EPA-R4-73-028a
June  1973
Environmental Monitoring  Series

                                                     Jiiiftji
                                             XsvXvX-x •'•'. •". \ • :• ;• A \ •.•/!•.•!•.•.•.vl
                                             >:?•:•:::::•::





                                             Ill
                                                         I

                                                         55
                       \
                        UJ
                        O


-------
                                        EPA-R4-73-028o
 GUIDELINES  FOR  DEVELOPMENT

             OF  A  QUALITY

       ASSURANCE  PROGRAM

 Reference Method for the Continuous  Measurement

      of Carbon Monoxide in the Atmosphere


                      by
        Franklin Smith and A.  Carl Nelson, Jr.
            Research Triangle Institute
     Research Triangle Park, North Carolina 27709
             Contract No. 68-02-0598
            Program Element No. 1H1327
     EPA Project Officer:  Dr. Joseph F. Walling

Quality Assurance and Environmental Monitoring Laboratory
       National Environmental Research Center
     Research Triangle Park, North Carolina 27711
                  Prepared for

          OFFICE OF RESEARCH AND MONITORING
        U.S. ENVIRONMENTAL PROTECTION AGENCY
              WASHINGTON, D.C.  20460

                   June 1973

-------
This report has been reviewed by the Environmental Protection Agency




and approved for publication.  Approval does not signify that the




contents necessarily reflect the views and policies of the Agency,




nor does mention of trade names or commercial products constitute




endorsement or recommendation for use.
                                ii

-------
                         PREFACE

      Quality control  is  an  integral part of any viable
environmental monitoring activity.  The primary goals of
EPA's quality control program are to  improve and document
the credibility of environmental measurements.  To
achieve these goals,  quality control  is needed in nearly
all segments of monitoring  activities and should cover
personnel, methods selection, equipment, and data
handling procedures.  The quality control program will
consist of four major activities:
         • Development and  issuance of procedures
         • Intra-laboratory quality control
         • Inter-laboratory quality control
         • Monitoring program evaluation and
           certification
All these activities are essential  to a successful  quality
control  program and will be planned and carried out
simultaneously.
     Accordingly, this first manual  of a series of five has
been prepared for the quality control  of ambient air
measurements.  These guidelines for the quality control
                           ill

-------
of ambient carbon monoxide measurements have been produced
under the direction the Quality Control Branch of the
Quality Assurance and Environmental  Monitoring Laboratory
of NERC-RTP.  The purpose of this document is to provide
uniform guidance to all EPA monitoring activities in the
collection, analysis, interpretation, presentation, and
validation of quantitative data.   In accordance with
administrative directives to implement an Agency-wide
quality control program, all EPA monitoring activities
are requested to use these guidelines to establish intra-
laboratory quality assurance programs in the conduct of
all ambient air measurements of carbon monoxide.  Your
comments on the utility of these guidelines, along with
documented requests for revision(s), are welcomed.
     All questions concerning the use of this manual  and
other matters related to quality control of air pollution
measurements should be directed to:
           Mr. Seymour Hochheiser, Chief
           Quality Control Branch
           Quality Assurance and Environmental
             Monitoring Laboratory
           National Environmental Research Center
           Research Triangle Park, North Carolina  27711
                           iv

-------
     Information on the quality control  of other

environmental media and categorical measurements can be

obtained by contacting the following person(s):

          Hater

            Mr. Dwight Ballinger, Director
            Analytical Quality Control  Laboratory
            National Environmental Research Center
            Cincinnati, Ohio  45268

          Pesticides

            Dr. Henry Enos, Chief
            Chemistry Branch
            Primate and Pesticide Effects Laboratory
            Environmental Protection Agency
            Perrine, Florida  33157

          Radiation

            Mr. Arthur Jarvis, Chief
            Office of Quality Assurance-Radiation
            National Environmental Research Center
            Las Vegas, Nevada  89114

     During the months ahead, a series of manuals will

be issued which describe guidelines to be followed during

the course of sampling, analysis, and data handling.  The

use of these prescribed guidelines will  provide a uniform

approach in the various monitoring programs which allows

the evaluation of the validity of data produced.  The

implementation of a total and meaningful  quality control

program cannot succeed without the full  support of all

monitoring programs.  Your cooperation is appreciated.

-------
                           TABLE OF CONTENTS

Section                                                               Page
  1.0  INTRODUCTION                                                     1

                      PART I.   OPERATIONS MANUAL

  2.0  GENERAL                                                          3

       2.1  Operating Procedures                                        4

                 ANALYZER CALIBRATION                                   8

                 SAMPLING                                              20

                 OPERATIONAL CHECKS                                    21

                 DATA PROCESSING                                       27

       2.2  Special Checks for Auditing Purposes                       35

            A.  Measuring Control Samples                              35

            B.  Water Vapor Interference Check                         36

            C.  Data Processing Check                                  38

       2.3  Special Checks to  Detect and/or Identify Trouble           39

            A.  Zero Drift Check                                       39

            B.  Flow Rate Variation Sensitivity Check                  41

            C.  Temperature Variation Sensitivity Check                41

            D.  Voltage Variation Sensitivity Test                     42

       2.4  Calibration of Sample Flow and Sample Cell
            Pressure Indicators                                        45

            A.  Flow Rate Calibration                                  45

            B.  Sample Cell Pressure Gauge Calibration                 45

       2.5  Facility and Apparatus Requirements                        48

            A.  Facility                                               48

            B.  Apparatus                                              48
                                   vi

-------
                      TABLE OF CONTENTS (CONT'D)



Section

                       PART II.  SUPERVISION MANUAL

  3,0  GENERAL                                                         50

       3.1  Assessment of NDIR Data                                    52

            A.  Required Information                                   52

            B.  Collection of Required Information                     52

            C.  Treatment of Collected Information                     55

       3.2  Suggested Standards for Judging Performance                57

       3.3  Collection of Information to Detect and/or
            Identify Trouble                                           57

            A.  Identification of Important Variables                  59

            B.  How to Monitor Important Variables                     62

            C.  Suggested Control Limits                               63

       3.4  Procedures for Improving Data Quality                      66

       3.5  Procedures for Changing the Auditing Level to Give
            the Desired Level of Confidence in the Reported Data       70

            A.  Decision Rule - Accept the Lot as Good If No
                Defects Are Found                                      71

            B.  Decision Rule - Accept the Lot as Good If No More
                Than One (1) Defect is Found                           71

       3.6  Monitoring Strategies and Cost                             72

            A.  Reference Method                                       72

            B.  Reference Method with Sample Diffusion Chamber         73

            C.  Reference Method Plus Sample Diffusion Chamber
                and Shelter Temperature Control Unit                   73
                                  vii

-------
                      TABLE OF CONTENTS (CONCL'D)


Section                                                               Page

                     PART III.  MANAGEMENT MANUAL

  4.0  GENERAL                                                         75

       4.1  Data Quality Assessment                                    76

            A.  Assessment of Data Quality                             78

            B.  Assessment of Individual Measurements                  80

       4.2  Auditing Schemes                                           80

            A.  Statistics of Various Auditing Schemes                 83

            B.  Selecting the Auditing Level                           88

            C.  Cost Relationships                                     91

            D.  Cost Vs. Audit Level    .                               94
                                    i
       4.3  Data Quality Versus Cost of Implementing Actions           96

       4.4  Data Presentation                                         102

       4.5  Personnel Requirements                                    104

            A.  Training and Experience                               104

       4.6  Operator Proficiency Evaluation Procedures                105

  REFERENCES                                                          107
  APPENDIX 	 REFERENCE METHOD FOR THE CONTINUOUS MEASUREMENT
                     OF CARBON MONOXIDE IN THE ATMOSPHERE
                    (NON-DISPERSIVE INFRARED SPECTROMETRY)             108
                                 viii

-------
                            LIST OF FIGURES


Figure                                                                Page

  1    Operational Flow Chart of the Measuring Process                 5-6

  2    Carbon Monoxide Monitoring System Flow Chart                     7

  3    Sample Calibration Curve                                        15

  4    Table for Converting Trace Deflection in Percent of
       Chart to Concentration in PPM                                   16

  5    Sample Daily Check Sheet                                        18

  6    Sample Form for Reporting Results of Quality Control Checks     23

  7    A Sample Graph of the Mean (c) and 3a Limits of Hourly CO
       Concentrations for a 24-Hour Period                             28

  8    Sample Sheet for Recording Hourly Averages                      29

  9    Sample Trace of 24-Hour Sampling Period with Zero and
       Span Calibrations                                               31

 10    SAROAD Hourly Data Form                                         33

 11    Calibration Set-Up for Pressure Gauges                          47

 12    Data Qualification Form                                         56

 13    Critical Values of Ratio s./a. Vs. n                            82

 14    Data Flow Diagram for Auditing Scheme                           84

 ISA   Probability of d Defectives in the Sample If the
       Lot (N=100) Contains D% Defectives                              86

 15B   Probability of d Defectives in the Sample If the
       Lot (N=50) Contains D% Defectives                               87

 16A   Percentage of Good Measurements Vs. Sample Size
       for No Defectives and Indicated Confidence Level                89

 16B   Percentage of Good Measurements Vs. Sample Size
       for 1 Defective Observed and Indicated Confidence Level         90

 17    Average Cost Vs. Audit Level                                    97

 18    Costs Vs. Precision for Alternative Strategies                 101

 19    Sample QC Chart for Evaluating Operator Proficiency            106

                                   ix

-------
                            LIST OF TABLES
Table                                                                 Page
  1    Analyzer Evaluation Data                                        44

  2    Apparatus Used in the NDIR Method                               49

  3    Suggested Performance Standards                                 58

  4    Methods of Monitoring Variables                                 62

  5    Suggested Control Limits for Parameters and/or Variables        64

  6    Quality Control Procedures or Actions                         67-69

  7    Critical Values of s /a.                                        81

  8    Required Auditing Levels n for Lot Size N=100
       Assuming Zero Defectives                                        88

  9    Costs vs.  Data Quality                                          91

 10A   Costs If 0 Defectives are Observed and the Lot is  Rejected      92

 10B   Costs If 0 Defectives are Observed and the Lot is  Accepted      92

 11    Costs in Dollars                                                93

 12    Overall Average Costs for One Acceptance -
       Rejection Scheme                        .                        95

 13    Assumed Standard Deviations for Alternative Strategies         100

-------
                             ABSTRACT
Guidelines for the quality control of ambient CO by the Federal
reference method are presented.  These include:

     1.   Good operating practices
     2.   Directions on how to assess data and qualify data
     3.   Directions on how to identify touble and improve data quality
     4.   Directions to permit design of auditing activities
     5.   Procedures which can be used to select action options and
          relate them to costs

The document is not a research report.  It is designed for use by
operating personnel.

This work was submitted in partial fulfillment of Contract Durham
68-02-0598 by Research Triangle Institute under the sponsorship of
the Environmental Protection Agency.  Work was completed as of May 1973.
                                xi

-------
1.0  INTRODUCTION

     This document presents guidelines for implementing a quality
assurance program for the continuous measurement of carbon monoxide in
the atmosphere using non-dispersive infrared (NDIR) spectrometry.
     The objectives of this quality assurance program for the NDIR method
of measuring atmospheric carbon monoxide are to:
        1)  provide routine indication, for operating purposes,
            of unsatisfactory performance of personnel and/or
            equipment,
        2)  provide for prompt detection and correction of
            conditions which contribute to the collection of
            poor quality data, and
        3)  collect and supply information necessary to describe
            the quality of the data.
     To accomplish the above objectives, a quality assurance program must
contain the following components:
        1)  routine training and evaluation of operators,
        2)  routine monitoring of the variables and/or
            parameters which may have a significant effect on
            data quality,
        3)  development through auditing procedures, statements
            and evidence to qualify data and detect defects, and
        4)  action strategies to increase the level of precision
            in the reported data and/or to detect instrument
            defects or degradation and to correct same.
     Implementation of a quality assurance program will result in data
that are more uniform in terms of precision aud accuracy.  it will enable
each monitoring network to continuously generate data that approach the
highest level of accuracy attainable with the NDIR method.

-------
     This document is divided into three parts.  They are:
        Part I, Operations Manual - The operations manual sets forth
recommended operating procedures, instructions for performing control
checks designed to give an indication or warning that invalid or poor
quality data are being collected, and instructions for performing certain
special checks for auditing purposes.
        Part II, Supervision Manual - The Supervision Manual contains
brief directions for 1) the assessment of NDIR data, 2) collection of
information to detect and/or identify trouble, 3) applying quality control
procedures to improve data quality, and 4) varying the auditing or
checking level to achieve a desired level of confidence in the validity
of the outgoing data.  Also, example monitoring strategies and costs as
discussed in Part 111 are summarized in this manual.
        Part III, Management Manual - The Management Manual presents
procedures designed to assist in 1) detecting when.data quality is
inadequate, 2) assessing overall data quality, 3) determining the extent
of independent auditing to be performed, 4) relating costs of data
quality assurance procedures to a measure of data quality, and 5) selecting
from the options available the alternative(s) which will enable one to meet
the data quality goals by the most cost-effective means.  Also, discussions
on data presentation and personnel requirements are included in this
manual.
     The scope of this document has been purposely limited to that of a
field document.  Additional background information is contained in the
final report under this contract.

-------
                      PART I.  OPERATIONS MANUAL

2.0  GENERAL

     This operations manual sets forth recommended operating procedures
for the continuous measurement of carbon monoxide in the atmosphere
using non-dispersive infrared (NDIR) spectrometry.  Quality control
procedures and checks designed to give an indication or warning that
invalid or poor quality data are being collected are written as part of
the operating procedures, and are to be performed by the operator on a
routine basis.  In addition, the performance of special quality control
procedures and/or checks as prescribed by the supervisor may be required
of the operator on certain occasions.
     The accuracy and/or validity of data obtained from this method
depends upon instrument performance and the proficiency with which the
operator performs his various tasks.  Deviations from the recommended
operational procedure may result in the collection of invalid data or at
least reduce the quality of the data.  The operator should make himself
familiar with the manufacturer's operational instructions and with the
rules and regulations concerning the NDIR method as written in the
Federal Register, Vol. 36, No. 84, Part II, April 30, 1971 (see Appendix
of this document).
     For illustration purposes, directions throughout this document are
written in terms of a 24-hour sampling period (i.e., 24 hours between
zero and span calibrations), and an auditing or checking level of 7 checks
out of a lot size of 100 sampling periods.  Sampling period durations and
auditing levels are subject to change by the supervisor and/or manager.
Such change would not alter the basic directions for performing the
operation.  Also, certain control limits as given in this manual represent
best estimates for use in the beginning of a quality assurance program and
are, therefore, subject to change as field data are collected.
     It is assumed that an analyzer which meets reference method specifi-
cations has been set up and checked out according to the manufacturer's
directions by an experienced technician.

-------
2.1  Operating Procedures
     The sequence of operations to be performed during  each  sampling
period is given in Figure 1.  Each operation or step  in the  process is
identified by a block.  Quality checkpoints in the measurement process,
for which appropriate quality control limits are assigned, are represented
by blocks enclosed by heavy lines.  Other checkpoints involve go/no-go
checks and/or subjective judgments by the operator with proper guidelines
for decision making spelled out in the procedures.  Theie. Op&LCULioYlA  and
chzchA c.uA& one. pfWQ>i&>&&> &tvp by  t>t
-------
                  ANALYZER CALIBRATION

                  1.  Verify concentration of
                     calibration gases when
                     first purchased and any
                     time desired performance
                     standards cannot be met.
                     Check cylinder pressure
                     daily.
Perform multi-
point/zero and
span calibra-
tions as
scheduled .


Multipoint
Calibration
                                                               Zero and
                                                            Span Calibration
                      Record  settings of zero
                      and span controls after
                      each calibration.
                  SAMPLING
                      Prepare analyzer for
                      sampling.
                      Check and adjust sample
                      flow and sample cell
                      pressure to specified
                      values for sampling.
                  6.  Visually check recording
                      system for proper operation.
                      Period between successive
                      zero  and span calibrations
                      No adjustments are made on
                      analyzer or recorder
                      controls during sampling
                      period.
                  OPERATIONAL CHECKS

                  8. After  each sampling period
                     check  and compare control
                     settings with settings from
                     Step 3.
 Check
Control
Settings
Figure  1:    Operational  Flow Chart  of the Measuring  Process

-------
13.
                                                                                  16
    Read and  compare sample
    flow rate with value from
    Step 5.
                                     Check
                                   Flow Rate
10.
     Read and  compare sample
     cell pressure to the value
     from step 5.
                                                       Visually check the
                                                       recorded data after
                                                       each sampling period
                                                       for signs of equipment
                                                       malfunction and
                                                       unusual pollutant
                                                       levels °r patterns.
                                                      DATA PROCESSING
11.
12.
     Check shelter temperature
     control  for proper opera-
     tion.  Check maximum
     temperature variation from
     the set  value.
     Visually check the water
     vapor  control unit for
     proper operation daily.
                                                       17.
                                                       Remove recorded data
                                                       from  recorder and edit
                                                       in preparation for
                                                       data  reduction.
                                                       Convert instrument
                                                       response to concentra-
                                                       tion in ppm as hourly
                                                       averages.
                                                                                   Data
                                                                                 Reduction
     Replace  filter monthly or
     at any sign of filter
     plugging or particulate
     buildup.
                                                                                  19
                                                  19.
                                                       Complete SAROAD form
                                                       for hourly averages and
                                                       document results of any
                                                       quality control checks.
                                                       Forward to supervisor.
14.
Visually check sample
introduction  system dally
for breakage, leaks, and
particulate deposits.
15.
     Visually check recording
     system  for proper operation
     over  the past sampling.
 Figure  1:   Operational Flow  Chart  of  the Measuring Process  (cont'd)

                                          6

-------
SAMPLE  INTRODUCTION SYSTEM
ANALYZER SYSTEM
DATA  RECORDING
     AND
DISPLAY SYSTEM
 ZERO GAS
                      SPAN GAS
                    Figure 2:  Carbon Monoxide Monitoring System Flow Chart

-------
                          ANALYZER CALIBRATION

Step 1.  Calibration Gas Check
         A multipoint calibration requires calibration gases with
concentrations corresponding to approximately 10, 20, 40, and 80 percent
                                                             3
of full scale and a zero gas containing less than 0.1 rag CO/m .  It is
further recommended that calibration gases certified to be within
+; 2 percent of the stated value be purchased in high pressure cylinders
with inside surfaces of a chromium-molybdenum alloy of low iron content
or other appropriate linings.  Store the cylinders in areas not subject
to extreme temperature changes (e.g., do not expose to direct sunlight).
It is recommended that CO in synthetic air be used for all calibration gases.
     It is recommended that at least three (3) control gas samples,
assayed and certified to be within + 1 percent of the stated level of CO,
be obtained for use in the auditing process for assessing data quality.
(Sections 2.2, 3.1, and 4.1 discuss the auditing process.)  The CO concen-
tration of the control samples should be distributed to cover the
range of about 5 to 40 ppm.*  These auditing gases (control samples)
could be purchased in size 3 cylinders to allow for portability and to
insure that the sample is exhausted before the CO concentration has
changed significantly due to deterioration with time.

A.   Concentration Verification
     When a quality assurance program is first started, the concentration
of all calibration gases on hand should be verified.  Two verification
procedures are discussed herein.  The first method represents the minimum
action necessary to verify concentration values.  It is possible that the
certified concentration values of calibration gases and auditing gases
obtained from the same supplier may be equally in error and, consequently,
be accepted as good by this method.  A second and somewhat more thorough
procedure which eliminates this possibility is given using gases from
different sources.
 The factor for converting CO from volume (ppm) to mass (mg/m^) units is:
 1 ppm = 1.145 mg/m3 at 25°C and 760 mmHg.

