Environmental Monitoring Series
EVALUATION OF THE EPA REFERENCE  METHOD
    FOR THE MEASUREMENT OF NON-METHANE
               HYDROCARBONS  - FINAL  REPORT
                        Environmental Monitoring and Support Laboratory
                               Office of Research and Development
                              U.S. Environmental Protection Agency
                         Research Triangle Park, North Carolina 27711

-------
                RESEARCH REPORTING SERIES

Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology.  Elimination of traditional grouping was  consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:

      1.  Environmental  Health Effects Research
      2.  Environmental  Protection Technology
      3.  Ecological Research
      4.  Environmental  Monitoring
      5.  Socioeconomic Environmental Studies
      6.  Scientific and Technical Assessment Reports (STAR)
      7.  Interagency Energy-Environment Research and Development
      8.  "Special" Reports
      9.  Miscellaneous Reports

This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of  environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia  22161.

-------
                                            EPA-600/4-77-033
                                            June 1977
        EVALUATION OF THE EPA REFERENCE
           METHOD FOR MEASUREMENT OF
           NON-METHANE HYDROCARBONS
                      by
                J. W. Harrison
                M. L. Tinunons
                R. B. Denyszyn
                C. E. Decker
          Research Triangle Institute
      Research Triangle Park, N. C. 27709
            Contract No. 68-02-1800
                Project Officer
                E. Carol Ellis
           Quality Assurance Branch
Environmental Monitoring and Support Laboratory
      Research Triangle Park, N. C. 27711
    NATIONAL ENVIRONMENTAL RESEARCH CENTER
      OFFICE OF RESEARCH AND DEVELOPMENT
     U. S. ENVIRONMENTAL PROTECTION AGENCY
      RESEARCH TRIANGLE PARK, N. C. 27711

-------
                               DISCLAIMER
     This report has been reviewed by the Quality Assurance Branch,
Environmental Monitoring and Support Laboratory, U. S, Environmental
Protection Agency, and approved for publication.  Approval does not
signify that the contents necessarily reflect the views and policies
of the U. S, Environmental Protection Agency, nor does mention of trade
names or commercial products constitute endorsement or recommendation
for use.

-------
                                   CONTENTS
                                                                         Page
Foreword	,	,	
Abstract	
Figures	,	
Tables	
Acknowledgements	,. ,	
     1.  Introduction  and Summary	  1
     2.  Conclus ions	,	,	  9
     3.  Recommendations.. ,	 13
     4.  Flame lonization Detector	 15
              Design Factors	 15
              Evaluation.	 18
     5.  Gas Ghromatograph  for NMHC Measurement.	 29
              Design and Construction.	 29
              Evaluation and Peaking of Methane Responses	 31
              Effect of Sampling Conditions and Oxygen Content of
              Reagent  Gases	 33
              Comparison to Guideline Specifications	 35
              Effects  of Water Vapor	 38
              Instrumental Response to NMHC	 40
              Summary  of RTI FID-GC Evaluation	 45
     6.  Review of Commercial Instruments	 47
              Design	 47
              Fabrication	 52
              Selection of Operating Conditions and Calibration	 54
     7.  Comparative Evaluation	 59
              Test Arrangements	,	 59
              Test Procedure	 64
              Operational Summary	 65
              Data Analysis	 66
              Results  and Discussion	 68
              Conclus ions	 98

-------
                                  CONTENTS
                                                                         Page
     8.  Sources of Error with Present Method	  101
              Instruments	°	  101
              Reagents	  106
              Procedure	  Ill
              Operators	,	  113
References	,	  115
Appendices
    A.  FID and GC Statistical Test Plans and Results	  119
    B.  Comparative Evaluation Data	  129
    C.  Evaluation of Strip Chart Recorders Used in Comparative
        Instrumental Evaluation	  159
    D.  Proposed Procedure for the Measurement of Non-Methane
        Hydrocarbons in the Atmosphere	,	  163

-------
                                  FOREWORD



     Ambient monitoring of the criteria pollutants as well as other harmful

trace contaminants is necessary for an assessment of the degree of improve-

ment or deterioration of the environment.  The Quality Assurance Branch of

the Environmental Monitoring and Support Laboratory contributes to the

ambient monitoring efforts by:

        evaluating analytical and monitoring techniques to determine

        their precision and accuracy, and

     -  testing the performance of analytical laboratories involved

        in air pollution measurements.

     These activities are aimed directly at the estimation of the quality of

ambient monitoring data collected by various local, state and federal pollu-

tion control agencies.

     Several problems with the measurement of non-methane hydrocarbons (NMHC)

using the EPA reference method have been reported.  The nature of these

problems are such that the validity of ambient measurements at the National

Ambient Air Quality Standard level of 0.24 pptn is questionable.  The research

program reported herein provides an analysis of the sources of error in the

measurement of NMHC by the reference method and makes recommendations for

minimization of these errors with the current instrumentation.
                                      John B. Clements, Ph.D.
                                      Chief
                                      Quality Assurance Branch
                                      Environmental Monitoring and
                                        Support Laboratory

-------
                                  ABSTRACT

     Many problems have been reported with the method and instruments
presently used to monitor ambient air hydrocarbons.  The study reported
here was carried out to determine, if possible, the sources of error
inherent with the present technique and to make recommendations as to
what modifications can be made to eliminate or minimize these errors.
A flame ionization detector and gas chromatographic instrument were
designed, built and evaluated to determine the effects of operating
parameters on hydrocarbon response and the differences in response to
various non-methane hydrocarbon (NMHC) species.  This instrument was
then used in a comparative evaluation with five commercial instruments.
The evaluation included determinations of calibration stability and
response to various NMHC species.   Following measurements with gases from
cylinders from commercial sources, the commercial instruments were further
compared in a three day test using ambient air.  Calibration stability
was found to be reasonable, with span shifts of greater than 5% the biggest
problem.  There were wide differences in responses to different NMHC
species.  These differences were somewhat reduced by using propane response
as the basis of calibration rather than methane.  When ambient air was
analyzed there were large discrepancies between analyzer readings which
appeared to be related to atmospheric water vapor content.  Recommendations
are made for changes in technique to minimize analyzer discrepancies.

-------
                                  FIGURES
Number                                                               Page
  1    Instrument requirements of present Federal Reference Method...   3
  2    Reagents and calibration procedure of present Federal
       Reference Method ............................... . ..............   4
  3    RTI field experience with NMHC reference method ...............   5
  4    Independent contractor survey problems with NMHC Reference
       Method [[[   7
  5    EPA in-house evaluation of NMHC Reference Method ..............   8
  6    Design of flame ionization detector .............. , ............  16
  7    Block diagram of FID electronics ..............................  17
  8    FID test chamber ..............................................  20
  9    Block diagram of FID test set-up ............... , ..............  21
 10    Effect of hydrogen flow rate on response ............... . ......  22
 11    Effect of FID pressure on response .............. , .............  22
 12    Effect of carrier air flow on response ........................  23
 13    Linearity of FID response to methane ..........................  23
 14    Effect of flame tip voltage on response .......................  24
 15    Effect of support air flow rate on response ...................  24
 16    Design of gas chromatograph ...................................  30
 17    Response characteristics of FID-GC ............................  32
 18    Effect of Fuel-Carrier ratio on response ......................  34
 ]9    Effect of sample inlet pressure on response ...................  34
 20    Dependence of response on carrier air oxygen content ..........  36
 21    Experimental test setup .......................................  39
 22    Peak area response as a function of sample dew point ..........  41
 23    Peak height response as a function of sample dew point ........  42
 24    Schematic diagram of flow system used for comparative
       evaluations [[[  62
 25    Linearity of propane response .................................  71

-------
                                  FIGURES

Number                                                               Pa§f
 27    Average CH, response showing ratios of individual instrument
       readings to average, variation of average and relative
       spread	•	• • • «	 84
 28    Average THC response showing ratios of individual instrument
       readings to average, variation of average and relative
       spread	•	 86
 29    Average NMHC response and relation of MSA 11-2 and Bendix
       8200 readings to average	 88

 30    Ratios of individual instrument CH, readings to Bendix 8200
       reading and relative spread to readings	  89
 31    Ratios of individual instrument THC readings to Bendix 8200
       reading and relative spread	 •  90
 32    Plot of meteorological data showing relationship to relative
       spread of CH, reading	   95

 33    Propagation of error in a calibration hierarchy	 107
 34    Relation of instrument scale, calibrated on the basis of
       apparent concentrations for zero gas and span gas, to true
       hydrocarbon concentration	 109
 35    Mean and standard deviation of total hydrocarbon indication
       of  zero air for each twenty-four hour period of long term
       stability test; Instrument:  RTI Prototype	 130

 36    Mean and standard deviation of total hydrocarbon indication
       of  zero air for each twenty-four hour period of long term
       stability test; Instrument:  Beckman 400	  131
 37    Mean and standard deviation of methane indication of zero
       air for each twenty-four hour period of long term stability
       test; Instrument:  Beckman 6800	,	,  132

 38    Mean and standard deviation of total hydrocarbon indication
       of  zero air for each twenty-four hour period of long term
       stability test; Instrument:  Beckman 6800	   133

 39    Mean and standard deviation of methane indication of zero
       air for each twenty-four hour period of long term stability
       test; Instrument:  Bendix 8200	   134
 40    Mean and standard deviation of total hydrocarbon indication
       of zero air for each tweny-four hour period of long term
       stability test; Instrument:  Bendix 8200	   135
 41    Mean and standard deviation of methane indication of zero
       air for each twenty-four period of long term stability
       test; Instrument:  Bendix 8201.	    36

-------
                                  FIGURES
Number
 42    Mean and standard deviation of total hydrocarbon indication
       of zero air for each twenty-four hour period of long term
       stability test;  Instrument:  Bendix 8201	 137

 43    Mean and standard deviation of methane indication of zero
       air for each twenty-four hour period of long term stability
       test; Instrument:  MSA 11-2	 138

 44    Mean and standard deviation of total hydrocarbon indication
       of zero air for each twenty-four hour period of long term
       stability test; Instrument:  MSA 11-2	 139

-------
TABLES
Number
1
2
3

4

5

6
7
8
9

10

11
12
13

14

15

16

17

18

19

20

Determination of Precision and Span Drift of RTI Instrument..
Zero Drift and Noise Data for RTI GC-FID 	
Comparison of RTI GC-FID to Some Specifications in EPA
Equivalency Guidelines 	
Data Indicating Water Vapor Interference of RTI Prototype
G.C 	
The NMHC Response of RTI Instrument in Various Modes of
Operation (CH^ = 1.0 as Basis) 	
List of Manufacturers and Representatives Surveyed 	
List of Questions in Survey of Manufacturers 	 , 	 	
Instruments Used in Comparative Evaluation 	
List of Reagent Gases and CH,/NMHC Mixtures for Comparative
Evaluation 	
List of Stock Hydrocarbon/Air Reagents with Reported and
Measured Concentrations 	
Zero Drift Performance of Instruments 	
Span Drift Performance of Instruments 	
Precision of Instruments on Methane and Total Hydrocarbons
Channels 	
Effect Carbon Numbers of NMHC with CHA = 1.0 as Basis,
Applying Initial Calibration over 14 Day Period 	
Results of Correcting NMHC Response to Propane, Applying
Initial Calibration over 14-Day Period 	
Effect of Methane on ECN of NMHC, Applying Initial Calibra-
tion over 14-Day Period 	
Ratio of ECN (no CH4)/ECN (with CH4>, Using Daily
Calibration over 14-Day Period 	
Effective Carbon Numbers of NMHC with CH = 1.0 as Basis,
Applying Daily Calibration 	
Results of Correcting NMHC to Propane, Using Daily
Calibration 	 • 	
Effect of Methane on ECN of NMHC, Using Daily Calibration...
Page
35
37

37

38

44
55
56
60

61

63
68
69

69

72

73

74

75

76

77
78

-------
                                   TABLES
Number                                                               Page
 21    Ratio of ECN (no CH,)/ECN (with CH4), Using Daily
       Calibration	,	* 79
 22    Data on Meteorological Conditions	 92
 23    Comparison of Specifications in EPA Equivalency Guidelines,
       with Those of Federal Reference Method.	, ,. .	112
 24    Test Variables  and Levels for FID Evaluation	119
 25    Test Plan for FID Evaluation	120
 26    Statistical Analysis  Results	123
 27    Test Variables  and Levels for FID-GC Evaluation	127
 28    Design  of GC  Evaluation  Test and Resulting Responses	128
 29    Comparative Evaluation Ambient  Air  Data	140
 30    Average Methane Concentration and Ratios  of Individual
       Instrument Readings to Methane	152
 31    Average Total Hydrocarbon Concentration and Ratios  of
       Individual Instrument Readings  to Average	155
 32    Results  of  Recorder Evaluation	162

-------
                              ACKNOWLEDGEMENTS

     The authors would like to acknowledge the very gracious help of many
people during the course of this study.  Personnel from the Environmental
Protection Agency who were willing to share experiences and provide critical
evaluation of our progress included V. Thompson, F. McElroy, R. Baumgartner,
W. Lonneman, F. Black, T. Clark, J. Margeson, J. Clements and G. Ortman.
Personnel from industry who were willing to share experiences and answer
our questions included R. Chapman and R. Villalobos of Beckman  Instruments;
0. Cano, J. Scales and E. Leesberg of Bendix Instruments; B. Behr of Bryon
Instruments; C. Burgett of Hewlett-Packard; P. Conner of Meloy  Laboratories;
and W. Dailey and A. Poli of Mine Safety Appliances.  Personnel from RTI
who gave their best effort included D. Brooks, D. Denton, D. Hardison,
K. Robbins, K. Straube and A. Turner.
     The authors would particularly like to acknowledge the interest and
patient guidance of C. Ellis throughout the extensive course of this research.

-------
                                  SECTION 1
                          INTRODUCTION AND SUMMARY

     The EPA Reference Method for non-methane hydrocarbon (NMHC) measure-
ment (1) is summarized in Figures 1 and 2.  Numerous operating problems
have been reported in the field application of this method.   Some of
these problems are summarized in Figures 3, 4 and 5 which gives the
experience of Research Triangle Institute (RTI) personnel (2), an inde-
pendent contractor survey (3), and an in-house EPA evaluation (4).
     In response to these difficulties EPA initiated the research program
reported here to determine experimentally the critical instrumental
variables which affect the response of the present Reference Method for
NMHC measurement and to make recommendations to improve the capability
of obtaining a uniform response to NMHC when performed by different
operators using different instruments at various geographical locations.
     The research was carried out in several phases.  First a flame ioni-
zation detector (FID) was designed.  The design was based upon a literature
review, prior RTI experience, and a comparison to existing commercial
designs.  This FID was then evaluated to determine the effect of various
operating parameters upon response to methane (CH,); Section 4 details
the design and evaluation.
     The next phase was to design, fabricate and evaluate a gas chromatograph
(GC), used with the well characterized FID, employing both the stripper
column technique of the Federal Reference Method (1) and a backflush technique.
Response to methane and various NMHC species were determined as well as the
effect of oxygen and water vapor partial pressure variations.  Section  5
details the design and evaluation of this phase.
     Following this evaluation the results were summarized in a report
distributed to six manufacturers of commercial instruments for NMHC analysis
along with a list of questions relative to operating parameters and calibration

-------
procedures to which answers were solicited.  Responses to these are dis-
cussed and compared in Section 6.  Also presented in this section is a
review of the design and fabrication features of representative commercial
instrument models.
     A comparative evaluation of five commercial analyzers as well as the
RTI GC-FID was carried out.  The objective was to determine calibration
stability and response to specific NMHC species as well as to determine
operational characteristics and problems.  This evaluation is described
in Section 7.  Based on this comparative evaluation, the manufacturers
survey, the evaluation of the RTI instrument as well as prior experience
of RTI and others, an analysis of sources of error is presented in
Section 8.
     Conclusions based on this research are presented in Section 2 and
recommendations are given in Section 3.  Part of these recommendations
involve a proposed Standard Method for Non-Methane Hydrocarbon Measurement,
which is given in an Appendix to this report.
     As a separate task on this project a draft Technical Assistance
Document for Non-Methane Hydrocarbon Measurement was prepared.  It will
serve as a basis for a document to be subsequently released through EPA.

-------
                   REFERENCE METHOD FOR NMHC MEASUREMENTS
                           Instrument Performance
Principle
     1.  Inject  sample air directly into  FID to measure THC.
     2.  Inject  sample air through a stripper column  into FID  to
         measure CH,.   (Measurement of CO via methanator.)   Backflush
         stripper column  to  atmosphere following  elution of  CH, and CO.
Available  Instruments
     Readout  Mode -  a)  chromatogram
                     b)  bargraph with automatic zero  and span
                        adjustment
     Data  Retrieval  - NMHC = THC  (peak height) -  CH4  (peak height)
     Range -  Atmospheric:  THC  - 20 ppm Full Scale  (F.S.)
                           CH,  - 10 ppm Full Scale
              Special:      THC  - 2 ppm Full Scale
                           CH,  - 2 ppm Full Scale
     Minimum Detectable
     Concentration - Atmospheric:  0.1 ppm equiv.  CH,
                     Special:      0.025  ppm equiv. CH,
Precision  0.5%  F.S. (0.1  ppm)  on 20 ppm  range
Accuracy   1% F.S.  (0.2 ppm) on  20 ppm range
           2% F.S.  (0.04 ppm) on 2 ppm range
Stability  Should meet specification with ambient temperature  variations
           of +3°C.
Figure 1.  Instrument requirements of present Federal Reference Method  (1)

-------
                    REFERENCE METHOD FOR NMHC MEASUREMENTS
Reagents
     FID Support Gas
     Fuel
     Carrier Gas
     Zero Gas
     Calibration Gases

     Span
- Air with <2 ppm equiv. CH^
- Hydrogen with <0.1 ppm equiv. CH^
- Air, N2, H2 or He with <0.1 ppm equiv. (
- Air with <0.1 ppm equiv.  CH^
- "Gases" with 10, 20, 40 and 80% F.S. of
  CH, with guaranteed certified analysis.
- 80% F.S. CH4 in "gas".
Calibration Procedure
     Zero - Introduce zero air and electronically zero recorder
            pen (i.e. set baseline)
     Span - Introduce span gas (nom. 80% F.S.) and set span
            control to proper scale reading
     Calibration Curve - Introduce nom. 10, 20, 40% F.S. and check
                         smooth curve.
Figure 2.  Reagents and calibration procedure of present Federal
           Reference Method (1).

-------
                 OPERATIONAL PROBLEMS WITH REFERENCE METHOD
                            RTI Field Experience

Variable FID/GC Response

     1.  Laboratory comparison of MSA, Beckman and Bendix analyzers
         indicated different response factors for reference samples
         of NMHC and CH, and for ambient air samples on instruments
         calibrated by procedure of the Reference Method.
     2.  NMHC measurements at different geographical locations were
         not consistent on instruments calibrated by Reference Method
         Procedure.
Calibration Gases
     Various suppliers of compressed gases are not able to provide
accurately known calibration mixtures containing methane and/or other
components in hydrocarbon free air.  The oxygen content of calibration
gases  is crtical i.e., blended air mixtures that deviate from the oxygen
concentration of ambient air (21% CL, 79% N2) adversely affect the
subsequent measurement of total hydrocarbon.

Effect of Moisture
     The total hydrocarbon measurement is depressed by moisture, often
producing negative non-methane hydrocarbon measurements.

Accuracy
     Inaccuracies are inherent with the subtraction of large numbers
(methane) from slightly larger numbers (total hydrocarbon); with the
unavailability of calibration mixtures; and with instruments for which
the minimum detectable level is 25 to 50 percent of the NAAQ standard.

Gas Chromatograph
     1.  Changes of FID reactant gas pressure, resulting in flame
         instability.
     2.  Use of pressurized calibration gas versus sample of
         atmospheric pressure.

     3.  Deterioration of analytical column.
                    I
     4.  Inability to maintain stable hydrogen, combustion air, and
         carrier gas flow rates.
     5.  Mechanical problems involved with operating valves.

     6.  Sample loss in inlet lines and plumbing.
        Figure 3.  RTI field experience with NMHC Reference Method  (2)

-------
                 OPERATIONAL PROBLEMS WITH REFERENCE METHOD
                      RTI Field Experience (Continued)
Reliability
     1.  Frequent electronic and/or mechanical failures.
     2.  Questionable operational performance due to complexity
         of system—60 percent valid /data recovered during recent
         field study.
Operational Requirements
     1.  Requires daily attention by highly competent operator.
     2.  Frequent calibration required due to span drift in excess
         of specifications.
     3.  Maintenance and/or  repairs generally require skilled
         personnel knowledgeable in electronics and gas chromatography.
     4.  Expensive and costly to operate.

   Figure 3 (Continued).   RTI field experience with NMHC Reference Method (2),

-------
                 OPERATIONAL PROBLEMS WITH REFERENCE METHOD
                          Scott Environmental Report
Coverage
Survey of 16 users of Reference Method.
     10 Beckman 6800
      2 Bendix  8201
      1 Bendix  8200
      1 MSA     2472
      1 AID     514
      1 Bryon   230
Results
For an NMHC mixture near 0.24 ppm equiv, CH,
                  Error Range               Number of Users
                    0-10%                         1
                   10-20%                         3
                   20-50%                         2
                   50-100%                        4
                   >100%                          6
Reasons
1.  Failure of operators to understand or follow the instrument
    manufacturers' operating instructions and the reference method
    procedures for NMHC as specified in the code of Federal Regulations.
2.  Span gases containing unknown amount of higher hydrocarbons.
3.  Span gases not in air.
4.  Span gases incorrectly analyzed for methane.
5.  Zero errors due to sampling system contamination and lack of
    adequate checkout procedures.
6,  Excessive instrument zero and span drift during unattended operation.
         Figure 4.  Independent contractor survey problems with
                    NMHC Reference Method  (3).

-------
                   OPERATIONAL PROBLEMS WITH REFERENCE METHOD
                         EPA Comparative Evaluation
                           (McEIroy and Thompson)
Coverage
   1     Bendix 8201
   1     MSA U~2               0-10 PPm scale
   1     Beckman 6800
   2     Beckman 400
Procedure
     CH, zero and span (80%)
     Artificial atmosphere C^/CH^ (1.97/0.9 ppm) and C2H4/C2H2 (1.9/1.9 ppm)
     Side-by-side on ambient  air at Durham Air Monitoring and Demonstration
          Facility
Results
     C2H6/CH, mix (4.84 equiv.  CH ) :   4.97 (Bendix 8201) to 6.3 (Beckman 6800)
     C2H4/C2H2 mix (7.6 equiv.  CH^) :   4.8 (MSA) to 7.55 (Beckman 400)
     Ambient Air:  1.   Pair std.  deviation ranges from 0.217 to 0.454 ppm
                       and pairs give statistically different readings.
                       Individual std.  deviations estimated as 0.23 ppm,
                       comparable to ambient air standard of 0.24 ppm.
Conclusions
         Differences in FID response to various NMHC species are not the
         overwhelming source of response differences.
         Drift,  instability, precision, and repeatability errors are
         apparently the important causes of discrepancies.
         The 5-10% (of 10 ppm F.S.  range) measurement discrepancies
         commonly observed are large with respect to the normal NMHC
         range of 0-2 ppm and very large with respect to the 0.24 ppm
         standards.
         Figure  5.   EPA in-house evaluation of NMHC Reference Method (4).

-------
                                 SECTION 2
                                 CONCLUSIONS

     Evaluation of the RTI designed flame ionization detector  (FID) and gas
chromatograph  (GC) for NMHC measurement established the following:
     1.  Response to methane was insensitive to support air flow rate over
                                 3
         a range of 100 to 300 cm /min.
     2.  Normal variations in atmospheric pressure of +5% about one
         atmosphere do not affect response more than +2%.
     3,  Response to methane is maximized over a relatively narrow range
         of hydrogen and carrier air flow rates.  If flow rates are set
         at maximum response values, fluctuations in flow of +5% result
         in less than 2% variation in response.  If these flow rates are
         initially set appreciably below the maximum response values,
         fluctuations in flow of +5% can result in response variations of
         up to 8-10%.
     4.  There were significant variations in response to NMHC, on a per
         carbon atom basis, compared to the response to methane.  These
         depended on the method of separation used.   Although typically
         NMHC response was 20 to 30% lower than methane response, it was
         in some cases as low as 50%,
     5.  Response to methane varied with oxygen content of synthetic air.
         Below about 19% 0- the response decreased rapidly with decreasing
         0_.   Above about 20% 0» the response increased gradually with
         increasing 0~.
     6.  Response to methane increased with increasing sample dew point.
     Following prototype evaluation, a comparative evaluation of this prototype
and five hydrocarbon analyzers was conducted.  Calibration stability and
response to individual NMHC species, with and without methane background, were
determined using cylinder gases.  Immediately after this characterization the
instruments were placed into service monitoring ambient air for three days.

