature-controlled chamber for maintaining the candidate standard at a constant,
           NIST-traceable temperature between weight measurements is required.

                                         3-24

-------
        4.  A source of clean, dry zero gas is required.

3.4.3   Assay Apparatus

3 4.3.1 Analytical Balance-
        Choose  a balance with  adequate  vibration-stabilization  control  and  appropriate
specifications for total weighing capacity, accuracy, precision, and readability. The balance should
be chosen such that the manufacturer's specified uncertainty (i.e., three times the standard
deviation or "reproducibility") of the balance divided by the weight of the candidate standard does
not exceed 0.001. The balance must be calibrated annually using NIST-traceable mass reference
standards  by the manufacturer or a manufacturer's representative (see Subsection 3.1.2).

        If possible, locate the balance in a climate-controlled, draft-free room, preferably dedicated
to the use of balances.  If this is not possible, the general guidelines listed below should be followed
to control environmental factors that may affect balance performance:

        •  Locate the balance away  from potential sources of drafts such as doors, windows,
           aisles with frequent traffic, ventilation ducts, and equipment with fans or moving parts.

        •  Locate the balance out of direct sunlight and away from local heating or cooling sources
           such as open flames, hot  plates, water baths, ventilation ducts, windows, and heat-
           producing lamps.

        •  Locate the balance on a sturdy base (ideally, a stone weighing table) and away from
           any equipment that produces vibrations. If this is not possible, isolate the balance from
           such equipment by placing  a stabilizing slab under the balance or composite damping-
           pads under the balance legs.

        •  Ensure that the balance-support is sufficiently level to permit leveling of the balance
           according to the  manufacturer's instructions.

3.4.3.2 Temperature-controlled Chamber—
        A temperature-controlled chamber is required for storing the candidate standard between
weighings. One possible design for the chamber is depicted in Figure 3-3.23 Clean, dry zero gas
enters the  chamber at the bottom after passing through the heat exchanger tubing (i.e., several
turns of copper tubing).  The zero gas1 flow rate must be sufficient to purge the chamber thoroughly.
The chamber and the heat exchanger are immersed in a thermostatted bath to the level shown in
the figure.  The  bath must control the mean temperature of the chamber to within 0.05 °C of the
setpoint with a temperature stability of ±0.05 °C. The temperature of the bath or the chamber must
be measured and recorded in the laboratory's records on at least a daily basis. A NIST-traceable,
liquid-in-glass thermometer or a temperature-sensing device must be used for this measurement
(see Subsection 3.1.2). A temperature-sensing device must be calibrated annually using NIST-
traceable temperature reference standards and must have an uncertainty similar to that of these
reference standards. The output of a temperature-sensing device may be recorded by a strip chart
recorder, data logger, or other data acquisition system.
                                          3-25

-------
                                        Air Outlet
        Thin Tube
  Water Level
          Perforated
            Spacer
Permeation Tube in
     Position
             1"
                                                  Removable Cap
Air Inlet
                Air Entering Heat
                Exchanger Tubing


                  Water Level
                                                     Permeation Tube
                                                          Holder
                                                        Heat
                                                     Exchanger
                                                       Tubing
                                                           Perforated Disk
                                                            Bottom Plate
   Figure 3-3. Chamber for maintaining permeation tubes at constant temperature.
                                                                     23
                                   3-26

-------
3.4.3.3 Electrostatic Charge Neutralization-
        Electrostatic charge buildup will prevent an analytical balance from operating properly.
Static charge is the accumulation of electrical charges on the surface of a nonconductive material,
which could be the permeation device or part of the analytical balance.  Common symptoms of this
problem include noisy readout, drift, and sudden readout shifts.

        To reduce static charge within the balance,  it may be necessary to place a radioactive
antistatic strip containing a very small amount (i.e., 500 picocuries) of Polonium-210 (Po210) in the
weighing chamber.  It may also be necessary to put each permeation device on an antistatic strip
before it is weighed.  Po210 antistatic strips are used to reduce electrostatic buildup in the analytical
balance's weighing chamber and on individual permeation devices by charge neutralization. They
will neutralize electrostatic charges on items placed within an inch of them. These antistatic strips
are safe, commonly available, and very inexpensive. Po210 has a half-life of 138 days. Change the
antistatic strips semiannually and dispose  of the old strips according to the manufacturer's
recommendations.

        Antistatic solutions are  available for coating  (and recoating at appropriate and relatively
infrequent intervals) the interior and exterior nonmetallic surfaces of the chamber. This coating
facilitates the draining of electrostatic charges from these surfaces to a common electrical ground
to which the metallic conductive surfaces are connected. Earth-grounded conductive mats placed
on the weighing table surface and under the analysts shoes are used to reduce electrostatic charge
buildup.  Do not assume that  the electrical grounding of the analytical balance eliminates all
electrostatic buildup because the ground may not be perfect.

        Even though a permeation device's weight might stabilize within 60 seconds and no weight
drift is observed during that period, the balance may still be influenced by electrostatic  charge
buildup. It may still be necessary to repeat the neutralization procedure and to use antistatic strips
inside  the weighing chamber. One may reduce the effect of electrostatic buildup on permeation
devices by putting them in an aluminum foil boat on the balance pan during weighings.

        Charge neutralization times may need to be longer than 60 seconds. Electrostatic charge
buildup becomes greater as the air becomes drier. A 60-second neutralization may work sufficiently
in ambient indoor air conditioned to 37 percent relative humidity and 23°C, but not in zero nitrogen.
This latter environment may require that the permeation device sit for more  time on the antistatic
strip. The longer neutralization period may have to be done inside the weighing chamber or a
second small chamber, which is used just for charge neutralization.

3.4.4  Weighing Interval

       The minimum time period between weighings of the candidate standard is a function of the
expected permeation rate, the specified uncertainty for the rate,  and the analytical balance's
readability (i.e.,  the smallest scale division). The following equation  is based on a ±1 percent
uncertainty specification for the  permeation rate:
                                       •   100 (Readability)
                           Weighing  =     -- - - -
                            Interval    (Expected Permeation Rate )
                                          3-27

-------
where the weighing interval is in minutes; the readability is in grams; and the expected permeation
rate is in grams per minute.

3.4.5   Assay Procedure

       1.  Turn on the balance and allow it to warm up for the period specified in the operator's
           manual. To ensure maximum stability, it is recommended to keep the balance turned
           on at all times.  This procedure enables the balance to be operational at all times and
           eliminates the need for a warmup period, before analyses. Newer balances are always
           turned on (except for their displays) when they are plugged in.

       2.  Check  the  balance level  and,  if necessary, adjust the level  according to the
           manufacturer's instructions.

       3.  Ensure that the balance room temperature is within 15 to 30 °C or, if given, within the
           balance manufacturer's specifications and that  the  balance and mass reference
           standards are equilibrated to the balance room temperature. Record the temperature
           in the laboratory's records.

       4.  Zero (i.e., tare) and calibrate the balance according to the manufacturer's directions.
           Record the tare reading in the laboratory's records. Many newer balances calibrate
           themselves automatically or only require a key to be touched to calibrate themselves.

       5.   On each  day that the candidate standard is to be  assayed,  verify the balance's
           calibration using at least two NIST-traceable mass  reference standards. Use smooth,
           nonmetallic forceps to handle the standards. This standard must have a mass that is
           similar to that of  the candidate standard.  Record the date, balance identificatipn,
           standards identification, certified weight of the standard, and the measured weight of
           the standard in the laboratory's records.  Calculate the relative difference (in  percent)
           between the certified and measured weights as follows:


                    Relative    100 (Measured Weight  -  Certified Weight)
                    Difference              (Certified Weight)
          Record the relative difference in the laboratory's records.  If the relative difference is
          >0.1  percent, the balance cannot  be used under this protocol until it has been
          recalibrated or repaired and until a subsequent verification has a relative difference of
          sO.1 percent.