-------
     Once the calibration gases have been initially verified, new gases
can be verified at the time of purhcase in a routine fashion by measuring
against the old gases.
     1.  Method I
         a)  Set up and check out the analyzer.
         b)  Calibrate the analyzer with the auditing gases
             (see Step 2A, page 11 for calibration procedures).
         c)  Construct a calibration curve from at least four (4)
             points, i.e., zero, span, and two (2) upscale points
             (see Step 2A, Procedure 24 on page 14 for guidance in
             constructing a calibration curve).
         d)  Check the calibration curve and if any of the measured
             points deviate from the smooth curve by more than
             + (1.0 + 0.01 C )*ppm,  have that cylinder of auditing
             gas reanalyzed.  In some cases a subjective decision
             will have to be made by the supervisor as to whether it
             is the span gas or one of the upscale gases that is in
             error.  If all measured points are within the above
             limits, use the best fit curve as the correct
             calibration curve.
         e)  Measure the calibration gases.
         f)  Have all calibration gases whose measured value differs
             from its certified value by more than + (1.0+ 0.02 C )**ppm
             reanalyzed until an acceptable set of calibration gases is
             obtained.
         g)  Obtain and verify new calibration gases before old ones
             are exhausted by calibrating the analyzer with the old
             calibration gas.  Accept the new gas as good if the
             measured and certified values are within
             + (1.0 + 0.02 C )ppm of each other; reject the gas
             otherwise.
 *
  1.0 ppm is based on the 3o value for repeatibility from a collaborative
  test (Ref. 1), 0.01 is the stated accuracy of the auditing gas, and C
  is the certified concentration of the auditing gas.
**
  0.02 is the stated accuracy of the calibration gas, and GC is the certified
  concentration of the calibration gas.
                                    9

-------
2.   Method II
    a)   Set up and check out the analyzer.
    b)   Have two (2) sets of calibration gases (+ 2%) from
        different suppliers available.
    c)   Have one set of auditing gases (+ 1%) from another
        supplier if possible; or when available, "standard
        gases" of EPA's recommendation.
    d)   Calibrate the analyzer with the auditing or standard
        gases (see Step 2A, page 11 for calibration procedure).
    e)   Construct a calibration curve from at least four (4)
        points, i.e., zero, span and two (2) upscale points
        (see Step 2A, Procedure 24 on page 14 for guidance
        in constructing a calibration curve).
    f)   Check the calibration curve and if either one of the
        two upscale points deviates more than + (1.0 + 0.01 C )ppm
                                                             3.
        from the smooth calibration curve, have that cylinder
        of auditing gas reanalyzed.  In some cases a subjective
        decision will have to be made by the supervisor as to
        whether it is the span gas or one of the two upscale
        gases that is in error.  If both upscale points are
        within the above limits, use the best fit curve as the
        correct calibration curve.
    g)   Measure both sets of calibration gases.
    h)   If both sets of calibration gases disagree (i.e.,
        measured and certified values differ by more than
        +_ (1.0 + 0.02 C )ppm for cylinders of each set),  have
        all calibration and auditing gases reanalyzed.
    i)   If one set of calibration gases agrees (i.e., measured
        and certified values agree within + (1.0 + 0.02 C )ppm
        for each cylinder),  accept that set as good and have the
        other set reanalyzed.  If both sets agree, accept both
        sets as good.
                              10

-------
        j)  Obtain and verify new calibration gases before the
            old gases are exhausted by calibrating the analyzer
            with the old gases and measuring the new calibration
            gas.  Accept the new gas as good if the measured and
            certified values are within + (1.0 + 0.02 C )ppm of
            each other; reject the gas otherwise.

B.   Cylinder Pressure Check
     Before each calibration, check the cylinder pressure of each
calibration gas to be used.  Order replacement for any cylinder with less
             6    2
than 2.1 x 10  Nm   (300 psi) pressure.
Step 2A.  Multipoint Calibration
A.   Frequency of Calibration
     A multipoint calibration is required when:
        1)  the analyzer is first purchased,
        2)  the analyzer has had maintenance which could
            affect its response characteristics, or
        3)  when results from the auditing process show that
            the desired performance standards are not being met
            (see A of Section 2.2).
B.   Calibration Procedures
     Follow the manufacturer's detailed instructions when calibrating a
specific analyzer.  General procedures are:
        1)  Turn the power on and let the analyzer warm up by
            sampling ambient, air.  This usually requires several
            hours (as many as 24 to 48 hours) depending on the
            individual analyzer.
        2)  Connect zero gas to the analyzer.
        3)  Open the gas cylinder pressure valve (see Figure 2,
            page 7).  Adjust the secondary pressure valve until
            the secondary pressure gauge reads approximately
                    4   -2
            3.4 x 10  Nm   (5 psi) more than the desired sample
            cell pressure.  Caution:  Do not exceed the pressure
            limit of the sample cell.
                                   11

-------
 4)  Set the sample flow rate as read by the rotameter
     (read the widest part of the float) to the value that
     is to be used during sampling.
 5)  Let the zero gas flow long enough to establish a stable
     trace.  Allow at least 5 minutes for the analyzer to
     stabilize.
 6)  Adjust the zero control knob until the trace corresponds
     to the line representing 5 percent of the strip chart
     width above the chart zero or baseline.  The above is
     to allow for possible negative zero drift.  If the strip
     chart already has an elevated baseline, use it as the zero
     setting.
 7)  Let the zero gas flow long enough to establish a stable
     trace.  Allow at least 5 minutes for this.  Mark the
     strip chart trace as adjusted zero.
 8)  Disconnect the zero gas.
 9)  Connect the span gas with a concentration corresponding
     to approximately 80 percent full scale.
10)  Open the gas cylinder pressure valve (see Figure 2,
     page 7).  Adjust the secondary pressure valve until the
     secondary pressure gauge reads approximately
             4   -2
     3.4 * 10  Nm   (5 psi) more than the desired sample cell
     pressure.
11)  Set the sample flow rate, as read by the rotameter, to
     the value that is to be used during sampling.
12)  Let the span gas flow until the analyzer stabilizes.
13)  Adjust the span control until the deflection corresponds
     to the correct percentage of chart as computed by
     Cg(ppm)
             x 100 + 5 (% zero offset) = correct percentage
                                          ,  ,
                                         of chart
     where
               C_ = concentration of span gas in ppm,
     and
               Cf = full scale reading of analyzer in ppm.

                            12

-------
     As an example see Figure 3, page 15, where the % zero
     offset is 5 and the correct percentage of chart for
     the span gas of 40 ppm would be
                         x 100 + 5 = 85.
                  50 ppm
14)  Allow the span gas to flow until a stable trace is
     observed.  Allow at least 5 minutes.  Mark the strip
     chart trace as adjusted span and give concentration
     of span gas in ppm.
15)  Disconnect the span gas.
16)  Repeat Procedures 2 through 8 and
     a)  if no readjustment is required, go to Procedure 17;
     b)  if a readjustment greater than 1 ppm is required,
         repeat Procedures 9 through 16.
17)  Lock the zero and span controls.
18)  Connect the calibration gas with a concentration
     corresponding to approximately 10 percent full
     scale to the analyzer.
19)  Open the gas cylinder pressure valve (see Figure 2,
     page 7).  Adjust the secondary pressure valve until
     the secondary pressure gauge reads approximately
             4   -2
     3.4 * 10  Nm   (5 psi) more than the desired sample
     cell pressure.
20)  Set the sample flow rate to the value used during
     sampling.
21)  Let the calibration gas flow until the strip chart
     trace stabilizes.  Note:  No adjustments are made at
     this point.
22)  Disconnect the calibration gas.
23)  Repeat Procedures 18 through 22 for each of the
     calibration gases with concentrations corresponding
     to approximately 20 and 40 percent of full scale in
     that order.
                            13

-------
       24)  Fill in the information required on a calibration sheet
            and construct a calibration curve of deflection as
            percent of chart versus concentration in ppm as illus-
            trated in Figure 3.  Draw a best fit, smooth curve
            passing through the zero and span points and minimizing
            the deviation of the three remaining upscale points from
            the curve.  The calibration curve should have no inflec-
            tion points, i.e., it should either be a straight line
            or bowed in one direction only.  Curve fitting techniques
            may be used in constructing the calibration curve by
            applying appropriate constraints to force the curve
            through the zero and span points.  This procedure becomes
            quite involved, however; and the most frequently used
            technique is to fit the curve by eye.
       25)  Recheck any calibration point deviating more than
            + (1.0 + 0.02 C )ppm from the smooth calibration curve.
            If the recheck gives the same results, have that cali-
            bration gas reanalyzed.  Use the best fit curve as the
            calibration curve.
       26)  Fill in the calibration conversion sheet (see Figure 4,
            page 16)  from the calibration curve.
       27)  In certain situations the supervisor may request that
            the calibration be repeated (replicated).  In this case
            obtain both sets of data and follow his instructions for
            preparing a calibration curve.

Step 2B.   Zero and Span Calibration
A.   Frequency of Zero and Span Calibration
     A zero and span calibration is performed before and after each sampling
period (taken as every 24 hours here) or as directed by the supervisor.
                                   14

-------
Location Date Operator
Analyzer No. Range Flow Rate Cell Pressure
Zero Gas Cylinder Pressure Cylinder No.
Upscale gas (80%) Cylinder Pressure Cylinder No.
(10%) Cylinder Pressure Cylinder No.
(20%) Cylinder Pressure Cylinder No.
(40%) Cylinder Pressure Cylinder No.
Zero Control Setting Span Control Setting
Recorder Type
	 1 1 	
:::::::: :j::::i|:| ::::::::::
1 ;i!i|i!i:|::i:i;
;;;;:||i;;;;l: ;;;;:;;; ;
:::|:±::::t:S|:|::^d
±i:Sffi:;itNffiP
:;;;;EEE;;E;E;E|P|||E
iffS5^rE£fen~rt±r3
Serial No.
::::::::::::::::::::::::::::::::::::::: ::::: ±: :::::::::::::::::::::::
i;;;;=i=;|;;;iiiiii!|ii:;;jE;;J;lg;E||;;||;|;iiE
	 1 -;irf + 	 W-+J---ftifTfIt:::::: + i:lif !
i|;;|!|!;;g;;;i;|;;i!^;piPg;^i|||i;;|;:E
E:ig'-ME|:::l:::::i:::^^^±:*mS::s;:3
^?||:::::::::::|:::::::l:::gg:|:||::|:J^: + g:::
PSS^feBsli^i^^^iSi^
^^p4*^^^^%f^^^i^i f^p
y^^^gy

SrS:::|:::::S::::±|::::|::
BiTiBI|l*i|iH
|ffi;g;pp^|±^
l4u|4|i!ll !':T[[tf!^pJJ±gj]
4Jlfr4r-^H • :'4'- Lji!l'[l'i:p^|
-SH^^'Cferr^^t^
Figure 3:  Sample Calibration Curve
                15

-------
Analyzer No.
Date of Calibration
% Chart
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0

5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0

10.5
11.0
11.5
12.0
12.5
13.0
13.5
14.0
14.5
15.0

15.5
16.0
16.5
17.0
17.5
18.0
18.5
19.0
19.5
20.0
PPM











































% Chart
20.5
21.0
21.5
22.0
22.5
23.0
23.5
24.0
24.5
25,0


25.5
26.0
26.5
27.0
27.5
28.0
28.5
29.0
29.5
30.0

30.5
31.0
31.5
32.0
32.5
33.0
33.5
34.0
34.5
35.0

35.5
36.0
36.5
37.0
37.5
38.0
38.5
39.0
39.5
40.0
PPM











































% Chart
40.5,
41.0
41.5
42.0
4?. 5
43.0
43.5
44.0
44.5
45.0


45.4
46.0
46. 5
47.0
47. 5
48.0
48.5
49.0
49.5
50.0

50.5
51.0
51.5
52.0
52.5
53.0
53.5
54.0
54.5
55.0

55.5
56- P
56.5
S7.0
S7. S
58.0
58. 5
SQ.n
SQ. S
60. n
PPM











































% Chart
fiO.S
61.0
61r5
69. n
fi9.S
63.0
63.5
f>4. n
64. S
6s.n


6S_ S
6^,n
66. S
67. n
67.5
68. n
6R S
69. n
6Q. S
70.0

70. S
71 .0
71 . 5
79. n
7?. 5
73.0
73.5
7A.O
74.5
75 0

7S. S
7fi.n
7fi. S
77 n
77 S
7R 0
78 ^
7Q n
7Q 5
«n n
PPM











































% Chart
fin. s
si .n
81 .S
ft? n
R9 . •;
R^ n
R-^,5
R4,0
R^ R
R=; n


RS, S
flfi n
86.5
R7 n
87,5
RR n
RR «;
RQ ,n
RQ <;
on o

on =;
Ql.n
QI. •;
q?.n
Q9. •;
q^.n
QV •;
04 n
9i 5
QS n

Q1; =;
QA 0
Q6 5
°7 0
07 5
98 0
98 5
9° 0
Q9 5
100 0
PPM











































  Figure  4:  Table for Converting Trace Deflection in Percent of Chart to
            Concentration in PPM
                                 16

-------
B.   Zero and Span Calibration Procedures
        1)  Connect the span gas with a concentration value
            corresponding to 80 percent of full scale, or other
            values as directed by the supervisor, to the analyzer.
        2)  Open the gas cylinder pressure valve and adjust the
            secondary pressure valve (see Figure 2, page 7) until
            the secondary pressure gauge reads approximately
                    4   -2
            3.4 x 10  Nm   (5 psi) more than the desired sample
            cell pressure.
        3)  Set the sample flow rate as read by the rotameter (read
            the widest part of the float) to the value to be used
            when sampling.
        4)  Let the span gas flow long enough to establish a stable
            trace on the strip chart recorder; allow at least 5 min-
            utes.  Mark the chart trace as an unadjusted span.
            Record unadjusted span reading in ppm on form in Figure 5,
            page 18, under column entitled "Unadjusted Calibration."
        5)  Disconnect the span gas.
        6)  Connect zero gas to the analyzer.
        7)  Open the gas cylinder pressure valve and adjust the
            secondary pressure valve until the secondary pressure
                                              4   -2
            gauge reads approximately 3.4 * 10  Nm   (5 psi) more than
            the desired sample cell pressure.
        8)  Set the sample flow rate as read by the rotameter to
            the value that is used when sampling.
        9)  Let the zero gas flow long enough to establish a stable
            zero trace on the strip chart recorder; allow at least
            5 minutes.  Mark the chart trace as an unadjusted zero.
            Record the unadjusted zero reading in ppm on form in
            Figure 5, page 18 under column titled "Unadjusted Cali-
            bration."  A supervisor could use this data to compute a
            mean and standard deviation in ppm for zero drift using
            unadjusted calibration data from at least 25 sampling
            periods (see Section 4.1 for computing standard deviations)
                                   17

-------
                                             CO  ANALYZER DAILY  CHECK  SHEET
        Station Name




        Analyzer Number
Location
Date












Operator












Sample Flow Rate
U/min)
Initial












Final












Sample Cell
Pressure (Nm~2)
Initial












Final












Cylinder
Pressure(Nm~2)
Zero












Span












New Control
Knob Setting
Zero












Span












Unadjusted
Calibration
Zero












Span












00
                                            Figure  5:  Sample Daily Check Sheet

-------
     If the unadjusted zero trace is more than + 3a ppm
     from the true zero value, check the temperature control
     for the analyzer and/or shelter for proper operation and
     other likely causes.  Report the situation to the super-
     visor.  Continue with the calibration.
10)  Adjust the zero control knob until the trace corresponds
     to the true zero setting.  Let the zero gas flow until a
     stable trace is obtained.  Mark the chart trace as an
     adjusted zero.
11)  Disconnect the zero gas.
12)  Reconnect the span gas and let flow until analyzer has
     stabilized; then adjust the span control until the
     deflection on the strip chart corresponds to the span gas
     concentration in ppm using the calibration conversion
     table as illustrated in Figure 4, page 16.  Let the strip
     chart trace stabilize.  Mark the chart trace as an
     adjusted span with the span gas concentration in ppm.  A
     supervisor could compute a mean and standard deviation in
     ppm for span drift using data from at least 25 sampling
     periods (obtain span drift in ppm by subtracting the true
     span gas concentration from the unadjusted span reading as
     recorded in the last column of the form in Figure 5).  If
     the required adjustment is more than +_ 3a ppm, try to
     determine the cause and report the situation to the
     supervisor.  Continue with the calibration.
13)  Disconnect the span gas.
14)  If a span adjustment greater than +1.0 ppm (2% of chart)
     is required, repeat Procedures 6 through 12 until no
     adjustments are required.
15)  Lock the zero and span controls.
                            19

-------
Step 3.  Record Analyzer Control Settings
         Record the following information on the check sheet (see Figure 5,
page 18).  Record 1 and 2 under "New Control Knob Settings," and 3 and 4
under "Cylinder Pressure."  Include units of pressure if other than newtons
                    _2
per square meter (Nm  ) are used.
        1)  Zero control knob position,
        2)  Span control knob position,
        3)  Zero gas cylinder pressure (read first stage pressure gauge),
        4)  Span gas cylinder pressure (read first stage pressure gauge).

                                SAMPLING

Step 4.  Place Analyzer in Sampling Mode
         Connect analyzer to sample introduction system.  Allow time for
the analyzer to stabilize.

Step 5.  Sample Flow and Sample Cell Pressure Check
         Check and, if necessary, adjust the sample flow to the desired
value.
     Record the sample flow (include units if other than A/min) and sample
cell pressure on the daily check sheet as "Initial" value (see Figure 5,
page 18).

Step 6.  Recording System Check
         Check the strip chart recorder for proper operation including:
        1)  chart speed control setting,
        2)  gain control setting,
        3)  ink trace for readability,
        4)  signs of excess noise, and
        5)  the recorder's deadband (according to manufacturer's
            directions about once a month).
     Automatic data acquisition systems incorporating magnetic tape
recorder or punched paper tape are checked for proper operation according
to the manufacturer's instructions.
                                   20

-------
Step 7.  Sampling Period
         The sampling period is defined as the time interval between
successive zero and span calibrations (usually 24 hours).
     Do not change control settings on the analyzer or recording system
during the sampling period.

                           OPERATIONAL CHECKS

Step 8.  Zero and Span Control Settings
         Compare the zero and span control settings to the values
recorded on the check sheet (Figure 5) under "New Control Knob Settings."
     If the settings before and after the sampling period do not agree,
note the difference in the data log book and
     1)  perform the normal zero and span calibration.  If the
         required zero and/or span correction is less than +1.0 ppm,
         continue in the usual manner (this assumes that the
         original settings were recorded wrong or that the change
         in setting was not large),
     2)  if the required zero and/or drift correction is greater
         than +1.0 ppm, mark the data void and report the situation
         to the supervisor.  Continue normal operations.

Step 9.  Sample Flow Rate
         Read the sample flow rate from the rotameter.  Record flow rate
(£/min) on daily check sheet (Figure 5) as "Final" value.  Compare initial
and final readings.
     If the change is greater than + 20 percent of the initial value, check
the particulate filter for plugging (see Step 13) and the sample air pump
system for proper operation.  Take corrective action.
                                   21

-------
     Compute percent difference by

                          Q±-Qf
                          —	 x 100 = percent

where
          Q. = initial flow rate (£/min)
and
          Qf = final flow rate
Step 10.  Sample Cell Pressure
          Check the sample cell pressure and compare with the initial
pressure recorded on the daily check sheet (Figure 5, page 18).  If the
pressure varied during the sampling period and
                      final pressure   x     <
                      initial pressure       —     '
try to determine the cause and initiate corrective action.  Report the
change to the supervisor and record the magnitude and direction of change
on the quality control check sheet (Figure 6, page 23) under "Data Quality
Statement."
     If the pressure varied by more than 10%, the supervisor should void
the data and locate and correct the cause before sampling is resumed.

Step 11.   Temperature Control Check
          Each shelter should be equipped with a temperature-indicating
device such as a wall thermometer or a maximum and minimum registering
thermometer.  Check the thermometer to verify that the temperature control
system is operating within limits.  Control limits on allowable tempera-
ture variations are determined by a supervisor from the temperature
variation sensitivity check in Section 2.3 and in conjunction with desired
accuracy.  If a larger than usual or allowable temperature variation is
observed, record cause and corrective action in the maintenance log book
                                   22

-------
City_
         QUALITY CONTROL CHECKS

                       Pollutant
Site Location_

Site Number
                       Analyzer Number

                       Date
Supervisor Responsible for Checks (Signature)
 Auditing
  Level
 n checks
N sampling
Type of
Quality
Control
 Check
Result
  of
Check
Corrective
  Action
  Taken
 Operator
Performing
  Check
Data Quality Statement:
  Figure 6:  Sample Form for Reporting Results of Quality Control Checks
                                  23

-------
maintained in the shelter.  If no cause is identified, or if the cause is
determined but cannot be corrected immediately, report it to the supervisor.
Reset the maximum and minimum registering thermometer after checking
temperature variation for a sampling period.