-------
 The conclusions from these  comparative  tests  are as  follows:
      1.   Zero drift of  these instruments  is apparently not  a  significant
          problem.
      2.   Span drift in  excess  of  5% over  a 24-hour period occurred  with
          every instrument,  ranging from about  15% to  70% of the  time
          during a  13 day period.
      3.   Precision ranged from 0.03 ppm to 0.11  ppm  on the methane  channel
          and  from  0.03  ppm  to  0.16 ppm  on THC  for the instruments evaluated.
      4.   There were significant differences in response to NMHC  both  on
          an intra-instrument and  inter-instrument basis.  Typically for
          a given instrument  the maximum deviation of  NMHC response, compared
          to the average response  of that  instrument  to 5 NMHC  species,  was
          about 20  to 30%.   Response deviations to a  given species were  about
          the  same  on an instrument to instrument basis.
      5.   The  NMHC  response  differences  between instruments are improved
          somewhat  by using  propane response as a basis for correction,  but
          are  nevertheless still significant.
      6.   The  field analyzers are  of complicated  design and construction  and
          are  subject to a variety of maintenance  problems.  As expected,
          one  of the components most susceptible  to malfunction is the automatic
          switching valve,
      7-   Materials of construction of gas lines  as well as filters  and
          columns can have a  significant effect on some of the  heavier or
          more reactive  hydrocarbon species.
      8.   Instrument  response to methane and NMHC  is very sensitive  to
          operating condition, primarily through  fuel  and carrier air  flows.
          Changes can alter the response pattern.
      9.   Instruments made to agree in the laboratory  through calibration
         with  dry  CH,/air mixtures immediately start  to display differences
         in response when exposed to the atmosphere.
    10.  Variations  in instrument response to  both CH, and THC are apparently
         related to atmospheric moisture content.  The mechanisms which  cause
         this relationship are not, at present, understood.
     Very obvious and significant disagreements between analyzers monitoring
ambient air cast doubt that  the present design generation can provide reliable
                                      10

-------
NMHC data, particularly in the lower ranges near the EPA air quality standard
of 0,24 ppm  (equivalent CH,).
     Several problems occur with the Federal Reference Method as written.
The restriction to one method, which in its application to several different
designs of commercial instruments has not demonstrated satisfactory consistency
and credibility of data, does not allow for the possibility that another,
even several other, methods might develop which provide the basis for better
measurements.  Second, although some instrument performance spcifications are
given, they are not definitive enough.  For example at present range, lower
detectable limit, precision and accuracy are specified.  However, they are not
specific as to how long these are to be maintained and under what conditions -
whether prior to delivery and in the manufacturer's calibration set-up or
on-site, continually.
     A third problem is in the specification of reagent gases.  There is no
definition of "air".  There is no requirement that the calibrating methane
concentrations be in air.  Although there are specifications on the total
hydrocarbon content of the fuel, carrier and support gases, there are no
indications as to how these gases can be checked to assure that specifications
are met and maintained.  Other than the THC content and the methane content,
there are no specifications on the presence, absence or concentration levels
of other species apt to be found in the atmosphere.
     A fourth problem occurs with calibration.   Presumably only dry gases are
used whereas in operation on ambient air appreciable, and highly variable,
water vapor concentrations occur.  The calibration gas is specified as methane,
although response to different NMHC species has been shown to be significantly
different both on an intra- and inter-instrument basis.  Furthermore the
frequency of calibration checks is not specified.
     A fifth problem occurs with the data reduction.  Although the method
depends upon a strip chart recorder there are not specifications on recorder
performance or calibration.  There are no specifications on record formats
such as recording of zero and span levels on the strip chart, or with the
strip chart, to facilitate re-checks of the data reduction.  Furthermore,
the method is implicitly limited to graphical analog recording and does not
recognize the advent of economical, reliable and much more convenient digital
data processing options.
                                      11

-------
     A sixth problem is in the failure to provide for inter-station comparisons,
or "round-robin" measurements in order to ascertain variability of data quality.
     It should be pointed out in closing that the observations and conclusions
about difficulties with the present Reference Method are made with the benefit
of several years operating experience of many people (hindsight).  The problems
are not at all unusual in the translation of a technique developed under
controlled laboratory conditions with highly individualized instruments into
a field environment with mass produced instruments.   That there are problems
does not reflect invidiously on either the authors of the method or instrument
designers and vendors.   Only extensive experience could provide the groundwork
for improving both the measurement technique,  instrumental design and data
reduction and reporting.
                                     12

-------
                                  SECTION 3
                               RECOMMENDATIONS

     Based on results of this research the following recommendations are
made:
     1.  The cause of the effect of atmospheric water vapor (or other
         interferents) on NMHC analyzer response should be determined
         and counteracted, if possible.
     2.  Equivalency testing of different design approaches should include
         actual ambient air evaluations and not rely just on cylinder
         gases.
     3.  Digital data manipulation, presentation and logging should be
         used to improve accuracy and decrease data reduction time.
     4.  The present Federal Reference Method for NMHC measurements
         (40 CFR Part 50, Appendix F and appropriate portions of Part 53)
         should be revised to allow for the use of properly certified
         reference and equivalent methods.  A proposed draft for such a
         document is included as Appendix D of this report.
     5.  The training of qualified NMHC analyzer operators should be
         augmented by a Technical Assistance Document on NMHC measurement.
         A draft TAD has been submitted under separate cover for issuance
         by EPA when suitable at some later date.
     6.  Calibration for NMHC response should use a correction factor
         based on instrument response to propane and include water vapor.
     There still remains the question of whether existing commercial analyzer
designs can be operated under conditions other than those now used which will
improve agreement.  Work, cited in Section 8, on automotive hydrocarbon
emissions analysis indicate that with the proper selection of fuel mixture
and flow rates response to various hydrocarbons can be made more uniform
than with pure H» fuel and, further, that correlation between individual
instruments of the same type can be greatly improved.  It is recommended that
this be investigated for ambient air monitoring instruments.  If this can
be established, the next step would be to see whether or not this optimi-
zation can be extended to instruments of different design.
                                      13

-------
14

-------
                                  SECTION 4
                         FLAME IONIZATION DETECTOR

DESIGN FACTORS
     A thorough literature search was made before the design of the FID
was decided.  The parameters most influential in affecting response were
determined.  The design chosen allows variation of these parameters in
order to assess their relative effect on hydrocarbon response.
     Response of an FID to hydrocarbons is strongly affected by the
magnitude and spatial distribution of the electrical potential difference
between the electrodes and by the spatial configuration of the flame (5).
The spatial distribution of electrical potential is determined by the
geometry of the electrodes and by the presence of any insulating layers
on the electrodes.  The flame configuration depends upon the jet size,
fuel feed and support gas flow.  The basic design of the prototype FID,
shown in Figure 6, is a modification of a model which has been used for
work under controlled pressures  (6).  As shown in the circuit diagram in
Figure 7 the flame tip is made positive, referenced to system ground.
A positive potential on the jet reduces the noise caused by electronic
emission from overheated jet tips (7).  In addition it has been reported
that the plateau of the current-voltage curve is more easily attained with
positive jets (8), although when actually in the saturation region,
response is the same for either polarity voltage on the jet (9).  Both
of these phenomena, electronic emission from hot jets and attainment of
saturation, were verified recently by McWilliam (10).
     The jet tip itself is subject to corrosion and should be constructed
from an inert material.   Corrosion products may be sufficiently insulating
to change the potential distribution.  The jet should have a high thermal
conductivity to effectively conduct heat away from the flame area, and
utilization of a heat sink is recommended (7).  Stainless steel meets both
                                      15

-------
                METERING VALVE
FLAME OUT SENSOR
          TO PUMP
        OR AMBIENT
                                                                    Y///A TEFLON
 SAMPLE &
CARRIER AIR
                                                                   ELECTROMETER CONNECTION
                                                                     SHELL  MADE OF ALUMINUM
                                                                     COLLECTOR AND TIP ARE
                                                                        STAINLESS STEEL
                                                                              IGNITION VOLTAGE
                                                                              POLARIZING VOLTAGE
                        Figure 6 .   Design of flame ionization detector.

-------
FID
               FLAME OUT
               DETECTOR
               & ALARM
                             LAMBDA
                             MODEL
                              30
                             POWER
                             SUPPLY
                                                KEITHLEY
                                               MODEL 610B
                                              ELECTROMETER!
       Figure 7.  Block diagram of FID electronics.

-------
 of  these  requirements.  The diameter  of  the jet opening  for  the  RTI  design
 was kept  to  less  than 0.51 mm  (0.020  inch) because  response  has  been
 reported  to  fall  off sharply with  larger sized openings  (11).
      The  collector  electrode was chosen  to be cylindrical  for  several
 reasons.   Obviously, it offers an  electrical potential distribution  with
 the same  rotational symmetry as the flame.  The flame tip  is at  a  positive
 potential, and  the  FID shell is at approximately the same  potential  as
 the collector.  In  order  to avoid  leakage currents  from  the  flame  tip to
 the shell walls,  the collector is  lowered below the level  of the flame,
 producing an effective electrical  shield for the flame and assuring  a high
 collection efficiency.  This geometry is also useful in  preventing
 recirculation of  combustion products  (12).  Response of  the  FID  is a function
 of  both the  length  and inside diameter.   The length of the collector affects
 the range of linear response if the length is too small  (5); consistent with
 results from Gill and Hartman (11), a collector of 25 mm length  with inside
 diameter  of  11 mm was selected.  In a recent article (13), a trumpet-shaped
 collector was claimed to  give linear response over a much wider  range than
 cylindrical  electrodes and saturation at lower voltages.    This type  geometry.
 may offer an  alternative  to the straight collector design.
      The  method of  introducing support air was chosen so that air  is admitted
 in  a  ring about the base  of the jet, giving uniform flushing of  the  jet
 from  all  sides.   This provides a non-turbulent flow, which reduces noise  (5).
      Both electrodes were well insulated with Teflon  to minimize  electrical
 leakage and  the modular design allows rapid disassembly  of the FID and easy
 replacement  of components for experimental variations.  At the relatively
 low temperatures used (70°-105°C) there was no problem with  degradation  of
 this  polymer.
 EVALUATION
      With the design of the FID determined, evaluation of  operating  char-
 acteristics was begun.   The objective was to determine optimum parameter
 values.   A factorial test design approach (14,15) was used to select the
 optimum operating point.   Such a test design is useful when the effect of
 a large number of operating parameters, many of them interacting, must be
assessed without devoting an excessive amount of time to  doing so.   Many
standard statistical test designs are available (16).   A suitable evaluation
                                      18

-------
sequence was designed  (17) using the following variables as experimental
parameters:
     1,  polarizing voltage,
     2.  sample concentration,
     3.  hydrogen flow rate,
     4.  carrier air flow rate,
     5.  support air flow rate, and
     6.  pressure.
Details of the test design are given in Appendix A.  All of these factors
have been shown to influence the response of the FID.  The gas flow and
polarizing voltage effects can be found in any comprehensive study of FID
performance  (5,10,17, for example).  The role of pressure in optimum FID
operation has been less well studied although it has been shown that normal
atmospheric  fluctuation can result in response variations as large as 6% (6).
Provisions were made, therefore, to control the pressure by placing a needle
valve in the exit gas port of the FID.  Subambient pressures were achieved
by connecting the exit gas port through this valve to a vacuum pump.  The
sample size  dependence is a measure of the linearity of a particular FID
and does vary with design from manufacturer to manufacturer (14).
     The physical arrangement used for the FID evaluation is shown in
Figure 8.  The chamber was controlled at a constant  (+0.5°C) temperature
of 70°C.  Flow and electrical circuit arrangements were as sketched in
Figure 9.  Support and carrier air were obtained from breathing grade
compressed air from which hydrocarbons were removed by passing over a
catalyst bed containing palladium coated on 3 mm (0.125 inch) alumina
spheres maintained at 280°C.  Previous experience with this type of clean-up
had yielded THC contents of less than 50 ppb.  Water vapor and NO  were
                                                                 X
removed by passing the exit stream over drierite and potassium permanganate
beds.  Hydrogen was obtained from ultra-pure grade compressed H~ with a
certified THC content less than 100 ppb.  A needle valve restrictor was used
in the sample inlet line to the FID to suppress pressure surges.
     Coded experimental values of the variables and the resulting response
values that resulted are given in Appendix A.  These were used in a
multiple regression analysis program.  Plots of the mean values of FID
signal current versus the variable values are shown in Figures 10 through 15.
                                      19

-------
                  NOTE:   DOOR NOT SHOWN
NJ
o



Al OUTER 	 ^
SHELL < >

MARINITE 	 	
T>JnTTT ATTOT1.! ^— .


STEEL INNER 	 -,
SHELL ^ 	
ZERO AIR ^
IN ^

HEATER w

VOLTAGE ^







1
si
*
a
3
tr
s
i
•.
;i
^
•y
?'
*;
|
'/;
pj
i
V-
:i^
«.i
v
j



/ y
r y
'/t
/


7
/
/
/
/
I/


t
/
/
/




A
















/




fjrT) y )












HEATER
BLOCK
k\\VI
K^l
'/////




/ '77 7 7
















/ / / / /




"7~/
















/ /




^ j
















/ 1




/
















t (




77








F
I
D






/ )




^7~7


0^













/ / /




77!















*V^T
/ / /




J J y ^
















////




' >
















V




r

*
r
/
i
/
f
/
/
)
.
/
/
f
/
/

/
/
* A

f
/



	 W TO
ELECTROMETER


., _ 	 THERMOCOUPLE
, (CHAMBER T^MP- )


K^
1 " > EXHAUST
+S
	 THERMISTOR
(FLAME OUT SENSOR)





. fERAMTC
( r'pA'M'n OT7T7


^ H t* PiAMPT.F. TN
	 POT ABT7Twr: uni TAHV
                                                                                                          IGNITION VOLTAGE
                                               Figure 8.  FID test chamber.

-------
AIR SUPPLY












TF^fP
CONTROL
#1




TEMP.
CONTROL
-'•O


\ \
«~^;
i
» '
•





l>
1
1 §






_^_

-------
   03
   01
   M
   01
       13
  CN
  H
  I
   O
   H

   X
   •M
   0
   0)
   n
   n

   O
   n)

   &
   •H
   C/3

   O
    12
      11
                 30

           Hydrogen  Flow
                                 40
50
Figure 10.


      16
w
a)
H
0)
fX
 CN
 iH
  I
  4-J
  c
  U
  Figure 11.   Effect of FID pressure  on  response.


                       22

-------
      20
 (1)   W
 I-i   01
 M   H
 3   (U
U   O.
 60rH  -, n
•H  I   10
C/2  O
       15
25
35
45
                                        3
                   Carrier  Air Flow (cm /min)
55
     Figure 12.  Effect  of  carrier air flow on response.

      30
      20
c  ^-^
a)  tn
M  CD
M  M
3  
-------
     20,-
 G  -^
 01   CO
 M   01
 »j   Lj
    0)
   1
       B-
                                        TT
                                                     D
•H  I
C/3  O
0
  0

  50
100         150          200
  Flame Tip Bias Voltage
                                                        250
Figure 14.

 20
                 Effect of flame tip voltage on response.
4-1

0)   CO
M   oi   r
M   M   I
3   0)

0   & 10
S3
•H  I
C/)  O
   H
P
0
200
                                                    ID-
                   250
                           300
                         350
400
450
                      Support Air  Flow Rate (cm /min)
        Figure 15.   Effect of support  air flow rate on response.
                            24

-------
     A mathematical model for the response surface which includes second
order and interaction terms, was obtained from a multiple regression analysis,
described in Appendix A.  Application of a standard Fisher test for statistical
significance indicated that of all of the linear, square and crossterms
(27 total), only ten or eleven could be considered influential in the range
of variables used.  Accordingly a reduced model was tested, using ten
variables.  The R-square value, a measure of how well the mathematical model
fits the data of this model, was 0.98535, compared to 0.98957 for the full
model, indicating that the fit was still good.  For this simple model the
response is given by
                         159.87 - 3.24X1 + 62.34X3 + 66.66X&
                                  6.61XnX0 - 4.43X,X, - 5.79XTX,
                                       13        1 J        ID
                                - 2.15X9XC. + 29.95X.JC, - 4.81X1X1
                                       z 5         36        i 1
                                                                         (1)
where the scaled variable values are given by:
                                           3
                    X  = [hydrogen flow  (cm /rain) - 35] /7                (2)
                    X  = [air flow cm3/min) - 333.5]/66.5                (3)
                     ^                   2
                    X  = [carrier flow cm /rain) - 35]/10                 (4)
                    X5 = [pressure (torr) - 760]/10                      (5)
                    X., = 3X[sample size  (ppm) - 1] .                      (6)

In this model the Fisher test indicated  that the X^X,. term was not signi-
ficant and could probably be excluded.   The Fisher test also indicated
that an even more simple model could be  used, employing only X_, X, and
their interaction.   This approximation gave an R-square value of 0.9720,
compared to the 0.98535 for the ten-term model, which is still a reasonable
fit.  This model is
                 R  = 153.73 + 63.23X0 + 68.08X, + 29.49X-X,  .           (7)
                  o                  3636
                                      25

-------
These response numbers can be converted to equivalent current values by
                                    -14
multiplying by a factor of 7.87 x 10    Ampere.
     Sensitivity of the response to changes in hydrogen flow, carrier air
flow or a change in concentration can be obtained by taking the partial
derivatives of R  in equation (1) with respect to the variables X^, X^
and Xc to obtain
     o
3R
AR-. - sY
4- dA-
3R1
A V J.
Axl + ax3
op
3R1
AV _L
AX., T „.,
— * O A f
op
                                                        AX6 ,            (8)
                                                     op
The partial derivatives are evaluated at the operating point values of the
parameters.  Assuming that the operating point is at a hydrogen flow of
     3                                                               3
32 cm /min selected for maximum response, a carrier air flow of 32 cm /min,
                                                               3
an operating pressure of 760 torr, a support air flow of 350 cm /min, an
electrode voltage of 100 volts and a sample size of 1 ppm CH,, changes
in response due to changes in the three operating variables can be calculated.
Three cases will be considered.
                                                               3
     Case 1  There is about a 10% change in hydrogen flow, 3 cm /min,
             with no change in carrier air flow or input concentration.
             Using the appropriate operating values in equations  (1)
             and (8) gives a negligible change in response, about 0.3%.
                                                                  3
     Case 2  There is about a 10% change in carrier air flow, 3 cm /min,
             with no change in hydrogen flow or input concentration.  This
             gives a 13.2% change in response, which is quite significant.
     Case 3  Both hydrogen and carrier air flow are stable at the selected
             operating point.  The sensitivity to changes in input hydro-
             carbon concentration is 1.42 x 10    amperes/CH,.
     The response surface of equation (1) has quadratic terms for both the
hydrogen flow and carrier air flow.   The plot of mean values of response
versus H2 flow rate, Figure 10, shows this very clearly, but it is not
apparent in the plot of mean values  of response versus carrier air flow
in Figure 12.   However, as will be shown in the next section, quadratic
dependence of  response on carrier air flow was also observed.
                                      26

-------
     To summarize the results of this investigation of FID response:
     1.  There is an optimum value of H,-, flow which will maximize response
         and minimize response variations due to flow fluctuations.
     2.  Response monotonically increases with carrier air flow.
     3.  Normal atmospheric pressure variations will cause about +2%
         response variation.
     4.  Response is linear for methane over 0.2-2 ppm range.
                                                                3
     5.  Variation of support air flow over the range 200-470 cm /min did
         not affect response.
     6.  Variation of bias voltage of flame tip over 50-250 volt range
         did not affect response.
Following this evaluation this FID was incorporated into a prototype gas
chromatograph design for an evaluation of operating characteristics.  This
is described next in Section 5.
                                      27

-------
28

-------
                                  SECTION 5
                   GAS CHROMATOGRAPH FOR NMHC MEASUREMENT

DESIGN AND CONSTRUCTION
     Having determined the operating characteristics of the FID, the design
for the gas chromatograph  (GC) was implemented as shown in Figure 16.
Before fabrication was begun, however, chromatographers from Bendix
Corporation and EPA were consulted to locate obvious design flaws, but
none were indicated to RTI.
     Valves #1 and #2 are zero volume, pneumatically operated valves (18).
They provide the necessary switching between sample loops, L, and L2, and
the stripper column and restrictor, S and R.  The volumes of L  and L? are
               3
matched at 3 cm  each, and R is used to match the pneumatic impedance of
S when a THC determination is being made.  The impedance matching is
necessary to prevent baseline shifts and gas flow alteration which cause
these shifts.  Valve #3 allows directing the backflush air stream from the
stripper column to the FID for the backflush mode of operation.  Valve #4
permits selection of sample or span gas lines.  Gas flows are controlled
by pneumatic flow controllers which give constant gas flow with constant
upstream pressure (19).
     The stripper column is mounted inside the instrument oven, which
is contrary to the Reference Method guideline (1).  The reason for this
is the nature of the stripper material, a carbon molecular sieve (20).
This material has the ability to separate CO, CO- and air from methane,
is not deactivated by moisture as are zeolites and is stable to over
120°C in oxidizing environments.   The only disadvantage of the carbon
molecular sieve is that higher molecular weight hydrocarbons (^Cf.) are
difficult to remove in the backflush mode of operation.  For most of the
compounds evaluated in our studies, however, this was not a serious problem.
                                      29

-------
          ZERO AIR
          SPAN GAS
        SAMPLE AIR
u>
o
   SYSTEM EXHAUST
         HYDROGEN
' TEMP.-CONTROLLED
I CHAMBER-INSIDE
1 DASHED LINE
                                                 BELLOWS
                                                  PUMP
                                          NOTE:   ARROWHEADS INDICATE
                                                 DIRECTION OF GAS FLOW.
 FLOW
CONTROL
    L^, L2 - SAMPLE LOOPS

    S      - STRIPPER COLUMN

    R      - RESTRICTOR

    N      - NUPRO VALVES

    P      - PRESSURE GAUGE
                                     Figure 16.  Design of gas chromatograph.

-------
     The zero air supply which was used as the carrier air and support air
and for sample dilution was obtained by passing breathing quality air from
tanks over a palladium catalyst bed, through a potassium permanganate bed
for removal of N0x species, and finally through a drying bed to remove
moisture.  A similar system, in routine operation at RTI, was found to
produce air with less than 50 ppb THC.   Hydrogen with <0.1 ppm THC was
supplied by cylinder.
EVALUATION AND PEAKING OF METHANE RESPONSES
     An evaluation procedure similar to the FID evaluation was used to
determine the optimum methane response of the GC-FID combination.   However,
only three parameters, hydrogen flow, carrier flow, and support air flow,
were varied.  Because FID evaluation had shown no effect on response for
polarizing voltage and little effect for pressure variations, the bias
voltage was set at 100 V, and the FID was vented to the atmosphere.  This
test sequence is shown in Table 28 of Appendix A along with the response,
(millimeters of peak height) to a 2.0 ppm sample of CH, in HC-free air (21).
Each parameter being studied was varied at three levels (27 total experiments)
The mean of the FID signal current was calculated at each of the levels for
the three parameters, and these data are presented in Figure 17.   From the
data shown in Figure 17, it can be seen that FID response to methane has the
following properties:
     1.  response is relatively insensitive to changes in support air flow in
                                      3
         the range studied  (100-300 cm /rain) for this instrument design,
                                                                     3
     2.  response as a function of hydrogen flow peaks at about 37 cm /min,
         and
     3.  response as a function of carrier air flow peaks at about 47 cm /min.
Comparing these results to those presented in the FID evaluation, the
support air insensitivity was the same above 100 cm /min.  (See Figure 15).
Hydrogen effects on the response curve were identical, but the maximum
                        3                               3
shifted from about 32 cm /min for the FID alone to 37 cm /min for the FID-GC
combination (Figure 10).  The response as a function of carrier air flow
showed no maximum when only the FID was studied (Figure 12), but peaked at
     o
47 cm /min with the FID-GC combination.
     The nature of the hydrogen and carrier air effects on the response
suggested that there should be a peak in the methane response of the GC,
occurring at about a hydrogen:carrier air ratio of 0.79.  However actual

                                      31

-------
      15
    CO
    
-------
measurement of the peaking in response, displayed in Figure 18, showed the
value to be closer to 0,70.  While operating at these peak response condi-
tions (for methane), the FID-GC methane sensitivity was calculated to be
1.39 x 10    amperes/ppm CH  which compared to 1.42 x 10    amperes/ppm CH
for the FID alone.
     For all practical purposes it appears that the characteristics of the
GC-FID are determined by the detector.
EFFECT OF SAMPLING CONDITIONS AND OXYGEN CONTENT OF REAGENT GASES
     To characterize the inlet parameters two experiments were performed.
                                                             3
In the first, the pressure at the inlet was held to 6.89 x 10  Pa (1 psi)
                                                             3
above the ambient, and sample flow rate was varied from 10 cm /min to
      3
100 cm /min using a constant concentration of 2.0 ppm methane.  The GC was
operated with the optimum hydrogen:carrier ratio (0.70) for peak methane
                                                               3
response indicated in Figure 18 with support air flow at 150 cm /min.   The
response increased only slightly (<2%) with increasing methane flow rate.
In the second test, the methane sample flow rate was held constant at
     3                                                      +3
30 cm /min, and the inlet pressure was varied from 6.89 x 10   Pa to
         +4
7.58 x 10   Pa (1 psi to 11 psi) above ambient.  The response increased
                                                     +4
linearly with pressure (see Figure 19) until 6.2 x 10   Pa was reached.
At this pressure there was an abrupt reduction in slope but the increase
in response with pressure remained linear.  The breakpoint most probably
occurred when sample line pressure matched carrier line pressure.
     The effect of using a nitrogen carrier instead of air was briefly
examined.  With the same operating conditions which yielded maximum in
the methane response, 2.0 ppm methane sample gave a response lower than had
been observed with air.  However, by increasing the nitrogen flow an
identical peak height to that with air was again attained, but the peak
area was only 50 to 70% as large as with air.
     The question of the effect of oxygen concentration in synthesized zero
air on the GC-FID response to methane was then addressed.  Air streams
ranging from 12% 0? to 27% 0~ were blended from nitrogen and oxygen
(<0.1 ppm THC each) and were used as the carrier gas.  Operating parameters
were selected as previously to give near optimum methane response.  The
logarithm of the background current versus the oxygen percentage gave a
linear relationship, as expected.
                                      33

-------
(0
01
l-l
01
o
H
 d
 01
 n)
 a
•H
W3
«
H
PH
    20-
    18
    16
    12
    10
                                 -H  =
          42  cm /min
          40  cm-Vmin
+ - Ho =  32  cm /min
               3
o - HO =  20  cm /min
0.6         0.8          1.0

   H2:Carrier Flow Ratio (—
                                                       1.2
                                    1.4
                                             cm /min
 Figure 18.   Effect of Fuel:Carrier Ratio  on  response.
d  ^
01  W
M  01
M  M
3  QJ
--  o,

     16
ed
d 
-------
     The effect of oxygen content on methane response is shown in Figure 20
where the peak area for each injection  (obtained by digital integration) vs.
% 02 is plotted.  The large variation in response at the lower percentage
02 content suggests the need to specify a minimum 0  content of about 19%.
However, in the range of 19% to 23% 02, the response was found to vary only
2%.  No attempt was made to peak the response at each of the individual CL
concentrations by flow rate adjustment as was done when investigating the
effect of nitrogen as a carrier gas.
COMPARISON TO GUIDELINE SPECIFICATIONS
     The precision, span drift, noise and zero drift were determined as
the final step in quantifying the operating characteristics of the prototype.
To get information on precision and span drift a five day test was performed
in which 2.0 ppm methane injections were made at various times during the
day (approximately hourly) until a total of seven injections per day had
been made.  These injections were made with the GC operating with optimized
response for methane.  The data, shown in Table 1, include the mean and
standard deviation of the peak heights of the seven daily injections measured
to the nearest one-half mm.  The span drift which is reported as a percent
was calculated according to the definition given in EPA performance guidelines
(22); and the precision was calculated as two times the standard deviation,
the procedure used in the Scott Report  (3).  Both the precision and span
drift are within the guideline values of 0.3 ppm and +5%.  The zero drift

     TABLE 1.  DETERMINATION OF PRECISION AND SPAN DRIFT OF RTI INSTRUMENT
Day No.
1
2
3
4
5
Response (mm)
Mean Stan. Dev.
93
94
95
94
94
+2.0
.5 +1.0
+2.0
+0.5
.5 +1.0
Span Drift
(%)
	
+1.6
+0.5
-1.1
+0.5
Precision
(ppm)
0.09
0.04
0.08
0.02
0.02
and noise values were determined by allowing the instrument to operate
unattended at the methane-response optimum for two days.  Using the methane
                                       35

-------
      600
      500--
  CO
   o
   X


   ro
   OJ
   en
   E
   O
   Q_
LO  to
<7>  OJ
   CSL
      400--
       soa
                                                                 1    22     23
                                                                                4-
1
14
15
16
17
18    19    20    21     2
  %  Oxygen in  Carrier Air
                                                        24     25    26



Figure 20.  Dependence  of response on carrier air oxygen  content.
28

-------
current/concentration conversion factor of 1.39 x 10 *  araperes/ppm CH,
determined previously, the zero drift and noise values were calculated
and are shown in Table 2,  The zero drift is broken down into 12-hour and
24-hour readings in ppm, and the noise is the maximum noise what was noted
during the two, 24-hour periods.  These values are also within the performance
guideline (22) values of +0,2 ppm and 0.05 ppm for zero drift and noise,
respectively.
               TABLE 2.  ZERO DRIFT AND NOISE DATA FOR RTI GC-FID

12
12
12
12
24
24

hr
hr
hr
hr
hr
hr
Period 2
morning,
evening,
morning,
evening,
, day 1
, day 2
Zero Drift (ppm)
day
day
day
day


1
1
2
2


+0.
+0.
+0.
+0.
±0.
+0.
06
11
10
05
14
1
Noise (ppm)
	
	
	
	
0.03
0.01
       comparison of RTI prototype performance with the performance guide-
line specifications for hydrocarbon monitors is shown in Table 3.
TABLE 3.  COMPARISON OF RTI GC-FID TO SOME SPECIFICATIONS IN EPA PERFORMANCE
          GUIDELINES (22)
Performance Parameter
Range
Noise
Lower Detectable Limit
Interference Eq.
Each Interferent
Total Interferent
Zero Drift, 12- and 24-Hour
Span Drift, 24-Hour
Lag Time, Rise Time, Fall Time
Precision
Units
ppm
ppm
ppm

ppm
ppm
ppm
percent
minutes
ppm
Document Value
0-5
0.05
0.1

+0.1
+0.2
+0.2
+5
10
0.3
RTI Value
Spec, met
0.05
0.1

not determined

+0.11
+2.8% maximum
within specs.
0.09
                                      37

-------
EFFECTS OF WATER VAPOR
     The final step in the evaluation of the RTI prototype was the determination
of the effects of water vapor on the response of the instrument.  The equipment
configuration for the test is shown in Figure 21.  A cylinder of 1.99 ppm
methane in HC free-air was connected to a bubbler fitted with a by-pass.  The
bubbler, enclosed in a heating mantle, was filled with distilled, deionized
water.  The bubbler outlet was connected to the GC by stainless steel tubing
which was heated to 110°C by tape heaters; the GC oven was also held at 110°C.
The dew point of the gas stream was measured at the inlet of the GC.  The
pressure at the GC inlet was also measured and maintained at 6.08 x 10* Pa
(0.6 atm.) above ambient.  This pressure was chosen to match the pressure in
the sample and carrier lines, a condition which yielded minimum baseline
disturbance when valve #1 was switched.  The instrument was operated near the
methane optimum, and seven injections were made at each dew point setting.  The
injections were made in THC mode so that the stripper column was by-passed.
Peak area (by digital integration) and peak height were recorded for all
samples.  The initial injections with the bubbler by-passed gave the response
of the "dry" methane, and the final set of injections, made with the bubbler
by-passed again, assured that the instrumental response to the "dry" methane
had not changed during the test.  The data are presented in Table 4 where
mean and standard deviation of both peak area and peak height are shown.