       6.  Review the  recorded bath or chamber temperature readings since the most recent
          weighing  of the candidate standard,  or since the standard was first put into the
          temperature-controlled chamber.  Record the minimum and maximum temperatures in
          the laboratory's records. The minimum and maximum temperatures must not have
          deviated from the setpoint by more than 0.1  °C.  If these temperatures deviate by more
          than this amount, the current assay and all previous assays are invalidated.

       7.  'Record the current bath or chamber temperature in the laboratory's records.

                                         3-28

-------
       8.  Verify that the candidate standard has been in the temperature-controlled chamber for
           a long enough time for its permeation rate to have stabilized.

       9.  Remove the candidate standard from the temperature-controlled chamber and place
           it on the balance's pan using stainless steel forceps or a similar noncontaminating
           device.  Note that Teflon* permeation tubes may have an electrostatic charge buildup
           due to the passage of the dry gas over them between weighings. Such charges should
           be removed from the candidate standard before weighing by Po210 antistatic strips or
           similar devices.  Note that electronic force balances may require that candidate
           standards be thermally equilibrated before they can be weighed.

       10. Record the date, time and the candidate standard's identification number and current
           weight in the laboratory's records.

       11. Return the candidate.standard to the temperature-controlled chamber. The standard
           should be outside the chamber only for a long enough time to weigh it using reasonable
           laboratory technique.

3.4.6  Number of Weighings of the Candidate Standard

       The candidate standard must be weighed at least six times after its permeation rate has
stabilized at the certification temperature. After the six or more weighings, the analyst may assess
the stability and uncertainty of the permeation rate by using the spreadsheet described in Appendix
B  or equivalent  statistical or graphic  techniques.   The analyst may  calculate a  provisional
permeation rate from the measured weights and the time between weighings using the following
equation:
                        Pemeation  =  (Previous We'9ht - Current We'9ht)
                                     Elapsed Time Between Weighings
Based on this data analysis, the analyst may perform additional weighings to reduce the uncertainty
or to  replace data that are discarded because they were obtained before the permeation rate
stabilized.

3.4.7   Calculation of Certified Permeation Rate

       The certified permeation rate for the candidate standard is the slope of the least squares
regression line for data from at least six weighings of the candidate standard after the permeation
rate has stabilized. This statistical analysis technique produces permeation rate estimates that are
more precise than those calculated from weight differences between individual weighings.  Although
the minimum  number of weighings is six,  more  precise estimates will be obtained  for more
weighings.
                                          3-29

-------
        Calculate the certified permeation rate and its uncertainty using the spreadsheet given in
Appendix B or using equivalent statistical techniques (e.g., the worksheet for linear relationships
given in Chapter 5 of Reference 15).  Figure 3-4 presents an example of the graphic output of the
spreadsheet This rate is the slope (bt) of the least squares regression line where the x-values are
the elapsed times from the initial weighing and the y-values are the measured weights of the
permeation device.  The spreadsheet also  calculates the predicted initial weight  (b0) of the
permeation device and 95-percent confidence limits for b0 and b,.

       After the data  from the six or more weighings have been entered in  the spreadsheet,
examine the 95-percent confidence  limits (CLs) for b0. If the measured weight from the initial
weighing falls outside  of these  limits, the permeation device may not have  been  completely
equilibrated at the initial weighing.  The analyst may  elect to discard the data from the initial
weighing to reduce the uncertainty of the certified permeation rate.

       Examine the upper and lower 95-percent CLs for bv They should differ from b1 by no more
than 1 percent of its value.  That is,
                                 upper CLfbJ-b,  * (b^/100 .
b, -lower
                                            -   <.
           4.36
                                              Adjust the Measured Weight scale
                                              minimum and maximum as needed
                                              to view the finear regression tine and
                                              Its confidence limits.
          4.18
                       15,000
  30,000     45,000   •   60,000
        Elapsed Time (min)
75,000
90,000
                       Figure 3-4. Example of spreadsheet graphic
                         output for calculating permeation rates.

                                           3-30

-------
       If these two criteria are met, the permeation device can be certified with a permeation rate
equal to b, and an uncertainty equal to the larger of the two values. If the criteria are not met, the
analyst must make additional weighings of the candidate standard as described in Subsection 3.4.5.
These additional measurements will be pooled with the previously collected measurements.  The
pooled data will be used to obtain new estimates of the b1 and its uncertainty. When an acceptable
value  for  the uncertainty is obtained,  record it and the slope in the laboratory's records.  If an
acceptable value is not obtained, the candidate standard cannot be certified under this protocol.

       The analyst should investigate any of the measurements that appear to be outliers. Such
data may be discarded if the analyst can demonstrate that the experimental conditions were
inappropriate during these measurements. Data cannot be discarded just because they appear to
be outliers but may be discarded if they satisfy statistical criteria for testing outliers.  The analyst
must record any discarded data and a brief summary of the investigation in the laboratory's records.

3.4.8  Uncertainty of Certified Permeation Rate for Candidate Standard

       The total analytical uncertainty of the certified permeation rate includes the uncertainty of
regression slope and the uncertainty of the  mass reference standard that was used to verify the
balance's calibration. The two components are  combined using the following equation for the
propagation of errors:
             Uncertainty (Total)
              Permeation Rate
( Uncertainty (Slope) Y f ( Uncertainty (Mass)"!
       Slope      J    {      Mass     J
                                                                       2
3.4.9  Certification Documentation

       See Subsections 3.1.5 and 3.1.6.

3.4.10 Recertification Requirements

       See Subsection 3.1.7.3.
                                          3-31

-------
                                      SECTION 4

                                    REFERENCES
 1.     U.S. Environmental Protection Agency.  Quality Assurance Handbook for Air Pollution
       Measurement Systems, Volume II. Ambient Air Specific Methods.  EPA-600/4-77-027a
       1977.

 2.     U.S. Environmental Protection Agency.  Quality Assurance Handbook for Air Pollution
       Measurement Systems, Volume III. Stationary Source Specific Methods.  EPA-600/4-77-
       027b, 1977.

 3.     U.S. Environmental Protection Agency. Code of Federal Regulations.  Title 40, Chapter I,
       Subchapter C, Part 50. "National Primary and Secondary Ambient Air Quality Standards."
     •  Washington, DC.  Office of the Federal Register.  July 1,1993.

 4.     U.S. Environmental Protection Agency. Code of Federal Regulations.  Title 40, Chapter I,
       Subchapter C, Part 58. "Ambient Air Quality Surveillance." Washington, DC. Office of the
       Federal Register.  July 1,1993.

 5.     U.S. Environmental Protection Agency. Code of Federal Regulations.  Title 40, Chapter I,
       Subchapter C,  Part  60.   "Standards of Performance for New Stationary Sources."
       Washington, DC.  Office of the Federal Register.  July 1,1993.

 6.     U.S. Environmental Protection Agency. Code of Federal Regulations.  Title 40, Chapter I,
       Subchapter C, Part 75. "Continuous Emission Monitoring." Washington, DC. Office of the
       Federal Register.  July 1,1993.

7.     F.R. Guenther,  W.D.  Dorko, W.R. Miller, and G.C. Rhoderick.  The NIST Traceable
       Reference Program for Gas Standards. National  Institute of Standards and Technology.
       Special Publication 260-126.1996. 40 pp.