Step 12.  Water Vapor Control Check
          Several techniques may be used to control water vapor interference.
Refrigeration and drying agents are two of the most commonly used methods.
Daily checks for these two methods are:
        1)  Drying Agents - The color of the drying agent (i.e.,
            silica gel or other indicating dessicants) is checked
            daily and replaced or rejuvinated when necessary as
            indicated by a change in color.
        2)  Refrigeration - Check the moisture trap between the
          /  refrigerator unit and the analyzer for condensed moisture.
            Any sign of moisture indicates a malfunction in the control
            unit.  Check the compressor for proper operation by
            measuring the temperature of the cooling coil (units are
            designed to operate at a specified temperature).  Drain
            condensate from the cold trap after each sampling period
            and before each calibration (or leave the drain cock open
            during sampling).  Report any period(s) of time that the
            sample air dewpoint was lower than the refrigerator dewpoint.
     Take corrective action and/or notify the supervisor at any sign of
malfunction or inadequacy of the water vapor control unit.  Report with
the data, by recording on the strip chart record, any period of time for
which the moisture control unit was not operating or its effectiveness was
not certain.
     Document malfunctions and corrective actions in the maintenance log
book.

-------
Step  13.  Particulate Filter Check
          A filter with a porosity of 2 to 10 micrometers is used to keep
large particles  from reaching the sample  cell.
      With no filter in the system observe the sample flow rate.  Place a
clean filter in  the filter holder and read the new flow rate.  Any drop
in flow rate is  due to the clean filter.  Record the magnitude of the
drop  in the maintenance log book.
      Initially measure the flow rate with and without the "dirty" filter
once  a month.  Replace the filter if the flow rate drop for the dirty
filter divided by the drop caused by the clean filter is more than 2.
Experience will  suggest how often such checks need to be made for a given
site.

Step  14.  Sample Introduction System Check
          A sample introduction system usually consists of an intake port,
trap  for moisture and large particulates, horizontal sampling manifold,
and exhaust blower as illustrated in Figure 2, page 7.
      Check the moisture trap for accumulated water and large particles.
Remove,  clean,  and replace the trap if any moisture and/or particulates
are present.
     Visually check the sample introduction system for breakage, leaks,
foreign objects in the intake port (e.g., spider webs, wasp nests), and
deposited particulates or excess moisture in the horizontal sampling
manifold.
     The above checks are made each sampling period (usually daily).
Conditions such as a break in the manifold or a leaky joint in the sampling
manifold network which could affect data quality are reported with the data
by a brief description of the condition under "Data Quality Statements" on
the form in Figure 6, page 23, and by marking the strip chart trace as
void and reporting the situation to the supervisor.  Take corrective
action and document in maintenance log book.
                                   25

-------
Step 15.  Recording System Check and Servicing
          Check the recording system for signs of recorder malfunctions
occurring during the past sampling period.  Specific procedures for
checking and servicing will depend on the type of recording system used.
     Check and service automatic data acquisition systems according to
the manufacturer's instructions.
     For a strip chart recorder check to see that:
        1)  the recorder did not run out of chart paper,
        2)  there is a continuous inked narrow trace for the
            entire sampling period, and
        3)  there was a uniform advancement of the chart paper
            by checking the start and end times on the chart and
            comparing with actual start and end times.
     Malfunctions in the recording system resulting in loss or invalida-
tion of data are corrected and documented in the maintenance log book.
The sampling interval affected by the malfunction is identified on the
strip chart record for that sampling period.
     Service the recorder for the next sampling period;
        1)  Check the ink supply and refill if less than 1/4 full.
        2)  Install a new roll of chart paper as necessary.

Step 16.  Visual Check of Recorded Data
          Check and edit the strip chart record for the past sampling
period to detect signs of monitoring system malfunctions and to validate
the data.
     Typical points to look for which may indicate system problems are:
        1)  A straight trace for several hours (other than minimum
            detectable).
        2)  Excess noise as indicated by a wide solid trace, or
            erratic behavior such as spikes that are sharper than
            is possible with the normal instrument response time.
            Noisy outputs usually result when analyzers are
            exposed to vibration sources.
        3)  A long steady increase or decrease in deflection.
                                   26

-------
        4)  A cyclic pattern of the trace with a definite time
            period indicating a sensitivity to changes in
            temperature or parameters other than CO concentration.
        5)  Periods where the trace drops below the zero baseline.
            This may result from a larger-than-normal drop in the
            ambient room temperature or power line voltage.
     If any of the above conditions are detected, data should be flagged,
troubleshooting done, and the supervisor informed.  Data should be declared
invalid only if malfunction of the instrument is detected; otherwise, it
should be reported.
     Also, for data validation, a graph could be prepared by the supervisor
from previous data (e.g., 1 year of data) containing the average and + 3o
values for the hourly averages for reference when editing data.  Figure 7,
page 28, is an illustration of such a graph.  The occurrence of any one
or more of the conditions listed below should be investigated for possible
causes (e.g., extra heavy traffic, shift in peak traffic hours, or periods
of atmospheric stagnation with high pollution levels):
        1)  an estimated 1 hour average falls outside the
            + 3a limits for that specific hour,
        2)  the daily pattern has shifted to left or right
            by 2 or more hours, and
        3)  abnormal pattern such as no peaks.
     Document any causes known or suspected or the absence of any known
causes on the form in Figure 6, page 23, under "Data Quality Statement."

                             DATA PROCESSING

Step 17.  Data Handling
          At the end of each sampling period the operator should make
certain that the strip chart contains the following information:
        1)  Sampling station number, location, pollutant being
            measured, and operator.
        2)  Starting time and date.  Ending time and date.
        3)  Proper identification of unadjusted zero and adjusted
            zero traces.

                                   27

-------
   7.0
                               8    10    12    14
                              Time of Day  (Hours)
16   18
20
22
24
    Figure 7:   A Sample Graph of the Mean (c)and 3a Limits of Hourly
               CO Concentrations for a 24-Hour Period
        4)  Proper identification of unadjlisted and adjusted span
            traces, and the concentration in ppm of the span gas.
        5)  Editing information identifying any periods of invalid
            data due to equipment failure or other known causes.

Step 18.  Data Reduction

A.   Procedure for Reading Hourly Averages from Strip Chart Records
     To determine the hourly average concentration from a strip chart
record, the following procedures are used:
        1)  Obtain the strip chart record for the sampling period
            in question.  The record must have adjusted span and
            zero traces at the beginning of the sampling period
            and unadjusted span and zero traces at the end of the
            sampling period.
        2)  Fill in the identification data called for at the top
            of an hourly averages sheet (see Figure 8, page 29).
                                     28

-------
CITY
SITE LOCATION_
DATE
SITE NUMBER_
POLLUTANT	
OPERATOR	
CHECKER
Hour
0-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16
16-17
17-18
18-19
19-20
20-21
21-22
22-23
23-24
Reading Zero Baseline
Original
























Check
























Original
























Check
























Difference
Original
























Check
























Add +5
Original
























Check
























PPM
Original
























Check
























         Figure 8:   Sample Sheet  for Recording Hourly Averages
                                     29

-------
3)  Using a straight edge, draw a straight line from the
    adjusted zero at the start of the sampling period to
    the unadjusted zero at the end of the sampling period
    as illustrated in Figure 9, page 31.   This line repre-
    sents the zero baseline to be used for the sampling
    period.
4)  Read the zero baseline in percent of chart at the
    midpoint of each hour interval and record the value on
    the hourly averages sheet in Figure 8, page 29.
5)  Determine the hourly averages by using a transparent
    object, such as a piece of clear plastic, with a
    straight edge at least 1 inch long.  Place the straight
    edge parallel to the horizontal chart division lines.
    For the interval of interest between two vertical hour
    lines, adjust the straight edge between the lowest and
    highest points of the trace in that interval, keeping
    the straight edge parallel to the chart division lines,
    until the total area above the straight edge bounded by
    the trace and the hour lines is estimated to equal the
    total area below the straight edge bounded by the trace
    and hour lines.  See Figure 9 for an illustrated example.
         Read and record on the hourly average sheet the
    percentage of chart deflection.
         Repeat the above procedure for all the hour intervals
    for which the analyzer was sampling and which have not
    been marked invalid.  Record all values on the hourly
    averages sheet in the column headed Reading under "Original."
6)  Subtract the zero baseline value (Column 2) from the reading
    value (Column 1) and record the difference in Column 3.
7)  Add the percent zero offset (Column 4) to the difference.
8)  Convert percentage chart values to concentration in ppm
    using the calibration conversion table (in Figure 4)
    developed from the calibration curve.  Record the ppm
    values in Column 5 on the hourly averages sheet.  The
    "Check" columns will be used in the auditing process and
    will be discussed in Section 2.2.
                           30

-------
                                       SAMPLING PERIOD  END
ia
c
-s
CO


T3

n>
o>
o
o
-h
o

-5

CO
EU
3
IQ
n>
T
o'
D.
M
tt>

o

CU
3
a.

CO
T3
O
OJ
cr
o
Jlllliiil I  I! Hi ill!  hi  llll !  i  il
i
1                   OVERNIGHT  PERIOD  NOT SHOWN
          1  PM
         12 N
                                      SAMPLING PERIOD  BEGIN

-------
Step 19.  Data Reporting
          Transcribe information and data from the hourly averages sheet
to a SAROAD Hourly Data Form (see Figure 10).
     Basic instructions for filling out the SAROAD Hourly Data Form are
given below.  If the data are to be placed in the National Aerometric
Data Bank, further instructions can be obtained from the SAROAD Users
Manual APTD-0663.
        1)  The SAROAD Hourly Data Form is an approved form for
            the recording of data observed on averages at intervals
            of less than 24 hours.  In this case the form is to be
            used for recording hourly averages of carbon monoxide
            observations.
        2)  Entries on the upper left of the form (see sample form,
            Figure 10) provide indentification (many of these items
            may already filled in by the time operators receive the
            cards).  These are:
            (1)  Agency - group recording the observations.
            (2)  City - city in which instrument is operated.
            (3)  Site - specific location of the sampler within city.
            (4)  Project name, if any.
            (5)  Parameter observed - carbon monoxide.
            (6)  Time Interval - Hourly.
            (7)  Method - Instrumental Nondispersive Infrared.
            (8)  Units of Observation - parts-per-million.
        3)  In the upper right hand corner of the SAROAD Hourly Data
            Form appears three lines of blocks for coding identifying
            information.  These correspond to the card columns of the
            numbers beneath each box when punched on an 80-column
            Hollerith card.  EPA will assign codes for the first line
            of blocks to the reporting agency when Site Identification
            Forms are initially submitted.   They consist of a two-
            digit code for state (SS) , a four-digit code for the area
                                  32

-------
CO
u>
                       LESS THAN 24-HOUR SAMPLING INTERVAL

                       LTL
1
Agency
City Name
Site Address
ENVIRONMENTAL PROTECTION AGENCY
National Aerometrlc Data Bank
P.  0. Box  12055
Research Triangle Park
North Carolina  27711
  State
Area
                   Site
                                                             Parameter observed
                                                            Time interval of obs.
                    Method
                      Of
  ri  r n  n  m
   2   31*56789  10
 Agency  | Project   Time   Year    Month

  Tl       12 13   m    15 16   17 18
  Parameter  code   Method   Units    DP
•LIMN   m   on  D
Day
19 20









-


















































St Hr]
21 22





























































Project
Rdg 1
33 31* 35 36









-















































































































Rdg 2
37 38 39 i«0







--



















































































































Rdg 3
itl i*2i»3 m*




























































































































Rdg 4
451*6 1*7 1*8









































































^-,






























*



















Rdg 5
1*9 50 51 52













k














































































































Rdg 6
53 Si* 55 56




























































































































Rdg 7
57 58 5960




























































































































Rdg 8
61 62 6361*




























































































































Rdg 9
6566 67 68




























































































































t*> 2.1 i
Rdg 10
69 70 71 72




























































































































8 29 J
Rdg 11
73 71*75 76




























































































































J 31 32
1 Rdg 12
77 78 79 80



























i '
j





















































|


i










!
I
















                                                       Figure 10:   SAROAD Hourly Data  Form

-------
    of the state in which the sampler is located (CCCC), and
    a three-digit number specifically identifying the site
    (XXX).  For the remaining two lines of blocks, codes are
    assigned for each study as follows:
    (1)  Agency - Agency Code
    (2)  Project - Project
    (3)  Time
    (4)  Year
    (5)  Month - 01 to 12 for example, as appropriate,
         a.  July - 07
         b.  August - 08
         c.  September - 09
         d.  October - 10
         e.  November - 11
    (6)  Parameter Code - 42101
    (7)  Method - 11
    (8)  Units - 07
    (9)  DP - 1 (designates the number of places to the
              right of the decimal point in the value entries)
4)  On the body of the form, the two-block first column,
    "Day", is the calendar day of the month (e.g., 01, 02).
    "ST HR" (start hour) calls for either 00 or 12 to denote
    the starting hour for which data on that line are
    recorded.  Two lines are used for each day's obser-
    vations.  The first line gives "00" (midnight) for
    "ST HR" and lists the a.m. observations.  The next line
    gives "12" (noon) for "ST HR" and lists p.m. observations.
                            34

-------
        5)  Record the hourly averages in the "Rdg" columns:
            "Rdg 1" would be for either the 0 to 1 hour reading
            or the 12 to 13 hour reading; "Rdg 2" would be for
            either the 1 to 2 hour reading or the 13 to 14 hour
            reading; etc.  In entering the hourly averages, the
            decimal point is located between the first and second
            column.
                 For example:
                    1.0 ppm would be entered as |   |   11 I 0
                    2.5 ppm would be entered as |   |   |2 |5 |  .
     Report the results of any special quality control checks performed
on special form in Figure 6.  Attach the special form for quality
control checks to the SAROAD form and give to the supervisor.
     File the hourly averages sheet in the data log book.

2.2  Special Checks for Auditing Purposes
     In making special checks for auditing purposes,  it  is  important that
the check be performed without any special check or adjustment of the
system (see section 3.2 for further discussion).  Three  special  checks
are required to properly assess data quality.  A checking or auditing
level of 7 checks out of 100 sampling periods is used here  for illus-
tration purposes.  The supervisor will specify the auditing level to be
used according to monitoring requirements.  Each of the  three checks is
discussed separately.

A.  Measuring Control Samples
    The operator, when given a control sample (auditing  gas) to  measure,
should proceed as follows:
        1)  Make no checks or adjustments on the system in
            preparation of the measurement.
                                    35

-------
         2)   Connect the control sample bottle in  the system  in
             the same manner that  the regular calibration gases
             are connected  (i.e.,  the control sample should pass
             through all the analyzer system including the water
             vapor control  unit and the particulate filter).
         3)   Let the sample gas flow until a stable trace is
             obtained.  Mark the trace with the code number of
             the control sample and the measured concentration
             in ppm.
         4)   Disconnect the control sample gas and connect the
             regular zero gas to the analyzer.
         5)   Perform a zero and span calibration as in Step 2B.
         6)   Remeasure the  control sample.
         7)   Return the control sample bottle and  the two (2)
             measured values, properly identified, to the
             supervisor.  After the supervisor evaluates the
             results, he may request that the operator perform a
             multipoint calibration and measure the control
             sample again.
         8)   The supervisor fills out and signs the form in
             Figure 6, page 23.

B.   Water Vapor Interference Check
     Water vapor checks should be independent and random; that is, a
qualified individual other than the regular operator should make the
check.  Also, the regular  operator should not know in advance when the
check is to be made.  The  individual making the check should not adjust,
replace, or  in any way change the water vapor control unit before per-
forming  the  check.  The exact procedure for performing the check will
depend on the type of control being used.  Two procedures are given
below.
     Drying Agent - When a scrubber column filled with a drying agent is
used to control water vapor interference, the check can be performed in
the following manner.
                                   36

-------
        1)  Connect the dry zero gas directly to the analyzer
            inlet, bypassing the scrubber column.  Let the
            zero gas flow until a stable trace is obtained.
            Mark the trace as dry zero gas.
        2)  Remove the segment of line bypassing the drying
            agent.  Place a 50-ml impinger containing 25 ml
            of distilled water, at room temperature, in the
            sample line such that the zero gas passes through
            the impinger, drying agent, and analyzer in that
            order.  CAUTION:  With NDIR analyzers having a
            pressurized cell, this impinger may have to be
            pressurized.  If so, it will have the full cell or
            pump pressure and will be subject to explosion.  Use
            a pressure vessel of adequate pressure capacity,
            and avoid the use of glass, if possible.
                 Let the gas flow until a stable trace is
            obtained.  Mark the trace as response to saturated
            zero gas.
        3)  Determine the difference in the two traces as an
            equivalent CO concentration in ppm.  Always subtract
            the dry measurement from the saturated measurement.
            In some cases negative values will result due to
            normal measurement error.  Document the check on the
            form for reporting quality control checks in Figure 6,
            page 23, and give to the supervisor for his signature.
     Replace or rejuvinate the drying agent if the interference is as
large as + 0.5 ppm.
     Other Methods - Other methods include refrigeration, refrigeration
preceded by humidifier, filter cells, or optical filters.  When refriger-
ation alone is used, the cold trap should be thoroughly drained before the
test and the dry zero gas allowed to flow at least 30 minutes before
reading.  Otherwise, all these methods can be checked in the following
manner.
                                   37

-------
        1)  Pass zero gas through the system in the same
            manner as is done in zero and span calibrations.
            Mark the trace as dry zero gas.
        2)  Insert a 50-ml impinger filled with 25 ml of
            distilled water at room temperature into the sample
            inlet line on the inlet side of the water control
            unit.  CAUTION:  With NDIR analyzers having a pres-
            surized cell, this impinger may have to be pressurized.
            If so, it will have the full cell or pump pressure and
            will be subject to explosion.  Use a pressure vessel
            of adequate pressure capacity, and avoid the use of
            glass, if possible.
        3)  Pass the zero gas through the impinger and system and
            mark the trace as saturated zero gas.
        4)  Determine the apparent change in concentration in ppm
            by subtracting the response to dry gas from the
            response to saturated gas.
        5)  Allowable interference levels for different control
            methods could be specified by the supervisor from
            method specifications or from the desired accuracy
            of the reported data.
        6)  Corrective action should be taken for a measured
            interference exceeding the allowable level for a
            particular control unit.
Document the results of the checks on the form in Figure 6, page 23, and
forward to the supervisor for his signature.

C.   Data Processing Check
     In auditing data processing procedures, it is convenient and allows
for corrections to be made immediately if checks are made for each sampling
period.  Hence, rather than check all 24 hourly averages for 7 days out of
every 100 days, it is suggested that 2 one-hour averages be checked each
                                   38

-------
24 hour sampling period.  Also, it is suggested that the 2 highest hourly
averages or the 2 hours for which the strip chart trace is most dynamic
in terms of spikes be selected for checking by scanning the strip chart
record.  The check must be independent;  that is, performed by an indi-
vidual other than the one who originally reduced the data.  The check is
made starting with the strip chart record and continuing through the
actual transcription of the concentration in ppm on the SAROAD form.
This, then, would include reading, calculation, and transcribing or
recording errors.
     The check is performed in the same manner as the original data were
processed as described in Section 2.1, Steps 17 through 19.  Values are
recorded on the form in Figure 8, page 29, in the "Check" columns.  If
either one of the two checks differ by as much as + 1* ppm from the
respective original value, all hourly averages for that sampling period
should be checked and corrected.  In cases where all hourly averages
have been checked, the two original, randomly selected checks should be
clearly identified on the hourly averages sheet.

2.3  Special Checks to Detect and/or Identify Trouble
     The following checks may be required when:  1) a quality assurance
program is first initiated in order to determine the analyzer's perform-
ance capabilities and to identify potential problem areas, and 2) at any
later time when it becomes increasingly difficult to meet the performance
standards of the auditing program to identify and/or evaluate trouble
areas.  Procedures for performing a zero drift check, flow rate variation
sensitivity check, temperature variation sensitivity check, and a voltage
variation sensitivity check are discussed individually.