    TABLE 4.  DATA INDICATING WATER VAPOR INTERFERENCE OF RTI PROTOTYPE G.C.
Sample
Dew Point
(°C)
9
16
18
21
26
41
10
9 i_
Correlation (R )
% Water
by Volume
0.70
1.12
1.27
1.53
2.07
4.80
0.75

Response
Areaa
(Pulse Count)
299285 + 5050
298826 + 5700
306232 + 3650
308740 + 6352
314789 + 2764
323803 + 4040
300065 + 2219
0.929
Peak Height
(mm)
79.8 + 1
77.9 + 1
81.7 + 1
82.4 + 2.5
81.9 + 2
85.0 + 0.5
82.7 + 1
0.430
a.  Mean value and standard deviation (N-l weighting) of seven injections.
b.  Correlation coefficients determined by two-variable linear regression analysis.
                                      38

-------
S
M
T
P
D
B
Sample
Heating Mantle
Heating Tape
Pressure Guage
Dew Point Sensor
Bubbler
©
                                                            GC
                  Figure 21.  Experimental test setup.

-------
The results are plotted in Figures 22 and 23 and can be summarized as follows:
     1.  As a function of sample dew point in the range of 9°C to 34°C
         (0.70% to 4.80% water vapor by volume), the peak area increased
         6.8% with identical methane injections.
     2.  In the same range peak heights increased by 5%.
     3.  The goodness of linear fit was measured by the square of the
                                   2                       2
         correlation coefficient, R ; peak area showed an R  = 0.929, and
                                2
         peak height showed an R  = 0.430.
INSTRUMENTAL RESPONSE TO NMHC
     The next phase of the evaluation was to determine, relative to methane,
the response to NMHC species.  Propane, propylene,  acetylene, 2-methyl-2-butene,
toluene, and acetaldehyde were chosen for the study.  These compounds, with
methane, are representative of the classes of NMHC's which may be encountered
in ambient air monitoring.  All were supplied as gases in HC-free air by
a commercial supplier (21) with the exception of propane.  This was a
Standard Reference Material grade supplied by the National Bureau of Standards
with an analysis of 94.2 +0.9 ppm, which was verified by analysis at RTI.
Sample gas streams were prepared by a single dilution of the NMHC with 2 ppm
methane in HC free air.   All flows were regulated by metering valves and
were measured with a bubble flow meter.  The dilution process yielded a
sample stream of approximately 2 ppm NMHC and 2 ppm methane.   The 2 ppm
level for the NMHC was chosen to minimize the possible errors in dealing
with low sample concentrations.  Flow conditions for hydrogen, carrier air
and support air were set for peak methane response as determined previously
because the optimum operation point for each of the individual NMHC had not
been determined.  The sample line was pressurized at 1.01 x 10  Pa (1 atm.)
above ambient.
     Three different modes of operation were used in the evaluation of NMHC
response.  The first was a chromatographic mode.  A carbon molecular sieve
(20) column operated at 95°C was used to separate methane from the NMHC
component.   The FID was operated also at 95°C.   Carrier air,  support air,
and hydrogen fuel flows were set to give maximum methane response.  The
second mode of operation was the strip-subtract method of the Federal
Register (1).  The stripper column was again a carbon molecular sieve,
operated this time at 105°C to assure adequate backflushing prior to
                                      40

-------
  m
   I
   o
   rH

    X
    c
    o
    o
    CO
    0)
-O   >->
K-   <
    TO
    0)
   C-H
300 —
       290 —
       ISO
                10
                                           -?()                                30

                                                  Dew Point (">.')

                    lrij;nre 29.   Peak  jirt-n  rc-sponsc as  a luru'tion of  sample  dew point
                                                                                                                       40

-------
75
                                                Dew Point  (°C)




                     Figure 23,  Peak height response as a function of sample  dew  point.

-------
introduction of a new sample.  The FID was also operated at  105°C.  The same
air and hydrogen flow rates were used as  in the chromatographic mode.  The
third mode was the backflush method.  In  this mode methane was allowed to
elute to the FID from the stripper column.  Following this the NMHC is
backflushed from the column into the FID  for analysis.  Again the carbon
molecular sieve column was operated at 105°C as was the FID, with the same
air and hydrogen flow rates.  Because 2-methyl-2-butene and  toluene were
apparently partially retained on the carbon molecular sieve, these two
compounds were re-run using 10% carbowax  on a commercial support  (20).
     The results of these evaluations are shown in Table 5.  The numbers
which are presented are the effective carbon numbers  (ECN) determined on
a per-atom-of-carbon basis in the following manner:
                                ECN =
                                      CNMHC
      is the concentration of NMHC's indicated by the instrument being used;
CLfMHr is the actual concentration; N is the number of carbon atoms per
molecule.  R^MU^ calculation depended on which of the three mdoes of opera-
tion was being used; the chromatographic and backflush approaches yielded a
direct value, but the strip method required subtraction of the methane
component from the THC valve.
     The ECN values for methane, propane, propylene, and acetylene are
reasonably consistent with literature values as can be seen by examining
the data from Ref. (23).  Note however that careful attention must be paid
to analysis conditions when making such comparisons.  These conditions
include, for example, sample pressure, sample flow rate, premixed- or
diffusion-type flame systems, and fuel carrier ratio, to mention a few,
and can significantly alter the ECN values for HC species.
     The response of toluene and acetaldehyde was quite low, and the 2-methy]
2-butene response was marginal.  Analysis of the toluene and 2-methy 1-2-buteue
cylinders on a Perkin Elmer 900 GC showed that the concentration certified
by the supplier (21) was correct; however, the aretaldehiyde had degraded to
60% acetaldehyde and 40% acetic acid.  One possible explanation is that there
was some condensation and adsorption problem in the line connecting the gas
mixtures to the RTI instrument.  Toluene, 2-methyl-2-butene and acetaldehyde

                                      43

-------
 TABLE 5 _.   THE NMHC RESPONSE OF RTI INSTRUMENT IN VARIOUS MODES OF OPERATION (CH, = 1.0 AS BASIS)

Methane
Propane
Propylene

2-Methyl-2-butene
Acetylene

Toluene
Acetaldehyde
Chromatographic
1.0
0.79
0.70

0.48
0.95
f] \
	
__(D
Strip
1.0
1.0
0.75

0.43
1.16

0.27
0.31
Backflush
1.0
0.80
0.50
(?">
0.66U;
0.79
(3)
0.50V '
— (4>
                                                         (1)  These compounds not eluted with carbon
                                                              molecular sieve stripper column.

                                                         (2)  This resulted from changing to 10%
                                                              carbowax column; ECN = 0.52 with
                                                              carbon molecular sieve.

                                                         (3)  Resulted from changing to 10% carbowax
                                                              column; ECN = 0.13 with carbon molecular
                                                              sieve.

                                                         (4)  Not tested in backflush mode.
Note:  "Chromatographic" measurements were made with GC oven, containing column and FID, at 95°C.

       "Strip" and "Backflush" measurements were made at 105°C.  Flow rates of hydrogen, carrier

       air and support air were maintained at the same levels for all three sets of measurements.

-------
are normally liquid at room temperature and near one atmosphere pressure.
The dilution and mixing system contained about 1.22 m  (4 feet) of stainless
steel tubing, 0.64 cm  (0.25 inch) diameter, as well as a stainless steel
mixing volume of 7.6 cm (3 inch) diameter and 10.2 cm  (4 inch) length.  This
was all at room temperature.  It is possible that some of the sample gas
was "lost" in this system.  However, there did not appear to be any induction
period required to achieve steady state response to any of these compounds
with the instrument.  Furthermore, the same sample mixing system was used
for the tests in all three modes of operation.
SUMMARY OF RTI FID-GC EVALUATION
     The prototype FID-GC exhibited a  response to methane that was very
similar to that for the FID alone.  Response was independent of support air
                             2
flow rate over the 100-300 cm /min range investigated.  Response to methane
peaked for both the hydrogen and carrier air flow rates.  However, there is
a peak in methane response versus the ratio of hydrogen to carrier air flow
rate which differs slightly from the flow ratio using values of both hydrogen
flow rate and carrier air flow rate which individually provide a peak response.
     Response to methane increased linearly with inlet pressure at about
30% per atmosphere over the range of 1.15 x 105 Pa to 1.70 x 105 Pa.   This
is about one-third the increase expected on the basis of the ideal gas law
and probably is due to the pressure drop through the sampling valve-sample
loop path, which reduced the sample loop pressure below the inlet pressure.
     Using synthetic carrier "air" of varying oxygen content, response to
methane was found to vary only slightly with 0- content above 20% 0~-
However below about 19% 0,., FID response to methane decreased at about 10%
per 1% change in oxygen content.
     For the RTI instrument, the strip mode of operation was superior to
the backflush mode.  This result cannot be generalized, however, since it
is a function of instrument design and column selection.
     Based on the results of the evaluation, the RTI Instrument was
demonstrated to conform to the specifications of the EPA performance
guidelines (22).   Response to several different NMHC species varied,
depending upon the method used to separate methane from the NMHC.  Although
it is well known that compound response is very sensitive to operating
parameters [see Ref.  25, for example], ostensibly the same FID operating
                                      45

-------
conditions were used.  Slightly different temperatures were used, but the
apparent ECN values for the condensible NMHC species appeared to be more
dependent on column effects than anything else.
     The effect of water vapor on response was somewhat surprising.  The
peak area and peak height increases which were noted on the RTI instrument
are contrary to much of the literature (24).  When comparing however, attention
must be paid to experimental conditions.   As an example, the presence of
1% of water has been reported to double the ionization-current in a pure
hydrogen flame while retarding ionization by 30% in other flames (25).
Definitive information on water vapor effects is not available for those
engaged in making NMHC measurements and is a topic certainly worthy of
further research, for water vapor will always be present in ambient sampling.
It should be noted that the effects reported in this section were for
water vapor carried directly into the FID with the sample.   Presumably
column effects such as site deactivation or displacement were not involved.
One is directed to Section 7 of this report for further evidence of the
impact of water vapor on ambient air monitoring.
     Quite typically more humid atmospheric conditions are  associated with
lower atmospheric pressures.  As reported in Section 4 the  FID response
changed on the order of 2% over the normal atmospheric pressure range.  The
pressure-induced deviation in response seems to operate in  a direction
which would augment the moisture-induced response increases.  It appears
possible to have an error source approaching 10% solely from normal
atmospheric pressure and moisture variation, if it is assumed that all
instruments respond as the RTI prototype.
     It should be pointed out that this summary and the conclusions drawn
from it are specific to the RTI design.  The effect of the various operating
parameters on response, however, are indicative of similar effects with
other designs, although the quantitative measures would be expected to vary
in detail with the particular design.  The purpose of this phase of the
evaluation is to illustrate that some operating parameters are more critical
in their effect on response than others.  This brings into question whether
or not these critical parameters are known by the instrument designer and are
sufficiently controlled over the operating life of the analyzer to maintain
a stable response pattern.  It also serves to illustrate to the users of
these instruments that proper instrument set-up and maintenance are necessary
to obtain good quality data.
                                      46

-------
                                   SECTION 6

                      REVIEW OF COMMERCIAL INSTRUMENTS

     Reviewing commercial instruments, particularly in a comparative
manner, poses a fundamental problem because different instruments possess
different operational philosophies and techniques.  Basically, a non-
methane hydrocarbon analyzer (or total hydrocarbon analyzer) is composed
of three separate subsystems:  the pneumatics, the detector, and the
signal processor.  By partitioning the commercial analyzers in this manner,
a comparison can be made which illustrates the operational nuances of a
particular instrument, while not losing sight of the overall specifications
prescribed by EPA.  The purpose of this section is to analyze the design
and fabrication of several common commercial analyzers, which were used in
the side-by-side, comparative evaluation described in Section 7.  The
instruments compared are the following:
          1.  a Beckman 6800 Environmental Gas Chromatograph,
          2,  a Beckman 400 Total Hydrocarbon Analyzer,
          3.  a Bendix 8200 Environmental Gas Chromatograph,
          4.  a Bendix 8201 Hydrocarbon Analyzer, and
          5.  an MSA 11-2.
According to manufacturer's specifications, all these instruments meet the
requirements for performing at least the total hydrocarbon measurement of
the Federal Reference Method (1).
DESIGN
     For analysis these commercial instruments are partitioned into sub-
systems as follows.   The pneumatics subsystem includes valving, tab ing
columns, flow and pressure controls, and, in general, all necessary hard-
ware to transport a, desired sample from the inlet port to the detector.
The second subsystem is the detector.  The only detector found in the
                                     47

-------
instruments examined is the FID.  The signal produced in the FID detector
is conditioned and processed in an electronic subsystem.  This subsystem
also includes the control electronics for automatic operation of the
analyzer in four of the six instruments evaluated.
Pneumatics
     The pneumatic design philosophy of commercial instruments appear  to
be divided into four categories.
     Category I:
          The simplest configuration is that of a total hydrocarbon analyzer,
such as the Beckman 400 or Bendix 8400, which involves a minimum of hardware.
Because this type instrument produces a continuous signal proportional only
to the total hydrocarbon content of the sample, there are neither switching
valves nor separation column; this instrument is incapable of separating
methane from the non-methane component.  The sample is typically pumped into
the instrument, with sample pressure being controlled by a regulator and
sample flow by a capillary.  The fuel and burner air lines have pressure
regulators and restrictors.  It has been shown in the work with the RTI-
designed instrument, and other studies (10,23,31), that the FID response varies
dramatically with hydrogen and carrier flow rate, implying that the pressure
drop across the restrictors must be sufficiently great to assure constant
flow, and this appears to be a major requirement for stable operation.
     Due to the simplicity of design, the operation of this category of
instruments is usually not involved and the instrumental output is a single
continuous trace on a strip chart recorder.
     Category II:
          The second category is a natural progression in increasing com-
plexity and is capable of methane and non-methane determinations.  A
model representing this category is the MSA 11-2; it is a dual-FID, continuous-
output instrument.  The configuration is not unlike a parallel arrangement
of two total hydrocarbon analyzers, in which one of the analyzers has been
modified to respond to methane only.   Again, the continuous output obviates
the necessity for any type of switching valve.   However, the use of two
detectors requires that the sample stream be split and both flows be kept
stable.   The sample stream fraction to be used for a methane determination
requires some type of column to either adsorb or catalytically burn all
                                     48

-------
non-methane hydrocarbons.  Pressure and flow regulation are accomplished
by using pressure regulators and flow restrictors, as previously described,
although the plumbing necessarily reflects the increase in complexity of
the instrument.
     With this design a new problem, column contamination, has been
introduced.  Columns which adsorb the non-methane constituent have a
finite limit on the quantity of material which can be adsorbed, and a
catalytic burner or cutter column, such as the MSA 11-2 employs, also
must be periodically regenerated.  Failure to perform necessary regener-
ation degrades column performance and impedes complete methane-non-methane
separation leading to increasingly inaccurate measurements.  Another
possible problem is the degradation of column and catalytic surfaces by
the gradual accumulation of water.
     Category III:
          The environmental chromatograph represents the next category.
The level of complexity has increased significantly for several reasons.
Characteristic instruments, as the Beckman 6800, Bendix 8200, and Bendix
8201, must separate the methane and non-methane fractions; this requires
flow switching capability  (usually at least two switching valves) and
stripper and/or separation columns.  There now is a timing requirement
so that fractions are detected at the proper time.  The introduction of
switching valves adds more problems; these valves cannot switch instantan-
eously and may require several hundred milliseconds to switch.  This finite
time produces pressure surges which can seriously perturb the baseline;
regardless of preventative measures, the instruments will usually reflect
some of the switching induced pressure surge.  Also, the seals in the
switching valves can leak or seize and impair operation.  An additional
pressurized air supply is usually required to operate the valves.
     The flow control is generally the same as previously described -
regulators and restrictors - implying the continued need for large pressure
drops across restrictors to maintain constant flow.
     Category IV:
          The final category represents instruments which operate in the
"backflush" mode, meaning that a sample is injected onto a column which
separates the methane, allowing it to pass on to the FID detector, and is
then backflushed so that the non-methane component may be directed to the

                                      49

-------
detector.  These instruments, such as the Hewlett-Packard 5830A Gas
Chromatograph or the Carle Series-R Analytical Gas Chromatographss may
have single or dual detectors and usually have only one switching valve.
These instruments possess all the problems previously described plus one
addition.  The column obviously must give adequate resolution of  NMHC and
methane, and non-methane fraction must not adhere to the column material
so tenaciously as to make removal overly difficult when backflushed.
This can be accomplished with a multicomponent column packing.
     In summary, most commercially available instruments, intended
primarily for the use of total hydrocarbon or non-methane hydrocarbon
measurement, fall into one of the four pneumatic design categories
described.  These categories are based on increasing instrument (and
pneumatic) complexity, and it appears that the problems associated with
each level of complexity are transmitted to higher levels.
Detectors
     All instruments examined in the comparative evaluation used flame
ionization detectors.  With the exception of the MSA 11-2 these were all
essentially the same design - cylindrical collectors coaxial with the jet
tip.  The design advantages of this configuration were enumerated in
Section 4 when the RTI prototype was discussed and will not be discussed
further here.  The FID used in the Beckman instruments is geometrically
similar to the RTI design (see Fig. 6), with the bottom edge of the
collector fitting down below the level of the jet tip.  The FID used in
the Hewlett-Packard Model 5830A is also similar.  The Bendix FID differed
only in that the cylindrical collector was raised slightly above the jet
tip.  The MSA 11-2 has an unusual geometry; the jet tip points horizontally
and is coaxial with a Swagelok^ferrule (used as the collector), located
about 3 mm (0.125 inch) from the jet tip.   This geometry does not appear
in the literature but is claimed by MSA to be less noisy.  This design
requires a lower polarization voltage for saturation (26).
     The polarizing voltage in these instruments is typically applied to
the jet, which is a good design feature, as was pointed out in Section 4.
Beckman and Hewlett-Packard FID's employ positive jet potentials of about
100 volts.  The Bendix instruments employ a negative voltage (-lOOv) on the
                                      50

-------
jet tip.  The MSA 11-2 polarizes with a positive potential which  is much
lower  (15 volts) than the other instruments examined.
     To summarize, the designs of  the detectors are quite similar, except
for that of MSA, which employs a unique geometry and a much lower polarizing
potential on the jet than do the other designs.
Electronics
     The level of electronic complexity follows the level of pneumatic
complexity.  This is illustrated by  the following descriptions.
     1.  The electronics of total  hydrocarbon analyzers  (Beckman  400, Bendix
8400)  consist basically of an electrometer amplifier and an attenuator network
to condition the continuous output signal for the strip  chart recorders,
     2.  The MSA has essentially the same amplifier-attenuator configuration
on each of its two continuously monitoring channels and  includes an analog
subtracter (difference amplifier)  to give the non-methane response
electronically (THC-CH^ = NMHC).
     3.  The environmental chromatographs (Beckman 6800, Bendix 8200, Bendix
8201)  use an amplifier-attenuator  arrangement for continuous output but also
have memory elements (sample-and-hold amplifiers) so that outputs can be
transferred to recorders for operation in the bargraph mode when the appro-
priate timing signals are received.
     4.  The electronics of backflush instruments are essentially the same
as described for the environmental gas chromatograph with the appropriate
timing considerations for a methane and non-methane peak detection.  However,
because of the peak shape electronic integration is needed for peak quantification.
     When discussing the electronics of these instruments, the crucial point
is the manner in which the non-methane component is actually determined.  The
most prevalent method,  the one prescribed in the Federal Register, requires
subtraction of a methane determination from a total hydrocarbon determination.
Since  these numbers are typically  of similar magnitude,  any small errors in
either result in large errors in their difference.   Instruments in Category II
and Category III use this method.   An alternative approach is to make both
the methane and non-methane measurement directly from a  single sample injection
by the backflush technique.  This  eliminates the necessity of subtracting two
numbers of similar size.
     A second important consideration is the method of quantification of
detector signals.  ?eak area is recognized as generally  the most accurate
measure of concentration in batch  analysis (27).  Digital integration can
                                      51

-------
be the most accurate manner of quantifying peak area  (27).  Use of digital
data facilitates data manipulation and further has the advantage of being
in the format used for reporting and comparison.  Analogue integration is
somewhat less accurate than digital integration but usually superior to
peak height measurement, which is still frequently used for quantitation
(28).  In a study of the accuracy obtained from various methods, Gill
(29) listed the following order for the precision of peak-area determin-
tion, expressed in relative standard deviations.
          1.  planimetry, 4.0%;
          2.  triangulation, 4.0%;
          3.  height and width of half-height, 2.5%;
          4.  cutting and weighing, 1.7%;
          5.  disk integration, 1.3%;
          6.  electronic digital integrator, 0.4%.
Novak (30) has recently listed the conditions when peak height quantitative
determinations are applicable:
          1.  narrow, tall peaks,
          2.  flow rate stability not possible, (e.g.  temperature programming)
          3.  good peak symmetry, and
          4.  constant chromatographic conditions for a single material.
Of the instruments examined, none used digital integration.   Peak height
or signal level subtraction was used in all cases to make the quantitative
evaluations in spite of the limitations imposed by conditions 3 and 4 of
Novak's work.
FABRICATION
     In view of the significance of the comparative evaluation of the
commercial instruments, careful examination of the fabrication of the
instrument was necessary.  The fabrication materials and details of an
instrument can play a major role in determining instrument response and
maintainability.   This is particularly appropriate in the current discussion
because the designs of all the analyzers appeared adequate to perform the
non-methane measurement (keeping in mind that the subtraction of nearly
equal magnitude measurement numbers has a large error potential constraining
the instruments to very stable and reproducible operation for accurate
results).   What follows, then, is a summary and critique,  instrument-by-
instrument,  of the fabrication.  It should be noted that these evaluations
                                     52

-------
are based on the examination of only one instrument of each model.
     1.  Beckman 6800
              The materials of construction and layout of the Beckman
         6800 exhibit good design practice.  All sample lines, seals and
         valves are stainless steel, and the plumbing is accessible,
         allowing ease of maintenance and repair.  The FID was clean,
         with neither jet tip nor collector showing undue oxidation.
         The temperature control by the oven was stable.
     2.  Beckman 400
              Like the 6800, the construction of the Beckman 400 is good;
         the plumbing is stainless steel and accessible.  This was the only
         instrument which used a battery for the polarizing voltage, but this
         presented no obvious problems, because of the low current.
     3.  MSA 11-2
              The MSA 11-2 has several apparent materials problems.  The
         sample line is made of copper, which has a tendency to adsorb
         oxygenated compounds and aromatics.  This has been reported in
         the literature (Ref. 31, for example).  The MSA also used plastic
         filter canisters and pressure regulators containing Buna-N rubber.
         The gas distribution block and selector valve are made of block
         aluminum although stainless steel shows better resistance to
         corrosion and is less likely to present compound adsorption
         problems.  The geometry and fabrication of the FID made accurate
         positioning of the jet difficult.  The instrument requires 3
         liters/minute of sample which can be a large drain on cylinder
         supplies during calibration.
              The instrument is designed and constructed so that repair
         and maintenance are easy to perform.
     4.  Bendix 8201
              The fabrication of the 8201 is good; the flow lines are
         stainless steel,  and plumbing is  accessible  and well-conceived.
         Tne jet tip  and  collector did show some oxidation,  and  one of
         the switching valves had to be  replaced due  to  an  0-ring
         breakdown.
                                     53

-------
     5.  Bendix 8200
              The fabrication of the Bendix 8200 is also quite good.  The
         plumbing was stainless steel and accessible.  The jet and collector
         showed some oxidation.   Plastic prefilters were used in the sample
         line.
     It may be concluded, then,  that the commercial instruments examined were
all well made, and, with the possible exception of the copper sample line,
the construction of these instruments should not be the limiting factor on
their response characteristics.   Copper sample lines can be a problem with
certain classes of compounds; this fact was confirmed during the comparative
evaluation with toluene and acetaldehyde.  The plastic filters are a potential
source of error, but their actual effect was not determined experimentally.
Finally, many of the problems,  such as jet collector oxidation, should be
corrected during routine maintenance, and the valve problems, such as the
ones experienced, are not uncommon or unreasonable.
SELECTION OF OPERATING CONDITIONS AND CALIBRATION
     Following evaluation of the RTI designed FID and gas chromatograph as
described in Sections 4 and 5 of this report, a summary report was distri-
buted to six manufacturers of analyzers for non-methane hydrocarbons along
with questions about the selection of operating conditions and methods of
calibration for their instruments.  A list of individuals and firms contacted
is given in Table 6.  Response was generally good although one firm declined
to send in answers to the questions posed.  The questions are listed in
Table 7.  The answers were generally frank and it was apparent that a lot
of thought had been given to the replies.
     From the answers received the following information was drawn.
     1.  Selection of Flows.  One manufacturer selects hydrogen and carrier
         gas flow settings to optimize response to methane, but pointed
         out that this does not insure peak response for other hydrocarbons.
         A second uses settings selected to optimize response to the
         greatest number of HC species.  A third selects those flows that
         give a "stable, reproducible response" (presumably to all hydrocarbons)
         A fourth uses a carrier air flow that provides desired elution
         times from the analytical column.  A fifth sets the H2/carrier air
         ratio to prevent flame blowout from the air peak from the sample.
                                      54

-------
    TABLE 6.  LIST OF MANUFACTURERS AND REPRESENTATIVES SURVEYED
Mr. John Scales
Sales Engineer
Bendix Process Instruments Division
P. 0. Drawer 831
Lewisburg, WV  24901

Mr. D. W. Stevens
Market Development Manager
Analytical Instrumentation
Beckman Instruments, Inc.
2500 Harbor Boulevard
Fullerton, CA  92634

Mr. Pat Conner
Marketing Manager
Instruments and Systems Division
Meloy Laboratories, Inc.
6715 Electronic Drive
Springfield, VA  22151

Mr.  Byron Behr
President
Byron Instruments,  Inc
520  S. Harrington  St.
Raleigh, NC  27601

Dr.  Charles A. Burgett
Avondale Division
Hewlett-Packard
Route 41 and Starr  Road
Avondale, PA   19311

Mr.  William V. Dailey
Product Line Manager
Technical Products  Division
Mine Safety Appliances Company
400  Penn Center Boulevard
Pittsburgh, PA 15235
                                  55

-------
          TABLE 7.   LIST OF QUESTIONS IN SURVEY OF MANUFACTURERS

1.  Do the operating instructions for initial pressure (and/or flow) settings
    for hydrogen and carrier gas for the FID's in your instruments insure
    operation at or near the peak of the response curve for the FID?
2.  How were these settings determined, on a single prototype evaluation,
    a range based on randomly selected and tested production line units,
    individually for each instrument?
3.  Are instrument specifications on drift, precision, etc., obtained from
    measurements on a single prototype, randomly selected and tested
    production line units, individual production unit testing?
4.  How are precision and accuracy determined; i.e. what gases are used,
    are they analyzed and certified by supplier, are they spot checked
    (or exhaustively analyzed) by you, how many measurements are taken
    over what period?
5.  Have you evaluated your instrument's response to non-methane hydrocarbons
    species by species?
6.  Based on your experience (i.e., that of your field representatives),
    what do you consider to be the significant sources of error in each
    of the following aspects of the present Federal Reference Method for
    NMHC measurement:
              1.   Instruments (and instrumental method)
              2.   Reagents (operating and calibration gases)
              3.   Procedure specified in the Federal Register
              4.   Operator
    and can you suggest methods for minimizing these errors?
                                     56

-------
2-  Method for Selection of Flows.  One manufacturer chose settings
    based on a prototype design later modified somewhat "based on
    experience".  A second used data from several prototypes, with
    these initial parameters based on prior experience.  A third
    uses "research data and evaluation of randomly selected units".
    A fourth adjusts settings based on tests on prototypes, pilot
    runs and production units.  A fifth individually adjusts settings
    for each instrument produced  (with criteria not specified.)
3.  Basis for Specifications.  For all manufacturers noise, span and
    zero drift are checked for each instrument produced prior to
    delivery.  Precision is checked on random instruments by one
    manufacturer.  The others check each instrument, over time spans
    that vary.   (See answers to question 4, below).  Accuracy checks
    were qualified by statements that reduced to "the accuracy check
    depends upon the accuracy of the determination of the calibration
    gas used."
4.  Calibration Gases and Methods.  Three of the manufacturers depended
    on certified analysis by gas supplier.   One, who uses diluted
    gases, has experienced several cases where the certified analysis
    was substantially in error.  Consequently, this manufacturer
    uses an intercomparison technique to identify gross errors.  This
    manufacturer's zero, noise, span and precision checks are made over
    a 24 hour period.  A second manufacturer cited similar problems with
    "certified analyses" and also uses an intercomparison technique to
    check for reproducibility.  This manufacturer checks over a 48 hour
    period.   Another manufacturer blends gases in-house and analyzes
    to about 1% accuracy at 5 ppm level.   Their instrument calibration
    stability checks are made over an 8 hour period.  No data on the time
    period of the calibration stability tests were given by the other
    two manufacturers,
5.  Non-Methane Hydrocarbon Response.  Three out of five manufacturers
    responding have done little or no species-by-species evaluation
    of NMHC response of their instruments.   Another manufacturer has
    performed extensive species by species tests and reported agreement
    to within 12% of literature values, except for toluene where wall
    adsorption was a problem.
                                57