8.     Nederlands Meetinstituut. Gaseous Primary Reference Materials. Delft, The Netherlands,
       no date. 12 pp.

9.     B.N. Taylor and  C.E. Kuyatt. Guidelines for Evaluating and Expressing the Uncertainty of
       NIST Measurement Results: 1994 Edition. National Institute of Standards and Technology.
       Technical Note  1297.  1994.  Current edition available from NIST Calibration Program,
       Building 820, Room 232, Gaithersburg, MD 20899, (301) 975-2002.

10.    G.L. Harris.  State Weights and Measures Laboratories: State  Standards Program
       Description.  National  Institute of Standards and Technology. Special Publication 791.
       1994.130pp.

11.    V.R. White. National Voluntary Laboratory Accreditation Program: 1997 Directory. National
       Institute of Standards and Technology. Special Publication 810.1997. 225 pp.
                                         4-1

-------
 12.    R.C. Shores, F. Smith, and D.J. von Lehmden. "Stability Evaluation of Sulfur Dioxide, Nitric
       Oxide and Carbon Monoxide Gases in Cylinders." EPA/600/4-86, U.S. Environmental
       Protection Agency, Research Triangle Park, NC. 19847 52 pp.

 13.    S.G. Wechter. "Preparation of Stable Pollution Gas Standards Using Treated Aluminum
       Cylinders." In Calibration in Air Monitoring, ASTM STP 598, American Society for Testing
       and Materials. Philadelphia, PA. 1976. pp. 40-54.

 14.    National Council of the Paper Industry for Air and Stream Improvement, Inc, "An Investi-
       gation of the Stability of H£ in Air Cylinder Gases." NCASI Special Report No. 90-09, New
       York, NY. 1990. 13pp.

 15.    M.G. Natrella. Experimental Statistics, National Bureau of Standards Handbook No. 91,
       U.S. Government Printing Office. Washington, DC. 1963. pp. 5-1 to 5-46.

 16.    American Society for Testing and Materials. Standard Practice for Dealing with Outlying
       Observations. ASTM Standard Practice E 178-80,1980.

 17.    U.S. Environmental Protection Agency.   List of Designated  Reference and Equivalent
       Methods. Current edition available from U.S. Environmental Protection Agency, National
       Exposure Research Laboratory, Mail Code MD-77, Research Triangle Park, NC 27711,
       (919) 541-2622 or from the Ambient Monitoring Technology Information Center (AMTIC) at
       http://www.epa.gov/ttn.

 18.    American National Standards Institute (ANSI) and American Society for Testing and
       Materials (ASTM).   Standard Specification for Laboratory Weights and Precision  Mass
       Standards. ANSI/ASTM Standard E 617-91,1991.

 19.    W. Kupper. "High Accuracy Mass Measurements, From  Micrograms to Tons," Instrument
       Society of America Transactions. 29(4). 1990.

20.    G. Harris.   "Ensuring Accuracy and  Traceability of  Weighing Instruments," ASTM
       Standardization News. 21(4):44-51. 1993.

21.    J.A. Wise. NIST Measurement Services: Liquid-in-Glass Thermometer Calibration Service.
       National Institute of Standards and Technology Special Publication 250-23. 1988.

22.    J.A. Wise. A Procedure for the Effective Recalibration of Liquid-in-Glass Thermometers.
       National Institute of Standards and Technology Special Publication 819. 1991.

23.    E.E. Hughes et al.  "Performance of a Nitrogen Dioxide Permeation Device."  Analytical
       Chemistry. 49(12):1823-1829. 1977.

24.    W.J. Mitchell  et al.  "Simple Systems for Calibrating and Auditing SO2 Monitors at Remote
       Sites." Atmospheric Environment. 26A(1):191-194. 1992.

 25.    F.P. Scaringelli et al.  "Preparation of Known Concentrations of Gases and Vapors with
       Permeation Devices Calibrated Gravimetricany."  Analytical Chemistry. 42(8):871-876.
       1970.
                                         4-2

-------
26.    D.L. Williams. "Permeation Tube Equilibration Times and Long-Term Stability." Calibration
       in Air Monitoring.  ASTM Publication STP 598. American Society for Testing and Materials.
       1976. pp. 183-197.
                                          4-3

-------
             APPENDIX A: INSTRUCTIONS FOR CALIBRATION WORKBOOK

 1.  ReadMe Spreadsheet

 Purpose

 This worksheet supports linear, quadratic, cubic, and quartic models:
    Linear:    y = fa + $x+ e
    Quadratic: y = $, + #x + /^x2 +.e   .
    Cubic:     y = fa + $x + /Jc2 + &X3 + e
    Quartic:   y = & + #x + ^x2 + /^x3 4- /?4x* + e
              y = response
              x = concentration

The worksheet estimates the coefficients (ps) and the variance of the error term, e. The
workbook then performs the following functions:

    •   determine which model (linear, quadratic, etc.) is better
    •   determine the replication of unknowns needed for uncertainty control
    •   determine whether zero and span  responses are acceptable
    •   estimate the concentration and 95% uncertainty of candidate standards analyzed on the
       same day as the initial calibration or a subsequent day.

Organization

The workbook consists of several worksheets, which are displayed as tabs at the bottom of the
screen. The functions of these worksheets are described below:

    ReadMe                describes the workbook, explaining how to use the worksheets
    Measurement Data     allows for user input of calibration and other analytical data and
                           includes statistical calculations for polynomial regression
    Curves 1               displays the calibration data, the best-fit line, and its confidence
                           bands
    Residuals 1             displays the difference between the observed responses and
                           those estimated by the best-fit calibration line
    Curves 2               displays the calibration data, the best-fit quadratic curve, and its
                           confidence bands
    Residuals 2            displays the difference between the observed responses and
                           those estimated by the quadratic regression line
    Curves 3               displays the calibration data, the best-fit cubic curve, and its
                           confidence bands
    Residuals 3            displays the difference between the observed responses and
                           those estimated by the best fit cubic regression  line
    Curves 4               displays the calibration data, the best-fit cubic curve, and its
                           confidence bands
    Residuals 4            displays the difference between the observed responses and
                           those estimated by the best-fit quartic regression line

                                         A-1

-------
    Chart Data             includes the data used to create the Curves and Residuals charts.
                                                                       .imifc
 Conventions                       ...

 The Measurement Data worksheet includes instructions that guide the user through the steps in
 its use. The worksheet is also color coded to simplify use. Shaded cells that are bordered in
 blue lines are for input of data. These cells are unprotected, but all other cells of the
 Measurement Data worksheet are protected. The only other unprotected cell in the workbook is
 cell F4 of the Chart Data worksheet  That cell controls the width of the confidence bands plotted
 in the Curves 1 and Curves 2 charts.

 Derived values and statements are colored red. These cells contain formulas and are
 protected to prevent alteration.

 Spreadsheet background colors indicate the order of the polynomial supported by the
 calculations in the area.

                    Light green is used for the linear model.
                    Yellow is used for the quadratic model.
                    Gray is used for the cubic model
                    Light blue is used for the quartic model.

 Use

The Measurement Data worksheet guides the user through six steps.

 STEP 1   Enter Calibration Data

 In this step, up to 50 calibration points may be entered. Each calibration point has two parts-the
certified concentration of the calibration gas standard and the instrument response when testing
the standard. These values are entered in two columns. The spreadsheet performs
computations in columns I through P (linear), Q through X (quadratic), Y through AZ (cubic), and
BA and above-(quartic).