A.   Zero Drift Check
     If available, set up equipment for monitoring and recording on strip
chart the analyzer's power source voltage and the ambient room temperature;
If such equipment is not available, use a regular A.C. voltmeter capable of
 For monitoring sites or analyzer configurations where the recorded data do
not exhibit frequently occurring sharp spikes, it would be advisable to use
a value of + 0.5 ppm as the difference above which all hourly averages are
rechecked.
                                   39

-------
measuring between 100 and 130 V.A.C. and connect it across the analyzer
power plug.  Locate a thermometer or other temperature-indicating device
near the analyzer to give a representative reading of the ambient room
temperature.  Preferably a maximum-minimum thermometer should be used.
        1)  Connect the zero gas to the analyzer and adjust
            the trace to 5 percent of chart.
        2)  Start temperature and voltage recorders or read
            and record the temperature and voltage each hour
            for the duration of the test.
        3)  Let the analyzer operate unadjusted for 24 hours
            with the zero gas.
        4)  From the strip chart(s) and recorded data determine
            the following:
            a)  difference between lowest (may be negative)
                and highest values of the zero trace in ppm
                as AC,
            b)  difference between lowest and highest temper-
                atures in °C as AT,
            c)  difference between lowest and highest line
                voltages recorded during sampling period in
                volts as AV.
            d)  Document the values of AC, AT and AV on the
                quality control check form in Figure 6, page 23.
        5)  Compare the fluctuation of the zero trace with the
            temperature and voltage fluctuations for similarities
            (i.e., see if the peaks occur at  about the same time).
            If it appears that the zero trace is sensitive to
            voltage and/or temperature changes, document with a
            short explanation under data quality statement on the
            form in Figure 6.
                                   40

-------
B.   Flow Rate Variation Sensitivity Check
        1)  With the analyzer in normal operating condition, connect
            a span gas with a concentration corresponding to approxi-
            mately 80 percent of full scale to the analyzer.
        2)  Adjust the sample flow and sample cell pressure to the
            normal operating values and allow time to obtain a
            stable trace.
        3)  Mark the trace with flow rate and cell pressure values.
        4)  Adjust the flow rate to 1/2 of its previous value.  Do
            not readjust the sample cell pressure.  Allow time to
            obtain a stable trace.
        5)  Mark the trace with flow rate and cell pressure.
        6)  Adjust the flow rate to 3 times its present setting.
            Allow time to obtain a stable trace.
        7)  Mark the trace with flow rate and cell pressure.
        8)  Record the three flow rate values with corresponding
            cell pressures and measured concentrations in ppm on the
            form for quality control checks (Figure 6).  Give the
            form to the supervisor.

C.   Temperature Variation Sensitivity Check
     From the zero drift check, if AC <_ 1 ppm and AT >_ 6°C (11°F), do not
perform a temperature sensitivity check.  Report temperature sensitivity
as AC/AT, where AC and AT are the apparent change in concentration and
change in temperature, respectively, as observed from the zero drift
check.
     If however, the above conditions are not satisfied, perform a
temperature sensitivity test as follows:
        1)  The analyzer is placed in a room where the tempera-
            ture can be varied by at least + 6°C (11°F).
        2)  Let the analyzer warm up sufficiently to get a
            stable trace.
        3)  Set up temperature-measuring device such as a maximum-
            minimum thermometer near the analyzer.
                                   41

-------
        4)  Perform a zero and span calibration  at normal  room
            temperature.  Reconnect the  zero gas to  the  analyzer.
        5)  Turn the temperature  control down  6°C.   Allow  time for
            the room temperature  and the analyzer trace  to stabilize.
            Read from the thermometer and record on  the  strip  chart
            the actual temperature.
        6)  Turn the temperature  control up 12°C from  its  previous
            setting (i.e., 6°C above the normal  setting).   Allow
            time for room temperature and analyzer to  stabilize.
            Record actual temperature on the strip chart.
        7)  Calculate
                                  cx  - c
                             AC =  Xl    J
                             AT    Tn - T,
            where
                  C   = concentration measured at T  ,

                  C   = concentration measured at T9,
                    2                              l
                   T  = highest temperature (centigrade),
                   T2 = lowest temperature (centigrade),
            and
                  -r=-  = apparent change in concentration
                        per °C change in temperature.
     Document the test on the form for reporting quality control  checks
in Figure 6, page 23.

D.   Voltage Variation Sensitivity Test
     From the zero drift check, if AC <_ 1 ppm and AV >_ 10 volts,  do not
perform a voltage variation sensitivity check.  Report voltage sensitivity
as AC/AT, where AC and AV are the actual values observed from the zero
drift check.
                                   42

-------
     If however, results from the zero drift check do not fall in the
above category, perform a voltage variation sensitivity test as follows;
        1)  Plug the analyzer into a variac capable of adjusting
            the power line voltage by +_ 15 volts from the normal
            line voltage and plug the variac into the regular
            electrical outlet.
        2)  Connect a voltmeter across the variac output leads.
        3)  Perform a regular zero and span calibration with the
            variac adjusted so that the voltmeter reads 115 volts.
        4)  With the span gas still connected, adjust the variac
            until the voltmeter reads 105 volts.  Allow the
            analyzer to stabilize.  Identify that portion of the
            strip chart trace as being at 105 volts.
        5)  Adjust the variac until the voltmeter reads 125 volts.
            Allow the analyzer to stabilize and properly identify
            the trace.
        6)  Read the trace deflection at 105 and 125 volts and
            convert to concentration in ppm.
        7)  Calculate the change in concentration per unit change
            in voltage.

                            AC = C125 ~ C105
                            AV       20
            where
                           the measured concentration at 125 volts,
            and
                    Cir)C. = the measured concentration at 105 volts,
                      — = the change in concentration per unit
                           change in voltage.
     Results from the flow rate, temperature, and voltage sensitivity checks
and estimates (could be actual measurements, if available) of the maximum
expected variation of each of the parameters under normal operational

-------
conditions are recorded in Table 1.  The maximum expected error for each
parameter is computed as illustrated in Table 1.  Assumed values of
expected variation are given for ambient room temperature as + 4.5°C from
a set value and for voltage as + 12 volts from a normal 115 volt source.
Variation in flow rate will depend on the analyzer.  A maximum expected
variation in Q can be determined by taking the 3a value of flow rate
changes from 20 to 30 sampling periods.  A fourth important parameter is
water vapor interference.  There are only two values for water vapor
interference as tested herein.  They are dry and saturated (see Section 2.2,
B for this check); therefore, the value, —, represents the maximum inter-
ference that is expected to occur and may not be representative of error
in the measured data in areas characterized by low relative humidities.
Table 1 should be completed for each analyzer and made a part of the
operational data log book.
                    Table 1:  Analyzer Evaluation Data
Variable
Flow Rate
Measure
of
Sensitivity
AC
AQ
Maximum
Expected
Variation
AQ =
Maximum
Expected Error
*jx AQ=
Temperature         ^ =               AT=9°C          ^ x AT =
Voltage             H =               AV = .24 V
Water Vapor                                              —
                                                         AW

-------
2.4  Calibration of Sample Flow and Sample Cell Pressure Indicators

A.   Flow Rate Calibration
     Rotameters usually require cleaning every six months to a year.  It
is suggested that they be calibrated after having been cleaned or at any
sign of erratic behavior.  Calibration can be accomplished using a wet
test meter as a secondary standard or a rotameter which has been recently
calibrated against a secondary standard.
     The easiest method is to use a calibrated rotameter as follows:
        1)  Place the calibrated rotameter in series with the
            rotameter to be calibrated and adjust the flow rate
            as read on the calibrated rotameter to 80 percent
            of full scale.
        2)  Adjust, if necessary, the test rotameter until it
            reads the same as the calibrated rotameter and lock
            the adjustment screw or knob.
        3)  Work down scale stopping at 65, 50, 35, 20 and 5 percent
            of full scale.  Record corresponding readings from both
            rotameters on a rotameter calibration sheet.
        4)  Using the calibration curve for the calibrated rotameter,
            construct a calibration curve of rotameter reading
            versus flow rate for the test rotameter.

B.   Sample Cell Pressure Gauge Calibration
     It is suggested that initially the sample cell pressure gauge be
calibrated at 6-month intervals or at any sign of erratic behavior of the
                                              3   -2
gauge, such as a change larger than + 6.9 x 10  Nm   (1 psi) in the sample
cell pressure during a sampling period in which the sample flow rate did
                                3           3
not change by more than + .014 m /hr (0.5 ft /hr).  If after two 6-month
calibrations the gauge shows no sign of change (i.e., reads within
          3   -2
+ 6.9 x 10  Nm   (1 psi) of the calculated pressure), go to once a year
calibrations.  Repeat the process making the period shorter or larger
according to the magnitude of change of the gauge until an optimum
calibration interval is realized.

-------
     One means of performing the calibration is using a bottle of zero
gas, the gauge to be calibrated, and a mercury manometer in a set-up as
shown in Figure 11.  With this set-up the gauge can be calibrated from
                                    5   -2
atmospheric pressure up to 2.07 x 10  Nm   (30 psi), usually 100 percent
of its range, in the following manner:
        1)  Open cylinder pressure valve.
        2)  Adjust secondary pressure regulator until the
                                                    5   -2
            secondary pressure gauge reads 1.38 x 10  Nm
            (20 psi).
        3)  Open the manometer valve slowly and let manometer
            stabilize.
        4)  Open gauge valve slowly, let test gauge stabilize.
        5)  Read the difference in the mercury columns (h) in
            centimeters.
        6)  Determine ambient atmospheric pressure (P ) from a
            calibrated wall barometer, or other suitable
                                                      _2
            barometer, in Newtons per square meter (Nm  ).
                                                             _2
        7)  Compute the pressure at the test gauge (P ) in Nm
            by
              P (Nm~2)  = 7.5 x 10 4 x h(cm Hg) + P (Nm 2)
               m                                  r
                  (Pm(psi)  = 5.17 x h(cm Hg) + Pr(psi)) .
        8)   Record the computed value of P  and the actual reading
            of the test gauge.
        9)   Repeat Procedures 2 through 8 with the secondary
                                            5   -2
            pressure gauge reading 1.72 x 10  Nm   (25 psi) and
            then 2.07 x 1Q5 Nm~2 (30 psi).
       10)   Construct a calibration curve of gauge reading versus
            computed pressure for the gauge.

-------
                                        TEST
                                        GAUGE
                                  GAUGE
                                  VALVE
 PRESSURE
 REGULATOR
PRESSURE
  VALVE
         HX1—'
                                  VENT
                        {XI	CXh
                             MANOMETER
                              VALVE
                                   MERCURY
                                  MANOMETER
   SUPPLY
   BOTTLE
Figure 11.  Calibration Set-up for Pressure Gauges

-------
 2.5  Facility and Apparatus Requirements

 A.   Facility
     A weatherproof shelter or room is required for housing the NDIR
 analyzer.  Ideally the shelter or room would be equipped with an auto-
 matic all seasons air conditioning unit capable of maintaining a pre-set
 temperature within + 3°C (5°F).  It is desirable that the heating/cooling
 be done electrically to guard against the station's emitting pollutants and
 altering the ambient air quality.  A heat pump or a cooling unit wit-.h
 electric resistance heaters would be suitable.
     The shelter must be large enough to house the analyzer, any data
 acquisition equipment, and storage space for the calibration gases.  It
 should also have adequate working space for the inspection, calibration,
 and maintenance of the system.

 B.   Apparatus
     Items of equipment with approximate costs are listed in Table 2.
 Costs associated with the analyzer, sample introduction system, and
water vapor control unit vary according to the analyzer model and make,
 size of the sampling station, and type of water vapor control unit used;
hence,  only approximate ranges of cost are given for these items.
     The calibration gases are for size 1A cylinders, certified to an
 accuracy of + 2% of the stated CO concentration as determined by analysis.
     Audit gases used as control samples are certified to an accuracy
 of +_ 1% of the stated CO concentration and for mobility can be obtained
 in size 3 cylinders.
     Each item is checked according to whether it is 1)  required in the
 reference method, 2) used to control a variable or parameter, 3) required
 for auditing purposes, or 4) used to monitor a variable.
                                   48

-------
                                      Table 2:  Apparatus Used  in  the NDIR Method
Item of Equipment
Apparatus
1. Carbon Monoxide Analyzer
2. Sample Introduction System
3. Water Vapor Control Unit
4. Strip Chart Recorder
Reagents
5. Zero Gas (1 bottle)
6. Calibration gases (4 bottles)
Optional Equipment
7. 3 Primary Standards (Spiked
Samples)
8. 2-Stage Pressure Regulator
9. Midget Impinger
10. Temperature Control (Heating/
Cooling System)
11. Diffusion Chamber
12. Constant Voltage Regulator
13. A.C. Voltmeter
14. Maximum Minimum Thermometer
Approx
Cost
1972

2 to 4,000


1,000

13
204

393
96
35
1,000
10
270
50
35
Associated
Error






Calibration
Calibration

Total Measure-
ment Error

Water Vapor
Interference
Zero Drift
Data Reduction
Zero Drift
Zero Drift
Zero Drift
Reference
Method

/
/
/


/
J








Variable
Control











/
/
/


Auditing
Equipment









/
/



•

Variable'
Monitoring














J
'
-p-
VO

-------
                     PART II.  SUPERVISION MANUAL

3.0  GENERAL

     Consistent with the realization of the objectives of a quality
assurance program as given in Section 1.0, this manual provides the
supervisor with brief guidelines  and directions for:
     1)   the collection and analysis of information necessary
          for the assessment of NDIR data quality,
     2)   isolating, evaluating,  and monitoring major
          components of system error,
     3)   changing the physical system to achieve a desired
          level of data quality,
     4)   varying the auditing or checking level to achieve
          a desired level of confidence in the validity of
          the outgoing data, and
     5)   selecting monitoring strategies in terms of
          data quality and cost for specific monitoring
          requirements.
     This manual provides brief directions that cannot cover all situa-
tions.  For somewhat more background information on quality assurance
see the Management Manual of this document.  Additional information
pertaining to the NDIR method can be obtained from the final report
for this contract and/or from the literature referenced at the end of
the Management Manual.
     Directions are written in terms of a 24-hour sampling period and
an auditing level of n=7 checks out of a lot size of N^lOO for illus-
tration purposes.  Information on different auditing levels is given in
the Management Manual.
     Specific actions and operations required of the supervisor in
implementing and maintaining a quality assurance program as discussed in
this Manual are summarized in the following listing.
                                   50

-------
1)   Data Assessment
    a)  Set up and maintain an auditing schedule.
    b)  Qualify audit results (i.e., insure that checks are
        independent and valid).
    c)  Perform necessary calculations and compare to
        suggested performance standards.
    d)  Make corrections or alter operations when standards
        are exceeded.
    e)  Forward acceptable qualified data, with audit results
        attached, for additional internal review or to user.
2)   Routine Operation
    a)  Obtain from the operator immediate reports of suspi-
        cious data or malfunctions.  Initiate corrective action
        or, if necessary, specify special checks to determine
        the trouble; then take corrective action.
    b)  On a daily basis, evaluate and dispose of (i.e., accept
        or reject) data that have been identified as question-
        able by the operator.
    c)  Examine operator's log books periodically for complete-
        ness and adherence to operating procedures.
    d)  Approve data sheets, calibration data, etc., for filing
        by operator.
    e)  File auditing results.
3)   Evaluation of Operations
    a)  Evaluate available alternative monitoring strategies
        in light of your experience and needs.
    b)  Evaluate operator training/instructional needs for
        your specific operation.
                           51

-------
 3.1  Assessment of NDIR Data
 A.   Required Information
      A valid assessment of a batch or lot of NDIR data can be made at a
 given level of confidence with information derived from three special
 checks.  The three checks are:
         1)  measurement of control samples,
         2)  water vapor interference check, and
         3)  data processing check.
 Directions for performing the checks are given in the Operations Manual,
 Section 2.2.  Directions for insuring independence and proper random-
 ization in the auditing process and for the analysis of the results are
 presented in this section.

 B.   Collection of Required Information
      1)  Measurement of Control Samples
          Acquisition of Control Samples - obtain at least three
          audit gases, for use as control samples, that have
          been assayed and certified to be within + 1 percent of the
          stated level of CO.  The three levels should be selected to
          span the range from about 5 to 40 ppm.  The specific values
          should be varied as new control samples are purchased.
          Code each control sample and record the code and certified
          concentration value in a log book which is not accessible
          to the operator(s).  Use these coded audit gases as control
          samples.
          Procedure for Performing Check - From the next 100 sampling
          periods* randomly select 7 periods** (e.g., one period
          selected randomly from seven intervals of fourteen sampling
          periods each would be satisfactory).  Then randomly select
          an hour for each of the 7 periods (it is felt that one hour
          randomly selected from the 8-hour working day will adequately
          satisfy the requirements).
  One sampling period is defined as one 24-hour day.
**The extent of auditing, i.e., the number of checks, will be discussed in
  the Management Manual.            co

-------
   At the selected hour within the appropriate sampling
period, have the operator measure one of the control samples
(i.e., one of the three audit gases).  Instruct the operator
to make no checks or adjustments on the system before
making the measurement.  A control sample can be reused
several times as long as the operator does not know the
true concentration.  The operator performs the measurement
according to the procedures given in the Operations Manual,
Section 2.2.A.
Treatment of Data - Two values are reported from each check.
One value represents the measured value of the control sample
with no adjustments made to the system prior to measurement.
The second reported value is the measure of the control sample
obtained after a zero and span calibration has been performed.
Results of the second measurement (i.e., the measurement made
after a zero and span calibration has been performed) are
used to detect and identify trouble and are discussed in
Section 2.4.  Results of the first measurement are used in
assessing data quality and are treated below.
   For each measurement or check, compute the difference in
the true or certified concentration, C > and the measured
concentration, C , in ppm as
                    d_ . = C . - C .
                     li    ai    01
where
          i is
               during a given auditing period.
i is the i   time that the check has been made
2)  Water Vapor Interference Check
    Procedure for Performing Check - Using the same seven
sampling periods as were randomly selected in 1 above, conduct
an independent (i.e., done by someone other than the regular
operator) water vapor interference check without forewarning
of the operator.  Perform the check according to instructions
given in the Operations Manual, Section 2.2.B.  To insure
                           53

-------
unbiased checks, it is recommended that the individual
performing the checks be changed periodically and that the
results from successive checks be evaluated by the super-
visor for reasonableness in terms of individual magnitudes
and variations in magnitude between checks.

Treatment of Data - For each measurement or check, compute
the difference in instrument response to dry zero gas, C,,
and saturated zero gas, C , in ppm as
                         s
                  d2i = Csi - Cdi
where
          i is
               made during a given auditing period.
i is the i   time that the check has been
3)  Data Processing Check
    Procedure for Performing Check - Independent checks on
data processing errors are made as directed in the Operations
Manual, Section 2.2.C.  Data processing checks are made each
sampling period (24 hours).  To insure continuous unbiased
checks, it is recommended that the individual performing the
checks be changed periodically.
Treatment of Data - Two checks are made each sampling period.
For each check determine the difference between the check
value and the original value.  If either check differs by as
much as + 1 ppm from the original value, all hourly averages
for that period are checked and corrected.  For reporting data
quality, the value used for correcting all hour averages (e.g.,
+ 1 ppm) is reported.  In situations where the procedure in
Section 4.1 of the Management Manual is to be followed for
data quality assessment, compute a value for reporting on the
form in Figure 12 by
   a)  subtracting the check value in ppm from the original
       value in ppm for each of the two hourly averages
       that were originally selected for checking.
                           54

-------
           b)  adding the two values computed in a) above
               algebraically (i.e., keep track of the signs)
               and dividing by 2, and
           c)  reporting the result as
                                  d3i
               where
                         i is the i   audit performed during
                           the auditing period.
C.   Treatment of Collected Information
     1)  Identification of Defects
         One procedure for identifying defects is to evaluate auditing
checks in pairs, i.e., d11d2i» di2d22  d!3d23 	' d!7d27'  If °ne °r both
members of the pair are defective, it counts as one defect.  No more than
one defect can be declared per set.  Data processing errors should be
corrected when found, and are not, therefore, discussed here.
     Any set of auditing checks in which the value of d.. . or d~. is
greater than +2.2 ppm or + 1.7 ppm, respectively, will be considered a
defect.  These values are assumed to be the 3o values and are discussed
in Section 3.2.  As data become available, these limits should be
reevaluated and adjusted, if necessary.  Small (e.g., less than 0.4 ppm)
negative values of d?. may occur as a result of measurement error.
     2)  Reporting Data Quality
         Each lot of data submitted with SAROAD forms or tapes should be
accompanied by the minimum data qualifying information as shown in
Figure 12.  The individual responsible for the quality assurance program
should sign and date the form.  As an illustration, values from Section 3.2,
Suggested Standards for Judging Performance, are used to fill in the
blanks in Figure 12.  The reported auditing rate is the rate in effect at
                                    55

-------
Supervisor's Signature
              Reporting Date
Auditing Rate for Data Errors:  n =  7  ,    N = 100
Definition of Defect:  ld-,.,1 > 2.2 ppm,   d9. > 1.7 ppm
                         li
          2i
Auditing Rate for Data Processing Errors:  n =  2  ,   N = 24
                    *   i    i     **
Definition of Defect :  |d_.|  >_ I	 ppm
Number of Defects Reported
(should be circled in the table below)
Audit
1. Measurement of Control
Samples (d )
2. Water Vapor Interference
Check (d2±)
3. Data Processing Check (d_.)
Check Values (ppm)
dll
d21
d31
d!2
d22
d32
d!3
d23
d33
_ _ 	 _ _




dli
d2i
d3i
	
	
	
dln
d2n
d3n
  Data processing errors are corrected when found and are, therefore, not<
  reported as defects.
**
  This is actually the value of one check while d_. is the average of

  two checks.
                   Figure 12:  Data Qualification Form
                                   56

-------
the beginning of the auditing period.  An increase or decrease in auditing
rate during the auditing period will be reflected by the total number of
checks reported.  The reason for change should be noted on the form.
     Check values (i.e., d.-.'s, d  's and d  *s) are calculated as directed
in Section 3.1.B and reported in ppm.  Values of d_. need be reported only
if requested by the Manager.  All reported check values exceeding the defini-
tion of a defect should be marked for easy recognition by circling on the form.
     Attach the data qualification form to the SAROAD form and forward for
additional internal review or to the user.