-------
6.  Problems with Present Federal Reference Method.  The responses were
    generally organized by the categories of instrument, reagents, procedure
    and operators.   These are summarized as follows:
    a)  Design and construction difference probably introduce significant
        differences in species by species NMHC response as well as
        linearity of response.  Three out of five respondents thought
        that the present Reference Method should be revised to include
        design options other than the present strip/subtract technique.
    b)  Calibration gases were uniformly highlighted as a problem due
        to contamination and/or inaccurate analysis by the gas supplier.
        Another problem cited was the instability of low ppm gas
        mixtures.
    c)  One respondent cited the need for more specificity of reagents
        and calibration gases.  Two explicitly stated that the procedure
        was too restrictive with respect to the instrumental method
        specified.
    d)  All respondents agreed that more operator training was necessary,
        citing the relative sophistication of the instruments contrasted
        with the (usual) inexperience of operators and maintenance
        personnel.
Along with these critiques of the present method there were several
recommendations for changes, which will be cited in the discussion of
Section 8 of this report.
     This review of commercial instruments provides the background for the
next section, which reports a side-by-side comparative evaluation of
several of the instruments performed to determine the significant source(s)
of discrepancy in field measurements of non-methane hydrocarbons.
                                      58

-------
                               SECTION 7
                        COMPARATIVE EVALUATION

     Following the evaluation of the RTI designed FID-GC instrument
reported in Sections 4 and 5, this instrument was prepared along with
five commercial analyzers, for a side-by-side comparative evaluation
in order to determine relative performance with regard to
          1.  calibration stability,
          2.  non-methane hydrocarbon response, and
          3.  operational characteristics.
This section describes the test arrangements and procedures and analyzes
the data obtained.  Detailed test data are given in Appendix B.
TEST ARRANGEMENTS
     A list of the instruments used in the comparative evaluation is
given in Table 8.  Instruments were located side-by-side in the same
laboratory and were subjected to the same diurnal temperature variation
(about 22° to 28°C).  Sample gases were introduced to a common manifold
which permitted simultaneous samples to be drawn by each of the instruments
under test.  The flow arrangements are shown schematically in Figure 24.
Commercial analyzers were placed into operation in strict accord with the
manufacturer's instructions.
     The list of reagent gases and concentration levels for the evaluation
are shown in Table 9.  The hydrocarbon stock mixtures were those shown
in Table 10, which gives the gas supplier's analysis and the RTI analysis.
The latter were carried out on a Perkin-Elmer Model 900 gas chromatograph.
This instrument was also used to determine the total hydrocarbon content
of the air and the hydrogen fuel for the FID.  The air was obtained from
a catalytic cleanup system-nickel oxide on firebrick operated at 450°C.
This supply was monitored periodically during the test sequence with the
highest THC concentration measured at less than 50 ppb.  The THC content
of the hydrogen was less than 100 ppb.
                                    59

-------
               TABLE 8.   INSTRUMENTS USED IN COMPARATIVE EVALUATION
a.  Instruments provided by EPA
          EPA No.           Serial No.
          083X70           33082
          102671
          090653
          083182
2014
28713
1001018
Instrument
Analyzer, Bendix Reactive
Hydrocarbon, Model 8201
Analyzer, MSA Non-Methane
Hydrocarbon, Model 11-2
Analyzer, Gas Chromatograph,
Bendix, Model 8200
Analyzer, Total Hydrocarbon,
Beckman Model 400
b.  Instrument provided by RTI
          RTI No.
Serial No.
1000389
Instrument
Analyzer, Gas Chromatograph,
Beckman Model 6800
Analyzer, Prototype Hydrocarbon
                                      60

-------
TABLE 9.
           LIST OF REAGENT GASES AND CH,/NMHC MIXTURES FOR COMPARATIVE
           EVALUATION
Zero air:  <50 ppb THC
Hydrogen:  <100 ppb THC
Span gas (80%);  8 ppm CH^ + <0.05 ppm NMHC
Span gas (20%):  2 ppm CH, + <0.05 ppm NMHC
            Mixes:
                                     Methane Background

Propane
(C3H8)


Propylene

Acetylene
Ethylene
Toluene
(C7H8)
Acetaldehyde
(CH3CHO)
2 ppm CH,
0.10
0.25
0.50
1.00
2.00
0.25
0.50
2.00
0.25
2.00
0.25
2.00
0.25
2.00
0.25
2.00
8 ppm CH,
0.1
0.25
0.50


0.25
0.50
2.00
0.25
2,00
0.25
2.00
0.25
2.00
0.25
2.00
                                    61

-------
                 Catalyst
                 Bed At
                  450°C
Catalyst
Bed At
 450°C
                                  I	I
 Bendix
  8833


Clean Air
 Supply
                                       Spherical
                                       Sampling Bulb
Figure 24.   Schematic diagram of flow system used for
            comparative evaluations.
                          62

-------
TABLE 10.   LIST OF STOCK HYDROCARBON/AIR REAGENTS WITH REPORTED AND

           MEASURED CONCENTRATION S

Compd.
C3H8
C3H6

C2H2
C7H8
C2H4
CH4 fl
CH4 #2

CH3CHO
Supplier's
Source Reported Cone, (ppm)
NBS 94.2 + 0.9
Scott 205 + 2%

Scott 203 + 2%
Scott 198 + 2%
In-house blend 	
In-house blend 	
In-house blend 	

Scott 183 + 2%
i
RTI (ppm)
95 + 0.7%
205 + 3.0%
3
203
206 ± 3.0%
260.1 + 4.
152.36 ± 2
152.62 + 3
2








5%
.0%
.0%


 All RTI analyses performed on a Perkin-Elmer Model 900 Gas Chromatograph.
2
 Acetaldehyde had oxidized to approximately 60% - CH3CHO


                                            40% - CH3 C02H.
 3
 Single gas chromatographic analysis.
                                    63

-------
     The concentration levels of hydrocarbons in Table 9 were obtained by
precision flow blending of stock hydrocarbon/air mixtures with the
catalytically purified air.  Flows were measured at least twice per day to
assure that the desired hydrocarbon levels were being maintained.  Daily
flow variations were typically less than 1%; only one day showed a large
variation (10%) and the sample concentrations were recalculated accordingly.
TEST PROCEDURE
     During the first part of the evaluation, which started July 28, 1976,
the instruments were examined for calibration stability.  Instrument zero
points and spans were set with hydrocarbon free (<50 ppb) air and with 8 ppm
methane in air, respectively.  All instruments were operated on a 10 ppm
full scale range throughout the entire test.  The bargraph response mode
was used for the Beckman 6800, Bendix 8200 and Bendix 8201.   The Beckman 400
and MSA 11-2 gave a continuous signal.  Zero and span stabilities over a
twenty day period were monitored.  No instrumental adjustments were made
unless absolutely required for analyzer operation (for example, relighting
an extinguished flame.)  All such adjustments were logged.
     In addition to zero and span stability, precision was ascertained
during this period by repetitive alternate injections of 2.2 ppm CH,  and
0.53 ppm propylene plus 2.2 ppm CH,.
     The second phase of the evaluation utilized varying levels of non-
methane hydrocarbons, on a species-by-species basis, with and without a
CH, background as shown in Table 9.  When a methane background was used,
it was at either a 2 ppm or an 8 ppm level.  The purpose of this was to
determine the effective carbon number (ECN) response of each instrument
on a species-by-species basis and to determine whether or not any synergistic
effects could be detected.  Also during this test phase the linearity of
response was determined using propane.
     The final phase was a 3-day sampling of ambient air in the Research
Triangle Park, N. C., conducted over the period of August 28 to August 31,
1976.
     Throughout the entire evaluation all instruments were operated from
a common source of hydrogen fuel and from a common air supply for both
the carrier and FID support air.  This was done to insure that all instru-
ments had the same bias (if any) due to reagent gases.
                                       64

-------
OPERATIONAL SUMMARY
     During the testing sequence various observations were made about the
operating characteristics and maintenance problems which were experienced.
These are summarized below for each instrument,
Bendix 8201
     This instrument flamed out 7/29/76 and a very difficult time was
experienced trying to get the instrument reignited.  A Bendix  representative
accomplished this 7/30/76.  An 0-ring on one of sampling valves began
leaking 8/15/76.  This had to be replaced and flows readjusted (8/17/76),
after which the instrument worked well.  During the NMHC response evalua-
tions the instrument did not respond to acetaldehyde.  Careful setting
of backflush air was necessary to keep an unperturbed baseline when the
stripper column was backflushed.
MSA 11-2
     This instrument had several flame outs during the early part of first
test phase due to pressure transients on the common reagent supply lines.
These occurred whenever another instrument was being connected or disconnected.
This was a unique phenomena for the test setup and would not be expected
in normal operation.  Trouble was experienced with the cutter oven beginning
July 30, and there was evidence of incomplete non-methane oxidation until
August 2, which resulted in an apparently low response for propane (due to
the higher output of the methane channel) until this date.  This problem
was rectified and the cutter oven temperature control worked properly
during rest of test, resulting in ECN values comparable to those of the
other instruments.
     Effects of sample line copper tubing (see discussion of Fabrication
in Section 6) were noted with toluene and acetaldehyde.  Both gave a time
lag response, described  by the following approximate equations, which
were determined graphically,

                          Toluene:  C = CQ  (1 - e"t/7.55)                (9)

                          CH3CHO:   C = CQ  (1 - e"t/7.60)                (10)

where t is in minutes elapsed, following a step change in concentration,
and C  is the concentration step magnitude for the space in ppm.  These
     o
                                       65

-------
long time constants, about 7.5 minutes, implied that rapid concentration
changes might go unnoticed, or be attenuated.
Beckman 6800
     Trouble was experienced with the slider in the CH,-CO sampling valve.
It would stick in mid-position.  This valve was replaced, and the 6800
worked well afterwards.  Also the instrument had a defective Auto-Manual
switch.  There appeared to be some trouble with electronics, with occasional
irregular peaks and spikes noted.  As with the MSA 11-2 some flame outs
occurred when work on other equipment affected line pressures.   As pointed
out before, however, this was a problem unique to this experimental set up,
     The sticking valve resulted in the loss of some of the data from the
long-term stability test.
Beckman 400
     The only real problem experienced was an occasional noisy signal, the
cause of which was never identified.   As with other instruments in the
test there was an occasional flame out when one of the other instruments
was put on-line or taken off-line, due to reagent gas pressure fluctuations,
Bendix 8200
     No operational problems were experienced with this instrument,   It
displayed good stability and reliable performance.
RTI Prototype
     Some flame outs occurred due to reagent gas pressure fluctuations
when work being performed on other instruments affected line pressure.
A solenoid valve failed in the air switch which operated the sample valve,
A hand operated system was fabricated and used for the remainder of the
test.  The baseline showed some downward drift due presumably to hydrogen
or carrier flow fluctuation toward the end of the stability test.
DATA ANALYSIS
     The electrical output signal from each instrument was fed to separate
strip chart recorders.  Details on these recorders are presented in
Appendix C which reports on evaluation to determine the calibration
stability and dynamic characteristics of these components of the measure-
ment systems.  It was concluded that these components contributed negligible
error to the overall measurement.
     Whenever trouble was experienced with an instrument due to equipment
malfunction or being temporarily decommissioned, such as by FID flame-out
                                       66

-------
due to a common supply line pressure surge, the data from the instrument
was discarded until proper operation was again attained.
     Zero and span drift were obtained by taking hourly data from each
chart and calculating the mean and standard deviation for each day.
These are plotted and presented, along with other test data, in Appendix B.
     Precision for CH^ (at 2.2 ppm) and THC (0.53 ppm) propylene/2.2 ppm
CH^) were calculated as twice the standard deviation of successive repeti-
tive measurements of these input samples over a period of 13 days.  It
should be noted that some instruments were inoperative during this period.
Only operating days were used to calculate precision for these instruments.
     The ECN values were calculated on a per-atom-carbon basis, as was
done in Section 5.  The ECN's which are reported represent averages of
all the ECN's for a particular NMHC.  These data were further broken down
into ECN's determined with a methane background and when only the NMHC was
being sampled; the differences in these ECN values are reported.   Also
reported are "corrected" ECN's which were obtained by multiplying the
"normal" ECN, determined on a CH, =1.0 basis, by the reciprocal of the
propane ECN for the instrument upon which the particular determination
was made.  Methane yielded a higher response than NMHC on a per carbon
atom basis in all of the instruments evaluated, and the "corrected" ECN
is an effort to compensate for this.
     Another important feature of the NMHC data illustrates the need for
frequent (daily or more often) calibration even though calibration frequency
is not presently specified in the Reference Method (1).   To develop this
two methods of data reduction are used.  The first method uses response
determined by a single 80% span calibration performed at the beginning
of the 14-day NMHC test period.   These data necessarily reflect all the
span and zero drifts which occurred during the test period.
     The second method takes advantage of the fact that the injection
sequence for a typical NMHC was begun with CH, only,  followed by CH./NMHC
mixture, and terminated with the NMHC only.   This sampling sequence gave
the effect of a daily (or more often, in some cases)  span calibration
either at the 20% or 80% span level.  The differences due to data reduction
based on the two different methods were found to be significant,  as will
be discussed later.
     For the ambient air measurements the initial calibration prior to the
start of the three day test was used for data reduction.
                                       67

-------
 RESULTS AND  DISCUSSION
      From the data analysis procedure described  in  the  previous  section
 the  following results are obtained.
 Zero Drift
      Table 11 shows  the  zero drift of each  instrument reported in  three
 categories.  The  first column  gives  the  % of  operating  days when the
 +0.2 ppm 24-hour  zero drift specification of  the EPA performance guidelines
 (22) was not met.  The second  column gives  the % of operating morning
 hours (0000-1200) when the +0.2 ppm  12-hour specification was not  met.
 The  third column  gives the % of operating afternoon-evening hours  (1200^
 2400) when the +0.2  ppm  12-hour specification was not met.  Instruments
 with both a  methane  channel and a THC channel have  two  entries per
 column.  The first is for the  CH, channel response  to zero air,  and the
 second is for the THC channel  response to zero air.
      As discussed in the OPERATIONAL SUMMARY  the Beckman 6800 instrument
 had  a faulty CH.-CO  sampling valve during the initial test phase.  This
 is believed  to be the cause for the  excessive deviations of zero drift
 shown; the sticking  valve altered the H« carrier flow rate thereby altering
 the  flame conditions and output levels.
	TABLE 11.   ZERO DRIFT PERFORMANCE OF INSTRUMENTS      	
             % of Operating Days
             when +0.2 ppm 24-
             Hour Zero Drift Spec.
             was Not Met
% of Operational
Morning Hours
(0000-1200) when
+0.2 ppm 12-Hour,
% of Operational
Afternoon Hours
(1200-2400) when
+0.2 ppm 12-Hour,


Bendix
Bendix 8201
Beckman 400
Beckman 6800
MSA 11-2
RTI


0,0
6,6
21
31,25
12,0
23
Zero Drift Spec.
Not Met

0,0
0,0
16
41,47
0,0
0
Zero Drift Spec.
Not Met

0,0
0,0
21
47,35
0,6
8
Note:  When two numbers appear, the first refers to CH, determinations of
       zero air and the second to THC determinations of zero air.

Note;  See note on Table 12 concerning Beckman 6800 data.
                                       68

-------
Span Drift
     Table 12 shows the span drift of the instrument reported as the
percentage of operating days when the span drift did not meet the EPA
performance guideline standard of +5% (22) .   All instruments exceeded
the specification a significant fraction of the time.
                  TABLE 12.  SPAN DRIFT PERFORMANCE OF INSTRUMENTS
Span Drift

Bendix 8200
Bendix 8201
Beckman 400
Beckman 6800
MSA 11-2
RTI
% of long term stability test when instrument did not meet
24-hour span drift requirement
2/13 x 100 = 15.4%
3/12 x 100 = 25%
9/13 x 100 = 69.2%
7/10 x 100 = 70%
6/13 x 100 = 46.2%
2/7 x 100 = 28.6%
Note:
Beckman 6800 met all
zero and span drift
requirement during first
3 days of test when valve
was functioning properly.
Data included solely to
illustrate effect of valve
failure.	
Precision
     The precision of the instruments is shown in Table 13.  These values
should be compared to the EPA performance guideline specification of 0.3 ppm
(22).  The precision on both the methane channel (measured at 2.2 ppm CH.)
and the total hydrocarbon channel (measured at 0.53 ppm propylene plus 2.2 ppm
CH,) is reported.  All instruments were within this specification.

TABLE 13.  PRECISION OF INSTRUMENTS ON METHANE AND TOTAL HYDROCARBONS CHANNELS


Bendix 8200
Bendix 8201
Beckman 400
Beckman 6800
MSA 11-2
RTI
f H ( T\T\TTt i
**J L L 1 V K M I" /
4
0.05
0.04
	
0.03
0.05
0.11

THC (ppm)
0.04
0.06
0.20
0.03
0.10
0.16
                                       69

-------
Linearity
     Propane was used to determine NMHC response linearity over the concentration
range 0.1 ppm to 2.0 ppm.  Data for five of the six instruments are reported in
Figure 25.  All of the instruments responses appeared linear in this range;
however, insufficient data were collected to determine accuracy and precision
in the test range.  The fact that a 1 ppm sample of propane, for example, does
not yield an equivalent 3 ppra CH, response, reflects the less efficient ionization
or ion collection of propane.
Non-Methane Hydrocarbon Response
     Two methods were used to analyze the data from this part of the evaluation.
The first method was to use the 80% span calibration response to methane, obtained
at the start of the 14 day test phase as the basis for determining the response
of each instrument to the various NMHC species.  The resulting data is presented
in Tables 14-17.  This data will not be discussed in detail except to point out
the much larger discrepancies in response to the different NMHC species, both
on an intra-instrument and inter-instrument basis, which appear due to the
shifting calibration of each instrument over the test period.
     The second method of data reduction was to use the daily calibration provided
by single injections of either 20% of span methane concentration or 80% of span
methane concentration, as the basis.  This data is presented in Tables 18-21.
Because of the more frequent calibration it is believed to be more representative
of instrument response characteristics and will be discussed in detail.
     In Table 18 the ECN values are shown using a methane ECN = 1.0 basis.  The
response of acetaldehyde was by far the lowest with a NMHC average of 0.39 (20%
lower than C2H^, the next lowest), omitting the Bendix 8201 which did not
respond.  The MSA exhibited almost three times the response of the Beckman 400
and 6800 and the Bendix 8200 and about 70% larger response than the RTI
instrument.  However it did take the MSA a considerable length of time to reach
this response level.  Regardless of the exact nature of the poor response, the
clear implication is that accurate CH^CHO determinations would be difficult
with the instruments used in the comparative evaluation in a reasonable time
frame.   Because of the low ECN, CH^CHO data was not included in the rest of the
analysis.
     Important information on the uniformity of NMHC response is found in
Table 18 in the row which lists the mean of the NMHC response for each
instrument and the column which lists the average for each of the NMHC species
for the instruments used in the study.  The average value for the NMHC's ranged
                                       70

-------
33
U


E
P.
CX
W
o
Pu
OS


<;
H
e
w


1
H
CO
2
                   BENDIX 8201

                   BENDIX 8200

                   BECKMAN 400

                   BECKMAN 6800
    1.0
    0.5 •
                              0.5                      1.0                     1.5


                                       PROPANE  CONCENTRATION (ppm)

                                    Figure  25.   Linearity of propane response.
2.0

-------
TABLE 14.  EFFECTIVE CARBON NUMBERS OF NMHC WITH CH  =1,0 AS BASIS, APPLYING INITIAL CALIBRATION OVER
           14 DAY PERIOD








Methane
(CH )
^T

Propane
(C.H )
3 o

Ethylene
(C H )
/- *T

Propylene
(C3H6)
Acetylene
(C2H2)
Toluene
(C7Hg)
(4)
Acetaldehyde
(CH3CHO)
(5)
Mean of NMHCV '
per instrument
CN
1
rH
rH
*3^
c/}
s


1.0


(2)
0.64



0.48



0.76

0.55

0.77

0.66


0.64

cJ
ttf
Jo
O O

g oo <:


1.0



0.51



0.37



0.45

0.46

0.43

0.27


0.44



Notes:


(1) Instrument not on-line when propane
response determined.

(2) Cutter oven not working properly;
based on propane— with-no— methane
samples.

(3) Calculated using toluene samples
only; toluene/methane wi-* gave
ECN = 0.1.

(4) Analysis showed decomposition to
60% acetaldehyde and 40% acetic
acid.

(5) Excludes CH3CHO ECN's.










-------
    TABLE  15.  RESULTS OF  CORRECTING NMHC RESPONSE TO PROPANE, APPLYING INITIAL CALIBRATION OVER 14 DAY PERIOD
LO

Methane
(CH4)
Propane
(C3H8)
Ethylene
Propylene
(C3H6)
Acetylene
Toluene
(C7Hg)
Acetaldehyde
(CH3CHO)
Mean of NMHC
CM
i
rH
rH
S
1.56
1.0
0.75
1.19
0.86
1.20
1.03
1.0
Beckman
400
2.08
1.0
0.52
0.65
0.71
0.52
0.25
0.68
Beckman
6800
1.67
1.0
0.70
0.97
0.93
0.87
0.87
0.89
3o
-O O
C CM
0) 00
PQ
2.17
1.0
0.70
0.70
1.30
0.85
0.43
0.91
•H tH
T3 O
C CM
CD oo
2.63
1.0
1.0
0.84
1.23
0.82
0
0.98
H
2.86
(1)
0.97
1.11
0.69
1.0
0.94
0.34
Notes :
(1) Use toluene ECN =1.0
(2) Notes from Table 14 apply.
(3) Method of Calculating Table 15 :
ECN of NMHC from Table 14
ECN of C0H0 for that Instrument
J O

-------
TABLE 16.  EFFECT OF METHANE ON ECN OF NMHC, APPLYING INITIAL CALIBATION OVER 14 DAY PERIOD

Propane With CH,
(C,HQ No CH.
jo 4
Ethylene With CH,
(C2H4) No CH4
Propylene With CH,
(C0H,) No CH.
36 4
Acetylene With CH,
(C2H2) No CH,
Toluene With CH,
(C?H8) No CH4
r-t
VI
s

0.64
0.64
0.48
0.73
0.76
0.64
0.55
0.87
0,77
Beckman
400
0.40
0.48
0.58
0.25
0.45
0.31
0.70
0.34
0.49
0.25
Beckman
6800
0.56
0.60
0.47
0.42
0.52
0.58
0.67
0.56
0.54
0.52
•H 0
13 0
0) 00
PQ
0.52
0.46
0.37
0.39
0.39
0.32
0.67
0.60
0.49
0.39
•H i-l
13 O
C CM
0) 00
0.54
0.38
0.37
0.38
0.35
0.32
0.50
0.47
0.31
H

	
0.42
0.34
0.46
0.39
0.51
0.24
0.50
0.35
Note: See notes on Table 14.

-------
TABLE 17. RAT]

Propane
(C3Hg)
Ethylene
Propylene
(C3H6)
Acetylene
Toluene
(C7Hg)
CN
1
T-H
iH
C/3
	
0.75
1.04
0.86
0.89
Beckman
400
1.20
0,43
0.69
0.49
0.51
ECN (NO CH.)
O OTi1
"U ^"^ T?P"M fT.TTT'U r*"U "\ '
CjvjJN 1 WXJ.ri Utl, )
4
Beckman
6800
1.07
0.89
1.12
0.84
0.96
Bendix
8200
0.88
1.05
0.82
0.90
0.80
"""O CD
(3 (N
01 OO
«
0.70
1.03
1.09
0.94
	
APPLYING INITIAL CALIBRATION OVER 14 DAY PERIOD
H
	
0.81
0.85
0.47
0.70
Note: ECN values presented in Table 16,

-------
TABLE 18.  EFFECTIVE CARBON NUMBERS OF NMHC WITH CH,  = 1.0 AS BASIS,  APPLYING DAILY CALIBRATION

Methane
(CH4)
Propane
(C3Hg)
Ethylene
(C2H4)
Propylene
(C3H6)
Acetylene
(C2H2)
Toluene
(C7Hg)
Acetaldehyde
(CH3CHO)
Mean of NMHC
Per Instrument
(ECN -ECN . )
max mxn
(ECN )
avg'
Per Instrument
CM
i
iH
iH

-------
TABLE 19.  RESULTS OF CORRECTING NMHC TO PROPANE (8), USING DAILY CALIBRATION





Methane
(CH )
4
Propane
(C3Hg)
Ethylene
(C~H, )
v 2 4'

Propylene
Acetylene
(C2H2)
Toluene
(C7Hg)
Acetaldehyde
(CH3CHO)
C6)
Mean NMHC ECN
(ECN -ECN . )
max mm
ECN
Avg
Per Instrument
CM
i

0
•H CO
O O M
s a. >
^ C/J ^


1.93


1.0
0.90


0.96

1.14

0.98

0.60

1.0



	
> en
•H T3 0)
-U CO -H
CO CU O
cu a. ai ex
pi C/} p* W


0.28


	
0.38


0.49

0.29

0.69

1.20

0.24



	

Notes:


(1) - (6) All notes from Table 18
apply.

(7) Use Toluene ECN = 1.

(8) Method of Correction
ECN of NMHC from Table 18
ECN of C0H0 for that Instrument
J 0
= ECN (corrected).