 STEP 2   Review the Parameter Estimates

 In this step, the user reviews the estimates of the intercepts (b0), slopes (b,) and other
coefficients (b2, b3, and b4) for the four models, examines their confidence intervals and the
 residual error variances (s2). The result of an F test indicates which of the models is best. The
 linear model is recommended unless  the quadratic or higher-order model  significantly reduces
the residual error.

 STEP 3   Review the Charts

 In this step, the user reviews the charts named Curves 1, Residuals 1, Curves 2, Residuals 2,
 etc.  These charts help the user understand why one model performs better than the other.
                                         A-2

-------
STEP 4   Assess Magnitude of Uncertainty

In this step, the user enters the assumed concentration of a candidate standard and selects a
replication number, r.  Based on the calibration results, the worksheet estimates the 95%
uncertainty that would result from measuring such a standard r times. The user can use this as
a guide for deciding whether to proceed with analysis, to produce additional calibration points, or
to take some corrective action.

STEP 5   Assay Candidate Standard on Same Day

In this step, the user enters the responses to a candidate standard that is tested on the same
day as the calibration of STEP 1. The worksheet provides an estimate of the candidate's
concentration and its 95% uncertainty. The worksheet also indicates whether the variability in
responses is larger than expected (unacceptable).

STEP 6   Assay Candidate Standard on Different Day from Initial Calibration

In this step, the user enters the responses to a candidate standard that is tested on a different
day from the calibration of STEP 1. The worksheet first assesses the zero and span responses.
If the zero and span responses are acceptable, the user proceeds to enter the results from
testing a candidate standard.  The results include those for zero and nonzero reference
standards. (The quadratic model requires the use of two different nonzero standards.)

The spreadsheet determines whether the regression curve has changed since the initial
calibration. The data are corrected for any change and the estimated concentration of the
candidate standard is provided together with its 95% uncertainty.

The spreadsheet also determines whether the standard error of the mean response is
acceptable (<1 % of the mean response). This additional check is meant to guard against
hysteresis or other errors that are not corrected by the spreadsheet's adjustments.

2.  Measurement Data Spreadsheet

STEP 1   Enter Calibration Data

Enter the calibration data in the shaded spaces below. The first column (I) simply counts the
calibration points that you enter. The second column (X) is for the certified concentrations of
the calibration gas standards. The third column (Y) is for the instrument responses
corresponding to the calibration standards. The number of points cannot exceed 50.
                                          A-3

-------
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
x,
0.000
0.500
1.000
1.500
2.000
2.500
3.000
3.500
4.000
4.500
5.000
5.500
6.000
6.500
7.000
7.500
8.000
8.500
9.000
Y,
0.2194
0.7141
1.2885
1.9132
2.5910
3.2866
4.1078
4.9446
5.8145
6.7230
7.7284
8.7566
9.8013
10.8818
12.0982
13.3122
14.5840
15.9238
17.3271
Color Code

red = derived value (protected)

blue = entered value (unprotected)

black = fixed text (protected)













STEP 2   Review the Parameter Estimates

Review the estimates of the coefficients (b0, (b,,...) for the linear and quadratic models, their
confidence, and the residual error variances (s2).
                                          A-4

-------
  Linear Model
95% Confidence Limits
Parameter
b° =
b,=
S2 =
s =
df =
t =
Quadratic Model
Parameter
b0 =
b1 =
b2 =
S2 =
s =
df =
t =
Estimate
-1.0778
1.9005
0.4986
0.7061
17
2.1098

Estimate
0.1964
1.0011
0.0999
0.0005
0.0220
16
2.1199
Lower
-1.7351
1.7757
0.2807
0.5298


95% Confidence
Lower
0.1960
1.0010
0.0999
0.0003
0.0164


Upper
-0.4204
2.0253
1.1205
1.0585


Limits
Upper
0.1968
1.0012
0.0999
0.0011
0.0335


Comparing the two models:
Fratlo= 1027
F^ca^ 2.3167
(5% significance
level)

The quadratic model produces a significantly smaller error variance. The quadratic model
appears to be the better choice.

If cubic or quartic models are supported by compelling scientific theory or data, then review the
following estimates for those models. Otherwise, go to Step 3.
                                         A-5

-------
   Cubic Model
                              95% Confidence Limits
Parameter
b0 =
b1 =
b2 =
b3 =
S2 =
s =
df =
t =
Estimate
0.1952
1.0030
0.0994
0.0000 •
0.0005
0.0227
15
2.1315
Lower
0.1593
0.9676
0.0901
-0.0006
0.0003
0.0168


Upper
0.2310
1.0385
0.1087
0.0007
0.0012
0.0352

•
Comparing quadratic and cubic models:
   'ratio" .

   '  critical =
0.9385
2.3849 (5% significance level)
The error variances are not significantly different at the 5% level. The quadratic model appears
to be a better choice than cubic.
 Quartic Model
                              95% Confidence Limits
Parameter
b0 =
b,=
b2 =
b3 =
b« =
S2 =
s =
df =
t =
Estimate
0.2206
0.9285
0.1390
-0.0069
0.0004
0.0003
0.0165
14
2.1448
Lower
0.1900
0.8786
0.1156
-0.0109
0.0002
0.0001
0.0121


Upper
0.2512
0.9783
0.1625
-0.0030
0.0006
0.0007
0.0261


                                         A-6

-------
Comparing quadratic and cubic models:

    Fratio=     1.8954
    F cm**! =    2.4630 (5% significance level)

The error variances are not significantly different at the 5% level.  The cubic model appears to
be a better choice than quartic.

STEP 3  Review the Charts

View the charts named Curves 1 and Residuals 1. Curves 1 shows confidence bands for the
estimated regression. Compare these bands with those of the quadratic regression, Curves 2.
(Note: You can change the width of the confidence band interval by changing the "p-value° in
cell F4 of the worksheet named Chart Data.) Residuals 1 shows how the calibration points
deviated from the calibration line. Look for a simple pattern (such as a quadratic curve) in the
chart.  If such a pattern appears, the quadratic model may be better. View Residual 2, the
deviations from the best- fit quadratic curve. If Residual 2 effectively removes the simple
pattern observed in Residual 1 and if the magnitude of the deviations has been significantly
reduced (as evidenced by a reduction in the estimate s2), then the quadratic model is superior.
An F-test can be run to determine if .the two error variances are significantly different.

    F=   1026.764               Prob. of greater F =  4.51 E-21

The quadratic model produces a significantly smaller error variance. The quadratic model
appears to be the better choice.

STEP 4   Assess Magnitude of Uncertainty

Enter the concentration at which you would like to evaluate the uncertainty of estimation and
prediction.  Also enter r, the number of assays to be performed. Increasing r tends to reduce the
prediction uncertainty, but with diminishing effect.

    Concentration .=   6
    r=               3

Review the estimated mean response (estimate that only takes into account the calibration
uncertainty), and the confidence intervals. Review the predicted mean response and its
confidence intervals. To satisfy the EPA protocol requirements, the 95% confidence limits for the
concentration should be s±1% of the concentration.
Estimates below are based on the quadratic model. Tab-Right to view estimates based on the
other model.
                                          A-7

-------
95% Confidence Limits

Instalment Response -
Estimate
9.8006
Lower Upper
9.7858 9.8155
95% Confidence Limits

Instrument Response =
Concentration =
95% uncertainty in prediction =
Prediction
9.8006
6.0000
Lower Upper
9.7698 9.8314
5.9860 6.0140
0.23%
STEP 5   Assay Candidate Standard on Same Day

Proceed with the analysis of candidate standards if their 95% uncertainties, as estimated above,
are <1%. Enter the responses from the repeated analyses of an individual candidate standard in
the spaces provided below.