3.2  Suggested Standards for Judging Performance
     Results from a collaborative test of the NDIR method (Ref. 1) show
that system precision is a function of the CO concentration.  The perform-
ance standard given below in Table 3 for measurement of control samples
was taken from the point of maximum system precision which occurred at a
concentration of about 17 ppm.  The value of + 2.2 ppm represents the
3a limit.  This standard should be reevaluated and adjusted for different
concentration levels when data collected from the measurement of control
samples, as directed in Section 2.2 of the Operations Manual, become
available.
     The suggested standards given for water vapor interference and data
processing errors are no more than rough estimates.  Reasonable performance
standards can be determined as data become available from the auditing
program.

3.3  Collection of Information to Detect and/or Identify Trouble
     In a quality assurance program one of the most effective means of
preventing trouble is to respond immediately to reports from the operator
of suspicious data or equipment malfunctions.  Application of proper
corrective actions at this point can reduce or prevent the collection of
poor quality data.  Important error sources, methods for monitoring
applicable variables, and suggested control limits for each source are
discussed in this section.
                                   57

-------
                Table 3:  Suggested Performance Standards
Standards for Defining Defects
   1.  Measurement of Control Samples;  d.. .  > + 2.2 ppm
                                         XX
   2.  Water Vapor Interference Check;  d   >_ 1.7 ppm
Standard for Correcting Data Processing Errors
   3.  Data Processing Check;   d».  _>_ 1 ppm
Standards for Audit Rates
   4.  Suggested minimum auditing rates for data error; number of
       audits, n = 7; Lot size, N = 100; allowable number of defects
       per lot, d = 0.
   5.  Suggested minimum auditing rates for data processing error;
       number of audits, n = 2; lot size, N = 24; allowable number
       of defects (i.e.,  ^_.
Standards for Operation
>_ 1 ppm) per lot, d = 0.
   6.  If at any time d = 1 is observed (i.e., a defect is observed)
       for either d.. or d?., increase the audit rate to n = 20,
       N = 100 until the cause has been determined and corrected.
   7.  If at any time d = 2 is observed (i.e., two defects are observed
       in the same auditing period), stop collecting data until the
       cause has been determined and corrected.  When data collection
       resumes, use an auditing level of n = 20, N = 100 until no
       defects are observed in three successive audits.
   8.  If at any time either one of the two conditions listed below is
       observed, 1) increase the audit rate to n = 20, N = 100 for the
       remainder of the auditing period, 2) perform special checks to
       identify the trouble area, and 3) take necessary corrective
       action to reduce error levels.  The two conditions are:
          a) two (2) d .  values exceeding +1.4 ppm, or
             three (3) dn.  values exceeding +0.7 ppm
                        Xx                  —
          b) two (2) d..  values exceeding +1.0 ppm, or
                      /x
             three (3) d«.  values exceeding +0.5 ppm.
                                   58

-------
A.   Identification of Important Variables
     A great many variables can affect the expected precision and accuracy
of measurements made by the NDIR method.  Certain of these are related to
analysis uncertainties and others to instrument characteristics.  Major
sources of error are discussed below.
     Inaccuracy and Imprecision in the Stated CO Concentration of
     Calibration Gases (Ref. 1) - There are two components of error
     involved; one is the error in the original assay, and the second
     is due to the deterioration of CO with time.
          Large errors in the original assay should be detected when
     the gas is first purchased by measuring with a properly cali-
     brated and functioning analyzer.  Changes in concentration
     occurring as a function of time will be detected at a given level
     when spiked samples can no longer be measured within given limits
     with a properly calibrated and functioning analyzer.
     Water Vapor Interference - Water vapor is a positive interference
     for all NDIR analyzers (Refs. 1-5).  The magnitude of the inter-
     ference is a function of the type of control equipment being used
     and the operational state of the equipment.
          Refrigeration and drying agents have proved to be effective
     in controlling water vapor interference (Ref. 1).  Refrigeration
     units should be preceded by a humidifier when used in locations
     where the dewpoint of the ambient air is frequently below the
     dewpoint in the refrigeration unit.  Drying agents have to be
     checked and replaced frequently when used in areas characterized
     by high relative humidities (Ref. 4).
          Error due to water vapor interference is not compensated for
     or corrected by the zero and span calibrations.  Its magnitude is
     monitored as part of the auditing program by performing periodic
     water vapor interference checks.
                                   59

-------
Data Processing Errors - Data processing, starting with
reducing the data from a strip chart record through the act
of recording the measured concentration on the SAROAD form,
is subject to many types of errors.  Perhaps the major source
of error is in reading hourly averages from the strip chart
record.  This is a subjective process and even the act of
checking a given hourly average does not insure its absolute
correctness.  The approach used in Section 2.2.C of the
Operations Manual means that one can be about 55% confident
that no more than 10% of the reported hourly averages are in
error by more than + 1 ppm.
     The magnitude of data processing errors can be estimated
from, and controlled by, the auditing program through the
performance of periodic checks and making corrections when
large errors are detected.  A procedure for estimating the bias
and standard deviation of processing errors is given in
Section 4.1 of the Management Manual.
Zero Drift - Zero drift is defined as the change in instrument
output over a stated period of time, usually 24 hours, of
unadjusted, continuous operation when the input concentration
is zero.
     Several variables contribute to zero drift.  Some variables
such as variations in ambient room temperature, source voltage,
and sample cell pressure result in a zero drift that is not
linear with time.  Therefore, performing a zero and span cali-
bration does not correct for the component of drift throughout
the sampling period but rather just at the time the calibration
is performed.
     Degradation of electronic components and increased accumu-
lation of dirt in the sample cell may result in a zero drift that
is linear with time.  Periodic zero and span calibrations allow
for correction of this component of zero drift for the entire
sampling period.
     The importance of zero drift to data quality can be deter-
mined from the results obtained from measuring control samples.
If a zero and span calibration is nearly always required in order
                              60

-------
to measure a control sample within desired limits (see Section 2.1),
a zero drift check as described in section 2.2 should be performed
to determine the characteristics and major causes of the drift.
For a drift that is generally linear with time, it is valid to
perform a zero and span before measuring control samples as part
of the auditing process.  However, if the drift is a function of
variations in temperature, voltage, or pressure, as can be deter-
mined by the special checks in section 2.2, zero and span calibra-
tions should not be performed before measuring control samples for
auditing purposes.  In this case meeting desired performance stan-
dards may require more frequent zero and span calibrations or more
rigid control of temperature, voltage, and pressure, as appropriate.
Span Drift - Span drift is defined as the change in instrument
output over a stated time period of unadjusted, continuous
operation when the input concentration is a stated upscale
vaJLue.  For most NDIR analyzers the major component of span
drift is zero drift and is corrected or controlled as dis-
cussed above.  The component of span drift other than zero
drift can be caused by either optical or electronic defects.
If this component of span drift is large or shows a continuous
increase with time, the manufacturer's manual should be
followed for troubleshooting and correction of the defect.  The
importance or magnitude of span drift can be determined from the
zero and span calibrations after each sampling period.
Excessive Noise - Noise is defined as spontaneous deviations
from a mean output not caused by input concentration changes.
Excessive- noise may result when an analyzer is exposed to
mechanical vibrations.  Other sources of noise include a high
gain setting on the recorder, accumulation of dirt on sample
cell walls and windows, or loose dirt in the sample cell
(Ref. 6).
     Excessive noise is evidenced by either an extra broad
strip chart trace or a narrow but erratic trace.  The manu-
facturer's manual should be followed for troubleshooting and
correcting the cause.

                              61

-------
B.   How to Monitor Important Variables

     System noise, zero drift, span drift, and sample cell pressure are
monitored as part of the routine operating procedures.   Implementing an

auditing program effectively monitors calibration gas concentration, water

vapor interference, and data processing errors.  Variations in ambient
room temperature and/or source voltage can be monitored with a minimum-
maximum thermometer and an a.c. voltmeter, respectively.  Table 4 summarizes

the variables and how they can be monitored.
                Table 4:  Methods of Monitoring Variables
          Variable
          Method of Monitoring
1.  Calibration Gas Concentration
2.  Water Vapor Interference
3.  Data Processing Errors
4.  Zero Drift
5.  Span Drift
6.  System Noise
7.  Sample Cell Pressure
    Variation
8.  Temperature Variation
9.  Voltage Variation
Measurement of control samples as
part of the auditing program.
Water vapor interference checks per-
formed as a part of the auditing
program.
Data processing checks performed as a
part of the auditing program.

Zero check and adjustment before each
sampling period as part of routine
operating procedure.

Span check and adjustment before each
sampling period as part of routine
operating procedure.
Check of strip chart record trace for
signs of noise after each sampling
period as part of routine operating
procedure.

Reading and recording sample cell
pressure at the beginning and end
of a sampling period as part of
routine operating procedure.

Minimum maximum thermometer placed
near the anlayzer, or any other tem-
perature-indicating device,  read
periodically throughout the sampling
period.  This would usually be done
as a special check.
A.C. voltmeter measuring the voltage to
the analyzer and read periodically
throughout the sampling period.  This
would usually be done as a special check.
                                   62

-------
C.   Suggested Control Limits
     Appropriate control limits for individual variables will depend on
the level of performance needed.  Table 5 gives suggested performance
standards for measuring control samples and water vapor interference.
The standards are given in terms of a mean (bias) and standard deviation.
     Standards given for the measurement of control samples were taken
from the results of a collaborative test (Ref. 1).  The standard
deviation, a , is actually a function of the CO concentration and should
be evaluated for different levels as the necessary data become available.
The value used here is probably adequate for concentration values between
8 and 30 ppm.
     In the table, error in measuring control samples has been divided
into four components.  They are: 1) error in calibration gas concentration,
2) zero drift, 3) span drift, and 4) noise.  The values given for the
various error components were arrived at in the following way.  Verifi-
cation of calibration gas concentrations can be made, at the 3a level,
within + (1.0 +0.02 C )ppm by measuring on a properly calibrated and
functioning analyzer.  This would result in an upper limit of + 1.2 ppm
for a calibration gas with a true concentration of 10 ppm.  Any deviation
larger than +1.2 ppm indicates that the CO concentration value has
actually changed with time from the certified value and the gas should be
reassayed.
     The nonlinear component of zero drift which can result from
variations in temperature, pressure, or voltage is not totally corrected
for by zero and span calibrations.  If the zero drift is randomly positive
and negative from sampling period to sampling period, the drift probably
has a large nonlinear component.  From previous experience with NDIR
analyzers* a + 1.2 ppm nonlinear zero drift over a 24-hour sampling period
is believed to be a reasonable upper limit.
     The effect of span drift, that component other than zero drift, is a
function of the CO concentration level being measured.  This component of
drift is normally small and is usually measured at about 80 percent of full
scale.  The effect,  then, in this case would be  the ratio of  the CO  concen-
tration being measured and 40 ppm (80 percent of scale) times the drift in ppm.
                                   53

-------
                 Table 5:   Suggested Control Limits  for Parameters and/or Variables
Parameter/Variable
1. Measurement of Control Samples
Calibration Gas Concentra-
tion Error
Zero Drift (Non-Linear
Component)
Temperature Variations
Voltage Variations
Cell Pressure Variations
Span Drift (other than zero
drift)
Noise
Total:
0! = Ja2 + a2 + a2 + a2
1 1 a b c d
2. Water Vapor Interference
Suggested Performance Standard
Mean
d =.025C *
X «*
d =.025C **
a c
v°

d =0
c
dd=o
d^=.025Co
J. 3
d"2=0.3
Standard Deviation
(ppm)
0^=0.72
a =0.4
Si
a =0.4
b

a =0.2
c
a =0.2
a
a|=0.67
a2=0.3
Upper Limit
(3a)
+ 2.16
± 1-2
± i-2

+ 0.6
+ 0.6
+ 2.01
+ 1.74
C  = concentration of control sample
 cL
C  = concentration of calibration gas

-------
It is estimated that this component of drift very seldom introduces an
error as large as +_ 0.6 ppm and on the average accounts for less than
+0.2 ppm error in the measured value.
     System noise can originate in the analyzer or recorder.  Specifications
on most analyzers quote a maximum noise level of + 1 percent of full scale
or + 0.5 ppm for a 0 to 50 ppm scale.  With proper maintenance the combined
noise levels of analyzer and recorder should seldom exceed an equivalent
concentration of +_ 0.6 ppm.
     Combining the means and standard deviations of component errors as
                         d1 = d
and
. +
^
CH ~*
t<
dc +
^2H
c
dd
2
^°d
shows that at this level of control the suggested performance standard for
measuring control samples is satisfied as is evidenced by
and
                           dl = dl = °'°25 Ct
                             I
                           0- -
     Water vapor is a positive interference for NDIR CO monitors.  Standards
given here are strictly estimates and should be reevaluated and adjusted
for different types of control units as data become available.  Here it
is assumed that interference errors have a negative exponential distribu-
tion whose mean, 
-------
3.4  Procedures for Improving Data Quality
     Quality control procedures designed to control or adjust data
quality may involve a change in equipment or in operating procedures.
Table 6 lists some possible procedures for improving data quality.  The
applicability or necessity of a procedure for a given monitoring
situation will have to be determined from results of the auditing process
or special checks as performed to identify the important variables.  The
expected results are given for each procedure in qualitative terms.  If
quantitative data are available or reasonably good estimates can be made
of the expected change in data quality resulting from implementation of
each procedure, a graph similar to that in Figure 18, Section 4.3 of the
Management Manual can be constructed.  The values used in Table 13 and
Figure 18 are assumed and were not derived from actual data.
     Equipment and personnel costs are estimated for each procedure.
Personnel costs were taken as 5 dollars per hour for operator time and
10 dollars per hour for supervisor time.  Equipment costs were prorated
over 5 years for continuous monitoring, i.e., sampling 365 days a year.
All costs are for a lot size of 100, that is, 100 days of sampling.
     A procedure for selecting the appropriate quality control procedure
to insure a desired level of data quality is given below:
        1)  Specify the desired performance standard, that is,
            specify the limits within which you want the devi-
            ation between the measured and the true concentration
            to fall a desired percentage of the time.  For
            example, to measure within + 3 ppm 95 percent of the
            time, the following performance standards must be
            satisfied:
                                     1 3 PPm-
        2)  Determine the system's present performance level from
            the auditing process, as described in Section 4.1
                                   66

-------
                            Table 6:   Quality Control Procedures or Actions
Procedure
Al. Verify concentration
of calibration gas




A2. Replicate calibration
curve

Description of Action
a) Verify concentration
of new calibration gas
as described in Section
2.1 of the Operations
Manual. If certified
and measured values
differ by more than
+(1.0 +0.02 Cc*)ppm,
reject the gas.
b) Verify concentration of
calibration gas anytime
a control sample cannot
be measured within
+(1.0 + 0.01 Ca**)ppm
after the analyzer has
been calibrated and is
in proper working order.
Repeat the calibration
process after one day and
use the average of each pair
(other than 0 and span) to
construct a calibration
curve .
Expected Results
Reduces likelihood of
calibration gas
errors exceeding
+(1.0 + 0.02 C )ppm.




Reduces random error
calibration points
(other than 0 and
span) by 1 and

detects large errors
made in original
replicate.
Costs
Equip
$70






Personnel
$20




40

Total
$90




40

**
    : certified concentration of calibration gas
C  = certified concentration of control sample

-------
                              Table 6:   Quality Control Procedures or Actions (Cont'd)
Procedure
A3. Perform multipoint
calibrations






A4. Perform a zero and
span calibration
every 8 hours
A5. Sample Diffusion
Chamber





A6. Temperature Control





Description of Action
Perform multipoint calibra-
tion when deviation between
measured and stated value of
control sample differs by
more than +(1.0 4- 0.01 Ca)
when measured immediately
after a zero and span cali-
bration.
Perform zero and span cali-
bration every 8 hours as
opposed to every 24 hours.
Place a diffusion chamber
in the sample inlet line
with sufficient capacity
to integrate or smooth out
peak concentrations of
less than 5 minutes
duration.
Install heating/cooling
system capable of main-
taining ambient room
temperature to within
+ 5°F (3°C) of a preset
value .
Expected Results
Reduces errors due
to a change in
instrument response
characteristics
between multipoint
calibrations.


Reduces error due to
the nonlinear com-
ponent of zero drift.
Reduces reading error
by eliminating sharp
spikes from the
strip chart trace.



Reduces zero drift
caused by voltage
variations and noise
spikes resulting
from sudden voltage
changes.
Costs
Equip








$100


1






55





Personnel
$ 40







500


10






20





Total
$ 40







600


11






75





00

-------
                             Table 6:  Quality Control Procedures or Actions (Concl'd)
Procedure
A7. Voltage Control





A8. Improve water vapor
interference
control



Description of Action
Install constant voltage
regulator capable of
maintaining line voltage
to within + 1% of a
preset value.

Improve water vapor
interference control by
equipment change and/or
increased maintenance.


Expected Results
Reduces zero drift
caused by voltage
variations and noise
spikes resulting
from sudden voltage
changes .
Maintains water
vapor interference
to a insignificant
level when compared
to normal measure-
ment error.
Costs
Equip
$ 15





25





Personnel
None





$ 10





Total
$ 15





35





VO

-------
            of the Management Manual by setting

                           T = d^ + d"2 + d3
            and
                             -V-T
°T =S  + S2 + S3
            If the relationship of (1) above is satisfied, no
            control procedures are required.
        3)  If the desired performance standard is not satisfied,
            identify the major error components.
        4)  Select the quality control procedure(s) which will
            give the desired improvement in data quality at the
            lowest cost.  Figure 18 in Section 4.3 of the
            Management Manual illustrates a method for
            accomplishing this.
     The relative position of actions on the graph in Figure 18 will differ
for different monitoring networks according to type of equipment being
used, available personnel, and local costs.  Therefore, each network would
need to develop its own graph to aid in selecting the control procedure
providing the desired data quality at the lowest cost.

3.5  Procedures for Changing the Auditing Level to Give the Desired
     Level of Confidence in the Reported Data
     The auditing process does not in itself change the quality of the
reported data.  It does provide a means of assessing the data quality.
An increased auditing level increases the confidence in the assessment.
It also increases the overall cost of data collection.
     Various auditing schemes and levels are discussed in Section 4.2.
Numerous parameters must be known or assumed in order to arrive at an
optimum auditing level.  Therefore, only two decision rules with two
levels of auditing each will be discussed here.
                                   70

-------
     For conditions as assumed in C of Section 4.2 of the Management
Manual, a study of Figure 17 gives the following results.  These conditions
may or may not apply to your operation.  They are included here to call
attention to a methodology.  Local costs must be used for conditions to
apply to your operation.
                 •
A.   Decision Rule - Accept the Lot as Good If No Defects Are Found
     (i.e., d = 0).
     1)  Most Cost Effective Auditing Level - In Figure 17 the two
         solid lines are applicable to this decision rule, i.e.,
         d = 0.  The cost curve has a minimum at n = 7 or an audit-
         ing level of 7 checks out of 100 sampling periods.  From
         the probability curve it is seen that at this auditing
         level there is a probability of 0.47 of accepting a lot as
         good when the lot (for N = 100) actually has 10 defects
         with an associated average cost of 234 dollars per lot.
     2)  Auditing Level for Low Probability of Accepting Bad Data -
         Increasing the auditing level to n = 20, using the same
         curves in Figure 17 as in (1) above, shows a probability
         of 0.09 of accepting a lot as good when the lot actually
         has 10 defects.  The average cost associated with this
         level of auditing is approximately 430 dollars per lot.

B.   Decision Rule - Accept the Lot as Good If No More Than One (1)
     Defect is Found (i.e.. d 4 1).
     1)  Most Cost Effective Auditing Level - From the two dashed
         curves in Figure 17 it can be seen that the cost curve has
         a minimum at n = 14.  At this level of auditing there is a
         probability of 0.51 of accepting a lot of data as good when
         it has 10 defects.  The average cost per lot is approximately
         340 dollars.
                                   71

-------
     2)  Auditing Level for Low Probability of Accepting Bad Data -
         For an auditing level of n = 20 the probability of accepting
         a lot with 10 percent defects is about 0.36 as read from the
         d _< 1 probability curve.  The average cost per lot is
         approximately 375 dollars.
     It must be realized that the shape of a cost curve is determined by
the assumed costs of performing the audit and of reporting bad data.  These
costs must be determined for individual monitoring situations in order to
select optimum auditing levels.