-------
                    TABLE 20.  EFFECT OF METHANE ON ECN OF NMHC, USING DAILY CALIBRATION
CD







Propane: with CH4
(C3Hg) no CH4
Ethylene: with CH4
(C2H4) no CH4
Propylene: with CH4
(C-H,) no CH,
jo 4
Acetylene: with CH4
(C2H2) no CH4
Toluene: with CH4
(C7Hg) no CH4
Acetaldehyde : with CH4
(CHoCHO) no CH4
CN
1
-i-l

*-^
CO
S
	
0.63
0.59
0.57
0.77
0.76
0.69
0.61
0.82
0.84
0.75
0.72

c
cfl
J 0
o o
0) 
-------
                                                  ECN (no CH.)
vo
' Ml»l ,IL 2.J.. IVAliU Ul' -p...-.-* / • I, PW '\ » UlJ-1-i*V-> WiliJ-lJ. >»/ II 1-l.UJ.m.J- -1-^1.1

Propane
(C3Hg)
Ethylene
(C2H9)
Propylene
(C3H6)
Acetylene
(C2H2)
Toluene
(CyH8)
Acetaldehyde
(CHLCHO)
rg
1
iH
TH

-------
from 0.42 for the Bendix 8201 to 0.70 for the MSA 11-2 with a mean value  of
0.53.  As a measure of the variance in NMHC response a relative spread defined
as the maximum ECN value minus the minimum ECN value divided by the average
ECN value is used.  A relative spread of 0.53 (53% of average) was seen in the
mean value of the NMHC response for these instruments.  By computing the  average
of the instrumental mean NMHC responses, an overall NMHC mean response of 0.53
was calculated.  Comparing this to the 0.44 value from Table 14, daily calibra-
tion resulted in a 20% increase in the ECN values.
     Considering the instruments individually, the MSA 11-2 had a uniformly
high NMHC response with a mean NMHC value of 0.70 and relative spread of  40%.
As was noted previously about 7.5 minutes was required for the toluene in this
instrument.  Response time was negligible for other species.  The two Beckman
Instruments, the Model 400 and 6800, gave ECN means of 0.52 and 0.57, respectively,
and slightly more narrow spreads than the MSA; compound-to-compound response of
the two instruments matched well.  Lower ECN values were exhibited by the
remaining three instruments, the Bendix 8200 and 8201 and the RTI prototype,
which responded approximately the same with mean ECN's of 0.48, 0.42, and 0.48,
respectively.  The spread in the NMHC response of the RTI model was the lowest
of any of the instruments.  It was quite interesting to note the similarity
between the two Beckman and two Bendix models since each of the two pair had
identical FID's.  The slightly higher response of the Beckman instruments may
be the result of the slight differences of detector design or to operating
conditions which the manufacturers suggest using - different flame conditions
producing different HC responses.  Time did not permit detailed study on the
instruments to determine the exact cause.
     Looking at the relative spreads on a per specie basis, toluene and acetalde-
hyde showed spreads greater than 100%.  The problem with acetaldehyde has been
discussed.  The large spread with toluene resulted from the unusually low
response of the Bendix 8201 and unusually high (compared to other instruments)
response of the MSA 11-2.  Propylene cause a similar response pattern although
not quite as severe as toluene.
     Methane is known to yield a higher response in FID's on a per-atom-carbon
basis than most NMHC's.  An effort to correct for NMHC response differences
was made on the ECN data of Table 18.  By noting that the mean value of the
NMHC for all the instruments was 0.53 and the NMHC species averages for all
the compounds (except CH.-.CHO) were close to this value, particularly propane
                                       80

-------
(0.54), propylene (0.51), and toluene (0.52), the possibility of correcting
the NMHC values presented itself.  The ECN values from Table 18 were
divided by the ECN of propane (since it is available as an NBS certified
SRM grade) per instrument.  The "corrected" ECN's are displayed in Table
19.  The mean NMHC ECN's have converged toward 1.0, as they should have,
and the relative spread of the mean NMHC ECN's is now 22% compared to the
53% for uncorrected ECN's.  The average of the NMHC means was exactly 1.0.
Note that the relative spreads per instrument and per specie have also
converged somewhat .
     The utility of using propane corrected response can be illustrated.
The "uncorrected" NMHC is given in ppm, with the response to CH,  = 1.0 as
basis, by

                            THC  - CH.  =  Z  (ECN )  •  C                     (11)
                                    4    n      n      n

 where ECN  is the  effective carbon number of the nth specie and  C   is
          n                                         —  r          n
 actual concentration of nth specie.   For  "correcting" with  the response
 to propane,  the ECN for propane,  determined  during calibration by using
 a propane injection, as well as methane,  would be used to divide into
 the difference obtained in  (11).   The maximum error would occur  if  the
 NMHC mixture consisted of a single component, the kth,  whose ECN showed
 greatest  deviation from propane.   Therefore,
                          Z  Cn x (ECN )  = 0  (n J> k)                      (12)
                          n           n
 and
                   C, ECN
                             =  "corrected"  NMHC                          (12)
                 ECN propane

 which would  give
                                  ECN  - ECN propane
                     %  Error    =  —^T	— x  100 .              (13)
                            max      ECN propane

 The  advantage  of this  correction  is  that CH,  calculation can continue as
 is currently prescribed  by the Federal Register.   The  only additional
 information  required is  the ECN value  for propane for  each particular

                                       81

-------
instrument, and reliable propane standards are available, Furthermore,
other NMHC ECN's than that of propane might be used if local situations
warranted.
     The possibility of methane affecting the NMHC response was examined.
In Table 20., the ECN's are compared with and without a methane background.
The ratio of ECN (no CH,):ECN (with CH^) is shown in Table 21, and no
definitive trends were perceived as most of the ECN averages were within
+10% of each other.
Ambient Air Measurements
     At the conclusion of the test phases on calibration stability and
non-methane hydrocarbon response the analyzers were operated at about
72 hours, from 1600 hours August 28, 1976 through 1600 hours August 31,
1976, sampling ambient air at the Research Triangle Institute in Research
Triangle Park, N. C.  Hourly average data for methane and total hydro-
carbons for the five commercial instruments are tabulated and presented
in Appendix B.
     The sampling arrangement for the ambient air analysis was as shown
schematically in Figure 26.  Air was inducted through an inverted funnel
located about 2.4 meters  (8 feet) above ground level.  Suction was from a
Metal Bellows Co. Model HB-151 pump through Teflon-^ tubing.  Pressurized
air from the pump, at a flow of about 20 liters per minute, was conducted
by 0.64 cm (0.25 inch) diameter PTFE tubing for a distance of about 9.1 m
(30 feet) prior to entry  into the building where the instruments were
housed.  From the entry point to the laboratory housing the instruments,
the tubing run was about  30.5 m (100 feet).  The sampling bulb manifold
was relieved to atmosphere in the laboratory through three unused ports,
which maintained the manifold pressure near one atmosphere.  Tubing runs
of 3.0 m  (10 feet) connected each instrument under evaluation to the
manifold.
     The RTI prototype was not used in this comparative evaluation because
it could not be used in any "automatic" unattended mode.
     All instruments spans were set just prior to this phase of the test,
using 8.0 ppm CH, in air.  In spite of this there is a discrepancy in
readings - even on methane - in the first hourly average.  Data on CH,
response has been plotted in Figure 27.  The curve at  the  top of
this figure is the average instrument reading (ensemble hourly average) ,
obtained from the responses of MSA 11-2, Beckman 6800,  Bendix 8200 .and
Bendix 8201.   The lower group of curves represent the ratio of individual
                                      82

-------
00
              SAMPLE
              FUNNEL
             PRESSURIZE
             SAMPLE TO
             BLDG.
   ///////    METAL BEL-
    GROUND   LOWS MOD.
             HB-151
         PUMP LOCATED
         UNDER PLATFORM
     DETAIL OF SAMPLE
     COLLECTION METHOD
                                V

                                            SAMPLE FUNNEL GIST *
                                                                 GUY WIRE

                                                                 PLATFORM


                                                                 TEFLON  "SMOG CHAMBER'
                    GUY WIRE
                                             x/
                                 RTI LABORATORY BUILDING #6
                         NMHC ANALYZERS
                         IN THIS AREA
                 APPROX.  9.1m (30 ft.)  of 0.64 cm
                 (0.25  in.)  TUBING EXTERNAL TO
                                 BUILDING
                                                   APPROX.  30.5m
                                                   (100  ft.)  OF
                                                   0.64  cm  (0.25  in.)
                                                   TUBING INTERNAL
                                                   TO BUILDING
BUILDING TEMPERATURE
20°C-21°C AT NIGHT
SAMPLING MANIFOLD
(RELIEVED TO ATMOSPHERE)
                             Figure 26.   Sampling arrangements for comparative  evaluation  using
                                         ambient air.

-------
      1.9
    e-  i
    rv  i. i
    •sT


   U  1.1

    Si
    cc
    c;
    G>
    ?
1.6
      1..5.
        30
          -.30
          -.10
               o
                                                      D
o o o  o o o o
000  O O O O
\J3 O ~3"  
       1.2
       1.1
       0.9
       0.8
                          BECK 6800
                                                    D  BECK 6800
                 BEN  8201
         MSA


        BEND IX-
         8200
                                                                  MSA 11-2

                                                                  BENDIX 8200
                                                   D
Figure 27.  Average  CH,  response showing ratios  of individual instrument

            readings to  average, variation  of  average and relative  spread.

                                        84

-------
instrument hourly average responses to the ensemble average plotted versus
time.  (The ensemble average ratio is taken as 1.0.)
     Several interesting features occur in this data presentation.  The
first is the immediate discrepancy in CH, readings when these analyzers
were exposed to ambient air.  The spread is initially about +10%, increasing
to +15%, -20% at about 6 hours after the start of the test, remaining in
the order of +15% for about 24 hours and then gradually decreasing to within
about +5% after about 65 hours exposure.  The second feature is that two
of the analyzers, Beckman 6800 and Bendix 8201, consistently gave higher
than average readings while the other two analyzers, MSA 11-2 and Bendix
8200 consistently gave lower than average readings.  A third feature is
that the spread in readings did not appear to be strongly dependent on the
ensemble average CH, reading.  The curve in the middle of Figure 27 shows
the relative spread, maximum reading minus minimum reading divided by the
ensemble average.
     Data on total hydrocarbon response is plotted in Figure 28.  The middle
curve shows the ensemble average THC reading.  This was obtained by using
hourly averages from all five analyzers, or however many were "on line" at
a particular hour.  The lower group of curves represent the ratio of
instrument hourly average THC readings to the ensemble average THC value
plotted versus time.  The ensemble average ratio is taken as 1.0.   Also shown
is the relative spread in THC readings.
     As was the case with the CH, response on at least two of the analyzers,
the Beckman 6800 and the Bendix 8201, there appears to be a larger deviation
from average during the first 8 to 12 hours after exposure to ambient air
then later when such exposure has become "routine".   The Beckman 6800 THC
response, initially some 23% below average gradually rose to about 5 to 6%
below average whereas the Bendix 8201 remained consistently at about 15%
below average.  Both the Bendix 8200 and Beckman 400 averages also fluctuated
with respect to the ensemble average, but were usually closer.  The Bendix
8200 average started at about 8% below ensemble average and rose to within
+5% for about 24 hours before drifting down to fluctuations around the -10%
level for the remainder of the test period.  The Beckman 400 average initially
                                       85

-------
  s
  P.
  p.
  w
g
u
H
H
     0,2
      2.2
     2.1
      2.0
     1.8
      1.3
      1.2
      1.1
     1.0
  8  0.9
     0.8
                   OOOO  OOO OO  OOO
                   OOOO  OOO OO  OOO
                   vO O  -3" "^  00 CN ^D O "-^  *vj" OO CN
                   iH CN  eJ^ O  O H rH CN d^ O O iH
                                               O  O O  O O O O
                                               o  o o  c o o o
                                               v^  f~^ sj1  "^ 00 CM ^O
                                               I—I  CN (^j  O O H rH
                   8/28
                            8/29
8/30
8/31
                                                                  MSA
               8200
                 6800
                                                8201
Figure  28.  Average THC response showing ratios  of individual  instrument
             readings to average, variation of  average and relative spread.

                                     86

-------
climbed to about 15% higher than the ensemble THC average.  Then  it dropped
to about 10% below ensemble average before returning closer  (5-10%) of  the
ensemble average over the last 40 hours of the test period.  The MSA response
was consistently higher than the ensemble average.
     The consequence of Beckman 6800 and Bendix 8201 consistently having
higher than average CH, readings and lower than average THC readings was
either negative or very small positive NMHC values.  This left only two
instruments for which NMHC readings could be compared, the MSA 11-2 and
Bendix 8200.  Figure 29 gives a plot of three NMHC related (?) values versus
time.  The first is a fictitious average reading, labeled A Avg., obtained
by subtracting the ensemble average CH. from ensemble average THC readings.
The second is the MSA 11-2 NMHC reading and the third is the Bendix 8200
NMHC reading.  The latter is usually in fair agreement with the A Avg.,
whereas the MSA 11-2 reading was at least twice the A Avg. or Bendix 8200
NMHC value.
        Several other forms were used to plot the data in order to see if
features could be found which would give some indication as to the cause
of discrepancies.
     Because the Bendix 8200 had exhibited relatively stable performance
during the laboratory phase and the NMHC readings derived from its (THC-CH.)
readings were in reasonable conformity with the (Average THC - Average CH,)
values (see Figure 29), the CH, data and THC data were replotted with respect
to the Bendix 8200 as a reference.  Plots for CH.  are shown in Figure 30,
and plots for THC are shown in Figure 31.  The spread in CH.  readings (lower
plot of Figure 30) still exhibits a strong downward trend over the test
period with a series of peaks, labeled A, B, C, D.  The Beckman 6800 and
Bendix 8201 readings converged toward those of the Bendix 8200, while the
MSA 11-2 CH, reading fluctuated back and forth with respect to the 8200
readings.
     The spread in THC readings with respect to the Bendix 8200 readings
shown in the lower curve of Figure 31, shows a trend of decline from 35%
to 45% at the start of the test to about 25% to 35% at the end of the test.
                                       87

-------
  0
      1.2
      1.1
      1.0
      0.9
      0.8
  o   0.7
      0.6
      0.5
      0.4
      0.3
      0.2
      0.1
                        MSA 11-2
                    AVG
                                      A AVG
                                                                  BENDIX
                                                                   8200
                   OOOOO  OOO OOO OOOOO  OO  O
                   ooooo  ooo ooo ooooo  ooo
                   L t-\ *^""i --*• -^- ^v-i  ^»o i rt  /•"N > -*• --*• /Vi ^N. I . rs ^^ ^ -*• . -*•  r^*\ /^ I  i r-\
                 8/28
                   H CN rs O O
                         *
8/29
                                          OOrH
                                                    cN
8/30
                                                         O  O rH  H
8/31
                                      TIME
Figure  29.  Average NMHC  response and relation of MSA  11-2 and Bendix 8200
             readings to average.
                                     88

-------
     1.3 (.
     1.2  -
     1.1
     1.0
     0.9
     0.8
                                        BECK
                                                          BECK
                                                                      = 1.0
O 0
o o
vO O
rH CN

0
0

cJ^.
1
0 O
o o
£>
rH rH

1 1
0 0
0 O
0 -*

1
0 0
0 0

-------
         1.4
         1.3
         1.2
       o
       PX
       en
       01
         1.0
         0.9
         0.8
                    ilSA
                   BECK
                   400
BECK
6800
                                                                  MSA
                                               BECK 400
                   BECK 6800

                   BENDIX
                    8201
       BENDIX
        8201
                      o o CD o o o  oooociooocicDcioo
                      ooooooooooooooooooo
                      rHcNCNOOtHtHCNCNCDOiHi-HCNCNCDCDi—liH
                          -H*-
                    8/28
             8/29
8/30
8/31
         0.5
         0-4
         0.3
         0.2
Figure 31.  Ratios  of individual instrument THC readings  to Bendix 8200 reading
            and  relative spread.
                                        90

-------
Inspection of the ratios of individual instrument THC readings to those of
the Bendix 8200, the upper group of plots in Figure 31, shows similar trends
to those noted previously in conjunction with Figure 28 with one exception.
This is the MSA 11-2, whose trend with respect to the Bendix 8200 is relatively
flat.
     There did not appear to be any significant advantage in taking the Bendix
8200 CH, and THC readings as the "standard" for comparison.
Comparison of Laboratory and Ambient Air Tests
     The higher THC  readings of the MSA 11-2 are somewhat consistent with
the data of Table 18 which showed that the MSA 11-2 gave higher response
than the other analyzers to the NMHC species used in the laboratory evalu-
ation.  The consistently low ambient air THC readings of the Bendix 8201
are in agreement with the low average ECN obtained for this instrument in
the laboratory tests.  After the initial "acclimatization" period the Beckman
6800 THC readings rose within 18% to 20% of the MSA 11-2 readings (cf. Figure
31).  This is consistent with the 23% higher average ECN response of the
MSA 11-2 with respect to the Beckman 6800.
     In the laboratory tests the MSA 11-2 average ECN values were about
50% larger than those for the Bendix 8200, whereas the ambient air values
of NMHC for the MSA  11-2 were about twice that of the Bendix 8200.  This
was explainable in part by the lower CH, values of the MSA 11-2 which were
obtained over the first 32-36 hours of the ambient air tests.  However
during the later portion of the ambient air evaluation,  the CH, values of
the MSA 11-2 increased to close to or slightly greater than those of the
Bendix 8200  (of Figure 27).
     The most obvious discrepancy between laboratory and ambient air readings
occurred immediately after the instruments were placed on line sampling
ambient air.  This was the immediate spread in CH, readings of about 21%
relative to the average although the instruments had just been made to
agree using CH, to set their spans.  This anomaly was pointed out in the
discussion of Figure 27.  It appears to be related to atmospheric moisture.
     Data on meteorological conditions during the ambient air sampling
test phase are given in Table 22.  They are from official records of the
                                       91

-------
TABLE 22.   DATA ON METEOROLOGICAL CONDITIONS
           August 28,  1976 - August 31,  1976 from U.  S.  Weather Station,
           Raleigh-Durham Airport,  North Carolina

Date Time
8/28 1553
1753
1952
2153
2353
Precip.
8/29 0153
0353
0553
0753
0954
1153
1353
1553
1752
1955
2154
2353

T (°C)
29.4
27.8
26,1
24.4
23.9
%
Rel
Humid
55
67
70
79
79
0.01 in equiv. RW
23.3
22.2
21.7
24.4
28.3
31,7
32.2
32.2
31.1
26.1
24.4
24.4
Zero Precip. Haze
8/30 0154
0353
0553
0752
0956
1155
1352
1553
1752
22.8
21.1
20.0
20.5
22.2
23.3
25.0
25.5
23.9
85
88
91
79
70
54
47
47
48
69
82
48
0245-0745
44
53
57
55
46
37
27
23
24

°C
Dew Point
19.4
21.1
21.1
20.5
20.0
1406 - 1423
20.5
20.0
20.0
20.5
22.2
21.1
19.4
19.4
18.9
20.0
20.0
12.8

10.0
11.1
11.1
11.1
10.0
7.8
4.4
3.9
2.2
Station
mmHg
Pressure
752.3
752.1
752.1
752.3
752.7
& 2158 - 2205
752.6
752.1
752.3
753.0
753.1
752,1
751.1
750.1
750.3
751.0
751.7
752.6

752.9
753.1
754.1
755.0
756.3
755.8
755.3
754.8
754.4

Sky
Cover %
100
100
100
100
100

100
50
20
20
0
50
50
40
30
80
80
30

40
50
90
100
60
20
0
40
90
                                   92

-------
                TABLE 22  (Continued)

Date
8/30



8/31












Time
1953
2154
2353
Zero
0153
0353
0553
0753
0953
1153
1353
1554
1753
1956
2153
2352

T (°C)
18.3
13.3
14.4
Precip. No
13.9
12.2
9.4
17.2
23.3
26.1
26.6
27.2
25.6
18.3
17.8
15.0
%
Rel
Humid
38
66
51
Haze.
53
66
77
58
36
30
31
29
32




°C
Dew Point
3.9

4.4

4.4
6.1
5.6
8.9
7.2
7.2
8.3
7.8
7.8
10.0
9.4
9.4
Station
mmHg
Pressure
755.0
755.1
755.0

755.4
755.5
756.3
756.8
756.9
756.3
755.4
754.9
754.5




Sky
Cover %
50
30
0

0
0
0
20
50
100
100
40
30



Zero Precip.  No Haze.
                        93

-------
U. S. Weather Service Station at Raleigh-Durham Airport  (RDU), which is
located some six to seven miles from the instrument test site on the
Research Triangle Institute campus.  For several days preceding the ambient
air sampling phase, the weather was typically hot with high humidity and
low circulation, characteristic of late summer conditions in  this area.
Because of the low circulation it appeared to be opportune to obtain
higher than normal hydrocarbon readings.  However, the day after the ambient
air sampling started, circulation increased, bringing unseasonably cool,
dry air into the area.  This circumstance fortuitously provided a change
in test conditions which appears to be significant.
     Air temperature  (I) and dew point (II) data for the period 1600 August 28
to 1600 August 31, 1976 are plotted versus time in Figure 32.  Also indicated
on the plot are the occurrence (at the RDU Weather Station) of two brief
rain showers, one about two hours prior to the start of the test and one
that occurred at about 2200 on August 28.  In addition the weather records
noted hazy conditions during the morning hours 0245-0745 on August 29, 1976.
Otherwise there was no other precipitation noted.
     Plotted in Figure 32 along with the weather data is the spread in
analyzer CH,  (III) readings taken from Figure 27.  The relationship between
the spread in readings and the dew point is obvious.  As the dew point
decreases the spread in readings decreases.  There is additional modulation
which gives peaks A, B, C, and D in the "CH, spread" curve.  Each of
these peaks appears when the air temperature is decreasing, a period typically
accompanied by the condensation of atmospheric water vapor.  The decreasing
air temperature at A' on the upper curve, during the night of August 28
was accompanied by a brief rainshower (at least at the RDU Weather Station)
and appears to be directly related to peak A on the "CH, spread" curve.
     The rising air temperature during the period 0600 to 1200 on August 29
is reflected inversely in the "CH, spread" curve.  As the weather pattern
began to change in the late afternoon and early evening of August 29, the
dew point dropped considerably, aborting the build-up of peak B.  However
after passage of the frontal system, the decline in air temperature over
portions B' and C' of the upper curve (Figure 32) apparently  fostered
sufficient condensation to result in peak C of the "CH, spread" curve.
     As the air temperature rapidly decreased 10.6°C (19°F) over four hours
(portion D1 of the upper curve, Figure 32) during the late afternoon and
                                      94

-------
       90
       80
       70
q,"   60
t-i
s

n)

tt   50
H

n   40
   60
       30
              AIR  TEMP
               DEW POINT
          -.30
              w
        -.20
          -.10
              u
      1.2
      1.1
   2
   ri
      0.8
                                                                 III
                    <3 O O O  O CD CT  O O O  O O C O O O O O O
                    ooooooo  oooooo ooooco
                      O*3"   O-JOOiHrH (NCvlCOr-H rH
                 8/28/76 '     8/29/76   '     8/30/76   '  8/31/76

                 *Light rain  1406-1423 and  2158-2205 8/28/76

                 +-H-*- Haze
                         BECK 6800
                                                    D  BECK  6800
                 BEN 8201
            MSA


           BEND IX-
            8200
                                                                       30
                                                                          25
                                                                          •H
                                                                          O
                                                                             o
                                                                          15  .
                                                                             01
                                                                             M
                                                                             D

                                                                          10 2
                                                                  MSA 11-2

                                                                  BENDIX 8200
Figure  32.   Plot of meteorological data  showing relationship to relative spread

             •a£ 6H^ Beading.

                                      95

-------
early evening of August 30, another peak of "CH, spread" occurred, indicated
as D on the Figure.  Correspondingly, as the air temperature rapidly increased
the next day, August 31, the "CH, spread" continued to decline until 1200
when the rate of change of temperature decreased as it approached the daily
maximum.
     To recapitulate the evidence of Figure 32, it appears that there is
a direct correlation between atmospheric moisture content and the discrepancy
of CH/ readings ("CH, spread") of the analyzers evaluated.  The operative
phenomena appears to be the condensation of atmospheric moisture, i.e., a
combination of dew point (moisture content) and either rising or declining
temperatures.
     Although the air temperature came near the dew point in only one case,
during the night of August 28-29, it is highly likely that the approach was
much closer in the sampling tubing than in the ambient air.   This is because
the approximately 30.5 m (100 feet) length of tubing inside the laboratory
building was at a considerably lower temperature (down to 20°C to 21°C) than
air temperature due to building air conditioning.   Some condensed moisture
was noted in the line on August 29, although there  was no evidence of condensed
moisture in the sampling manifold.
     Comparison of individual analyzer CH, readings ratioed to the average,
as in Figure 32, or to the Bendix 8200, as in Figure 30, indicates that
the MSA 11-2 exhibited the largest deviations during the periods qf rapidly
decreasing air temperature when moisture condensation was apparently occurring.
The Beckman 6800 and Bendix 8201 CH, readings appeared to be affected more
by the general dew point trend, as did the Bendix 8200, with all showing
marked fluctuations during the early evening of August 29, when the weather
pattern was drastically changing.
     Comparison of individual analyzer THC readings (see Figure 28) to the
air temperature and dew point data show decreasing deviations from average
for the MSA 11-2, which consistently read higher THC than average, and the
Beckman 6800, which consistently read lower THC than average.  Both of
these trends match the dew point trend, i.e. the lower the dew point the
smaller the deviation from average.  There did not appear to be a strong
correlation with rapidly decreasing or increasing air temperature such as
that noted with the CH, readings.
                                      96

-------
     A similar, but inverted, trend is detectable in the THC readings,
Figure 28, where the ratio of the MSA 11-2 and Bendix 8200 THC to ensemble
average THC tended to decrease with decreasing dew point whereas the
Beckman 6800 THC ratio to ensemble average THC tended to increase with
decreasing dew point.  The Bendix 8201 THC did not appear to follow this
trend.  None of the analyzers tended to show the pronounced peaks that
occurred with the CH, readings.
     Granted this, it remains to be determined how condensable moisture
affects analyzers in two different ways.  As was pointed out in connection
with Figure 27, two analyzers,  the MSA 11-2 and Bendix 8200, consistently
read lower CH, than average and two, the Beckman 6800 and Bendix 8201,
consistently read higher CH, than average.  In comparing these instruments
there appear to be more similarity than differences in design for those
which behave differently with respect to atmospheric moisture, and more
difference than similarity in design for those which behave the same
way with respect to atmospheric moisture.  There appears to be only one
seemingly insignificant item which the "low readers" shared and which
the "high readers" lacked.  This was an in-line particulate filter.  The
MSA 11-2 and Bendtx 8200 have built-in filters.  The other three instruments
do not.
     The condensation of water vapor on a surface provides a liquid film
which is capable of taking a variety of gases into solution.  The same
can be said for a droplet of water in the air.  The solubility of hydro-
carbons  in water  is low, depending on relative polarity and molecular
size, and it is well known that methane is soluble only to a limited extent
in water.  However, one hypothesis is that as atmospheric water vapor
condenses it absorbs CH,, among other gases.  As an ambient air sample is
drawn into an analyzer the steam will contain water vapor entrainment which
tends to condense on the walls of the sample lines.  If there is a particu-
late filter of high specific surface area in the gas stream, condensation
can coat the filter fibers, providing a liquid film, depleting the gas stream.
Alternatively, if there is no filter the condensation and absorption takes
place on sample line walls.  Thus an aliquot of sample in a, say 3 ml
volume is enriched by additional gas as the sample is drawn into the heated
                                       97

-------
sample loop.  However, the opposite effect of atmospheric mositure on THC
response, compared to CH, response, makes this hypothesis untenable.
     Another hypothesis is that somehow the presence or absence of water
vapor changes the flame dynamics.  However, for four of the five instruments
compared which could measure methane, the measurement includes stripping
on an adsorbing column, which would be expected to remove the water vapor.
In the MSA 11-2, of course, a catalytic oxidation step is used for NMHC
stripping.
     A third hypothesis is that the presence of condensation on the filter
increases resistance to flow and decreases the sample flow rate (in the
MSA 11-2) or sample pressure (in the Bendix 8200).
     At this point it appears that none of these hypotheses are tenable and
that further study must be undertaken to elucidate  the mechanism relating
response variations with atmospheric water.
CONCLUSIONS
     The comparative evaluation has yielded data which support the following
conclusions:
     1.  Zero drift of these instruments is apparently not a significant
         problem.
     2.  Span drift in excess of the EPA performance guideline 24-hour
         specification of 5% can be a problem, and expectation of calibration
         stability longer than 24 hours is unrealistic.
     3.  Precision is within EPA specifications when span drift is taken
         into account.
     4.  There are significant differences in response to non-methane
         hydrocarbons, both on an intra-instrument and inter-instrument
         basis.
     5.  The NMHC response differences are ameliorated by using propane
         response as a basis for correcting effective carbon number.
     6.  The field analyzers are of complicated design and construction
         and are subject to a variety of maintenance problems.  As
         expected, one of the components most susceptible to malfunction
         is the automatic switching valve.
     7.  Materials of construction of gas lines and values as well as
         filters and columns can have a significant affect on some of the
         heavier, or more reactive hydrocarbon species.