Note: This step applies only to candidate standards that are assayed on the same day as the
calibration.

Enter the instrument responses for up to 10 repeated analyses of a single candidate standard
below.
                                         A-8

-------
Analysis
Number .
1
2
3
4
5
6
7
8
9
10

Estimated
Response Concentration
4.500 3.2466
4.501 3.2473
4.499 3.2460
mean =
standard deviation =
df =
F =
Fsig? =
Pr(>F) =
95% Uncertainty =


NOTE: For Cubic
and Quartic Model
estimates, view the
Calculations in the
spreadsheet's
shaded regions.

3.2466
0.0006
2
0.0008
FALSE (The sample
variance is acceptable.)
0.9992
0.58%
                                  23.97% =
portion of uncertainty2
due to calibration
STEP 6   Assay Candidate Standard on Different Day from Initial Calibration

This step applies to candidate standards that are assayed on a different day than the initial
calibration. Before candidate standards are run, the measurement system is challenged with
zero and span checks.  Three or more discrete checks of the zero gas and three or more
checks of the span gas are made. Enter the results below:
                                          A-9

-------
Response to Zero gas
0.000
0.001 '
j^OOl
ni '3
meanj= 0.000
s = 0.001
Cal. Resp.= ^ 0.1 96
Span cone.
_';j 9.000
9.000
! 9.000
n =
mean =
s =
Cal. Response =
Response
to Span
16.010
16.000
15.990
3
16.000
0.010
17.301
                              Zero Gas Results
Span Gas Results
        Std. Error = s/sqrt(n) =  0.0006
                    Rrs/100=  0.1600
                              Std. Error is okay

    Relative Difference (RD) ^  1.14%
                              RD is okay
0.0058
0.1600
Std. Error is okay

-7.52%
RD is excessive
Following successful completion of the zero and span checks, the candidate standard is
measured together with reference standards. While the candidate standard is normally
interspersed with the reference standards, the analysis conducted in this sheet requires that the
results be entered separately. There are two ways to do this. One way is to enter an analysis
set (one candidate standard response and the responses from its zero and nonzero standard
analyses) below. Another approach is to enter all of the responses (multiple sets) below. Enter
zero and reference standard responses in the area to the left and enter the responses to a
single candidate standard in the are to the right, below.

More than one nonzero reference standard is needed for the quadratic and higher-order
models.

Estimates below are based on the quadratic model. Tab-Right to view estimates based on the
other model.
                                         A-10

-------
              Reference Standards (Enter 0 for
                    Zero Concentraton)
                                                Candidate Standard
Cone.
0.000
0.000
0.000
4.500
4.500
4.500
9.000
9.000
9.000



nn=
Response
0.218
0.219
1.220
6.693
6.723
6.773
17.317
17.327
17.337



9
Cone.2
p.ooo
0,000
0.000
20.250
20.250
20.250
81.000
81.000
81.000




Cone.3
0.000
0.000
• o.boo
91.125
91.125
91.125
729.000
729.000
729.000




Cone.4
0.000
0.000
0.000
410.063
410.063
410.063
6561.000
6561.000
6561.000




Observed
Response
4.010
4.000
3.990

nnn =
mean = 4.000
stdev =
std error =
df =
F =
F sig? =
Pr{>F}
The standard
Estimated
Cone.
2.9437
2.9374
2.9311

3
2.9374
' 0.0063
0.12%
2
0.1128
FALSE
0.8938
error is okay.
Coefficient are not significantly different.
Consider including thenew data as part of original calibration (Step 1).

Estimated Concentration of Candidate Standard
   2.9374
95% Uncertainty
  0.66%
       Portion of uncertainty2 due to calibration uncertainty
             45.68%
95% Confidence Limits for Candidate Standard Concentration
Lower
2.9181
Upper
2.9567
These upper-and lower limits are compared with the corresponding limits estimated on different
assay dates to establish that the candidate standard has not drifted.
                                        A-11

-------
           APPENDIX B: INSTRUCTION FOR PERMEATION RATE WORKBOOK


 1.  ReadMe Spreadsheet

 What this Workbook Is All About

 This workbook enables the user to estimate the rate at which the weight of a permeation tube
 decreases. A linear relationship between the tube's weight and elapsed time is established. If
 the estimated weight at time zero is significantly different from the actual weight at time zero,
 then at least the earliest data pair should be removed from the analysis. Uncertainty of the slope
 estimate (the rate of weight loss or drift) will be determined. The traceability protocol requires
 that this estimate have a relative uncertainty of less than 1%.

 How the Workbook Is Organized

 The workbook consists of several worksheets, which are displayed as tabs on the bottom of the
 screen. Each worksheet performs a distinct function as  described below.

    ReadMe                describes the workbook and explains how to use the worksheets
    Data                   allows for user input of the calibration data (elapsed time and
                           weight)
    ANOVA                performs analysis of variance and determines whether the
                           intercept term is needed
    Calibration Results     calculates the drift and its uncertainty
    Curve                  graphically displays the drift line together with its confidence bands
    Residual               graphically displays the vertical difference between the observed
                           and estimated weights
    Report                 summarizes the assay results for a permeation device
    Chart Data              includes the data used to create the curve and residual charts.

 How the Worksheets Are Set up

 Each worksheet contains instructions that guide the user through the steps in  using
the worksheet.  The worksheets are also color-coded to  simplify use.  Shaded cells that are
bordered in blue lines are cells whose contents you can  change (i.e., enter data). In other
sheets you can  change the following variables:

Sheet               Variable                 Location        Current Value

Data                Unit of Time             H22            m
Data                Unit of Weight           H24            g
Report              Device ID                F5              test data
Chart Data          significance level         D2             1.00E-05

 Derived values are colored.red. These cells contain formulas that should not be changed. The
cells are protected to prevent alteration.


                                         B-1

-------
 How to Use the Worksheets

 Step 1:       Enter the elapsed times (all in the same units) and corresponding tube weights in
            '  the Data worksheet. The worksheet will compute, the total weight loss for each
              observation.
                                                •

 Step 2:       Select the significance level (alpha) to be used in producing confidence limits for
     "'        theJSsiimated slope and intercept. Then review the results of the F-test and t-test
              to determine whether the intercept term Js needed. If the intercept term is
              significant, then determine which of the early data points should be removed.
              Removing those data, and correcting the elapsed times, repeat Steps 1 and 2.

 Step 3:        Examine the corresponding Curve chart. You  may need to adjust the chart's axis
              scaling.  The points should appear to fall virtually on top of the black line. The
              black line should be very close to its confidence bands (colored red and blue).

 Step 4:        Examine the corresponding Residual worksheet. The residuals should appear to
              be random in both magnitude and direction.  If they appear to follow a regular
              pattern, then the simple linear model is not appropriate. The device does not
              have a constant rate of weight loss. More time may be  required to establish and
              measure a linear relationship. Observations taken before the linear relationship
              is established should be discarded and not used in the statistical analysis.

Step 5:        Print the one-page report provided in the Report sheet.  The report summarizes
              the assay data and indicates the uncertainty of the estimate.
                                          B-2

-------
2.  Data Entry Worksheet
                                                   Data Entry Worksheet
Elapsed
Time
1 X,
Weight
Y,
Enter the data in the blue-bordered spaces. The first
~ column (X) is for the elapsed time. The time of the first
entry should be zero. The second column (Y) is for
the tube weights.
1
0
1
2
3
4
5
6
x,
0
8641
18722
40322
64802
74882
84962
Y,
4.354206
4.33745
4.316766
4.273494
4.224514
4.20378
4.18439
                                     n = number of weighing. This can't exceed 50.