3.6  Monitoring Strategies and Cost
     Selecting the optimum monitoring strategy in terms of cost and data
quality requires a knowledge of the present data quality, major error
components, cost of implementing available control procedures, and poten-
tial increase in system precision and accuracy.
     Section 4.3 illustrates a methodology for comparing strategies to
obtain the desired precision of the data.  Table 6 of Section 3.4 lists
control procedures with estimated costs of implementation and expected
results in terms of which error component(s) are affected by the control.
     Three system configurations identified as best strategies in Figure 18,
Section 4.3 of the Management Manual are summarized here.

A.   Reference Method
     Description of Method;  This refers to a, sampling system as illustrated
in Figure 2, Section 2.2 of the Operations Manual.   Routine operating pro-
cedures as given in the Operations Manual are to be followed with special
checks performed to identify problem areas when performance standards are
not being met.   An auditing level of n = 7 out of a lot size of N = 100
is recommended for this strategy.  This strategy is identified as AO in
Table 13 and Figure 18 in the Management Manual.
     Costs:  Taken as reference or zero cost.
     Data Quality:  Combining the assumptions made concerning water vapor
interference and data processing errors with the standard deviation of
measuring control samples (Ref.  1), the data quality is described by

                                   72

-------
                  C  = C  - (0.025 CT + 0.30) + 3(0.93)  .
                   t    m            t         —

For a true concentration, C , of 10  ppm the measured value, C  , will be
                           t                                 m
within the following limits
                             7.8 < C  < 13.3
                                    m
approximately 99.7 percent of the time.

B.   Reference Method with Sample Diffusion Chamber (A5)
     Description of Method;  Identical with (A) above except a diffusion
chamber large enough to integrate or smooth out sharp spikes of less than
5 minutes duration is used.  This reduces the chance of large errors in what
can be a highly subjective process in the measurement method.
     Costs;  Estimated average cost per lot in excess of the costs of (A)
above is 10 dollars.
     Data Quality:  From Table 13 and Figure 18 the data quality would
be described by

                  C  = C  - (0.025 C,. + 0.3) + 3(0.82).
                   t    m           t        —

For a true concentration, C , of 10 ppm the measured value, C , would
fall within the following limits
                             8.1 < C  < 13.0
                                    m
approximately 99.7 percent of the time.
C.   Reference Method Plus Sample Diffusion Chamber (A5) and
     Shelter Temperature Control Unit (A6).
     Description of Method;  Identical to (B) above with the addition of
operating the analyzer in a shelter where the temperature is controlled
to within + 5°F of a set value.
                                   73

-------
     Costs:  From Figure 18 it is seen that the average cost per lot of
data in excess of the cost of (A) above is about 85 dollars.
     Data Quality;  The combination of A5 and A6 as shown in Figure 18
has a standard deviation of about 0.7 ppm.  Neither A5 nor A6 affect
system bias; therefore, data quality can be reported as
                  CL = C  - (0.025 C.  + 0.30) + 3(0.7)
                   t    m           t         —

For a true concentration, C , of 10 ppm the measured value, C  , will
fall within the following limits
                             8.5 < C  < 12.7
                                    m
approximately 99.7 percent of the time.
                                   74

-------
                     PART III.  MANAGEMENT MANUAL

4.0  GENERAL

     The objectives of a data quality assurance program for the NDIR
method of measuring the atmospheric carbon monoxide were given in Sec-
tion 1.0.  In this section of the manual, procedures will be given to
assist the manager in making decisions pertaining to data quality based
on the checking and auditing procedures described in Sections 2 and 3.
These procedures can be employed to:
     1)   detect when the data quality is inadequate,
     2)   assess overall data quality,
     3)   determine the extent of independent auditing to be
          performed,
     4)   relate costs of data quality assurance procedures
          to a measure of data quality, and to
     5)   select from the options available to the manager
          the alternative(s) which will enable him to meet
          the data quality goals by the most cost-effective
          means.
Objectives 1 and 2 above are described in Section 4.1.  The determination
of the extent of auditing is considered in Section 4.2.  Finally the
Objectives 4 and 5 are discussed in Section 4.3.  The cost data are
assumed and a methodology provided.  When better cost data become
available, improvements can be made in the management decisions.
     If the current reference system is providing data quality consistent
with that required by the user, there will be no need to alter the physical
system or to increase the auditing level.  In fact, several detailed pro-
cedures could be bypassed if continuing satisfactory data quality is
implied by the audit.  However, if the data quality is not adequate,
i.e. either a large bias and/or imprecision in the reported data, then
(1) increased auditing should be employed, (2) the assignable cause is
to be determined, and (3) the system deficiency corrected.  The correc-
tion can take the form of a change in the operating procedure, e.g.
increased frequency of calibration, such as every month; or it may be
                                   75

-------
improved instrumentation to control environmental variations during the
sampling period, i.e. between zero-span calibrations.  Another possi-
bility is  to increase the auditing level and hence increase the confidence
in  the reported results.  These alternatives will be  considered in
Section 4.2.

4.1  Data  Quality Assessment
     As a  result of the audits suggested in the Supervision Manual, one
can (1) compare the estimated variations in the measured concentrations
with suggested standards, (2) make an overall data quality assessment,
and (3) detect when the data quality may be inadequate.  It is important
that the audit procedure be independent of previously reported results
and be a true check of the system under normal operating procedures.
Independence can be achieved by providing a control sample of unknown
concentration to the operator and requesting that he measure and report
the concentration of the sample or having another person perform the
check.  To insure that the check is made under normal operating procedures,
it is required that the audit be performed without any special check of
the system prior to the audit other than that usually performed each
sampling period, such as a zero and span calibration.
     Assume for convenience that an auditing period consists of N = 100
days (or sampling periods).   Subdivide the auditing period into n equal
periods or nearly equal periods.  Make one audit during each period and
compute the deviations (differences)  between the audit values and the
stated values (or previously determined values as determined by the
operator)  as indicated in the Supervision Manual.  For example, if
seven audits (n = 7) are to be performed over 100 sampling periods (N = 100),
the 100 periods can be subdivided into 7 intervals (6 with 14 periods and
1 with 16 periods).  The audit scheme for the data processing errors will
be slightly different from the above and will be discussed under 3)  below.
The checks are to be combined for the selected auditing period and the
mean difference or bias and the standard deviation of the differences are
to be computed as indicated below.
                                   76

-------
     1)  Measurement of Control Samples
                                        n
                                        I
                           Bias = d,   1 i
                                   1      n

where

          d1. = deviation of measured concentration of
                CO from the stated value for the control sample.
               Standard Deviation = s.
                                     'I   f      n-1

where
          d- - the average bias,
and
               the estimated standard deviation of the measured
               concentrations corrected for the average bias d .
The level of sampling or auditing n will be considered as a parameter to

to be selected by the manager to maintain the quality of data as required.

     2)  Error Due to Water Vapor Interference


                                       n

                          R-           -
                          Bias
                                  2

where

          d_.  = deviation of the measured concentrations under
                saturated and dry conditions.
                                           I. M21 - 32)2
              Standard Deviation = s0 =
                                    2   f       n-1


     The formulas for average bias and the estimated standard deviations

are the standard ones given in statistical texts (e.g., see Ref. 7).

                                   77

-------
     3)  Errors in Data Reduction and Recording
         An auditing procedure for data processing errors was described in
the Supervision Manual, Section 3.1.  This procedure suggested that two
hours be selected at random during each sampling period, that an indepen-
dent check of the concentration of CO be obtained for these hours, and
the differences between the hourly averages obtained by the operator, C  ,
and the corresponding check values, C ., i.e., C . - C  ., be computed.
These differences are treated as a go/no-go check, i.e., if either one of
the differences is larger than + 1 ppm, all hourly averages for that period
(day) are rechecked and corrected; otherwise, no corrections are made.
The value + 1 ppm is a suggestion only; experience or results from the
auditing process will indicate a more appropriate limit to use.
     In order to compute an overall bias and standard deviation associated
with the data processing, request that the values of d_. be reported with
the data (see Section 3.1.B in the Supervision Manual) and calculate
                                     n
                                     S  d3i
                               -    i=l
                        Bias = d_ = 	 , and
              Standard Deviation = s_ =
A.   Assessment of Data Quality
     The above values completely describe the variation of the reported
data from the true or stated concentrations in that all the operator,
instrument, environmental, and data reduction errors are contained as
components of one of the three errors.  Hence, an overall statement of
data quality can be obtained by combining these results as follows:

                     Overall Bias T = cL + d_ + cL

where the individual biases are set equal to zero if they are negligible
or not significantly different from zero.
                                   78

-------
                                    f~2
Overall Standard Deviation = d  = Vs  + s_ +
                                                    2    2
                                                    _ + s_ ,
and hence the true concentration should fall in the following interval
where C  is the measured concentration,
       m
                                  "    ~ *
                             C  - T + 2o_ ,
                              m     —   T '
approximately 95 percent of the time, or within the interval
                             C  - T + 35
                              m     —   T
approximately 99.7 percent of the time.  The value 2a_, is actually dependent
on the number of audits conducted.  If n is large, say about 25 or larger,
the value 2 is appropriate.
     In reporting the data quality, the bias, overall standard deviation,
and auditing level should be reported in an ideal situation.  (See
Section 4.4 for further discussion on data presentation.)  More restricted
information is suggested in the Supervision Manual as a minimal reporting
procedure.
     In summary, the data provided by these three audits is sufficient
to provide an overall estimate of the data quality.  One assumption has
been made in this analysis, which can be checked with the accumulation
of data, i.e., the values of s.., s2, and s~ have been assumed to be
independent of the concentration level.  In practice a slight dependence
on the concentration level is expected, and thus the data should be
logically grouped to measure the variation at low, intermediate, and high
concentration levels within the range of concentrations normally measured.
     If the overall reported precisions/biases of the data meet or
satisfy the requirements of the user of the data, then a reduced auditing
level may be employed; on the other hand, if the data quality is not
 A positive bias in the measurement must be subtracted from the measured
value when estimating the true concentration.
                                   79

-------
adequate, assignable causes of large deviations should be determined, and
appropriate action taken to correct the deficiencies.  This determination
may require an increased checking or auditing of the measurement process
as well as the performance of certain quality control checks, e.g., monitor
temperature and voltage variations over 24-hour sampling period, check
zero and span calibration procedures, determine adequacy of calibration
curve, etc.

B.   Assessment of Individual Measurements
     Individual checks on the standard deviations of the three audits
can be made by computing the ratio of the estimated standard deviation,
s., to the corresponding suggested standard, a., given in Table 7.  If
this ratio exceeds values given in Table 7 for any one of the audits,
this would indicate that the source of trouble may be assigned to that
particular aspect of the measurement process.  Critical values of this
ratio are given in Figure 13 as a function of sample size and two levels
of confidence.  Having assessed the general problem area* one then needs
to perform the appropriate quality control checks to determine the
specific causes of the large deviations.

4.2  Auditing  Schemes
     Auditing  a measurement process  costs  time  and money.  On  the other
hand, reporting poor quality data also  can be very costly.  For example,
the reported  data might be used to determine a  relationship between
health damage  and concentrations of  certain pollutants.  If poor quality
data are reported, it is possible that  invalid  inferences or standards
derived from  the data will cost many dollars.   These implications may be
unknown to the manager until some report is provided to him referencing
his data; hence, the importance of reporting the precision and bias with
the data.
                                   80

-------
                 Table 7 :  Critical Values of s./o.
Level of
Confidence
90%
95%
Statistic
s./a1
s./a1
n=5
1.40
1.54
n=10
1.29
1.37
n=15
1.23
1.30
n=20
1.20
1.26
n=25
1.18
1.23
           estimated standard deviation


           hypothesized or suggested standard deviation
Audit
Control Sample
Water Vapor
Data Reduction
Overall Standard Deviation
Suggested
a1 = 0.72
a2 - 0.30
03 = 0.50
CTT = 0.93
Standard
(at


(at
10 ppm)


10 ppm)*
For concentrations different from 10 ppm, use a, = 0.072 (concentration
in ppm), and
until further information concerning the dependence of a  on the concen-
tration of CO is obtained.
                                   81

-------
1.60
1.10
     0    5     10    15     20    25     30    35    40




                  Sample  Size (n)
Figure 13:  Critical Values of Ratio a la  Vs. n
                           82

-------
     As a result of the cost of reporting poor quality data* it is desirable
to perform the necessary audits to assess the data quality and to invali-
date unsatisfactory data with high probability.  On the other hand, if the
data quality is satisfactory, an auditing scheme will only increase the
data measurement and processing cost.  An appropriate tradeoff or balance
of these costs must be sought.  These costs are discussed in Section C
below.
     Consider the use of a control sample to check the major errors in the
NDIR measurement process.  Using the suggested standard deviation of 0.72
ppm for the measured concentration of CO at about the 10 ppm level and a
range of 1% in the deviation of the stated concentration of the control
sample from the true concentration, a single measured concentration
should fall between 10 + 0.1 + 2(0.72)ppm or 10 + 1.54 ppm approximately
95% of the time, 10 + 0.1 + 3(0.72)ppm or 10 + 2.26 ppm approximately
99.7% of the time.  A deviation outside the 2a (or 30) limits is
considered a defect in enforcement (or routine) monitoring of air quality.
The number of defects can be determined from the results of the reported
audits.
     Now consider the implication of an auditing scheme to determine or
judge the quality of the reported data in terms of an acceptance sampling
scheme.  Let the data be assembled into homogeneous lots of N = 50 or 100
sampling periods.  Suppose that n = 7 (10, 15, or 20) periods are sampled
in the manner suggested in Section 3.1.  That is, one day is selected at
random during each 14 periods, and for 100 periods a sample of size 7
would be obtained.  Figure 14 gives a diagram of the data flow, sampling,
and decision making process.

A.   Statistics of Various Auditing Schemes
     Suppose that the lot size is N = 100 periods (days), that n = 7 periods
are selected at random, and that there are 5% defectives in the 100, or 5
defectives.  The probability that the sample of 7 contains 0, 1, ..., 6
defectives is given by the following.
                                   83

-------
Data Flow
Lot 1
N - 100
Days


Lot 2
N - 100
Days


                                             Sample
                                             n - 7
                                        Periods  (days)
       Observe
    d - 0 defects
  Observe
d • 1 defect
                           Calculate Costs of
                             Accepting and
                           Rejecting the Lot
    Accept Data If
    Cost Comparison
    Favors This Action
    Data Quality Is
    Acceptable
                      Reject Data
                       Otherwise
       Figure 14:  Data Flow Diagram for Auditing Scheme
                                 84

-------
                     p(0  defectives)   =

 and  for  d  defectives

                     p(d  defectives)   =
n
The values are tabulated below for d = 0, 1,  ..., 6 and for the  two
data quality levels.

                                     Data Quality
d
0
1
2
3
5
6
D=5% Defectives
0.6903
0.2715
0.0362
0.0020
0.00004
= 0
D=15% Defectives
0.3083
0.4098
0.2152
0.0576
0.0084
wO
     Figure 15A gives the probabilities of d = 0 and d <_ 1  defectives  as
a function of sample size.  The probability is given for lot size N =  100,
D = 5 and 15% defectives, for sample sizes (auditing levels) from 1 to 20.
For example, if n = 10 measurements are audited and D =  5%  defectives, the
probability of d=0 defectives is 0.58.  Figure 15B gives the probabilities
for lot size N = 50, for D = 6, 10, and 20% defectives,  and for d = 0
and d <_ 1.  These curves will be used in calculating the cost  relationships
of Section C.
               N°!5'/V!88!/  =  9^:"89.=  0.6903.
     r!0d\          / 1001\        100-99'"94
                  (7!93! )
                                   85

-------
    1.0
 t
.3
 og
 ,
•s
.0
o
n
0.8
0.6
    0.4
    0.2
                                                              d i 1, D - 5%
                                                            - 0. D - 5Z
                                                              d S 1, D - 15Z
                                                              d - 0, D - 15%
                               10           15

                              Sample Size  (n)
                                                    20
25
    Figure  15A:   Probability of d Defectives in the Sample If


            the Lot  (N - 100) Contains D% Defectives
                                  86

-------
 %
 4J
•8
•§
 M
0.2
                                10          15
                                 Sample Size (n)
                                                   20
   Figure 15B:   Probability of d Defectives in the Sample If the

                Lot (N • 50) Contains DX Defectives.
This graph is for a  lot  size  of N  =  50.   Only whole numbers of defectives
are physically possible;  therefore,  even  values  of D (i.e., 6, 10, and
20 percent) are given rather  than  the  odd values of 5 and 15 percent as
given in Figure ISA.
                                      87

-------
B.   Selecting  the Auditing Level
     One consideration in determining an auditing level n used in assessing
the data quality is to calculate the value of n which  for a  prescribed
level of confidence will imply that the percent of defectives in the  lot is
less than ten percent, say, if zero defectives are observed  in the  sample.*
Figures 16A and 16B give the percentage of good measurements in the lot
sampled for several levels of confidence, 50, 60, 80,  90, and 95%.  The
curves in 16A assume that 0 defectives are observed in the sample,  and
16B, 1 defective observed in the sample.  The solid curves on the figures
are based on a lot size of N = 100; two dashed curves  are shown in
Figure 16A for N = 50; the differences between the corresponding curves
are small for the range of sample sizes considered.
     For example, for zero defectives in a sample of 7 from a lot of
N = 100, one is 50% confident that there are less than 10% defective
measurements among the 100 reported values. For zero defectives in  a
sample of 15 from N = 100, one is 80% confident that there are less than
10% defective measurements.   Several such values were obtained from
Figure 16A and placed in Table 8 below for convenient reference.
                 Table 8:  Required Auditing Levels n
                           for Lot Size N = 100
                           Assuming Zero Defectives
Confidence Level
50%
60%
80%
90%
95%
D = 10%
7
9
15
20
= 25
15%
<5
6
10
15
18
20%
<5
<5
8
11
13

Obviously,  the definition  of  defective need not  always  be  the  same and
must be  clearly stated  each time.  The definitions  employed  herein are
based on results of  collaborative  test programs.
                                   88

-------
   100
en
4J

g
3
cd
"8
3
u
(-1
Oi
    40
    20
                               10          15


                                Sample Size (n)
 Figure 16A:  Percentage  of  Good Measurements Vs. Sample Size



       for No Defectives  and Indicated Confidence Level
                                  89

-------
  100
   80
I
§  60
0)
o
3
o  40
g
o
   20
           :  60%
             80%
            444-
             90%
             I 1 I
             95%
                            10           15
                              Sample Size (n)
                                                     20
                                                                 25
Figure 16B:  Percentage of Good Measurements Vs.  Sample Size
   for 1 Defective Observed and Indicated Confidence Level
                       Lot Size = 100
                               90

-------
C.   Cost Relationships

     The auditing scheme can be translated into costs using the costs

of auditing, rejecting good data, and accepting poor quality data.
These costs may be very different in different geographic locations.
Therefore, purely for purposes of illustrating a method, the cost of

auditing is assumed to be directly proportional to the auditing level.
For n = 7 it is assumed to be $155 per lot of 100.  The cost of rejecting

good quality data is assumed to be $600 for a lot of N = 100.  The cost
of reporting poor quality data is taken to be $800.  To repeat, these
costs given in Table 9 are assumed for the purpose of illustrating a

methodology of relating auditing costs to data quality.  Meaningful
results can only be obtained by using correct local information.