                                      98

-------
     8.   Instruments made to agree in the laboratory through calibration
         with dry CH,/air mixtures, immediately start to display differences
         in response when exposed to the atmosphere.
     9.   Variations in instrument response to both CH,  and THC are
         apparently related to atmospheric moisture content.  The
         mechanisms which cause this relationship are not understood at
         present.
     Using the information developed in this and prior sections of this
report some conclusions can be drawn about the sources of error which mili-
tate against measurement accuracy with the present Federal Reference Method.
These are presented in the next section.
                                       99

-------
100

-------
                                  SECTION 8
                    SOURCES OF ERROR WITH PRESENT METHOD

     The information and data developed in previous sections of this report,
combined with that obtained from prior studies  (3, 4), provide the basis
for discussion of the sources of error in the measurement process of the
present Federal Reference Method for the determination of non-methane
hydrocarbon concentration in ambient air.  For  convenience this analysis
is divided into four categories which correspond to the interacting components
of the measurement procedure:  instruments, reagents, procedure and operators.
INSTRUMENTS
     Presently available commercial analyzers designed for measurement of
non-methane hydrocarbons in ambient air via a subtractive technique (i.e.,
total hydrocarbon "reading" minus methane "reading") are subject to several
problems which contribute to errors in ambient  air NMHC measurement.  These
will be discussed in the categories:  calibration stability, NMHC response,
signal processing, and instrument specifications.
Calibration Stability
     Electronic zero (baseline) for the instruments evaluated in this study
was relatively stable and within EPA performance guideline specifications.
However, span drift of greater than + 5% per 24 hours was experienced for
significant fractions of operating time.  Precision for repeated injections
over relatively short periods of time (typically 3 to 4 hours) was generally
very good, well within EPA specifications.  This performance indicates that
the automatic baseline correction feature of the commercial analyzers works
well, but the FID ionization process and/or the electronic gain of the
signal amplifiers is not as stable as desired.
     In Section 6 when the response of manufacturers to the question about
selection of H2 and carrier air settings was summarized, it was apparent that
few (if any) have carried out detailed studies  of the effects of flow varia-
tions on the response of their FID designs.   One notable exception is for the
Beckman 400, the operating manual for which has data showing the effect of
                                      101

-------
H  and carrier air flows (actually given in terms of supply pressures)  on
methane response.  The common method of obtaining constant flow  control is  to
use a pressure regulator in conjunction with a section of capillary  tubing  main-
tained at essentially constant temperature or to use a porous resistor.   Instruc-
tion manuals give a range of pressure settings for a given gas stream flow;
although in some cases specific pressure values are given for the  individual
instruments.  Alternatively a range of flows may be given, to be measured
with a flow meter (such as a "bubble" flow meter),  Once these settings  are
made, presumably they are fixed until some event such as a major overhaul,
de-activation, storage and re-activation, or some similar occurence  requires
a set-up "from scratch".
     As shown in Section 4, the response of an FID to methane (and presumably
other hydrocarbons) depends upon the H./carrier flow ratio.  If  H£ and carrier
air flows are not set in the vincinity of maximum response - which is a  relatively
broad peak for each, the response may vary significantly with variations  in
the flow of either gas stream.  Therefore, it appears that when  FID  response
is not peaked with respect to H« and carrier gas flows, stable control of
these flows becomes an essential factor in maintaining stable instrument
response.
     For those cases where flow control is based on the attainment of critical
flow conditions—i.e., the attainment of sonic velocity (32) in  a capillary
tube—the gas mass flow rate is given by (33)

              Qmax = C0A2P1 [ff (^l) ?=T ] ^ X ^'^^ams per second  (14)

where (]„ is a dimensionless discharge coefficient,  A~ is the cross sectional
area of the capillary, P..  the upstream pressure, g the acceleration  due  to
gravity, M the gas molecular weight, R the gas constant, T the gas absolute
temperature, and y the ratio of specific heat at constant pressure to specific
heat at constant volume for the gas.  For a given gas and capillary  size, the
flow rate is directly proportional to the upstream pressure, PI , and inversely
proportional to the square root of temperature.   Therefore, to a good approxi-
mation, the fractional change in flow rate is directly proportional  to the
fractional change in upstream pressure and is directly proportional  to one-half
the fractional change in gas stream temperature.  Therefore, if R represents
                                     102

-------
FID response to a given species—e.g., methane—since
then
            /.   v     /          \       /      \      /
                                           )R    \ „    I AP
    (\     /          \
^H / \ V^~~ ^/H
  n0/   /.            * H,
       AR-ISr   10.  (;?--^l   +  I f5    IQ          2
                                    _    ^  Air'      x          ' Air

where the partial derivatives with respect to H2 and carrier air flows repre-
sent the sensitivities of response to these factors, which, of course depends
upon the operating point values of the variables.
     When critical flow conditions do not occur, the flow depends upon pressure
drop.  This situation usually occurs when the sample is being inducted into
sample loops or is being routed through the stripper column or through the
tubing leading to the FID.  In these cases the flow rate is given by

                  Q = ^  (Prp2)  L1 + ^p^"J                         (17)

where G is the conductance of the flow path (34).  Condutance depends upon the
flow geometry and the viscosity of the gas, which depends on the temperature.
As with the case of the critical orifice, it is essential that both the gas
temperature and the pressure be held constant if the flow is to be constant.
     With present analyzer designs there is no provision for direct flow
measurement with feedback control to maintain constant flow.  The Beckman
6800 and Bendix 8200 both have rotameters for measurement of sample, carrier
air and support air flow rates.  Hydrogen flow is not monitored.   While the
accuracy of rotameters leaves a lot to be desired, they are useful as flow
indicators and should be routinely checked to assure that flow rates are
stable.  The present Federal Reference Method, as well as equivalent procedures,
specify an overall calibration stability without specifying the component
drifts which influence this stability.  There is no need to specify the
latter if the overall stability is as specified.  However, at present there
does not appear to be any simple, short method to check stability.
     In addition to electronic factors and flow control factors in response
stability, there are other effects which influence flame kinetics.  Atmospheric
pressure variations were found to have a small effect on the CH,  response of

                                      103

-------
the RTI-designed FID, ranging about 2% for the normal range of atmospheric
pressures (cf. Figure 12).  This is probably representative for the instruments
now available.  A more serious problem is that of atmospheric water vapor.  A
change from 10°C to 30°C dew point gave about 5% increase in CH, response of
the RTI GC-FID design (cf. Figure 23) and atmospheric moisture variations
apparently significantly influenced commercial analyzer response, as was
discussed in the previous section of this report.  Any recommended measurement
method should take such possible effects into account.
Non-Methane Hydrocarbon Response
     The laboratory evaluation reported in Section 7 demonstrated that there are
significant differences in response of the commercial analyzers to non-methane
hydrocarbons, even when propane response is taken as the basis for comparison
rather than methane.  If the NMHC species distribution remained relatively
constant, a weighted average response factor could be used and agreement main-
tained.  Since the species distribution does not remain constant,  however, large
discrepancies occur.  One alternative approach would be to convert all of the
NMHC to methane and measure the converted CH,  to obtain directly NMHC as equivalent
CH,.  In view of the experiences reported in Section 7, it would appear that if
such alternative approaches are to be evaluated, they should be done so with
realistic synthetic gas mixtures which include water vapor and COp.
     Work has recently been reported (35, 36)  in which the effect of sample flow
rate and fuel composition on FID response to NMHC relative to that for CH, was
studied.  Reschke (35) undertook a study to improve instrument-to-instrument
correlation for Beckman 400 hydrocarbon analyzer instruments, studying the effect
on response to methane, ethane, propane, ethylene, propylene, benzene, toluene,
and acetylene in air of burner flow parameters, fuel type and composition.  Because
this work was directed to automobile emissions study the hydrocarbon concentrations
were very large, 150-200 ppm range, compared to those in ambient air.   As a
result of his investigation Reschke recommended that the sample flow rate be
set to a minimum value (about 5 cm /minute), that a mixed composition fuel be
used (40% H,, 60% He), that the fuel flow rate to the burner be set high (100-
      3    z
120 cm /minute) and that the support air flow rate be set high (four times the
fuel flow rate).  He points out the importance of setting the burner flow rates
of each instrument to the same value to achieve good correlation.
                                       104

-------
     Black and High (36) reported a similar study, comparing the effect
on NMHC analysis uniformity of various fuel mixtures, using exhaust gases
from 22 vehicles, and comparing their results to those obtained with a species-
by-species analysis by a laboratory gas chromatograph.  These authors also
concluded that a 40% H2> 60% He fuel mixture was the most universally applicable
for the THC analyzers, Beckman 400 and Beckman 108, used in their work.
     These studies indicates that the proper choice of fuel composition and
burner flow conditions will greatly improve uniformity of NMHC response and
instrument-to-instrument correlation for instruments of the same type.  Whether
or not a single fuel mixture composition can be found which will improve
agreement between FIDs of different design is a matter that can only be determined
by further research.  However, it should be pointed out that the achievement of
NMHC response uniformity at the higher concentration levels of automotive
sources is obtained in a tradeoff with sensitivity.  It is possible that the
reduced sensitivity might not be practicable for ambient air monitoring.   This
also is a subject for further research.
     The two references cited are just a small indication of the intensive work
that has been done to determine and correct the source of NMHC measurement
discrepancies in automotive emissions monitoring.   The reader is referred to
the large body of literature generated by the automotive emissions  measurement
specialists, reported chiefly through EPA,  the Society of Automotive Engineers
(SAE) and the Bureau of Mines.
Signal Processing
     The present method of using strip chart data  and graphical conversion
to obtain NMHC is highly inaccurate and  time consuming.  Although  strip chart
recorders probably contribute  the smallest errors  in  the measurement system
(cf. Appendix C), the graphical method used to extract data  from the charts
and  convert  to concentration  units is  subject  to error because  of  the
imprecision  of scale measurement and  the tedium which can  lead  to  errors  in
interpretation.  The difficulties in  quantitating  the difference in two
numbers of approximately the  same magnitude are well known.
     Because changes in column conditions  can cause peak shape  changes,  the
use of peak height rather than peak area as a measure of concentration is
subject to error.  The availability of high stability, accurate analog
                                       105

-------
integrators makes the measurement of peak area routinely achievable.   Further-
more, the availability of relatively inexpensive analog-to-digital  (A/D)
converters and digital arithmetic circuitry offers an economical  as well  as
reliable method of converting analog signals from the FID  to digital  form
for arithmetic manipulation.  The resulting digital signals can be used to
drive printing data loggers or be transmitted by line to a central data
collection system or be stored on magnetic tape for later processing.  If it
is desired to use peak shape as a diagnostic tool to check on proper  column
operation, an auxiliary analog output can be provided with a signal obtained
prior to the A/D converter for presentation on a strip chart or memory trace
cathode ray oscilloscope.  Digital printout has advantages of accuracy.
speed and presentation of the data in the format in which it is normally
reported.
     Any recommended standard practice endorsed or promulgated by EPA should
include these signal processing improvements.
Instrument Specifications
     As with most items instrument specifications are subject to caveat emptor.
Instruments should be set up in strict accordance with the instruction manual
and tested to assure conformity with specifications.  Generally any serious
problems will be apparent within a short time.   An equally important  practice
is to periodically check to make sure that specifications are still being
met.
REAGENTS
     Problems with gas mixtures used to calibrate air monitoring instruments
are not confined to those for hydrocarbon analysis alone.  A recent symposium
(37) has dealt with various aspects of this problem.  In regard to so-called
zero air it was pointed out (38) that this could contain varying, and unspeci-
fied, amounts of water vapor and C0«.  Additionally the analysis provided by
the gas supplier may be faulty.   As an example, it has been reported  (38)
that in zero gas certified by suppliers as having 0.1 ppm max THC as methane,
four out of six cylinders contained significantly more, with one having greater
than 0.5 ppm THC content.
     Analysis error can be significant for span gas mixtures as well.   Such
analyses are usually carried out on commercial THC instruments which  are
calibrated with NBS propane SRM's (since a methane SRM is no longer available
from NBS).   As has been demonstrated abundantly in this report, FID response
                                     106

-------
depends upon the HC species and the operating conditions.  In addition  to
this source of error, the use of contaminated gas lines and regulators  can
contribute a highly variable background.  Scott and Marrin (38) have
reported that uncleaned regulators can  give  (apparent) methane concentration
from 0,07 to 1 ppm.  One further problem  is  the stability of calibration
gas mixtures, particularly in the low ppm range.  This has recently been
discussed by Wechter and Greico (39).
Calibration Gases
     The errors in "zero" air and span  gas concentration determinations
introduce bias error into the measurements performed on instruments calibrated
with these gases.  In order to assess this contribution to the overall  error,
it will be assumed that the errors discussed fall on a normal distribution
curve when expressed as percent of full scale (40).  The "calibration
hierarchy" (in Dieck's terms  (40)) involves  three stages, which are shown
in Figtlfe 33.  The total uncertainty is given by
                                     - b + 2S
                                               (18)
where b is the NBS measurement bias and 2S  is  the quadruture sum of the
precision errors due to instrument variation that includes 92% of data.
This uncertainty provides an estimate of  the measurement error due to span
gas inaccuracy,
         CALIBRATION
          HIERARCHY
NBS


NBS SRM
Cal. Gas


Reference
Gas
COMPARISONS INVOLVED  ERROR PROPAGATION
                      NBS Primary Standard
                      vs. NBS SRM Cal. Gas.
                      NBS SRM Cal. Gas vs.
                      Supplier Reference Gas
                      Supplier Reference
                      Gas vs. Instrument
                                        True Value
         Figure 33.  Propagation of error in a calibration hierarchy.

-------
     The effect of THC content of zero air is to shift the calibration as
                                                                      *
shown in Figure 34.   The relationship of the apparent concentration, C ,
of an unknown determined by using the span gas of "certified concentration"
                  *
(i.e., apparent) C  and a zero air containing THC at a concentration Cin,
                  S                                                   -LU
is
                                         10
or
                                   (1 - C10/Cu)
                      Cu(1 + Eu> = (1 - c|°/C )  Cs & + £s>
                                         J.U  S
where e  is the error in the measurement of the true concentration of the
       u
unknown, C  and e  is the error in the span determination, given by
          u      s
equatidtl (18).  It should be pointed out that this portrayal of calibration
shift does not take into account any differences in response to different
NMHC components in the zero and span gas.  Such differences add to the
error when readings from different instruments are compared.  From equation
(20) it can be seen that the error for a single measurement will depend
on the residual THC in the zero air, the error in analysis of the gas
concentration used to span the instrument, and the relative location of
the measured concentration on the scale.  For small concentrations of the
unknown the error is dominated by the zero gas THC content.  For concen-
trations at or above the concentration used to set the instrument span, the
dominant error source is that due to the span gas analysis error.  This
is for a single determination.
     In the strip-subtract method two measurements are made and the
difference used to determine NMHC, or

                CfHC - CTHC - 4= (CTHC - CCH > U + "^ '         W)
                                  4             4
The error term is the combined error from the two separate measurements.  If the
span of both the CH, and THC channels have been set using the same calibration
gas source the analysis of which is traceable through the hierarchy shown
in Figure 33, they will both have the same bias due to the NBS measurement
                                      108

-------
            Maximum  Scale  (Range)
             Span  Setting
-5
n>
00
n
Cu
    x  -•
              Scale Indication of
              Unknown
                                           Apparent Concentration
                                4-
                                             True Concentration
                                             *

                                            :s
                                                                         C

                                                                         C
    0  C
        10
 U
x c*
s Ls
               Figure  34.  Relation  of  instrument  scale,  calibrated  on the basis of apparent
                           concentrations  for  zero gas  and  span gas,  to true hydrocarbon
                           concentration.

-------
bias in the transfer from the NBS primary standard  to  the  SRM mixture used
by the gas supplier to calibrate his instrument.  This bias  is about  1%
(38).  Using twice the mean square error of  the combined measurements the
error term will be
                   = b + 2   e    + e    + eCH  + £THC                   (22)
where e,Tt>c represents the NBS analysis precision, ec ITT> the suppliers  analysis
       NDO                                         aUr
precision, emur, the measurement precision of the total hydrocarbon channel.
            1 rl(j
If all of these are 1%, the total error could be as high as 5%.   If,  as  is
typical, the gas supplier's analysis precision is 2%, the total error could
be as high as 6.3%.
     Some gas suppliers use synthetic air mixtures of oxygen and  nitrogen
for zero air as well as background air with hydrocarbon blends.   As was
reported in conjunction with Figure 20, when the oxygen content of "air"
drops below about 20%, hydrocarbon response of the FID begins rapidly
decreasing.  Above 20% the change in response for a given percentage  increase
in oxygen is more gradual.  Thus it appears that 20% oxygen in synthetic
air should be a certification threshold.
     Another problem with calibration gases is that moisture content  is
usually (purposefully) very low, whereas ambient air usually has  appreciable
moisture content in many areas of the country.  Since stability of gas
mixtures has been questionable and water vapor content could be expected to
vary with ambient conditions, it seems more reasonable to specify a method
of approximating the humidity range expected for the operating instrument
by a controlled humidification conditioning of the calibrating gas.
Regulators and Connecting Lines
     As previously cited (38), the use of uncleaned regulators and supply
lines can provide a residual hydrocarbon background which can be  highly
variable.   In order to prevent such contamination, scrupulous cleanliness
is needed.  Impromptu swapping of lines and regulators should be  discouraged.
A thorough cleaning procedure for lines and regulator should be specified
and operators or other personnel responsible for analyzer installation and
maintenance should be trained to avoid ad hoc or slip-shod methods.   None
of these cautions are provided in the present method.
                                      110

-------
Operating Reagents
     The present Federal Reference Method  specifies  the purity  of  operating
reagents—for example, support air,  carrier air  or other gases, hydrogen
fuel—in terms of their total hydrocarbon  content.   As has been discussed
at length above, gas supplier certifications should  be at least spot
checked to assure confidence.  The effect  of various levels of hydrocarbons
in operating reagents was not checked during the present investigation, so
no quantitative statements can be made about effects.  However, good prac-
tice would be to check calibration (zero and span) each time an operating
reagent supply is changed and to recalibrate if necessary.
PROCEDURE
     The present Federal Reference Method  is limited to one instrumental
approach only.  A less restrictive specification that allows for alternative
instrumental methods has been reported (22) but has not yet been promulgated
for hydrocarbon analyzers.  Several  instruments which embody alternative
methods are currently commercially available or nearing availability.  These
are:
     1.  MSA 11-2 which uses continuous reading dual FIDs,
     2.  Hewlett Packard which uses  a backflush method to
         obtain NMHC directly,
     3.  Byron Instruments which converts all hydrocarbons to CH,.
     Of these only the MSA 11-2, which has been available for several years,
was evaluated along with several strip/subtract type analyzers (the Federal
Reference Method) in this study.  With this exception,  the performance of
these alternative methods via-a-vis  the strip/subtract method using ambient
air has not been reported.
Specification of Instrument Performance
     The specifications for NMHC analyzers published in the performance
guidelines (22)  appear to be more realistic than those in the present
Federal Reference Method.   A comparison is given in Table 23.   There are
no specifications in the Federal Reference Method on noise, interference
equivalents,  zero drift, span drift  or response times,   These could be
construed to be inherent in the accuracy requirement of 0.2 ppm (1% of full
scale range of 20 ppm).  The latter  is unrealistic, however, in terms of
                                     111

-------
 calibration  reagent  concentration  errors,  span drift,  and  other error
 sources  discussed previously,

  TABLE 23.   COMPARISON OF SPECIFICATIONS IN EPA PERFORMANCE GUIDELINES (2,2)
             WITH THOSE OF FEDERAL  REFERENCE METHOD (1)
Performance Parameter
Range
Noise
Lower Detectable Limit
Interference Eq.
Each Interferent
Total Interferent
Zero Drift, 12- and 24-Hour
Span Drift, 24-Hour
Lag Time, Rise Time, Fall Time
Precision
Accuracy
Units
ppm
ppm
ppm

ppm
ppm
ppm
percent
minutes
ppm
ppm
Performance
Guideline Value
0-5
0.05
0.1

+0.1
+0.2
+0.2
±5
10
0.3
No
specification
Federal Reference
Method Value
0-20
No specification
0.1

Np specification

Np specification
No specification
No specification
0.1
0.2
Specification of Reagents
     One of the most apparent lapses in calibration gas specification is
the failure to specify air as the gas containing CH. at the desired calibra-
tion levels.  As noted in the Scott report (3) the use of methane in other
than air for calibration yielded large measurement errors.
     A second problem is the use of CH, to set the span of both the methane
and total hydrocarbon channels with no consideration given for NMHC response
differences.  As shown in Section 7, there is much better agreement between
analyzers when propane is used as the basis for comparison of NHHC response.
     A third problem is the reliance upon gas supplier certification for
the CH, and THC values.  Deviations can lead to significant errors.  Some
method is needed for verification of supplier's analyses.  One method is
to cross-check using the "old" (nearly depleted) reagents and then the
"new" to see if any obvious inconsistencies arise.  Another method is to
                                       112

-------
participate in periodic "round-robin" tests using specially prepared gases
to check calibration.  A third is to conduct analyses on all  incoming  gases
with a reliable analytical grade gas chromatograph.  The latter is the most
rigorous, but also the most expensive, and for most monitoring sites probably
out of the question.
Calibration
     The calibration procedure of the Federal Reference Method is based on
the use of strip chart recorders for readout.  It should be re-written to
take into account the possible use of digital readout.
     A noticeable lack is the failure to specify how often calibration should
be checked and the permissible deviation beyond which re-calibration would
be required.
Maintenance
     Except for suggesting stripper column replacement for every 2 months of
operation, there are no maintenance cautions or requirements.   As was pointed
out in Section 7, malfunctions are a common occurrence, particularly with
the moving parts of the automatic analyzers.  Manufacturers generally supply
maintenance and trouble-shooting procedures in their instruction manuals.
This information can be used to formulate a routine preventive maintenance
schedule.  Such procedures, plus routine calibration checks, could be very
advantageous in increasing the yield of credible data.
OPERATORS
     Even though NMHC analyzers are for the most part "automatic" and
"continuous"  (or continual), reliable data collection depends heavily upon
the presence of trained, pragmatic operators and the exercise of supervision
over monitoring operations.
Training
     Ambient air analyzers are sophisticated instruments which rely upon
a variety of physical and chemical principles of operation.   From the
basic sensing principle through electronic and electromechanical components,
a wide variety of technology is used to accomplish the measurement.  If
these instruments are to be properly used to gather reliable  data, they
must be placed into operation and be kept operating by personnel who under-
stand their functionality and idiosyncrasies.  Since few organizations are
                                      113

-------
 References

 1.   40 CFR Part 50, Appendix E—Reference Method for Determination of
     Hydrocarbons Corrected for Methane.  Federal Register 36  (84):
     8198,  April 30, 1971.

 2.   Decker, C.  E.   Research Triangle Institute.  Private communication.

 3.   Reckner, L.  Survey of Users of the EPA Reference Method  for
     Measurement of Non-Methane Hydrocarbons in Ambient Air.    Final
     Report on Contract 68-02-1206.  EPA-650-75-008, Environmental
     Monitoring Series.  Scott Environmental Technology Plumsteadville,
     Pennsylvania 18949.  December, 1974.

 4.   McElroy, F. and V. Thompson.  Hydrocarbon Measurement Discrepancies
     Among Various Analyzers Using Flame lonization Detectors.
     EPA-600/4-75-010, Environmental Monitoring and Support Laboratory,
     Research Triangle Park, N. C.  27711.  September, 1975.
 5.   Bruderreck, H,, W. Schneider and I. Halasz.  Quantitative Gas
     Chromatographic Analysis of Hydrocarbons with Capillary Columns
     and Flame lonization Detector.  Analytical Chemistry 36  (3):
     461-473, 1964.
 6.   Bocek, P.,  J.  Nova'k and J. Janak.  A Flame lonization Detector for
     Work under Controlled Pressures.  Journal of Chromtographic Science
     8_ (4):  226-228, 1970.
 7.   Littlewood, A. B.  Gas Chromtography.  Academic Press, New York
     1970,  pp. 301-307.
 8.   Sternberg,  J.  C., W. S. Gallaway and D. T. L. Jones, Gas  Chromatography
     1961,  Lansing.  N. Brenner, J. Callen, and M. Weiss, eds.  Academic
     Press, New York, 1961, pp. 231-267.
 9.   McWilliam,  I.  G.  Linearity and Response Characteristics  of Flame
     lonization Detector.  Jour, of Chromatography 6:  110-117, 1961.

10.   McWilliam,  I.  G.  A Study of the Flame lonization Detector.  Jour.
     of Chromatography 51:  391-406, 1970.
11.   Gill,  J. M. and C. H. Hartmann.  Characteristics of lonization
     Detectors and Gas Chromatography Electrometers.  Jour, of Gas
     Chromatography 5 (12): 605, 1967.
12,   Lucero, D.  P.  and P. H. Smith.  Pressurized Hydrogen Flame lonization
     Detectors for Operation at Reduced Atmospheres.  Jour, of Chromato-
     graphic Science 10 (9):  544-549, 1972.
13.   Kauffman, H.,  0. Kapp, and H. Straub.  A New Form of Construction
     for a FID.   Chromtographia 5:  558-563, 1972.

14.   Former, 0.  F.  and D. J. Haase.  A Statistical Study of Gas
     Chromatographic Systems Employing Flame lonization Dectors.  Anal,
     Chim.  Acta.  48:  63-78, 1969.
15.   Grant, D. W. and A. Clarke.  A Systematic Study of the Quantitative
     Effect of Instrument Control on Analytical Precision in Flame
     lonization Gas Chromatography.  Analytical Chemistry 43  (14):  1951-
     1957,  1971.
                                        115

-------
16.   Cochran, W.  G.  and G. M.  Cox.  Experimental Design, 2nd Ed.  John
     Wiley and Sons, New York, 1957.
17.   Hartwell, T.   Research Triangle Institute, Statistics Research Division,
     Private communication.
18.   Valco Instruments Co.  Zero-volume, switching valves; Model #LV-8-HTA.

19.   Matheson Gas Products.  Pneumatic flow controllers; Model 8287.

20.   Supelco, Inc.   Carbosieve-B (60/80 Mesh) and 10% Carbowax 20M on
     Supelcoport (80/100 Mesh).
21.   Scott Research Laboratories, Inc.  All cylinder samples except
     propane (NBS).

22.   	Guidelines for Determining Performance Characteristics
     of Automated Methods for Measuring Nitrogen Dioxide and Hydrocarbons
     Corrected for Methane in Ambient Air.   EPA-65Q/4-74-018, Environmental
     Monitoring Series.  Quality Assurance and Environmental Monitoring
     Laboratory,  NERC, Research Triangle Park, N.  C.  27711, November,
     1974.
23,   David, D. J.   Gas Chromatographic Detectors,   John Wiley and Sons,
     New York, 1974, pp. 42-75.

24.   Lucero, D. P.   Water Vapor Sensitivity Response of Hydrogen Flame lonization
     Detectors.  Jour, of Chromatograph Science 10(7):   463-467,  1972.

25.   Bocek, P. and J,  Janak.   Flame lonization Detection.  Chromatographic
     Reviews 15:   111-150, 1971.

26.   Poll, A. A,   MSA.  Private communication.

27.   Szepesy, L.   Gas Chromatography.  CRC Press,  Cleveland, Ohio,  1970.

28.   McNair, H. M.  and E, J.  Bonelli.  Basic Gas Chromatography.   Varian
     Aerograph Publications,  Walnut Creek,  California,  1968, p.  152.
29.   Gill, J. M.   Varian Aerography Previews and Reviews.  Varian Aerography
     Publications,  Walnut Creek,  California, August, 1968.
30.   Novak, J,  Quantitative Analysis by Gas Chromatography.  Marcel Dekker,
     Inc., New York, 1975, p.  162.

31.   Ettre, L. S.   Practical Gas Chromatography.  Perkin Elmer Corporation,
     Norwalk, Connecticut, 1973,  pp. 2-22.

32.   Lee, J. F. and F. W. Sears.   Thermodynamics,   Addison-Wesley Pub. Co.
     Inc., Reading,  Massachusetts, 1977, pp. 242-247-

33.   Nelson, G. 0.   Controlled Test Atmospheres Principles and Techniques.
     Ann Arbor Science Publishers, Ann Arbor, Michigan, 1971, pp. 45-49.
34.   ibid., p. 43.

35.   Reschke, G.  D.   Optimization of a Flame lonization Detector for Determination
     of Hydrocarbon in Diluted Automotive Exhausts.  Paper 770141,  Proceedings of
     the International Automotive Engineering Congress  and Exposition, Cobo Hall,
     Detroit, February 28-March 4, 1977,

36.   Black, F. and L.  High.  Automotive Hydrocarbon Emission Patterns and the
     Measurement of Non-Methane Hydrocarbon Emission Rates.  Paper 770144, ibid.

37.   American Society for Testing and Materials.  Calibration in Air Monitoring,
     Proceedings of a Symposium held at the University  of Colorado, Boulder,
     Colorado.
                                        116

-------
38.   Scott, W. E. and J. T. Marrin.  Problems with Zero Gases for Ultra-Low
     Level Measurements of Air Pollutants,   pp.  275-281 in Ref.  37.