                                     n = 7
                                     No data entry is required for derived values, which are
                                     colored red, such as n and I. These values are
                                     tabulated automatically and their cells are protected
                                     from alteration.

                                     Multiple weighings at a single point in time requires
                                     multiple entries in each column. Reenter the time in
                                     column X and enter the corresponding weight in
                                     column Y.

                                     Enter the time and weight units in the spaces below:

                                     Unit of Time = m

                                     Unit of Weight = g
This sheet derives the regression equation in the form: y = b0 + b, x + e. The intercept and
slope are estimated. The sheet determines whether the intercept (weight estimated for time
zero) is significantly different from the observed weight at time zero. It also estimates the
uncertainty in the slope estimate and compares this uncertainty with EPA's 1% limit.

STEP 7

Review the estimates of the intercept (b0), slope of the regression line (b,), and their confidence
limits along with the estimates of variance-covariance matrix (V) and the residual error variance
(Var).

                                          B-3

-------
 Derivation of the estimated intercept (b0) and slope (bt) of .the regression line

                                                             95% Confidence Limits
   X'X=  7
    <| : . --1 ' 'I I ' .1 ;_ I
          292331
292331

1.91E+10
         ^  0.396793   -6.1E-06

             -6.1E-06   i46E-10'
  det(X4X)=  4.81 E+10
      X'Y =  29.8946

             1234673
     VY =  127.6972

       df=  5

t(0.95, dO =  2.570578
                                       bc=  4.354403

                                       b,=  -2E-06
bc lower limit = • 4.353872

b0 upper limit =  4.354935
                                    b, lower limit =  -2E-06

                                    b1 lower limit =  -2E-06
Derivation of the error variance (Var) and variance-covariance matrix (V)
b'X'Y
b'X'Y-sum(Y)2/n
(Y'Y - b'X'Y)
Var =(Y'Y-b'X'Y)/df
V = Var * (X'X)'1
STEP 2
                       127.6972           SS(model), 2df
                       0.027619           SS(regression) 1df
                       5.38E-07           SS(residual)
                       1.08E-07           MS(residual), n-2 df
                       4.27E-08      -6.5E-13
                       -6.5E-13       1.57E-17
Examine the upper and lower limits for the intercept, b(
                                            'o-
   b0 lower limit =
                4.353872
bc upper limit =   4.354935
         y0 =   4.354206
                                       Conclusion: y0 is within the confidence limits for the
                                       intercept
If y, is within the confidence limits for the intercept, proceed to STEP 3. Otherwise, consider
removing the first observed weight from the analysis.  Re-enter the times and weights.
Remember that the first time (XJ should be zero. This will require adjustment of the other
elapsed times. After entering the data, return to STEP 1, above.

STEP 3

Examine the upper and lower limits for the slope, bv The limits should differ from the estimate
by no more than ±1% of the estimated slope.

   (b, upper - b,) / Ib,! = 0.51%    Conclusion: Uncertainty is acceptable.
   (b, lower - b,) / lb,l = -0.51%
                                         B-4

-------
If the uncertainty is unacceptable, consider collecting additional data. Also, view the Curve and
Residual plots.  They may reveal a nonlinear relationship for a portion of the data. The initial
measurements may not align with subsequent measurements if the device was in the process of
stabilizing or equilibrating during those times. If this is the case, the initial points of the Residual
chart would appear to be outliers. The residuals with the same sign (all positive or all negative)
and their magnitude will likely be greater than the magnitude of subsequent residuals.  If this is
the case, consider removing the initial points from the computations and re-enter the remaining
times and weights with the times adjusted so the first entry has time zero.

If the uncertainty is acceptable, print the Report spreadsheet and include it with the
certification documentation.

3.  Assay Results For Permeation Device

This sheet provides calibration information and assay results, including uncertainty estimates for
a permeation device identified as: test data

Test Results

Intercept (b0), slope (bt), and their confidence limits
X'X = 7 292331
292331 1.91E+10
(X'X)-1= 0.396793 -6.1E-06
-6.1E-06 1.46E-10
det(X'X)= 4.81 E+10
X'Y = 29.8946
1234673
Y'Y= 127.6972
df= 5
t(0.95, df) = 2.570578
b0 = 4.354403
b, = -2E-06
95% Confidence Limits
b0 lower limit = 4.353872
b0 upper limit = 4.354935
b1 lower limit = -2E-06
b, lower limit = -2E-06
Error variance (Var) and variance-covariance matrix (V).
b'X'Y
b'X'Y - sum(Y)2 / n
(Y'Y - b'X'Y)
Var =(Y'Y-b'X'Y)/df
V = Var" (X'X)'1
=  127.6972
=  0.027619
=  5.38E-07
= 1.08E-07
=  4.27E-08
   -6.5E-13
     SS(model), 2df
     SS(regression) 1df
     SS(residual)
     MS(residual), n-2 df
-6.5451 E-13
1.56725E-17
Upper and lower limits for the intercept, b0:

   b0 lower limit =    4:3538722
   b0 upper limit =    4.3549347
              y0 =    4.354206
                                          B-5

-------
Upper .and lower limits for the slope, b,:
       .-,-,   •      -    -or;      •    -.-•.-:-
   (b, upper - b,) / lb,l = 0.51%
   (b, lower-b^/lbtU -0.51%

Estimated rate of: weight loss, bt =  2.005E-06    g/m
                                        B-6

-------
          APPENDIX C: CALCULATION OF TOTAL ANALYTICAL UNCERTAINTY
ASSAY RESULTS
                                               i
In thfe sheet the results of two or three Assays are entered. Calibration dates are entered
so Assays having the same calibration uncertainty may be treated correctly.  (Assays having
a common calibration share the same calibration uncertainty.)

Enter the results for up to three separate assays in chronological order below.

ASSAY 1

500   = estimated concentration
0.005  = 95% uncertainty (expressed as percentage of estimated concentration)
0.5    = portion of 95% uncertainty^ due to calibration
35551 = date of prior calibration

ASSAY 2

500   = estimated concentration
0.005  = 95% uncertainty (expressed as percentage of estimated concentration)
0.5    = portion of 95% uncertainty due.to calibration
35582 = date of prior calibration

ASSAY 3 (if applicable)

500   = estimated concentration
0.005  = 95% uncertainty (expressed as percentage of estimated concentration)
0.5    = portion of 95% uncertainty due to calibration
35582 = date of prior calibration

Number of different calibrations represented by the above data:
       N =   ' 2     (If this value seems to be  incorrect, check the dates
                    entered for the three assays. The earliest data should
                    be for Assay 1. Trailing spaces may cause N's formula
                    to interpret identical dates as different.)
                                         C-1

-------
 COMPARISONS
                                      Upper Confidence Limits
                             Variance Components
Calibration
1
2
-2
Assay
1
2
3
Lower
Confidence
Limits
497.5
497.5
497.5
Assay 1
502.5
—
True
True
Assay 2
502.5
True
—
True
Assays
502.5
True
True
—
Calibration
3.125
3.125
3.125
> .
Imprecision
3.125
3.125
3.125
 "FALSE* indicates an inconsistency such as an upper confidence limit for one assay that
 is lower than a lower confidence limit for another (non-overlapping intervals). "FALSE" will
 appear for Assay 3 if no data have been entered for Assay 3.

 OVERALL ESTIMATE

       Note: Calibration Case = 15
                     Case
Cal. No.   Cal. No.    Cal. No.
4*
6*
9
12
15
18
1
1
1
1
1
1
1
2
1
1
2
2
—
—
1
2
2
3
                     *4 and 6 are cases where there is no 3rd assay. In
                     case 4, the two assays share a common calibration.
                     In case 6, the two assays have different calibrations.