                      Table 9: Costs vs. Data Quality
                                      Data Quality
                          "Good"

                          D <_ 10%
                    Incorrect Decision
                                         "Bad"

                                        D > 10%
                                   Correct Decision
Reject Lot of
      Data
Lose cost of performing
audit plus cost of reject-
ing good quality data.
(-$600 - $155)
Lose cost of performing
audit, save cost of not
permitting poor quality data
to be reported. ($400 - $155)
Accept Lot of
      Data
     Correct Decision

Lose cost of performing
audit.  (-$155)
     Incorrect Decision

Lose cost of performing
audit plus cost of declaring
poor quality data valid.
(-$800 - $155)
 Cost of performing audit varies with the sample size; is assumed to be
$155 for n = 7 audits per N = 100 lot size.
                                   91

-------
     Suppose that 50 percent of the lots have more than 10 percent
defective and 50 percent have less than 10 percent defective.  (The
percentage of defective lots can be varied as will be described in the
final report.)  For simplicity of calculation, it is further assumed
that the good lots have exactly 5 percent defectives and the poor quality
lots have 15 percent defective.
     Suppose that n = 7 measurements out of a lot N = 100 have been audited
and none found to be defective.  Furthermore, consider the two possible
decisions of rejecting the lot and accepting the lot and the relative costs
of each.  These results are given in Tables 10A and 10B.
Table 10A:  Costs If 0 Defectives are Observed and the Lot is Rejected

Reject Lot

D = 5%
D = 15%
Correct
Decision
	
P2 = 0.31
C2 = 400 - 155
Incorrect
Decision
P1 = 0.69
C1 = -600 - 155
	
Net Value ($)
PICX = -$521
P2C2 = $76
Cost =
                                                                 -$445
Table 10B:  Costs If 0 Defectives are Observed and the Lot is Accepted

Accept Lot

D = 5%
D = 15%
Correct
Decision
PI = 0.69
C3 = -155

Incorrect
Decision
—
P2 = 0.31
C4 = -800 - 155
Net Value ($)
P;LC3 = -$107
P2C4 = -$296
                                            Cost = p-j^C  + p2C4 = -$403
                                   92

-------
      The  value  P1(p2)  in the  above  table  is  the  probability  that  the
 lot  is  5% (15%)  defective given  that  0  defectives  have been  observed.
 For  example,
                    [probability  that  the  lot is  5% defective]
                    V	and_ 0 defectives are observed      /
              r°
              1°
lot is 5% defective and \
   defectives observed  )
lot is 15% defective and
/lo
1°
   defectives observed
                   0.5(0.69)
            0.5(0.69)  +  0.5(0.31)
                                      0.69.
                    [probability that the lot is 15% defective]
                   _\	and. 0 defectives are observed	I
               rlot is 5% defective and \ +
                0 defectives observed  ]
                  0.5(0.31)
            0.5(0.31)  +  0.5(0.69)
                                       /
                                    = 0.31.
                               lot is 15% defective and
                                0 defectives observed
It was assumed that the probability that the lot is 5% defective is 0.5.
The probability of observing zero defectives , given the lot quality is 5%
or 15%, can be read from the graphs of Figures ISA or 15B.
     A similar table can be constructed for 1, 2, ..., defectives and the
net costs determined.   The net costs are tabulated in Table 11 for 1, 2,
and 3 defectives.  The resulting costs indicate that the decision preferred
from a purely monetary viewpoint is to accept the lot if 0 defectives are
observed and to reject it otherwise.  The decision cannot be made on this
basis alone.  The details of the audit scheme also affect the confidence
which can be placed in the data qualification; consideration must be given
to that aspect as well as to cost.

                      Table 11:  Costs in Dollars

Decision
Reject Lot
Accept Lot
d = number of defectives
0
-445
-403
1
-155
-635
2
+101
-839
3
+207
-928
                                   93

-------
D.   Cost Vs. Audit Level
     After  the decision criteria have been selected,  an average  cost  can
be calculated.  Based on the results of Table 11,  the decision criterion
is to accept the lot if d = 0 defectives are observed and  to  reject the
lot if d =  1 or more defectives are observed.  All the assumptions of
the previous section are retained.  The auditing level is  later  varied
to obtain the data in Figure 17.
     One example calculation is given below and summarized in Table 12.
The four cells of Table 12 consider all the possible  situations  which can
occur, i.e., the lots may be bad or good and the decision  can be to
either accept or reject the lot based on the rule  indicated by Table
The costs are exactly as indicated in Tables 10A and  10B.   The probabilities
are computed as follows.

         q1 = (prob. that the lot is 5% defective  and 1 or
               more defects are obtained in the sample)
            = (prob. that the lot is 5% defective)(prob. 1 or
               more defectives are obtained in the sample
               given the lot is 5% defective)
            =  0.5 (0.31) = 0.155

Similarly q_, q_, and q. in Table 12 are obtained  as  indicated below.

                            q2 = 0.5 (0.69) = 0.345
                            q3 = 0.5 (0.69) = 0.345
                            q4 = 0.5 (0.31) = 0.155

The sum of all the q's must be unity as all possibilities  are considered.  The
value 0.5 in each equation is the assumed proportion  of good  lots (or poor
quality lots).   The values 0.31 and 0.69 are the conditional  probabilities
that given the quality of the lot, either d = 0 or d  = 1 or more defectives
are observed in the sample.   Further details of the computation  are given
in the final report of this  contract.
                                   94

-------
                 Table 12:  Overall Average Costs for One
                            Acceptance - Rejection Scheme
Decision
Reject any lot of
data if 1 or more
defects are found.
Accept any lot of
data if 0 defects
are found.
Good Lots
D = 5%
qx = 0.155
C1 = -$755
q3 = 0.345
C3 = -$155
Bad Lots
D = 15%
q2 = 0.345
C2 = $245
q4 = 0.155
C4 = -$955

qlCl + q2C2 = ~$ 32
q3C3 + q4C4 = ~$202
                                                    Average Cost = -$234

     In order to interpret the concept of average cost, consider a large
number of data lots coming through the system; a decision will be made
on each lot in accordance with the above and a resulting cost of the
decision will be determined.  For a given lot, the cost may be any one of
the four costs, and the proportion of lots with each cost is given by the
q's.  Hence the overall average cost is given by the sum of the product of
q's by the corresponding- C's.
     In order that one may relate the average cost as given in Table 12
to the costs given in Table 11, it is necessary to weight the costs in
Table 11 by the relative frequency of occurrence of each observed number
of defectivesj i.e., prob(d).  This calculation is made below.
No. of
Defectives
d = 0
1
2
3
4
Decision
Rule
Accept
Reject
Rej ect
Reject
Reject
Costs ($) from
Table 11
- 403
- 155
101
207
244

Prob(d)
0.50
0.34
0.1255
0.030
0.0042

Cost x Prob(d)
-$201.5
- 52.7
12.6
6.2
1.0
                                       Totals   0.9997
-$234.4
                                   95

-------
Thus the value -$234 is the average cost of Table 12 and  the weighted
average of  the costs of Table 11.  The weights, Prob(d),  are obtained
as follows:

    Prob(d=0) = Prob(lot is good and d=0 defectives are observed)
              + Prob(lot is poor quality and d=0 defectives are observed)

              = 0.5 (0.69) + 0.5 (0.31) = 0.50.

This is the proportion of all lots which will have exactly 0 defectives
under the assumptions stated.  For d = 1, 2, 3, and 4 the values of  the
probabilities in parentheses above can be read from the table on page 87.
     Based on the stated assumptions, the average cost was determined for
several auditing levels as indicated in Table 12.  These  costs are given
in Figure 17.  One observes from this figure that n = 7 is cost effective
given that one accepts the lot only if zero defectives are observed. (See
curve for d = 0.
     If the lots are accepted if either 0 or 1 defectives are observed,
then referring to the curve d _<_ 1, the best sampling level is n = 15.
The curve of probability of d = 0 (d <_ 1) defectives in a lot of N = 100
measurements if there are 10% defectives is also given on the same
figure.
     Another alternative is to accept all data without performing an
audit.   Assuming that one-half (50%) of the lots contain more than 10%
defectives, the average cost on a per lot basis would be 0.5(-$800) = -$400.
This, however, would preclude qualification of the data.  Regardless of cost
it would be an unacceptable alternative.

4.3  Data Quality Versus Cost of Implementing Actions
     The discussion and methodology given in the previous section was
concerned with the auditing scheme (i.e., level of audit or sample size,
costs associated with the data quality, etc.).   Increasing the level
of audit of the measurement process does not by itself change the
quality  of the data but it does  increase the information about the
                                   96

-------
                       Probability
                       if d - 0
                                                            0)
                                                            N
                                                             O

                                                             01
                                                            CO
                                                            
-------
quality of the reported data.  Hence, fewer good lots will be rejected
and more poor quality data will be rejected.  If the results of the
audit imply that certain process measurement variables are major contrib-
utors to the total error or variation in the reported concentration of
CO, then alternative strategies for reducing these variations need to be
investigated.  This section illustrates a methodology for comparing the
strategies to obtain the desired precision of the data.  In practice it
would be necessary to experiment with one or more strategies, determine
the potential increase in precision, and relate the precisions to the
relative costs as indicated herein.  Several strategies are considered but
only a few of the less costly ones would be acceptable as illustrated in
Figure 18.  The assumed values of the standard deviations and biases for
each type audit are not based on actual data, except for the reference
method.  In this case values were taken from Ref.  1.  These values are
probably smaller than those experienced in the field.
     Several alternative actions or strategies can be taken to increase
the precision of the reported data.  For example,  if the instrument
responses to voltage, temperature, and humidity variations are large
contributors to the variation of an observed instrument response, then
additional control equipment for one or more of the environmental effects
can reduce the variation of the measured responses by calculated amounts
and thus reduce the error of the reported concentrations.  In this manner,
the cost of the added controls can be related to the data quality as
measured by the estimated errors of the reported results.  It must be
recognized that these errors are dependent to some extent on the
concentration.
     Suppose that it is desired to make a statement that the concentration
of CO is within 2.1 ppm (3a  limit) for 1 hour averages and that the
minimal cost control equipment and checking procedures are to be employed
to attain this desired precision.  In order to determine a cost efficient
procedure, it is necessary to estimate the variance for each source of
error (or variation) for each strategy and then select the strategy or
combination of strategies which yields the desired precision with minimum
cost.  One such calculation is summarized in Table 13 with assumed costs
of equipment and control procedures.
                                   98

-------
     Examining the graph in Figure 18 of cost versus precision, one
observes that the combination of actions A5 and A6 is the least costly
strategy that meets the required goal of 2.1 ppm  (aT < 0.7 ppm) in the
                                                   i. """"
reported concentration.  Similarly A5 meets the goal of 2.5 ppm
(OT _<^ 0.83 ppm).  The assumed values of the standard deviations of the
measured concentrations of CO for the alternative courses of action are
given in Table 13.  The estimated costs for the various alternatives are
given in Table 6 of Section 3 and in Table 13.
     Two curves are given in Figure 18 to illustrate the assumed
relationship between cost of reporting poor quality data and the measure
of precision cr_.  In one curve it is assumed that there is 0 cost in
reporting data for which a_, £ 0.60 and the cost increases rapidly beyond
OT = 0.60.  The other curve assumes a_ = 0.80 is acceptable.
     Data processing or data reduction errors have been included in these
sample analyses for illustration purposes.  The value assumed for a, in
Table 13 is an estimate and was not derived from actual data.  If
information regarding the distribution of data processing errors is
desired, the field supervisor could be requested to forward actual values
for d_-, d«2, 	, d,7 with the data qualification form shown in Figure 12
of the Supervision Manual.
                                   99

-------
     Table 13:  Assumed Standard Deviations for Alternative Strategies
1. Control Sample d..
°1
2. Water Vapor d-
Interference
°2
3. Data Reduction d»
°3
Alternative Strategies
AO
0.25
0.72
0.30
0.30
0
0.50
Al
0.25
0.70
0.30
0.30
0
0.50
A2
0.25
0.69
0.30
0.30
0
0.50
A3
0.10
0.68
0.30
0.30
0
0.50
A4
0.25
0.54
0.30
0.30
0
0.50
A5
0.25
0.72
0.30
0.30
0
0.50
A6
0.25
0.57
0.30
0.30
0
0.50
A7
0.25
0.68
0.30
0.30
0
0.50
A8
0.25
0.72
0.10
0.10
0
0.50
  ** 2

       (overall variance)
    OT (overall std. dev.)



 ***      Max. Pos.
    Bias

          Max . Neg .
   Added Cost/100 Periods
0.86
0.93
0.55
0
$0
0.83
0.91
0.55
0
$90
1.82
0.90
0.55
0
$40
0.80
0.90
0.40
0
$40
0.63
0.79
0.55
0
$600
0.67
0.82
0.55
0
$11
0.66
0.82
0.55
0
$75
0.80
0.90
0.55
0
$15
0.78
0.88
0.35
0
$35
Alternative Strategies are given in Table 6, Section 3; the 0,,'s,


i=l, 2, and 3, are assumed values based on results given in Ref. 1, and


where data are not available they are engineering judgments.



      2.2,2
 ** 2
   °"T
       .FT"

'  aT =VaT •
***
   Bias  =  T  =  d1 + d_ + d«; the biases and standard deviations which are


   dependent on concentration level are determined at 10 ppm.  In order to


   estimate the true concentration, the estimated bias T must be subtracted


   from the measured concentration and then the appropriate 2a  or 3a


   error added, i.e., c  - T + 20_, gives the 95% limits.
                       m         JL
                                      100

-------
 •co-
 o
 u

 T3
 
-------
4.4  Data Presentation
     A reported value whose precision and accuracy (bias) are unknown is
of little, if any, worth.  The actual error of a reported value—that is,
the magnitude and sign of its deviation from the true value—is usually
unknown.  Limits to this error, however, can usually be inferred, with
some risk of being incorrect, from the precision of the measurement
process by which the reported value was obtained and from reasonable
limits to the possible bias of the measurement process.  The bias, or
systematic error, of a measurement process is the magnitude and direc-
tion of its tendency to measure something other than what was intended;
its precision refers to the closeness or dispersion of successive
independent measurements generated by repeated applications of the
process under specified conditions, and its accuracy is determined by
the closeness to the true value characteristic of such measurements.
     Precision and accuracy are inherent characteristics of the measure-
ment process employed and not of the particular end result obtained.
From experience with a particular measurement process and knowledge of
its sensitivity to uncontrolled factors, one can often place reasonable
bounds on its likely systematic error (bias).  This has been done in the
model for the measured concentration as indicated in Table 13.  It is
also necessary to know how well the particular value in hand is likely
to agree with other values that the same measurement process might have
provided in this instance or might yield on measurements of the same mag-
nitude on another occasion.  Such information is provided by the estimated
standard deviation of the reported value, which measures (or is an index
of) the characteristic disagreement of repeated determinations of the
same quantity by the same method and thus serves to indicate the precision
(strictly, the imprecision) of the reported value.
     A reported result should be qualified by a quasi-absolute type of
statement that places bounds on its systematic error and a separate
statement of its standard deviation, or of an upper bound thereto, when-
ever a reliable determination of such value is available.  Otherwise a
computed value of the standard deviation should be given together with
a statement of the number of degrees of freedom on which it is based.
                                   102

-------
     As an example, consider the example given in Section 4.3, Table 13.
In this case the estimated 2a limits of the reported concentration of CO
by the NDIR reference method are +2(0.93) or + 1.86 ppm.  Suppose that
a positive bias of 0.3 ppm results from the water vapor interference,
then the results could be reported as the measured concentration, C ,
with the following 2a limits and audit level, e.g.,

                         C  - 0.3 + 1.9 ppm, n = 20 .

     The replication error is a measure of the variation of successive
determination of CO with the same operator and instrument on the same
sample within a time interval short enough to avoid change of environ-
mental factors.  This replication error as given by the standard devia-
                                         q
tion s was measured to be about 0.17 mg/m  (.15 ppm).
     The repeatability error is a measure of the variation between test
results on the same sample on different days by the same laboratory.  The
standard deviation was estimated to be s = 0.57 mg/m  (.50 ppm).  This
measure of variation must by definition include the replication error, i.e.,

                              2                 2       1/2
         a(repeatability) = [a (replication) + a  (day)]
It is indicated (Ref. 1) that the a for repeatability varies significantly
between laboratories; it appears to depend on the concentration but the
results are erratic.
     The requirements for reporting data quality as outlined in the
Supervision Manual involves adopting a standard, performing an audit, and
comparing the audit result to the standard.  A defect is defined in terms
of the standard.  This approach does not make maximum use of the collected
data, but its simplicity should aid in the implementation of a quality
assurance program.  After experience has been gained in using the auditing
scheme and in calculating the results, it is recommended that the above,
more comprehensive method of data presentation be implemented.
                                   103

-------
4.5  Personnel Requirements
     Personnel requirements as described here are in terms of the NDIR
method only.  It is realized that these requirements may be only a minor
factor in the overall requirements from a systems point of view where
several measurement methods are of concern simultaneously.

A.   Training and Experience
     1.  Director
         The director or one of the professional-level employees should
have a basic understanding of statistics as used in quality control.  He
should be able to perform calculations, such as the mean and standard
deviation, required to define data quality.  The importance of and require-
ments for performing independent and random checks as part of the auditing
process must be understood.  Three references which treat the above-
mentioned topics are listed below:
         Probability and Statistics for Engineers, Irvin Miller
         and John E. Freund, published by Prentice-Hall, Inc.,
         Englewood, N. J., 1965.
         Introductory Engineering Statistics, Irwin Guttman and
         S. S. Wilks, published by John Wiley and Sons, Inc.,
         New York, N. Y., 1965.
         The Analysis of Management Decisions, William T. Morris,
         published by Richard D. Irwin, Inc., Homewood, Illinois,
         1964.
     2.  Operator
         There are or can be two levels of operation involved in the NDIR
method.
     First, an operator or technician who is involved in the preliminary or
initial setup and checkout or is responsible for troubleshooting and
repairing the analyzer should have technical training in electronics and/or
instrumentation as obtained in a technical or service school or
                                   104

-------
several years of on-the-job experience.  For a. specific analyzer it would
be desirable to have the technician checked out by a manufacturer's repre-
sentative or at least to have him participate, with the representative, in
the initial installation and startup.   The manufacturer's instruction
book should be available for study or reference by the technician.
     Routine operations involve the use of external controls only and
require no high-level skills.  A high school graduate with proper super-
vision and on-the-job training can become effective at this level in a
very short time.
     An effective on-the-job training program could be as follows:
        a)  Observe experienced operator perform the different
            tasks in the measurement process.
        b)  Study the operational manual of this document and
            use it as a guide for performing the operations.
        c)  Perform operations under the direct supervision
            of an experienced operator.
        d)  Perform operations independently but with a high
            level of quality control checks utilizing the
            technique described in the section on Operator
            Proficiency Evaluation Procedures  below to encourage
            high quality work.
Another alternative would be to have the operator attend an appropriate
basic training course sponsored by EPA.

4.6  Operator Proficiency Evaluation Procedures
     One technique which may be useful for early training and qualification
of operators is a system of rating the operators as indicated below.
     Various types of violations (e.g., invalid sample resulting from
operator carelessness, failure to maintain records, use of improper equip-
ment, or calculation error) would be assigned a number of demerits
depending upon the relative consequences of the violation.  These demerits
could then be summed over a fixed period of time of one week, month, etc.,
and a continuous record maintained.  The mean and standard deviation of
the number of demerits per week can be determined for each operator and
                                   105

-------
 a quality  control  chart provided for maintaining a record  of proficiency
 of each operator and whether any changes  in  this level have occurred.   In
 comparing  operators, it is necessary to assign demerits on a per unit
 work load  basis in order that the inferences drawn from the chart be
 consistent.   It is not necessary or desirable for the operator  to be
 aware of this form of evaluation.  The supervisor should use it as a means
 of determining when and what kind of instructions and/or training is
 needed.
     A sample QC chart is given in Figure 19 below.  This  chart assumes
 that the mean and  standard deviation of the number of demerits per week,
 e.g., are  5 and 1, respectively.  After several operators  have been evalu-
 ated for a few weeks, the limits can be checked to determine if they are
 both reasonable and effective in helping  to improve and/or maintain the
 quality of the air quality measurement.
     The limits should be based on the operators whose proficiency is
 average or slightly better than average.  Deviations outside the QC
 limits, either above or below, should be considered in evaluating the
 operators.   Identifying those operators whose proficiency  may have
 improved is just as important as knowing those operators whose proficiency
may have decreased.
     The above procedure may be extended to an entire monitoring network
 (system).   With appropriate definitions of work load, a continuous record
may be maintained of demerits assigned to the system.  This procedure might
serve as an incentive for teamwork,  making suggestions for improved
operation procedures, etc.
             •H
             C
             * 7
             M
             O c
             Q     1234   5  6  7  8  9  10 11 12 13
                          Time Intervals (Weeks)
   Figure 19:   Sample QC Chart for Evaluating Operator  Proficiency
                                   106

-------
                              REFERENCES
1.  Herbert C. McKee et al., "Collaborative Study of Reference Method
    for the Continuous Measurement of Carbon Monoxide in the Atmosphere
    (Non-Dispersive Infrared Spectrometry)," Southwest Research Insti-
    tute, Contract CPA 70-40, SwRI Project 01-2811, San Antonio, Texas,
    May 1972.

2.  Frank McElroy, "The Intech NDIR-CO Analyzer," presented at the llth
    Methods Conference in Air Pollution, University of California
    Berkeley, California, April 1, 1970.