39.   Wechter, S.  G. and H, A.  Greico.  Gas  Standards,  How Standard are  They?
     pp.  246-254  in Ref. 37.

40.   Dieck, R. H.  Gas Turbine Emission Measurement Instrument Calibration.
     pp.  16-39 in Ref. 37.
                                        117

-------
118

-------
                                 APPENDIX A
                  FID AND GC STATISTICAL TEST PLANS AND RESULTS

     The flame ionization detector was evaluated for the effect of six
variables at five levels.  The variables and levels are shown below in
Table 24.  The test plan included 65 observations at various combinations
of levels that were based on a modified fractional factorial design.   The
test plan is given in Table 25.
VARIABLES

H2 Flow (cm3 /min) X±
Air Flow (cm3 /min) X2
Carrier Air Flow X.,
(cm /min)
Voltage (volts) X,
Pressure (torr) X,-
Sample Size X,
(ppm CH^)

-2
21
200
15

50
740
2
(0.2 ppm)

-1
28
267
25

100
750
4
(0.5 ppm)
LEVELS
0
35
333
35

150
760
6
(1 ppm)

1
42
400
45

200
770
8
(1.5 ppm)

2
49
467
55

250
780
10
(2 ppm)
     The analysis program for this test plan included provision for the
fact that test variables might not be precisely at the indicated levels.
This made the actual setting of variables for each observation easier and
facilitated the experiments.  The actual (coded) values used for each
observation are shown in Table 26 along with the response (last colum of
data) in coded units of FID current that resulted for each combination of
variables at the given level.
                                      119

-------
TABLE 25.   TEST PLAN FOR FID EVALUATION
Run
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
H? Flow
Xl
-1 (28)
1 (42)
I
-1
1
_i
-1
1
1
-1
-1
1
-1
1
1
-1
1
-1
_i
1
-1
1
1
-1
-I
1
1
-1
1
-1
Air Flow
x2
-1 (267)
-1
1 (400)
1
-1
-1
1
1
-1
_i
1
1
-1
-1
1
1
-1
-1
1
1
-1
-1
1
1
-1
-1
1
1
_i
-1
Carrier Air Flow
X3
-1 (25)
-1
-1
-1
1 (45)
1
1
1
-1
-1
_i
-1
1
1
1
1
-1
-1
-1
-1
1
1
1
1
-1
-1
-1
-1
1
1
Voltage
X4
-1 (100)
1 (200)
-1
1
-1
1
_i
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
-1
1
1
1
Pressure
X5
-1 (750)
-1
-1
T-l
-1
-1
-1
-1
1 (770)
1
1
1
1
1
1
1
-1
-1
-1
-1
-1
-1
-1
-1
1
1
1
1
1
1
	 r—i 	 1 	 1 • 1 ' p 	
Sample Size
X6
^ fr~ •" "~ i ~ ——
-I (4)
-I
-rl
-1
"-1
-1
H
-1
-1
-1
-1
-1
-1
-1
-1
-1
1 (8?
1
1
1
1
1
1
1
1
1
1
1
1
1
                   120

-------
TABLE 25. (Continued)
Run
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
H Flow
Xl
-1
1
0 (35)
0
0
0
0
0
0
0
0
-2 (21)
-2
2 (49)
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Air Flow Carrier Air Flow Voltage Pressure
X2 X3 X4 X5
-1
1
0 (333)
0
0
0
0
0
0
0
0
0
0
0
0
-2 (200)
-2
2 (467)
2
0
0
0
0
0
0
0
0
0
0
0
1
1
0 (35)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-2 (15)
-2
2 (55)
2
0
0
0
0
0
0
0
-1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-2
-2
2
2
0
0
0
1
1
(150) 0 (760)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
(50) 0
0
(250) 0
0
-2 (740)
-2
2 (780)
Sample Size
X6
1
1
0 (6)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
                                       121

-------
TABLE  25,  (Continued)
Run

61
62
63
64
65
H0 Flow
2X
Xl
0
0
0
0
0
Air Flow
x2
0
0
0
0
0
Carrier Air Flow
X3
0
0
0
0
0
Voltage,
X4
0
0
0
0
0
Pressure
X5
2
0
0
0
0
Sample Size
X6
0
-2 (2)
-2
2 (10)
2
                                      122

-------
TABLE  26,  STATISTICAL ANALYSIS RESULTS
ObS
l
3
4
5
6
7
ft
9
10
	 -it-
12
13
16
1?
18
19
PD
	 21
.._ 22
23
25
	 27
.28
29
30
31
32
...33
34
35
36
37
38
39
40
41
42
43
44
45
46
_-_47
48
49
51
52
"'54
Xl-._
-0,98571
1 .25714
1,05714 .
-1.50000
1.10000
-0,94286
-1.01429
Q. 98571
1.42857
-1.04286
-0.95714 ._
1,18571
-0.98571
.1.05714 	
-1.00000
0,9S71U
-0.95714
-0.98571
-1.02857 	 	
0.94286 	 _
0.9857J
-0.94286
-0.94286
0.98571
1.02857 	
-0.98571 	 _
-0.95714
-0.97143
-0.28571 	
-0.21429 	 _
-0,04286
-0.42857
-0.25714
0.41429
0,01429
0.01429
0.02857
-1.95714
-1.97143
L.97143
2.07143
-0.02857
-0.02857
0.00000
0.00000
0,05714
0.15714
0.11429
0.28571
0,071^3
X2
-0.99850
-0*98045
0.96541
1.00000
-0.-98647
-1.03308
1.03759
0.-957_8-9_
-1.07519
-1.00000
1.00000
1.00000
-1.02556
-1 ,00602
	 1.06466
1.04812
-0.99098
0.9b992
0.97444
. .-0.9879 /
	 -0.99248 	
1.01203
0.98797
-0.99248
_ 0.97143
	 1.02105
-0.99248
-1.00000
0.98947
1,01654
	 0.02256
... 	 0.06165
_r^l ^05263
-0.00752
-0.02256
-0.1ZL82
-0.01353
-0.00752
-0.00752
-0.00451
-0.00752
-0.00752
-0.01955
-2,00752
-1.98496
1.99098
1.97744
-0.03759
0.06617
-0.00752
-0.03759
-0.00752
.07669
X3
-1.00
-1.0-0
rl.OO
0,89
1.08
0,97
--1.00
-I. 01
-U02
1.06
1.05
-1.01
0.99
-1 .<">?
-1.01
-1.02
-1 .00
1,04
1.01
1 .01
0.98
-1.00
-0,99
-1.02
-1.00
i.Ol
0,94
1.00
1 ,00
0.10
-0.07
0,02
0.00
0.03
-0^0.6
0.00
0.01
0.01
0.01
0.02
0.02
0.00
0,00
0.01
-0.02
0.01
-1.96
-1.94
1.93
1.95
-0.07
0.00
X4
-1
1
-1
1
-L
1
-1
1
-1
1
1
-1
-1
1
1
-1
1
-1
1
-L
1
-1
1
-1
1
1
-1
1
	 0
0
	 0
0
0
0
0
0
0
0
0
... 0..
0
0
0
0
0
0
0
0
0
-2
-2
X5 -
-1.0
-1.0
-1.0
-i.o
-1.0
-1.0
-1.0
-1 .0
1.0
1.0
_ ... 1.0 _ .
i.o
1.0
I .0
- 1,0
i.o
-1 .0
-i.o
-1.0
-1 .0
-1.0 ...
-1.0
-1.0
1.0
-UO
i.o
1.0
I.o
1.0
1.0
1,Q
0.0 ...
0.0
0,0
0.0
0.0
0,0
0.0
0.0
.. 0.0 	
0.0
0.0
	 	 0.0 	 	 _
0.0
0.0
-0.8
-0,8
-0.8
-0.8
-0.8
-0.8
-0.8 	
-0.8
-0.8
X6_
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1 	
-1
-1
-1
-I 	 _
-1
1
1
1
I
1
1
1 ._..
. 1 .
1
1
1
1
.. 0 	
. 0 	
0
0
0
0
0
0
0
0
0
0 	
0
0
0
0
0
0
0
0
0 .
0
0
RES
17.8
53.2-
.65.5
62.8
120*0-
110.4
112.5
1-3-3 ,-2-
49,2
55,3
56.0
37.6
121.5
t2.3.Q
120.0
118.5
135.0
130.5
106.5
300,0
315.0
297.0
140.0
..98.0
153.0
305.0
300.0
296.0
297,0
177.0-
156.0
1-59..JQ
160,0
156.0
L50.0
151.5
151,5
15.1,0.
157.5
154.5
127.5_
127.5
150.0
165.0
148.5
150.0
34.0
30.3
271.5
271.5
165.0
166,5

-------
TABLE  26.  (Continued)
_Daa_...
56
58
59
t>0 _
ol
62
. -6J5 ..
64
XI ._..
-.085714
__._.2d285J -
.271429
-.200000
.357143
.357143
-.185714
-.885714
-.314286
X?
.0312030
-..QZ66SLL7-
.0827066
-.0075188 ...
__, 052.63 L6
-.0526316
-.0616541
-^0526-3JJx____
-.0075186
X3
-.01
-,01
.00
-.03
-.01
.00
.19
,IU
.35
_X4
2
_2_
0
0
0.
0
0
o
0
X5
-0.8
-0.8
-4,0
-4.0
2.0
2.0
0.0
0.0
0.0
X6
0
o
0
0
o
0
-2
2
2
.... _RES
169,5
166..5
166.5
172.5
160 5
162.0
19.0
_V53 Q
360,0
                                      124

-------
     The coded variables in Table 26 can be translated to measured laboratory
values by using the relations

                          H2 Flow (cm3 /min) = 7 Xj_ + 35                  (23)

                 Support Air Flow (cm3/min) = 66.5 X2 + 333.5            (24)

                 Carrier Air Flow (cm3/min) = 10 X3 + 35                 (25)

                Electrode Potential (Volts) = 50 X  + 150                (26)
                            Pressure (torr) = 10 X5 + 760               (27)

The sample size in ppm CH, was as shown in Table 24, ranging from 0.2 ppm
to 2 ppm.  The response values shown in Table 26 were related to the FID
                                                                    -14
signal current by multiplication by a conversion factor of 7.87 x 10    Ampere
     An example of conversion from the coded values of Table 26 to
experimental values is provided by considering observation number 23.
An FID signal current of 2.7 x 10    amperes resulted when 1.5 ppm methane
                         3
was injected (from a 3 cm  sample volume) into the FID operated at 750
torr pressure, with an electrode potential of 100 volts.  The hydrogen
                          3                                           3
fuel flow rate was 41.9 cm /min, the carrier air flow rate was 45.1 cm /min
                                          3
and the support air flow rate was 400.8 cm /min.
     The mean values of response for the variables at each level were as
shown in Figures 10 through 15 in Section 4.
     The coded variable values X.. through X, and RES for each of the 64
observations were used as inputs to a statistical analysis program REGR
of the STATISTICAL ANALYSIS SYSTEM authored  by A. J. Barr and J. H. Goodnight
of the N. C.  State University Institute of Statistics.  This program is
in the library of routines available at The Universities Computing Center
(TUCC), Research Triangle Park, N. C.
     The output of the program gave an analysis of variance table and
related statistical measures for the dependent variable RES and each of
the independent variables used in the model.  The first trial used 27
                                       125

-------
independent variables, X.. through X,-, their squares and cross products.
In addition the program gave coefficient values ("B values") for the
regression model, of the form
                                                                         (28)
Goodness of fit was measured by the square of the multiple correlation
coefficient, "R-square", and the coefficient of variability, "C.V.",
expressed as a percent.
     The full model had an R-square value of 0.989567 and a C.V. of
7,3%.  Upon examination of the statistical measures of significance for
the independent variables it was apparent that many could be discarded
as having negligible effect on the model.  A second model, retaining
only the independent variables X., X3, X&, X^, X,X5? X^, X^, X-^Xg,
XXX  and XJC™, was tested.  The R-square value for this was 0,985352 and
the C.V. was 7.13%.  The mathematical expression for this model is given
in equation (1) in Section 4 of this report.
     A final trial was made using the variables X_, X, and X_X,.  The
R-square value for this was 0.972016 and the C.V. increased to 9.26%.
This still appears to be a reasonable fit to the data.  The mathematical
form for this simpler model is given in equation (7) in Section 4.
     The results of the modeling of the response surface may be interpreted
as follows.  To a good approximation the FID response to methane over
the range of 0.2 to 2 ppm is dependent only upon the methane concentration
and the carrier air flow.  The next higher level of modeling takes into
account the effect of hydrogen flow as well and the interactions between
JL and carrier air flow rates, H~ and sample size and carrier air flow
and sample size.  There is also a slight effect of pressure but this is
not as significant as the other factors.
     These results and the use of the mathematical model for sensitivity
analyses are discussed in Section 4.
     As was discussed in Section 4, the statistical analysis indicated
that the hydrogen flow and carrier air flow rates were the most influential
                                       126

-------
operating variables for a given sample sign.  Accordingly, when the FID
was installed in the RTI-designed gas chromatographic  (GC) analyzer, a
simpler test design employing only three variables at  three levels was
used to obtain optimum flow setting data.  The variables and levels are
given in Table 27 below.  The sample size used in all  observations was
2 cm  of 2.05 ppm CH, in air, and a temperature of about 87°C was used.

           TABLE 27.  TEST VARIABLES AND LEVELS FOR FID-GC EVALUATION
VARIABLES

3
Hydrogen Flow Rate (cm /min)
3
Carrier Air Flow Rate (cm /min)
3
Support Air Flow Rate (cm /min)

-1
20
30
100
LEVELS
0
32
40
200

1
42
50
300
     The test plan and resulting response values are given in Table 28,
Mean values of response were as plotted  in Figure  18 of Section 5.  The
"response number" values  in Table  28  can be  converted to FID signal
                                          -14
current values by multiplying by 7.87  *  10    Ampere.
     Because of  the  fewer number of variables and  the experience with
the FID evaluation no regression model was used to analyze the data.
                                       127

-------
TABLE 28.  DESIGN OF GC EVALUATION TEST AND RESULTING RESPONSES
H2 Flow
-1
-1
-1
-1
-1
-1
-1
-1
-1
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
Carrier Air
0
0
0
1
1
1
-1
-1
_n
0
0
0
1
1
1
-1
_1
-1
0
0
0
1
1
1
-1
-1
-1
Support Air
0
1
-1
0
1
-1
0
1
-1
0
1
_i
0
1
— 1
0
1
_1
0
1
-1
0
1
-I
0
1
-1
Response No.
114. 7
106.6
109.8
78.5
77.0
77.3
123.2
J.20.9
128.4
188.6
184.9
189.0
194.6
189.3
187.0
156.6
153.0
154.1
199.9
197.3
187.5
231.0
230.2
216.0
143.8
139.5
135
                              128

-------
                                 APPENDIX B
                         COMPARATIVE EVALUATION DATA

     Instruments zero points were set with air containing less than 50 ppb
THC and span was set with 8 ppm CH, in air to begin calibration stability
tests on July 28, 1976.  These tests continued until August 16, 1976.
During this period precision was measured daily by repetitive injections
of methane, approximately 2.2 ppm in air for the CH, channel, and 0.53 ppm
propylene plus 2.2 ppm methane in air for the THC channel.
     Plots of mean zero drift and standard deviation on a 24-hour basis
are shown in Figures 35 through 44.
     Span drift data is presented in Section 7, Table 12.
     Precision data as standard deviations of response to repeated
injections over a four-to-seven hour period daily have been presented
in Section 7, Table 13.
     The NMHC response under various conditions was presented in Section 7,
Tables 14 through 21.
     Ambient air data from the comparative evaluation over the period 1600
August 28, 1976 through 1500 August 31, 1976 are given in Table 29.  Using
hourly averages at increments of two hours, these data were used to cal-
culate the mean CH, and the ratio of individual instrument values to the
mean value.  These data are presented in Table 30.  Similar data for THC
readings are given in Table 31.
                                       129

-------
LO
O
   I.
   ft
   c
   n
   C
   0)
   U
   C
   O
           2.0
           1.9
           1.;
           1.7
           1.6
           1.5
1.4
           1.3
1.2
           1.1
           1.0
            .9
                O
                             THC - Mean & Standard Deviation
                      7/31  8/1   8/2   8/3   8/4   8/5   876   8/7   8/8    8/9    8/10  8/11  8/12  8/13

                       i	I	T	I	I	I	I	J	I	i     |	I	I      I
                Figure 35.  Mean and standard deviation  of  total hydrocarbon indication of zero air for

                            each twenty—four hour period of long term stability test;

                            Instrument;   R.TI

-------
U)
1.5


1.4


1.3


1.2


1.1
     |  1.0
     c
     o
     o
     c
     o
     o
         .9
         .7
         .6
         .5
         .4
                  Beckman 400 - Mean & Standard

                  Deviation of THC
                 --    C)
                             J_    C)
7/29  7/30  7/31  8/1
 •     i     I	L_
8/2
 '
                                                     C)
                                                                C)     0
                                                                                         0
                                                                                                     C)    -T-
                                                                                   C)
                                                                                                         I
                                                «
8/4
 I
8/5
 '
                                                                 8/6
8/7
 I
                                                                     8/8   8/9   8/10  8/11  8/12  8/13  8/14   8/15
                       Figure 36.  Mean and standard deviation of total hydrocarbon indication of zero air for each
                                   twenty-four hour period of long term stability test;

                                   Instrument:  Beckman 400.

-------
1.4
1.3

L.2
1.1
(
^ 1.0 _
e
^ -9
c
1-1
Concentrat
-~) CO
.6
.5
.4
Beckman 6800 - Mean & Standard _
Deviation of CH.
4
-
)
(
•

-
^^
-
_


)
_§>_
C







)
c







)
c







) c
— -~m


-

—.—


>

*
> ^

~



)
<


)

)
<


)

)


C
)

(
) O
() -
()

„
7/29 7/30 7/31 8/1 8/2 8/3 -874 8/5 8/6 8/7 8/8 8/9 8/10 8/11 8/12 8/13 8/14 8/15
i i i i i i i I i i I i i i i i i i
Figure 37.  Mean and standard deviation of methane indication of zero air for each twenty-four
            hour period of long term stability test;
            Instrument:  Beckman 6800

-------
      1.0
OJ
U)
.9
.8
.1
^ .6
6
ft
s 	 X f
c .5
o —
•H
U3
*-i /
-u .4
C
CJ
u
c
° 0
u . 3
.2
.1


Beckman 6800 - Mean & Standard _
Deviation of THC
-

) C

••»


^
-
_


; 5 3









> c
5





>
c









>
_ /
V







)







>


c





(
\
)



) ~
w

(


—
—



(

)



)
J
(



c

>

•~^~ ""*
-i- C

) <) C)
( )
-
:

7/29 7/30 7/31 8/1 8/2 8/3 8/4 8/5 8/6 8/7 8/8 8/9 8/10 8/11 8/12 8/13 8/14 8/15

i i i
i i i
LI 1 1 1 1
1 	 1 	 1 	 1 	 1 1 1
                    Figure 38.   Mean and standard deviation of total hydrocarbon indication of zero air for each
                                twenty-four hour period of long term stability test;

                                Instrument:  Beckman 6800.

-------
E
cx
a.
c
o
U
C
o
                 Bendix 8200 - Mean & Standard


                 Deviation of CH.
                                                   o
                                                                                 I
<>
              7/30  7/31  8/1   8/2   8/3   8/4   8/5    S/-6    8/7    8/8    8/9   8/10  8/11  8/12  8/13  8/14  8/15
            Figure 39.  Mean and standard deviation  of  methane -indication of zero air for each twenty-four

                        hour period -of .tong  term stability test;
                        Instrument-:  Bendix  8200.

-------
01


.4

.»— s
& 3
ex • J
Concentration (
K3
1 ^ 1
.1

Bendix 8200 - Mean & Standard
Deviation of THC
—


5]5 5
]£ (
)
-




r C
)





)
<
)


_
> -j-
 0
5 i <> J_ I i
I i
-
7/30 7/31 8/1 8/2 8/3 8/4 8/5 8/6 8/7 8/8 8/9 8/10 8/11 8/12 8/13 8/14 8/15
A J 	 1 	 A J 	 1_ A A 	 L 1 1 1 1 , , ,
                Figure 40.  Mean and standard deviation of total hydrocarbon indication of zero air for each
                            twenty-four hour period of long term stability test;

                            Instrument:  Bendix 8200.

-------
                                      BEST AVAILABLE COPY
     ,30
O,
cx
c
o
cfl
J-l
.u
C
Ol
O

O
u
     .25  -
     .20
             Bendix  8201 - Mean  &  Standard


             Deviation  of CH,
     ,10
     .05
                                                                          O

                                                                    O    _-
8/1   8/2   8/3   8/4   8/5   8/6

 I _ i _ I _ J _ I _ I
                                                       8/8   8/9   8/10  8/11 8/12

                                                        I _ I _ j _ i _ j
I    I
        Figure 41.   Mean and standard deviation of methane indication of zero air for each

                    twenty-four hour period of long term stability test;

-------
   I,
   a.
-  g
bJ -H
   a)
   u

   o
   u
                   Bendix 8201 - Mean & Standard


                   Deviation of THC
        .30
       .25
-*-     ()     ()
        .20
                         0    -L
        ,15
         10
        .08- -
7/29  7/30  7/31   8/1  8/2
8/3

  i
                                                    8/4   8/5
 8/6

	i
                                                8/7   8/8
8/9
  i
                   C)

                         <)





8/10  8/11  8/12  8/13  8/14
              Figure 42.   Mean and standard deviation of total hydrocarbon indication of zero air for each

                          twenty-four hour period of long term stability test;


                          Instrument:  Bendix 8201.

-------
   1.0
                                              MSA - Mean & Standard Deviation of
I
ex
c
o
•H
4J
to
    .9
    -8
            -T-    O
                                                 -^    C)
                                                                                     _    I
            7/29   7/30   7/31  8/1   8/2
             1     1     1     4     1
                                          8/3
                                            1
8/4   8/5
8/6
  1
8/7
  1
8/8
  1
8/9   8/10  8/11  8/12  8/13  8/14   8
l5
                Figure 43.  Mean and standard deviation of methane  indication  of  zero air for each twenty-four

                            hour period of long term stability  test;

                            Instrument:  MSA 11-2.

-------
  1.1
   1.0
t
ex
c
o
CO
*-l
                                                                          MSA - Mean & Standard Deviation - THC
                                           o
    .6
    .5
7/29  7/30  7/31  8/1   8/2   8/3   8/4   8/5
                                                            8/6
                                                              i
8/7
  i
8/8   8/9
8/10  8/11  8/12  8/13  8/14  8/15
  i      i      i      i      i      i
                Figure 44.   Mean and standard deviation of total hydrocarbon indication of zero air for each
                            twenty-four hour period of long term stability test;

-------
TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date
vO
r>-
oo
CN
OO
Time
1600
1700
1800
1900
2000
2100
Component
CH4
THC
NMHC
CH4
THC
NMHC
CHA
THC
NMHC
CH/4
THC
NMHC
™4
THC
NMHC
cn4
TKC
NMHC
MSA 11-2
1.46
2.37
0.91
1.49
2.32
0.83
1.45
2.26
0.81
1.38
2.32
0.94
1.31
2.37
1.06
1.29
2.52
1.23
Beckman
400

1.98


2.20


2.16


2.05


2.07


2.12

Beckman
6800
1.70
1.62
-0.08
1.72
1.65
-0.07
1.71
1.59
-0.12
1.76
1.68
-0.08
1.76
1.65
-0.11
1.81
1.86
O.05
Bendix
8200
1.38
1.95
0.57
1.39
1.92
0.53
1.39
1.91
0.52
1.42
1.97
0.55
1.39
1.99
0.60
1.45
2.13
0,68
Bendix
8201
1.66
1.63
-0.03
1.70
1.61
-0.07
1.68
1.58
-0.10
1.71
1.63
-0.08
1.71
1.66
-0.05
1.74
1.75
0.01

-------
                TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
                                August  28,  1976  -  August  31,  1976
Date
^o
r-~
00
CN
1
CO
o
r^
1
Ox
CN1
OO
Time
2200

2300
2400
0100
0200
0300
Component
CH4
THC
NMHC
CHA
THC
NMHC
CH.
4
THC
NMHC
CH4
THC
NMHC
CT4
THC
NMHC
CHA
THC
NMHC
MSA 11-2
1.30
2.59
1.29
1.29
2,57
1.28
1.50
2.68
1.18
1.47
2.47
1.00
1.47
2.40
0.93
1.47
2.35
0.88
Beckman
400

2.12


2.16


2.01


1.96


1.94


1.91

Beckman
6800
1.89
1.92
0.03
1.88
1.92
0.04
1.88
1.90
0.02
1.88
1.84
-0.04
1.88
1.75
-0.13
1.86
1.75
-0.11
Bendix
8200
1.52
2.11
0.59
1.55
2.14
0.59
1.52
2.02
0.50
1.53
2.05
0.52
1,54
2,01
0.47
1,52
2.00
0.48
Bendix
8201
1.80
1.77
-0.03
1.83
1.70
-0.07
1.80
1.72
0.00
1.82
1.67
-0.15
1.81
1.70
-0.11
1.81
1.67
-0.14
NOTE:  From hour date CH, readings of MSA 11-2 have been corrected  for  an unexplained,
       abrupt baseline shift.  Validity of data uncertain,  but seems consistent  with  others.

-------
TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date
VD
r^
1
cr.
r-j
00
Time
0400
0500
0600
0700
0800
0900
Component
CH4
THC
NMHC
CHA
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMIIC
CH4
THC
NMHC
MSA 11-2
1.43
2.35
0.92
1.42
2.31
0.89
1.42
2.30
0.88
1.47
2.28
0.81
1.43
2.26
0.83
1.48
2.25
0.77
Beckman
400

1.89


1.81


1.90


1,81


1.88


1.79

Beckman
6SOO
1.83
1.76
-0.07
1.83
1.73
-0.10
1.85
1.74
-0.11
1.83
1.74
-0.09
1.83
1.71
-0.12
1.83
1.74
-0.09
Bendix
8200
1.53
2.00
0.47
1.54
1.97
0.43
1.53
1.97
0.44
1.55
1.98
0.43
1.54
, 1.97
0.43
1.55
1.97
0.42
Bendix
8201
1.81
1.67
-0.14
1.77
1.63
-0.14
1.79
1.72
-0.07
1.81
1.67
-0.14
1.81
1.67
-0.14
1.77
1.67
-0.10

-------
TABLE  29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date
\D
r^
1
ON
CM
1
CO
Time
1000
1100
1200
1300
1400
1500
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH.
4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
MSA 11-2
1.48
2.23
0.75
1.48
2.21
0.73
1.47
2.19
0.72
1.45
2.23
0.78
1.45
2.20
0.75
1.42
2.20
0.78
Beckman
400

1.79


1.79


1.74


1.70


1.70


1.78

Beckman
6800
1.83
1.70
-0.13
1.81
1.67
-0.14
1.81
1.67
-0.14
1.78
1.70
-0.08
1.81
1.71
-0.10
1.77
1.70
-0.07
Bendix
8200
1.52
1.95
0.43
1.51
1.93
0.42
1.51
1.92
0.41
1.50
1.93
0.43
1.51
1.93
0.42
1.51
1.93
0.42
Bendix
8201















1.73
1.67
-0.06 ;

-------
TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August  31,  1976
Date








r-^
CM
1
CO








Time

1600


1700


1800


1900


2000


2100

Component
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CE4
THC
NMHC
CH4
THC
NMHC
MSA 11-2
1.38
2.21
0.83
1.37
2.23
0.86
1.38
2.21
0.83
1.40
2.21
0.81
1.38
2.25
0.87
1.48
r 2.35
0.87
Beckman
400

1.72


1.72


1.70


1.70


1.79


1.79

Beckman
6800
1.77
1.70
-0.07
1.77
1.69
-0.08
1.81
1.72
-0.09
1.81
1.70
-0.11
1.75
1.70
-0.05
1.81
1.78
-O.O3
Bendix
8200
1.48
1.93
0.45
1.49
1.93
0.44
1.51
1.93
0.42
1.51
1.94
0.43
1.51
1.95
0.44
1.51
2.04
0.53
Bendix
8201
1.74
1.66
-0.08
1.73
1.63
-0.10
1,73
1.63
-0.10
1.73
1.61
-0.12
1.74
1.67 .
-0.10
1.77
1.67
-0.10

-------
                             TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
                                            August 28, 1976 - August 31, 1976
L/l
Date
vO
1 —
1
ON
CSI
CO
\D
r^
1
0
m
oo
Time
2200
2300
2400
0100
0200
0300
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
cii4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
MSA 11-2
1.52
2.52
1.0
1.74
2.74
1.0
1.55
2.23
0.68
1.55
2.16
0.61
1.58
2.16
0.58
1.74
2.21
0.47
Beckman
400

1.92


2.16


1.83


1.85


1.85


1.93

Beckman
6800
1.58
1.94
0.36
1.98
2.12
0.14
1.86
1.74
-0.12
1.81
1.78
-0.03
1.78
1.76
-0.02
1.86
1.82
-0.04
Bendix
8200
1.55
2.13
0.58
1.71
2.37
0.66
1.50
1.93
0.43
1.56
1.85
0.29
1.55
1.81
0.26
1.63
1.86
0.23
Bendix
8201
1.76
1.77
0.01
1.95
1.95
0
1.80
1.70
-0.10
1.77
1.66
-0.11
1.76
1.66
-0.10
1.-85
1.72
-0.13

-------
TABLE 29.   COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date
\D
r-^
0
f>
00
Time
0400
0500
0600
0700
0800
0900
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CIIA
THC
NMHC
CH4
THC
NMHC
CHA
THC
NMHC
MSA 11-2
1.61
2.12
0.51
1.65
2.12
0.47
1.68
2.16
0.48
1.68
2.18
0.50
1.71
2.21
0.50
1.61
2.14
0.53
Beckman
400

1.96


1.90


1.90


1.90


2.04


1.91

Beckman
6800
1.78
1.76
-0.02
1.81
1.78
-0.03
1.82
1.80
-0.02
1.84
1.80
-0.04
1.81
1.88
0.07
1.81
1.82
0.01
Bendix
8200
1.59
1.78
0.19
1.59
1.76
0.17
1.61
1.83
0.22
1.59
1.85
0.26
1.60
1.91
0.31
1.58
1.80
0.22
Bendix
8201
1.82
1.67
-0.15
1.77
1.66
-0.11
1.83
1.70
-0.13
1.81
1.75
-0.06
1.32
1.72
-0.10
1.79
1.66
-0.13

-------
TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date
\o
r^
1
0
CO
00
Time
1000
1100
1200
1300
1400
1500
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH.
4
THC
NMHC
CH,
4
THC
NMHC
CH4
THC
NMHC
CH,
4
THC
NMHC
MSA 11-2
1.65
2.15
0.50
1.65
2.10
0.45
1.61
2.09
0.48
1.60
2.07
0.47
1.58
2.12
0.54
1.55
2.18
0.63
Beckman
400

1.83


1.79


1.79


1.79


1.88


1.96

Beclonan
6800
1.81
1.78
-0.03
1.76
1.74
-0.02
1.77
1.75
-0.02
1.77
1.75
-0.02
1.76
1.78
0,02
1.76
1.78
-0.02
Bendix
8200
1.58
1.81
0.23
1.59
1.74
0.15
1.56
1.77
0.21
1.56
1.77
0.21
1.55
1.73
0.18
1,57
1.76
0.19
Bendix
8201
1.77
1.72
-0.05
1.77
1.75
-0.02
1.74
1.63
-0.11
1.73
1.61
-0.12
1.73
1.61
-0.12
1.73
1.61
-0.12

-------
                            TABLE  29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
                                           August 28, 1976 - August 31, 1976
00
Date
^
r-^
0
CO
00
Time
1600
1700
1800
1900
-2000
2100
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
MSA 11-2
1.53
2.21
0.68
1.47
2.21
0.74
1.45
2.16
0.71
1.45
2.21
0.76
1.52
12.35
0.83
1,60
2.44
0.84
Beckman
400

1.87


1.88


1.88


1.88


1.96


2.16

Beckman
6800
1.71
1.80
0.09
1.75
1.82
0.07
1.76
1.74
-0.02
1.71
1.75
0.04
1.86
1.95
0.09
1.86
1.90
0.04
Bendix
8200
1.55
1.75
0.20
1.55
1.73
0.18
1.55
1.76
0.21
1.57
1.74
0.17
1.58
1.76
0.18
1.66
1.85
0.19
Bendix
8201
1.71
1.63
-0.08
1.71
1.63
-0.08
1.73
1.63
-0.10
1.71
1.63
-0.08
1.82
1.72
-0.10
1.82
1.80
-0 .02

-------
TABLE .29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date



VD
r~
1
o
m
i








r^
ro
00



Time

2200


2300


2400


0100


0200


0300

Component.
CHA
-THC
NMHC
CH4
THC
NMHC
CHA
THC
NMHC
CHA
THC
NMHC
CT4
THC
NMHC
CH4
THC
NMHC
MSA 11-2
1.68
2.35
0.67
1.68
2.37
0.69
1.78
2.46
0.68
1.93
2.48
0.55
1.69
2.41
0.72
1.87
2.43
0.56
Beckman
400

2.13


2.24


2.23


2.20


2.09


2.20

Beckman
68QO
1.91
1.97
0.06
1.94
2.05
0.11
1.94
2.05
0.11
2.01
2.09
0.08
1.86
1.95
0.09
1.94
2.03

Bendix
8200
1.78
2.05
0.27
1.70
1.93
0.23
1.79
1.99
0.20
1.75
2.05
0.30
1.71
1.80
0.09
1.71
2.00

Bendix
8201
1.88
1.80
-0.08
1.86
1.72
-0.06
1.88
1.77
-0.11
1.95
1.85
-0.10
1.89
1.75
-0.14
1.96
I
1.87


-------
                             TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.