The standard error of the estimate produced in an assay is equal to approximately Vz of
the "95% uncertainty." The inverse of the square of the standard error is the (raw)
weighting factor used in producing an overall estimate of the concentration. The raw
weights are adjusted (Adj. Wt.) so their sum is 1.00.
                                         C-2

-------
95%
Uncert.
'.005
'.0043301

Raw Wt.
40000
53333.333
0
Adj. Wt.
0.4285714
0.5714286
0
Wt.
*Conc.
214.28571
285.71429
0
Variance
ofWL
*Est.
1.1479592
1.5306122
0
  Calibration   Estimate

       1           500

       2           500
500          = overall estimate of the candidate standard's concentration
1.6366342    = 95% uncertainty (concentration units)
0.0032733    - 95% relative uncertainty

The standard error and 95% uncertainty displayed above do not account for uncertainty
in the reference standards used to calibrate the analytical instrument. In the space below,
enter the 95% uncertainty (typically 2 times the standard error) of the reference standards.
If different calibration standards had different uncertainties, enter the largest.

Example: If NIST SRMs were used.in the calibration and their certified concentrations
were 100 +/-1 ppm, 200 +/-1 ppm, 300 +/- 2 ppm, 400 +/- 3 ppm and 500 +/- 4 ppm,
then the largest 95% uncertainty is for the 100 ppm standard: 1/100 = 0.01 or 1%.
(SRM uncertainties are expressed as two-sigma limits which are 95% confidence
intervals.)
0.005
=  95% uncertainty (2 times the standard error) of the reference standard
0.0059761  =  95% uncertainty of the candidate standard (including the contribution of the
               reference standard)
                                          C-3

-------
                          APPENDIX D: MATRIX NOTATION
1.  Matrix Notation
Matrix notation is used to simplify the presentation of calculations that are performed in the linear
regression. A matrix is a rectangular array of numbers. Boldface capital letters represent
matrices, and lower case letters with subscripts represent individual numbers in the matrices. X,
below, is a 10 by 3 matrix. It has 11 rows and 3 columns. The rows are numbered 0,1, 2,...10
and columns are numbered 0,1, and 2. (Other texts may begin numbering with 1.)
                       X =
1
1
1
1
1
1
1
1
1
1
1
1.002
0.902
0.802
0.701
0.601
0.501
0.401
0.301
0.200
0.100
0.000
1.0040
0.8136
0.6432
0.4914
0.3612
0.2510
0.1608
0.0906
0.0400
0.0100
0.0000
Xy denotes the number that is found in the ith row and the jth column. X0., = 1.002. The first row
and column are numbered zero.

A matrix that has only one column is called a column vector, and a matrix that has only one row
is called a row vector.

             0.999
             0.915
             0.828
             0.738
       y=   0.644   is a column vector.
             0.549
             0.448
             0.346
             0.237
             0.122
             0.001
                                         D-1

-------
 Subscripts following vector names denote the row or column of the vector. For example, y1 is
 the number in the second row of y, 0.915.  (Remember that we begin counting rows with zero.)

 Matrix operations that come into play for calibration include multiplication, transposition, and
 inversion. The rules for these operations can be found in any introduction to matrices. We will
 use the following notation for these operations:

 X* denotes the transpose of X (the ith column of X becomes the ith row of X')

 For the matrices X and Y above,

                  1            1           1           1           ...           1

        X'=    1.002       0.902       0.802       0.701         ...         0.000

                1.0040     0.8136      0.6432      0.4914        ...        0.0000

 X'Y denotes multiplication of matrices X* and Y. X1 must have the same number of columns as
 Y has rows. For the matrix X above,

                       11        5.511       3.8658                     5.827

       X'X=        5.511       3.8658       3.0438   and     X'Y=     4.0132

                   3.8658       3.0438       2.5544                    3.1272

    det(X'X) =  1.0521        (the determinant of X'X)

(X'X)'1 denotes the inverse of the product of X' and X
0.5800
-2.196
1.7392
-2.1962
12.5026'
-11.5744
1.7392
-11.5744
11.5513
             (X-X)-1 =
2. Calibration by Linear Regression Using Matrix Notation - Example

The linear regression approach is illustrated below for the simple quadratic curve.

The starting point for regression analysis will be a matrix named X. This matrix will have 3
columns (one for each coefficient to be determined). The number of rows will be the same as
the number of calibration measurements that are performed by the measurement system. The
                                         D-2

-------
first column is a vector of 1s. The second column contains the certified concentrations of the
calibration standards. The third column contains the squares of the values appearing in the
second column.  When this matrix is multiplied by the vector of coefficients [b0, bv bj, the result
is a vector.of responses, so that:

   response, = 1 * b0 + concentration! * b, + concentration2 * b2

or, letting y represent response and x represent concentration,
Now, we're interested in estimating the the coefficients b0, b1( and b2, and we're also interested
in computing how much error is involved when we use the information to estimate the
concentration in an "unknown."

3. Determining the Calibration Equation

The coefficients of the calibration equation or curve are found by matrix multiplication and
inversion:
     = (X'X)-1X'Y = [b0,b1,b2]
Example

    1          1.002        1.0040              0.999               0.9967
    1          0.902        0.8136              0.915               0.9151
    1          0.802        0.6432              0.828               0.8297
    1          0.701        0.4914              0.738               0.7394
X=1          0.601        0.3612       y=    0.644        b'x=  0.6462
    1          0.501        0.2510              0.549               0.5491
    1          0.401        0.1608              0.448               0.4482
    1          0.301        0.0906              0.346               0.3434
    1          0.200        0.0400              0.237               0.2336
    1          0.100        0.0100              0.122               0.1210
    1          0.000        0.0000              0.001               0.0046

        0.0046
    b = 1.1837                    b'=   0.0046    1.1837   -0.1932
       -0.1932

    The quadratic calibration curve is:  response = 0.0046 + 1.1837 C + -0.1932 * C2
                                          D-3

-------
4.  Determining the Estimation and Prediction Error

One assumption that underlies the regression approach is that random error is constant across
the measurement range. Sometimes it may be necessary to apply a transformation in order to
achieve this characteristic, called homogeneity of variance.  An estimate of this variance is
obtained using matrix operations:

    Var = residual sum of squares / degrees of freedom = Y'Y - b'X'Y / df

    This estimate's "degrees of freedom" (df) is the number of calibration points less the number
    of coefficients  estimated for the calibration equation.

Another important output of the regression analysis is the "variance-covariance" matrix, V:

    V = Var • (X'X)'1

The variance of each coefficient is found in the principal diagonal of V.  For example, the
variance of b0 is V00.  Covariances are found as off-diagonal elements of V.

Hypothesis tests can  be performed and confidence intervals can be estimated for each
coefficient using the coefficient's  estimate, the coefficient's variance (contained in V), and the
degrees of freedom, df.