3.  Hezekiah Moore, "A Critical Evaluation of the Analysis of Carbon
    Monoxide with Nondispersive Infrared (NDIR)," presented at the
    9th Conference on Methods in Air Pollution and Industrial Hygiene
    Studies, Pasadena, California, February 7-9,  1968.

4.  Richard F. Dechant and Peter K. Mueller, "Performance of a Continuous
    NDIR Carbon Monoxide Analyzer," AIHL Report No. 57, Air and Industrial
    Hygiene Laboratory, Department of Public Health, Berkeley, California,
    June 1969.

5.  Joseph M. Colucci and Charles R. Begeman, "Carbon Monoxide in Detroit,
    New York, and Los Angeles Air," Environmental Science and Technology _3
    (1), January 1969, pp 41-47.

6.  "Tentative Method of Continuous Analysis for Carbon Monoxide Content
    of the Atmosphere (Nondispersive Infrared Method)," ija Methods of
    Air Sampling and Analysis, American Public Health Association,
    Washington, D. C., 1972, pp 233-238.

7.  John Mandel, The Statistical Analysis of Experimental Data, Interscience
    Publishers, Division of John Wiley & Sons, New York, N. Y., 1964.
                                  107

-------
                      APPENDIX
REFERENCE METHOD FOR THE CONTINUOUS MEASUREMENT
       OF CARBON MONOXIDE IN THE ATMOSPHERE
       (NON-DISPERSIVE INFRARED SPECTROMETRY)
Reproduced from Appendix C, "National Primary and Secondary Ambient Air
Quality Standards," Federal Register,  Vol 36, No. 84, Part II, Friday,
April 30, 1971
                         108

-------
                                                 RULES AND  REGULATIONS
                      Figure B2. Assembled sampler and shelter.
                                  Figure B3,  Orifice calibration unit.
APPENDIX  G—REFERENCE  METHOD  roa TUB
  CONTINUOUS   MEASUREMENT   or  CARBON
  MONOXIDE  IN  THE  ATMOSPHERE  (NON-
  DISPERSIVE INFRARED  SPECTROMETRT)

  1. Principle and Applicability.
  1.1  This method Is based on the absorp-
tion of  infrared  radiation by carbon mon-
oxide. Energy from a source  emitting radia-
tion In  the Infrared region Is split  Into
parallel  beams  and  directed through ref-
erence and  sample cells. Both  beams pass
Into matched cells, each containing a selec-
tive  detector and  CO. The  CO In the  cells
absorb infrared radiation only at Its charac-
teristic frequencies and the detector Is sensi-
tive to those frequencies. With a nonabsorb-
Ing gas In the reference  cell, and with no
CO In  the sample  cell,  the signals from
both  detectors are balanced  electronically.
Any CO Introduced Into the sample cell will
absorb radiation, which reduces the temper*
ature and pressure In the detector cell and
displaces a* dlaphram. This displacement Is
detected electronically and amplified to pro-
vide  an output signal.
  1.2  This method Is applicable to the  de-
termination of carbon monoxide In ambient
atr, and  to  the  analysis of  gases under
pressure.
  2. Range and Sensitivity.
  2.1  Instruments are available that meas-
ure In the range of 0 to 58 mg./m.J  (0-50
p.p.m.), which Is the range most commonly
used for urban atmospheric sampling. Most
Instruments measure In additional ranges.
  2.2  Sensitivity Is 1 percent of full-scale
response per 0.8 mg. CO/m.1 (0.5 p.p.m.).
  3. Interferences.
  3.1  Interferences vary between individual
Instruments. The effect of carbon dioxide
Interference  at  normal  concentrations  Is
minimal. The primary Interference is water
vapor, and with  no correction may give an
Interference equivalent to as high as 12 mg.
CO/m.*  Water vapor interference  can be
minimized by  (a)  passing  the air sample
through silica  gel or similar drying agents,
(b) maintaining constant humidity in  the
sample  and calibration  gases  by refrigera-
tion,  (c) saturating the air sample and cali-
bration  gases  to  maintain constant humid-
ity or (d)  using narrowband optical filters
In combination with some of these measures.
  3.2  Hydrocarbons  at  ambient levels do
not ordinarily interfere.
  4. Precision,  Accuracy,  and  Stability.
  4.1  Precision determined with calibration
gases  is  ±0.5 percent full scale In the 0-58
mg./m.1 range.
  4.2  Accuracy  depends on  Instrument
linearity and  the absolute  concentrations
of the calibration gases.  An accuracy of ± 1
percent of full scale In the  0-58  mg./m.1
range can be obtained.
  4.3  Variations In ambient room tempera-
ture  can cause  changes equivalent to  as
much as O.B  mg. CO/m."  per *C. This effect
can be minimized by operating the analyzer
In a  temperature-controlled  room. Pressure
changes between  span  checks  will  cause
changes In Instrument response.  Zero drift
Is usually less than ±1 percent of full scale
per 24 hours. If  cell  temperature and pres-
sure are maintained constant.
  5. Apparatus.
  5.1  Carbon Monoxide Analyzer. Commer-
cially available Instruments should be In-
stalled on location and demonstrated, pref-
erably  by the  manufacturer,  to meet  or
exceed  manufacturers  specifications  and
those described In this method.
  5.2  Sample introduction  System. Pump,
flow control valve, and fiowmeter.
  5.3  Filter (In-line). A filter with a poros-
ity of 2 to  10 microns  should be used to
keep  large particles from the sample cell.
  5.4  Moisture Control. Refrigeration units
are available  with  some commercial Instru-
ments for maintaining  constant humidity.
Drying tubes  (with sufficient capacity to op-
erate for 72  hours)  containing  Indicating
silica gel can be used. Other techniques that
prevent  the  Interference of  moisture  are
satisfactory.
  6. Reagents.
  6.1  Zero Gas. Nitrogen or helium contain-
ing less than 0.1 mg. CO/m.*
  6.2  Calibration  Gases. Calibration gases
corresponding  to 10, 20, 40,  and 80 percent
of full  scale  are used. Gases must be pro-
vided with certification or guaranteed anal-
ysis of carbon monoxide content.
  6.3   Span Gas. The calibration gas corre-
sponding to 80 percent of full scale Is used
to span the instrument.
   7. Procedure.
   7.1   Calibrate the Instrument as described
In 8.1.  All gases  (sample, zero,  calibration,
and span) must be Introduced Into the en-
tire  analyzer  system. Figure Cl shows  a
typical  flow  diagram. For specific  operating
Instructions,  refer  to the  manufacturer's
manual.
                                    FEDERAL REGISTER,  VOL .36,  NO. 84—FRIDAY, APRIL 30, 1971
                                                              109

-------
                                                 RULES AND  REGULATIONS
  8.  Calibration.
  8,1  Calibration  Curve.  Determine  the
linearity of  the  detector response at the
operating flow rate and  temperature. Pre-
pare a calibration curve and check the curve
furnished with  the Instrument.  Introduce
zero gas and set the zero control to Indicate
a recorder reading  of  zero. Introduce span
gas and adjust the span control to indicate
the proper value on the recorder scale  (e.g.
on 0-68 mg./m.1 scale, set the 46 mg./m.'
standard  at  80  percent  of  the  recorder
chart). Recheck zero and  span until adjust-
ments are no longer  necessary.  Introduce
Intermediate calibration gases and plot the
values obtained.  If  a  smooth curve  Is not
obtained,   calibration  gases  may   need
replacement.
  9.  Calculations.
  9.1  Determine the concentrations directly
from the  calibration curve. No calculations
are necessary.
  9.2  Carbon monoxide  concentrations in
mg./m.1 are converted to p.p.m, as follows:

       p.p.m. CO = mg. CO/m." x 0.873

  10. Bibliography.
  The Intech NDIR-CO Analyzer  by  Prank
McEIroy.  Presented at the llth  Methods
Conference  In Air  Pollution,  University of
California. Berkeley, Calif., April 1, 1970.
  Jacobs,  M. B. et  al., J.A.P.C.A.  9,  No.  2t
110*114. August  1959.
  MSA LIRA Infrared  Gas and Liquid Ana-
lyzer Instruction Book, Mine  Safety  Appli-
ances Co., Pittsburgh. Pa.
  B*eckman Instruction 1635B. Models 215A,
315A and 415A Infrared Analyzers. Beckman
Instrument  Company,  Fullerton. Calif.
  Continuous CO Monitoring System, Model
A 6011, intertech Corp., Princeton, N.J.
  Bendli—UNOR  Infrared  Gas  Analyzers.
Ronceverte. W. Va.
                 ADDENDA
  A. Suggested  Performance  Specifications
for NDIR Carbon Monoxide  Analyzers:

Range (minimum)	   0-58mg./m»
                            (0-50 p.p.m.).
Output (minimum)...,-   0-10, 100, 1,000.
                           6,000 mv.  full
                           scale.
Minimum detectable sen-   0.6 mg./m.«  (0.5
  Bltlvtty.
                           p.p-m.).
Lag  time  (maximum)—   15 seconds.
Time to 90  percent re-   30 seconds.
  sponse (maximum).
Rise  time,   90   percent   15 seconds.
  (maximum).
Fall   tune,   90   percent   15 seconds.
  (maximum).
Zero drift (maximum)...   3 percent/week.
                           not  to  exceed
                           1   percent/ 24
                           hours.
Span drift (maximum) —   3 percent/week,
                           not   to  exceed
                           1   percent/24
                           hours.
Precision  (minimum)---  ±O.Spercent.
Operational  period  (min-   3 days.
  Imum).
Noise (maximum)	   ±0.5 percent.
Interference   equivalent  1 percent  of full
  (maximum).             scale.
Operating  temperature   5-40° C.
  range (minimum).
Operating humidity range  10-100 percent.
  (minimum).
Linearity  (maximum de-   1 percent  of full
  vlatlon).                 scale.

  B. Suggested Definitions  of  Performance
Specifications:
Range—The minimum and nwrt"nim meas-
  urement limit*.
Output—Electrical signal which  Is propor-
  tional  to the measurement; intended  for
  connection to readout or data processing
  devices. Usually expressed as millivolts or
  mllllamps full scale at a given impedance.
Pull Scale—The maximum measuring limit
  for a given range.
Minimum Detectable Sensitivity—The small-
  est amount of input  concentration that
  can be detected as the concentration  ap-
  proaches zero.
Accuracy—The degree of agreement between
  a measured value and the true value; usu-
  ally  expressed as ± percent of full scale
Lag  Time—The  time interval  from  a step
  change In Input concentration at the  In-
  strument Inlet to  the first corresponding
  change In the instrument output.
Time to 90 percent Response—The time  In-
  terval  from a  step change  In the input
  concentration  at the  instrument Inlet to
  a reading of 90 percent of the  ultimate
  recorded  concentration.
Rise Time (90 percent)—The Interval  be-
  tween Initial response time and time to 90
  percent response after a step Increase in
  the inlet concentration.
Fall Time  (90 percent)—The  Interval  be-
  tween  Initial response time and time to
  90 percent  response after a  step decrease
  In the Inlet concentration.
Zero Drift—The change In Instrument out-
  put over a stated  time period,  usually 24
  hours,  of unadjusted  continuous  opera-
  tion,  when the  Input  concentration  is
  zero;  usually  expressed as percent full
  scale.

               SAMPLE INTRODUCTION
    SAMPLE IK
Span Drift—The change In instrument out-
  put over a stated time period, usually 24
  hours, of unadjusted continuous  opera-
  tion,  when  the  Input concentration Is a
  stated upscale value; usually expressed as
  percent full scale.
Precision—The degree  of agreement between
  repeated" measurements of the same con-
  centration, expressed as the average devia-
  tion of  the  single results from the mean.
Operational Period—The period of time over
  which the instrument con be expected to
  operate unattended  within speclflcations.
Noise—Spontaneous deviations from a mean
  output not caused by input concentration
  changes.
Interference—An undeslred positive or nega-
  tive output caused  by a substance other
  than the one  being measured.
Interference  Equivalent—The  portion  of
  indicated  input  concentration due  to the
  presence of an Interferent.
Operating  Temperature Range—The range
  of ambient  temperatures over which the
  instrument   will  meet  all performance
  specifications.
Operating Humidity Range—The  range of
  ambient relative humidity  over which the
  Instrument   will  meet  all  performance
  specifications.
Linearity—The maximum deviation between
  an actual Instrument reading  and  the
  reading predicted by a straight line drawn
  between  upper  and  lower  calibration
  points.

               ANALYZER SYSTEM
    SPAN
    AND
 CALIBRATION
                                                            I. R. ANALYZER
                           F!gur« C1. Carbon monoxide analyzer flow diagram.
APPENDIX  D — R
                 KEEUCK METHOD roa  TUB
                  Pl.tOTOCi-WJMICAL OXIDANTS
  CORRECTED  ros  iNTEBrpotF.NCES  PTTK  TO
  KmtOGEtf Oxrosxs AND SULFUR DIOXIUE

  1. Principle, and Applicability.
  1.1   Ambient  air  and ethylene  **e  de-
livered  simultaneously  to  a mixing  Ron*
whc-re tho ozone  in  the air reacts with the
ethyitine to emit  light which Is detected by
a plaotomultlpllw txibe. The resulting: photo-
ourrcut i/> amplified and Is either read tii-
rectiy car- displayed on a recorder.
  1.3   The nuH;iiod 13 applicable to the con-
tinuous m.i£Hmjr«zn«n.fc of taxxuo In  wxiblcnt;
air.
  2. Ktt?ipe ttnd Sensitii'ity,
  2,1   The range is 9.8 tt%~  CX/'m.' to greater
than  19(10  Ng. Cyin.* (0.005 p.p.m. O( to
greater than 1 p.p.m.  Os) .
  2.2  The sensitivity 13 9.8 Ag. Oa/m.' (0.005
p,p,m. Of).
  3. Interferences.
  3.1  Other oxidizing and reducing spociw
normally found In ambient air do not Inter-
fere.
  4. Precision and Accuracy.
  4.1  Tho ftverage deviation from  t3ie mcac
of repeated siwgle mewfmremerits docs not ex-
ceed 5 percent of the mean of the measure-
ments.
  4.2  The method IB accurate  within ±7
percent.
  5. Apparatus.
  5.1  Detector Cell. Figure Dl is a. drawing
of a typical detector coll showing flow pcithB
of gases, the mixing none, and placement of
the photomuUlpUcr  tube. Other flow paths
in, which the air and othylene streams .meet
                                    FEDERAL REGISTER, VOL 36, NO.  84—FRIDAY,  APRIL 30, 1971
   U. S. GOVERNMENT PRINTING OFFICE: 1973	746772/4195
                                                                 110

-------
 BIBLIOGRAPHIC DATA
 SHEET
1. Report No.
EPA-R4-73-028a
                                                                       3. Recipient's Accession No.
4. Title and Subtitle
 GUIDELINES  FOR DEVELOPMENT OF A QUALITY ASSURANCE PROGRAM
 Reference Method for Continuous Measurement of CO in the
 Atmosphere
                                                 5. Report Date
                                                   June 1973
                                                 6.
7. Author(s)
  Franklin  Smith and A.  Carl Nelson
                                                 &• Performing Organization Rept.
                                                   No.
9. Performing Organization Name and Address

  Research  Triangle  Institute
  Research  Triangle  Park, N.C.   27709
                                                  10. Project/Task/Work Unit No.
                                                  11. Contract/Grant No.
                                                   EPA- Durham
                                                   68-02-0598
12. Sponsoring Organization Name and Address

  Environmental  Protection Agency
  National  Environmental  Research Center
  Research Triangle  Park, N.C.   27711
                                                  13. Type of Report & Period
                                                    Covered
                                                   Interim contract
                                                   report-field document
                                                  14.
IS. Supplementary Notes
16. Abstracts
           Guidelines  for the quality control  of ambient CO by the Federal  reference
  method  are  presented.   These  include:
           1.   Good operating practices
           2.   Directions on how to assess  data and qualify data
           3.   Directions on how to identify trouble  and improve  data quality
           4.   Directions to permit design  of auditing  activities
           5.   Procedures which  can be used to select action options and  relate them
               to costs
  The document is not a  research report.   It is designed for use by operating personnel
17. Key Words and Document Analysis.  17a. Descriptors

  Quality Assurance
  Quality Control
  Air Pollution
  Quantitative Analysis
  Gas Analysis
  Carbon Monoxide
17b. Identifiers/Open-Ended Terms
17e. COSATI Field/Group
                               > Q-JQ ^
18. Availability Statement
                                      19.. Security Class (This
                                        Report)
                                      	UNCLASSIF1EC
                                                                   LASSIF1ED
                                                                   •Class (Thi!
                                      20. Security Class (Tl

                                           UNCLASSIFIED
21. No. of Pages

 128
22. Price
FORM NTIS-35 (REV. 3-721
                                                                                 USCOMM-OC 14A92-P72

-------
    INSTRUCTIONS  FOR  COMPLETING  FORM  NTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
   Guidelines to Format Standards  for Scientific and Technical Reports Prepared by or for  the Federal Government,
   PB-180 600).

    1.  Report (Dumber.  Each individually bound report shall carry a unique alphanumeric designation selected by the performing
       organization or provided by the  sponsoring organization.  Use uppercase letters and Arabic numerals only.  Examples
       FASEB-NS-87 and FAA-RD-68-09.

    2.  Leave blank.

   3.  Recipient's Accession Number.  . Reserved for use by each report recipient.

   4.  Title and Subtitle.  Title  should indicate clearly and briefly the subject coverage of the report, and be displayed promi-
       nently.  Set subtitle,  if used, in smaller type or otherwise subordinate it to main title.  When a report is prepared in more
       than one volume, repeat the primary title, add volume number and include subtitle for the specific volume.

   5-  Report Dote, l-'ach  re-port shall carry a date indicating at  least month and year.  Indicate the basis on which it was selected
       (e.g., date of issue, date of approval, date of preparation.


   6*  Performing Organ!zation  Code.  Leave blank.

   7.  Authorfs).  Give name(s)  in conventional order (e.g., John R. Doc, or J.Robert Doe).  List author's affiliation if it differs
       from the  performing organization.

   8.  Performing Organization  Report  Number.  Insert  if performing organization wishes to assign this number.

   9.  Performing Organizotion Name and  Address,  dive name, street, city, state, and zip code.   List no more than two levels of
       an organizational hierarchy.  Display the name of the organization exactly as it should appear in Government indexes such
       as  USGRDR-I.

  10.  Project/Task/Work Unit Number.  Use the project, task and work unit numbers under  which the report was prepared.

  11.  Contract/Grant Number.  Insert contract  or grant number under which report was  prepared.

  12*  Sponsoring Agency  Nome and Address.  Include  zip code.

  13.  Type of Report and Period Covered.  Indicate  interim, final, etc., and, if applicable, dates covered.

  14.  Sponsoring Agency  Code.   Leave blank.

  15.  Supplementary Notes.  Fnter information not  included elsewhere but  useful, such  as: Prepared in cooperation with . . .
       Translation of ...   Presented at conference of ...  To be published in ...  Supersedes . . .       Supplements

  16.  Abstract.   Include a brief  (200 words or less) factual summary of the  most significant information contained in the report.
       If the report contains a significant bibliography  or literature survey, mention it here.

  17.  Key  Words and  Document Analysis,  (a).  Descriptors.  Select from the- Thesaurus of Engineering and Scientific Terms the
       proper authorized terms that identify the major concept of the  research and are sufficiently specific and precise  to be used
       as index entries for cataloging.
       (b).  Identifiers ond Open-Ended Terms.   Use identifiers for project names, code names, equipment designators, etc.  Use
       open-ended terms written in descriptor form for those subjects for which no descriptor exists.
       (c).  COSATI Field/Group.   Field and Group  assignments are to be taken from  the  1965 COSATI Subject Category  List.
       Since the majority of documents  are multidisciplinary in nature, the primary Field/Group assignment(s) will  be the specific
       discipline, area of human  endeavor, or type of physical object. The application(s) will be cross-referenced  with secondary
       Field/Group assignments  that will follow the primary posting(s).

  18.  Distribution Statement. Denote  releasability to the public or limitation for reasons other than security for  example  "Re-
       lease unlimited".  Cite any availability to the public, with address and price.

  19 & 20.  Security Classification.  Do not submit classified reports to the National Technical

  21.  Number of  Pages.   Insert  the total number of pages, including this one and  unnumbered pages, but  excluding distribution
       list, if any.

  22.  Price. Insert the price set by the National Technical Information Service or the Government Printing Off ice, if known.
FORM NTIS-35 CREV. 3-72)                                                                                  USCOMM-OC I4982-P72

-------