Continuing our example

             Var = (Y'Y - b'X'Y) / df =      5.91 E-06

                 3.43E-06      -1.3E-05   1.03E-05

         V=      -1.3E-05     7.39E-05   -6.8E-05      df = 8

                 1.03E-05      -6.8E-05   6.83E-05     (df = degrees of freedom)

       95% Confidence Interval for b0= b0 +/- t(0.05,df) * sqrt(V00)

             95% Cl for b0 = 0.000324    to    0.008865

             t(0.05,  df) = 2.306006

       95% Confidence Interval for b, = b, +/- t(0.05,df) * sqrt(V1t1)

             95% Cl for b, = 1.163855    to    1.203512

       95% Confidence Interval for b2= b2 +/- t(0.975,df) * sqrt(V2i2)

             95% Cl for b2 = -0.21224    to    -0.17413
                                          D-4

-------
Another use of V is in computing the uncertainty in a regression predicted concentration of an
individual unknown. The analyzer is subjected to the unknown, and a mean response, R, is
produced. A solution for C is found. This is the estimated concentration of the unknown.
Deriving the confidence intervals for this estimate requires finding two alternative concentrations,
one higher and one lower than the estimate, such that the probability of having produced a
lesser or greater average response is sufficiently small. For a 95% confidence interval, the
lower bound is a concentration whose response would be less than the observed response with
97.5% probability; the upper bound is a concentration whose response would be less than the
observed response with 97.5% probability.

Unfortunately, for quadratic curves, this derivation is not so simple.

   R measurements of an unknown produce an average response resp:
                     R=   6
                   resp =   0.601

   The estimated concentration is found by solving the following quadratic equation:

   0.601 = b0 + b, C + b2 C2
   (b0 - 0.601) + b1C + b2C2 = 0

   The potential solutions are found using the quadratic formula:

       C =  0.553935       and   5.573267     (only the first of these is reasonable)

   Now, if the concentration really had been at this value, the 95% confidence interval for
   the mean response of six measurements would be symmetric about the observed response:


                As above, t =  2.306006

                         x=  1  0.553935  0.306843         = [1, resp, resp2]

                        xb =  0.601  (check)

   var(predicted mean response for x)  = [var/R + x' V x]

                       x'V=  -6.09E-07    6.97E-06    -6.7E-06

                      x'Vx=  1.2E-06

                     var/6=  9.86E-07

   var(predicted mean response for x)  = 2.19E-06

   95% confidence interval for predicted response = 0.597588    to     0.604412

             This is the observed response-/+:   0.003412     and   0.003412

   Solving for concentration, the interval is no longer perfectly symmetric:

                                         D-5

-------
                            0.550418
to    0.557456
              This is the estimated concentration -/+:

    As a percentage of the concentration, this is -/+:
             0.003516     and   0.003521

             0.006348     and   0.0063569
 Fortunately, even with the quadratic calibration curve, with good precision, the confidence
 intervals will be within a small enough region that the curve is close to linear and the interval will
 be very nearly symmetric. The uncertainty criterion for multipoint calibration requires the 95%
 confidence interval's half-width to be less than 1 %. The calibrated range of the analyzer extends
 across all concentrations for which the criterion is satisfied.

 Continuing our example
Estimated
Concentration Response
1.002 0.9967
0.902 0.9151
0.802 0.8297
0.701 0.7394
0.601 0.6462
0.501 0.5491
0.401 0.4482
0.301 0.3434
0.200 0.2336
0.100 . 0.1210
0.000 0.0046
0.210 0.2446
95% conf. interval
for response
0.9924
0.9121
0.8273
0.7371
0.6437
0.5466
0.4457
0.3411
0.2313
0.1181
0.0003
0.2423
1.0010
0.9181
0.8320
6.7417
0.6487
0.5517
0.4507
0.3457
0.2359.
0.1240
0.0089
0.2469
95% conf. interval for
concentration
0.9966
0.8985
0.7993
0.6985
0.5984
0.4984
0.3986
0.2988
0.1979
0.0974
-0.0036
0.2079
1.0074
0.9055
0.8047
0.7035
0.6036
0.5036
0.4034
0.3032
0.2021
0.1026
0.0036
0.2121
% error for
concentration
-0.53
-0.39
-0.33
-0.36
-0.43
-0.51
-0.60
-0.72
-1.06
-2.58
—
-0.9996
0.54
0.39
0.33
0.36
0.43
0.52
0.60
0.72
1.06
2.59
—
.1.0004
The calibration curve's uncertainty is acceptable for concentrations above 0.21 ppm.

5.  Stability Test

As discussed in Subsection 2.1.6.2, the stability test requires at least three initial measurements
of the candidate standard plus at least three additional measurements following a period of 7
days or more. The standard's concentration must be in the calibrated range of the analyzer per
Subsection 2.1.7.2.
                                           D-6

-------
 Concentrations are estimated using the calibration curve, producing at least three estimates for
 the initial concentration and at least three estimates for the concentration following the holding
 time. A students t-test is applied as follows:

    Initial Data              Final Data (after holding time)
        C1                        C4
        C2                        C5
        C3                        C6

    s, = standard deviation of (C1, C2, C3)         \, = (C1 + C2 + C3) / 3
    s2 = standard deviation of (C4, C5, C6)         % = (C4 + C5 + C6) / 3

            alpha = significance level of the test = 0.05
    t(1-alpha/2,df) = value of students t for which the distribution function value is 0.975
                   and degrees of freedom = number of observations - 2

    s = sqrt(s12 + s22)

    If 1x1 - x2l / s > t(1-alpha/2,df) then the difference is statistically significant and the candidate
    standard has failed the initial stability test.  The test can be repeated after an additional
    7 days or more, using  the second and third sets of results in the calculations, as above. If
    another significant difference is found, then the candidate standard is unusable and is
    disqualified for further  use.

 Example:

    Initial Data              Final Data (after 7-day holding time)
    0.995   ppm                     0.989  ppm
 .   0.996   ppm                     0.989  ppm
    0.992   ppm                     0.982  ppm

  s, =  0.0020817 ppm                x, =  0.9943333 ppm

  S2=  0.0040415 ppm                x^  0.9866667 ppm

   s=  0.004546 ppm        % difference =  0.77% ppm

      Ix1-x2l/s=   1.686442

   t(1-alpha/2,df)=   2.7764509

The difference is not statistically significant, so the standard can be certified as stable.

6.  Recertification

 Per Subsection 2.1.6.3, a  standard can be recertified if, after the certification period has elapsed,
the mean concentration of at least three assay results is within 1.0 percent of the original
certified concentration. Additionally, the difference between the estimated mean and the certified
concentration must not be statistically significant at the 1 % level.


                                           D-7

-------
 To determine whether the concentration of the standard has changed since the initial
 certification, new measurements are made using a measurement system that has been
 calibrated according to Subsection 2.1.7.5. Original certification data are used to provide an
 initial estimates of mean (x,) and standard deviation (s,). New data are used to estimate a
 second mean (xj and standard deviation (sj.  These are used in a t-test that is similar to that
 used in the stability test. A critical value for t is based on a significance level of 1% (alpha) and
 degrees of freedom equal to the number of initial and recertification data minus 2. A pooled
 estimate of the standard deviation (s) is derived from s, and Sj. If the difference between x, and
 Xj,, divided by s, is greater (in absolute value) than the critical value for t, then the initial and new
 concentrations are significantly different and the standard cannot be recertified.

 Example:

    Initial Data             Recertification Data
    0.995  ppm                0.989  ppm
    0.996  ppm                0.99   ppm
    0.992  ppm                0.994  ppm
    0.999  ppm
    0.999  ppm
    0.993  ppm

    s, = 0.0029439   ppm                 x, =   0.9956667   ppm

    S2 = 0.0026458   ppm                 x,,=   0.991        ppm

    s = 0.0028619   ppm         % difference =  0.47%


     Ix1-x2l/s=   1.6306179

  t(1-alpha/2,df)=   2.7764509
   The % difference is less than the 1% specification, and the difference in means is not found
to be statistically significant. The standard may be recertified.  The certified concentration of the
standard is the grand mean of the combined data set.

   Certified Concentration = mean (initial data + recertification data) = 0.994  ppm
                                          D-8

-------