                                            August 28, 1976 - August 31, 1976
Ul
o
Date
\O
r~-
i— I
ro
oo
Time
0400
0500
0600 i
0700
0800
0900
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CHA
THC
NMHC
MSA 11-2
1.81
2.35
0.54
1.80
2.35

1.87
2.37
0.50
1.82
2.36

1.84
2.55
0,71
1.81
1.94

Beckman
400

2.07


2.09


2.17


2.23

-
2.38


2.28

Beckman
6800
1.91
1.96
0.05
1.91
2.01
0.10
1.9-6
2.03
0.07
1.96
2.00
0.04
1.96
2.13
0.13
1.91
2.03
0.12
Bendix
8200
1.71
1.95
0.24
1.73
1.95

1.75
2.01
0.26
1.73
2.01

1.79
2.09
0.30
1.71
2.01

Bendix
8201 ••
1.91
1.87
j
1.95
1.87

1.91
1.82

1.92
1.82

1.91
1.90

1.85
1.85


-------
TABLE 29.  COMPARATIVE EVALUATION AMBIENT AIR DATA.
               August 28, 1976 - August 31, 1976
Date
vD
r>-
1
iH
rn
1
00
Time
1000
1100
1200
1300
1400
1500
Component
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
NMHC
CH4
THC
•NMHC
CH4
THC
NMHC
CH4
THC
NMHC
MSA 11-2
1.77
2.26
0.49
1.69
2.14

1.68
2.14
0.46
1.66
2.12

1.65
2.12
0.47
1.68
2.09

Beckman
400

2.01


1.90


1.83


1.79


1.83


1.77

Beckman
6800
1.81
1.99
0.18
2.01
2.03
0.02
1.77
1.85
0.08
1.76
1.82
0.06
1.77
1.83
0.06
1.77
1.75
-0 .02
Bendlx
8200
1.70
1.95
0.25
1.66
1.81

1.66
1.79
0.13
1.61
1.76

1.62
1.79
0.17
1.61
1.73

Bendix
8201
1.83
1.75

1.77
1.67

1.74
1.66

1.74
1.63

1.76
1.66

1.76
1.66


-------
TABLE  30.  AVERAGE METHANE CONCENTRATION AND RATIOS OF INDIVIDUAL INSTRUMENT
           READINGS TO METHANE
Date
8/28/76




8/29/76





8/29/76





Time
1600
1800
2000
2200
2400
0200
0400
0600
0800
1000
1200
1400
1600
1800
2000
2200
2400
CH^ Avg.
(ppm)
1.55
1.56
1.54
1.63
1.68
1.68
1.65
1.65
1.65
1.61
1.60
1.59
1.59
1.61
1.60
1.60
1.68
Ratios
MSA 11-2
0.94
0.93
0.85
0.80
0.90
0.88
0.87
0.86
0.87
0.92
0.92
0.91
0.87
0.86
0.87
0.95
0.92
(Instrument Value /Average)
Beckman
6800
1.10
1.10
1.14
1.16
1.12
1.12
1.11
1.12
1.11
1.14
1.13
1.14
1,11
1.13
1.10
0.99
1.11
Bendix
8200
0.89
0.89
0.90
0.93
0.90
0.92
0.93
0.93
0.93
0,94
0.95
0.95
0.93
0.94
0.95
0.97
0.89
Bendix
8201
1.07
1.08
1.11
1.11
1.07
1.08
1.10
1.09
1.10
	
	
..._
1.09
1.08
1.09
1.10
1.07
(1)
   Average from MSA 11-2, Beckman 6800, Bendix 8200, Bendix 8201 hourly
   average values.
                                      152

-------
TABLE 30.  AVERAGE METHANE CONCENTRATION AND RATIOS OF INDIVIDUAL INSTRUMENT
          ' READINGS TO METHANE (Continued)
Date Time
8/30/76 0200
0400
0600
0800
1000
1200
8/30/76 1400
1600
1800
2000
2200
2400
8/31 0200
0400
0600
0800
1000
1200
CH4 Avg.
(ppm)
1.67
1.70
1.74
1.74
1.70
1.67
1.66
1.63
1.62
1.70
1.81
1.85
1.79
1.84
1.87
1.88
1.78
1.71
Ratios
MSA 11-2
0.95
0.95
0.97
0.99
0.97
0.96
0.95
0.94
0.89
0.90
0.93
0.96
0.95
0.99
1.00
0.98
1.0
0.98
(Instrument Value /Average)
Beckman
6800
1.07
1.05
1.05
1.04
1.06
1.06
1.06
1.05
1.08
1.10
1.05
1.05
1.04
1.04
1.05
1.05
1.02
1.03
Bendix
8200
0.93
0.94
0.93
0.92
0.93
0.93
0.94
0.95
0.96
0.93
0.98
0.97
0.96
0.93
0.93
0.95
0.96
0.97
Bendix
8201
1.06
1.07
1.05
1.05
1.04
1.04
1.05
1.05
1.07
1,07
1.04
1.02
1.06
1.04
1.02
1.02
1.03
1.02
^Average from MSA 11-2, Beckman 6800, Bendix 8200, Bendix 8201 hourly
   average values.
                                      153

-------
 TABLE 30.   AVERAGE METHANE CONCENTRATION AND RATIOS OF INDIVIDUAL INSTRUMENT
	READINGS TO METHANE (Continued)	
 Date
       CH4 Avg
Time    (ppm)
                          (1)
                                  Ratios (Instrument Value/Average)
                               MSA 11-2
          Beckman
           6800
         Bendix
          8200
        Bendix
         8201
 8/31/76   1400
         1.70
0.97
1.04
0.95
1.04
' ^Average from MSA 11-2,  Beckman 6800,  Bendix 8200,  Bendix 8201 hourly
   average values.
                                      154

-------
TABLE  31.  AVERAGE TOTAL HYDROCARBON CONCENTRATION AND RATIOS OF INDIVIDUAL
           INSTRUMENT READINGS TO AVERAGE
Ratios (Instrument Value/Average)
THC Avg. (1)
equiv, ppm
Date Time CH4 MSA 11-2
8/28/76 1600
1800
2000
2200
2400
8/29/76 0200
0400
0600
0800
1000
1200
8/29/76 1400
1600
1800
2000
2200
2400
2.13
2.05
2.07
2.21
2.17
2.06
2.03
2.02
1.98
2.03
1.88
2.00
1.95
1.95
1.97
2.15
1.97
1.11
1.10
1.14
1.17
1.24
1.17
1.16
1.14
1.14
1.10
1.16
1.10
1.14
1.13
1.14
1.17
1.13
Beckman
400
0.93
1.05
1.00
0.96
0.93
0.94
0.93
0.94
0.95
0.88
0.93
0.86
0.87
0.88
0.86
0.90
0.89
Beckman
6800
0.76
0.77
0.80
0.87
0.88
0.85
0.87
0.87
0.86
0.82
0.89
0.97
0.99
0.99
0.99
0.99
0.98
Bendix
8200
0.92
0.93
0.96
0.95
0.93
0.98
0.99
0.98
0.99
0.96
1.03
____
0.85
0.84
0.85
0.82
0.87
Bendix
8201
0.77
0.77
0.80
0.80
0.79
0.83
0.82
0.85
0.84
	
— — — —
1.23
1.26
1.28
1.26
1.21
1.20
^ 'Average from all instruments operating.
                                       155

-------
TABLE  31.  AVERAGE TOTAL HYDROCARBON CONCENTRATION AND RATIOS OF INDIVIDUAL
           INSTRUMENT READINGS TO AVERAGE (Continued)
Ratios (Instrument Value/Average)
THC Avg. (1)
equiv. ppm
Date Time CH4
8/30/76 0200
0400
0600
0800
1000
1200
8/30/76 1400
1600
1800
2000
2200
2400
8/31/76 0200
0400
0600
0800
1000
1200
1.91
1.92
1.95
2.01
1.94
1.89
1.91
1.94
1.89
2.01
2.14
2.18
2.09
2,13
2.15
2.27
2.09
1,96
MSA 11-2
1.13
1.11
1.11
1.10
1.11
1.11
1.11
1.14
1.15
1.17
1.10
1.13
1.16
1.11
1,10
1.12
1.08
1.09
Beckman
400
0.97
1.02
T— -
1.01
0.94
0.95
0,98
0.97
1.0
0.97
1.00
1.02
1.00
0,97
1.01
1,05
0.96
0.94
Beckman
6800
0.92
0.92
0.92
0.93
0.92
0.93
0.93
0.93
0.92
0.97
0.92
0.94
0.94
0.92
0.94
0.94
0.95
0.95
Bendix
8200
0.95
0.93
0.94
0.95
0.93
0.94
0.90
0.90
0.93
0.87
0.96
0.91
0.86
0,92
0.93
0.92
0.93
0,92
Bendix
8201
0.87
0.87
0.87
0,86
0.89
0,86
0.84
0.84
0.86
0.85
0.84
0,81
0.84
0,88
0,85
0.84
0.84
0.85
(1)
   Average from all instruments operating.

                                       156

-------
TABLE 31.  AVERAGE TOTAL HYDROCARBON CONCENTRATION AND RATIOS OF  INDIVIDUAL
	INSTRUMENT READINGS TO AVERAGE  (Continued)
Date    Time
             THC Avg.' '
             equiv. ppm
                CH,
                              Ratios  (Instrument Value/Average)
                                    Beckman   Beckraan  Bendix  Bendix
                                     400       6800     8200    8201
8/31/76 1300    1.94
                           1.09
                                     0.94      0.94     0.92    0.86

    Average  from all operating instruments.
                                       157

-------
158

-------
                                  APPENDIX  C
                 EVALUATION  OF  STRIP  CHART RECORDERS USED  IN
                      COMPARATIVE  INSTRUMENTAL EVALUATION
INTRODUCTION
     During the comparative  evaluation of hydrocarbon analyzers, strip chart
recorders  (SCR) were used  to record  the experimental data.  To assure that
none of the data was unfairly biased by a faulty  SCR, the operating characteris-
tics of all the recorders  were examined.
     As recently as 1968,  about  28%  of those actively involved in analysis were
using peak height as the mode of  quantitation  (35).  The current Reference
Method for measuring non-methane  hydrocarbons  (NMHC) makes no mention of SCR
specifications, although an  accurate NMHC analysis is practically impossible
with a malfunctioning  or improperly  adjusted recorder.
     The recorders used in the evaluation were the following:
     1.  A Hewlett-Packard 680,
     2.  Two Hewlet-Packard  7101B  (they are differentiated by the
         final 4 digits of their  serial numbers,  -1295 and -1293),
     3.  A Varian G-4000,
     4.  A Linear Instruments 252A,
     5.  A Leeds and Northrup Speedomax XL610, and
     6.  A Honeywell Electrik.
The first five SCR's were  supplied by EPA for  the instrumental evaluation
and the Honeywell and  Leeds  and Northrup SCR's were supplied by RTI.
DESCRIPTION OF TESTS PERFORMED
     Five parameters were  of  particular concern in evaluation of the performance
of the recorders.   The first  of  these was an examination of the zero drift of
the SCR's by short circuiting the  inputs and allowing the recorders to operate
with no manual adjustments for a  68  hour test period.  This reflects any
ambient dependence of  the  recorder electronics and provides a check on the
linearity of the chart drives over a long-period.
                                      159

-------
     The rise and fall times, defined as 10% to 90% (and vice versa) of
full-scale transitions,  were measured by applying a full scale, 0.1 Hz
square wave from a Hewlett-Packard Model 3300A Function Generator to the
recorder while using fast chart speeds.  By measuring the difference in the
10% and 90% points and knowing the chart speed, good rise and fall time
approximations can be made.
     Deadband is defined as the amount of applied input necessary to pro-
duce a recorder deflection and was measured by connecting the recorder
input to a Fluke Model 731A DC Transfer Voltage Source (calibration traceable
to NBS).  The 731A is capable of 1 uv resolution, and by changing the input
voltage in small increments, the deadband was determined for the recorder
scales used in the instrumental evaluation most frequently.   Deadband is
reported as a percentage of full scale.
     Any recorder overshoot was noted when the full scale, square wave was
being applied.
     Finally, a small signal, 3dB frequency was determined by applying a
sine wave from the Hewlett-Packard 3300A to the recorder inputs,,  The peak-
to-peak excursions covered approximately 20% of the scale.  The frequency
was increased from 0.1 Hz to the point where the response was 0.7 that of
the 0.1 Hz response.  This frequency represents the 3dB frequency of the
entire recorder electromechanical system and represents recorder frequency
limitations.
RESULTS AND DISCUSSION
     The results of the evaluation procedure are given in Table 32.
     The zero drift for the HP680, Linear Instruments 252A,  and the Varian
G6400 was so slight that only upper limit approximations were made.   The
drifts of the HP7101B recorders and the Honeywell recorder were tneasureable
but were still very slight, being well within specifications.  The zero
drift of the Leeds and Northrup was impossible to determine because of the
ink system produced a thick,, smeared trace.  The value given in Table 32
is a gross upper limit and the actual value is probably considerably less.
     The rise time of the instruments is comparable (HP680 has only a 5 inch
chart; thereby giving the lower value) as are the fall times.  The Honeywell
recorder did not have a variable chart speed, and rise and fall times could
not be measured with the chart speed it did have.
                                      160

-------
     Deadband values are comparable.  The only recorder to show overshoot
was the HP680 and this value was within manfacturer's specification (no
effort was made to correct the overshoot).  The 3dB points ranged from
about 2 to 4.5 Hz.
     All chart drives were linear except the Varian G4000 which was in
need of maintenance.
CONCLUSIONS
     Based on the characteristics presented in Table 32 it appears that the
reocrders are all comparable and introduced no unfair bias to the instru-
mental data obtained in the comparative evaluation.
                                      161

-------
         TABLE  32.   RESULTS  OF  RECORDER EVALUATION
Instrument
HP 680
HP 7101B-1295
HP 7101B-1293
Linear Ins tr laments
25 2A
Varian GAOOO
Honeywell Electrik
Leeds & Northrup
Channel //I
Channel #2








Zero Drift
(% of Full Scale)
<0.005
0.06
0.08
_cp.006
< 0.005
0.08
£0.25
£0.25
Rise Time
(msec)
250
416
417
317
375
	
580
580
Fall Time
(msec)
125
333
416
127
250
	
580
660
Deadband
(% of Full Scale)
0.1
0.04
0.02
0.08
0.06
0.04
0.03
0.04
.
Overshoot
(% of Full Scale)
2
<0.1 in.)
None
None
None
None
None
None
None
I
3 dB Point
(Hz)
4.5
3.6
3.4
1.9
4.5
4.0
2.2
2,1
TABLE 32.  RESULTS OF RECORDER EVALUATION

-------
                              APPENDIX D
               PROPOSED PROCEDURE FOR THE MEASUREMENT OF
               NON-METHANE HYDROCARBONS IN THE ATMOSPHERE
1.0  PURPOSE
     In typical polluted air samples the principal hydrocarbon component,
methane (CH,), is usually more abundant than all other hydrocarbons
combined.   Methane levels range from a natural background of about
1.2 ppm (parts per million) to values as high as 10 ppm in highly
polluted urban environments.   Since methane is inert in photochemical
reactions, it is necessary to measure the methane background separately
to permit an estimation of the non-methane hydrocarbon (NMHC) fraction
which is reactive.  The National Ambient Air Quality Standard for NMHC
                                                         2
content in ambient air is 0.24 ppm as equivalent methane.    This
equivalent methane value is measured in terms of the response of a
flame ionization detector (FID) to NMHC relative to the FID response
       3
to CH,.   Hydrocarbon analyzers must be capable of measuring the low
NMHC levels relative to the high CH, background levels.
     The purpose of this standard is to specify performance criteria
and a uniform calibration procedure applicable to all analyzers designed
for the measurement of NMHC concentration in ambient air.   In addition,
specifications for reagents are given.  There is a brief discussion of
possible sources of error.   A more detailed discussion of error sources,
instrumental design principles and recommended good practices for
installation, maintenance and operation of NMHC analyzers is given in
reference 4.
2.0  PRINCIPLES OF MEASUREMENT
     As of the date of the issuance of this procedure, all instruments
commercially available tor measurement of methane and NMHC concentrations
in ambient air use the FID to develop an electrical signal current which
is related to the hydrocarbon content of the sample.  The response
factors—i.e.,  the unit current per unit of hydrocarbon concentration—
depend upon operating conditions and the type of hydrocarbon species.
In general, the response for a given hydrocarbon species is dependent
upon the specific physical conditions of the FID used—i.e., the geometry,
                                    163

-------
materials of construction, reagent composition and flow rates.  Usually
the instrument manufacturer specifies the reagent composition and the
flow rates, which are specified either directly in terms of flows
actually measuring during initial instrument setup or indirectly in
terms of pressure regulator settings.
     Several methods of separation, to distinguish methane from the
non-methane hydrocarbon components in a sample to be analyzed, have
been used or proposed.  The most widely ysed have been based upon a
difference technique, either batch pr continuous.  The batch technique
routes an aliquot of sample through a stripper column which delays the
passage pf all hydrocarbons except methane.  The methane fraction is
rapidly eluted and passed on to the FID for quantification.  Following
methane elution, the stripper column is backflushed to remove tT"3 NMHC
fraction, which is exhausted back to atmosphere.   During this part
of the cycle another aliquot of sample is introduced directly to the
(same) FID to obtain a signal related to the total hydrocarbon (THC)
content of the air.  Both the THC and CH, signals are recorded on a strip
chart recorder, usually in a bargraphic representation with recorder pen
deflection proportional to the peak height of the respective signals.
These are usually converted manually to equivalent concentrations,
using calibration curves, and the difference taken to determine NMHC
concentration in terms of equivalent CH, response.
     In the continuous difference technique a sample stream is divided
into two parts and routed to two FIDs to simultaneously measure the
total hydrocarbon content (THC channel) and the methane content (CH,
channel).  The methane concentration is obtained by oxidation of all
NMHC components on a catalyst bed prior to the FID for methane determination.
The electrical signals of the THC channel and CH,  channel are subtracted
by analog circuitry to obtain the NMHC concentration.  This NMHC signal
and the methane signal are available for continuous recording.
     Instead of graphical or analog subtraction to obtain NMHC content ,
a direct method may be used.  This employs a stripper cplumn to separate
CH, for determination in an FID.  In this design, however,  the backflushed
NMHC components are directed to the FID for a determination of the NMHC
                                   164

-------
concentration directly.  The CH, and NMHC signals are available in analog
form on a recorder or in digital form for printout or storage for subse-
quent use.
     Another method which avoids the problems of nonuniform response of
the FID to various NMHC species is to oxidize the NMHC backflush fraction
and then convert it to methane in a hydrogen stream over a hot catalyst.
3.0  PERFORMANCE SPECIFICATIONS
     Assurance of data quality is based upon the ability of NMHC analyzers
to meet or to exceed certain performance specifications  on range, lower
detectable limit, sensitivity, precision and calibration stability.   When
these specifications are met, the accuracy of the measurements will be
related in a well-defined manner to the accuracy of the analysis of
reagent standards used for calibration.
     3.1  Range
          The range of the instruments used will depend upon the maximum
concentrations of methane and NMHC normally encountered in the region to
be monitored.  For methane 10 ppm full scale is normally sufficient.
Instruments which measure NMHC directly should have a maximum range of
5 ppm equivalent CH,.
     3.2  Noise Level and Lower Detectable Limit
          Instrument noise, random fluctuations in output signal,
determines the lower detectable concentration level for hydrocarbon
concentrations.   The noise level is defined as the standard deviation,
S, of the fluctuations about a given signal level.   The lower detectable
limit is defined as twice the noise level, or 2 S.
     For non-methane hydrocarbon measurements a noise level of 0.05 ppm
equivalent CH, should be maintained.   This permits a lower detectable
limit of 0.1 ppm, or about 40% of the current air quality standard of
0.24 ppm for NMHC.   Those instruments which rely on a difference of THC
and CH,  readings require lower noise levels since
                                         2      2
              RNMHC = (RTHC ~ ^H* -   STHC + SCH.
                                 4                4
where R indicates the "reading" (signal value) for the respective
quantities and S indicates the standard deviation of fluctuations about R.

-------
For equal noise levels of the THC and CH, signals, a noise level of
0.05 ppm for the NMHC indication requires a noise level of 0,035 ppm
for these THC and CH, signals, with a consequent lower detectable
limit of 0.07 ppm for each.  Instruments which measure NMHC directly
either through backflush, oxidation and re-forming to methane or by
some qther methods should have for both the NMHC and methane channel
a maximum noise level of 0.05 ppm equivalent CH, and a lower detectable
limit of 0.1 ppm equivalent methane.
     3.3  Precision, Stability and Accuracy
          Measurement precision, defined as the standard deviation of
response to repeated injections of sample of the same concentration,
should be 0.1 ppm equivalent CH,, determined at an input sample concen-
tration at 80% of full scale (80% span).
     Zero drift is defined as the maximum average deviation from average
baseline setting measured over both 12-hour and 24-hour consecutive
periods with an input sample containing the minimum obtainable (i.e.
nominally "zero") hydrpcarbons in air (see Section 5.4) and with no
adjustment of instrument electronic or pneumatic controls during these
periods.   Maximum zero drift should be equal to or less than a signal
corresponding to 0.1 ppm equivalent CH,  when the temperature of the room
containing the instrument is maintained within the range of 20°C to 30°C.
     Span drift is defined as the percentage change in response to an
input sample at a concentration equivalent to ^0% of span measured
over a 24-hour period with no adjustment of instrument electronic or
pneumatic controls.   Maximum span drift should be equal to or less than
5% when the temperature of the room containing the instrument is
maintained within the range of 20°C to 30°C.
     The accuracy of measurement is dependent upon instrument precision,
calibration stability,  and the accuracy of analysis of the reagent gases
used for calibration.  At the upper end of the instrument scale the
accuracy is determined by instrument precision, span gas analysis error
and span drift and any analysis error for the "zero" reagent gas
hydrocarbon content.
                                    166

-------
4.0  APPARATUS
     Figure I, a schematic of typical calibration apparatus, shows the
suggested configuration of the components listed below.  All components
of the hydrocarbon transfer and dilution system should be of glass,
stainless steel or other non-reactive material and should be free of
trace hydrocarbon contaminants.
     4.1  Flow Contollers
          Devices should be capable of maintaining constant air flows or
hydrocarbon mixture flows within +2% of the required flow rate.
     4.2  Pressure Regulators
          Specially cleaned regulators for air cylinders and for
hydrocarbon cylinders which contain no trace hydrocabons or other
contaminants that would react with hydrocarbons should be used.
     4.3  Mixing Chamber
          The chamber should be constructed of glass or other nonreactive
material and designed to provide thorough mixing of hydrocarbon calibration
gases and diluent air.
     4.4  Humidifier
          A thermostated bubbler containing distilled water should be
employed to obtain air streams of constant humidity.
     4,5  Sample Manifold
          The manifold should be constructed of glass or other non-
reactive material and should be of sufficient diameter to insure an
insignificant pressure drop at the analyzer connection.  The system
must have a vent designed to insure atmospheric pressure at the manifold
and to prevent ambient air from entering the manifold.
5.0  REAGENT SPECIFICATIONS
     In the following specifications "air" is defined as a gas mixture
containing 21% + 1% by volume 02, with the balance, N2>  Rare gas
(Ar, He, etc.) and C0~ content should not significantly exceed normal
concentrations in the atmosphere at sea level (about 330 ppm by volume).
Total hydrocarbon content, methane pulse non-methane hydrocarbons, should
not exceed 0.1 ppm equivalent CH,.  Water vapor content should be at or
below that corresponding to -40°C dew point.
                                    167

-------
     5.1  Fuel for Flame lonizatlon Detector  (FID)
          Fuel for the flame ionlzation detectors used for analysis
should be hydrogen or a mixture of hydrogen and gas such as nitrogen
or helium containing less than 0.1 ppm total hydrocarbons  (THC) expressed
as equivalent methane.
     5.2  Support Air
          Support air for the FID should have the composition defined
as "air" and contain less than 0.1 ppm THC expressed as equivalent methane.
     5.3  Carrier Gas
          The composition of carrier gas depends upon the type instrument
design used for analysis.  Commonly either helium, nitrogen, air or
hydrogen is used.  Whichever gas is used, THC content should be less than
0.1 ppm equivalent methane.
     5.4  Zero Gas
          The gas mixture used for setting the baseline (zero level) of
the instrument should be "air" of the composition defined above, containing
less than 0.1 ppm THC equivalent methane.
     5.5  Calibration Gases
          The methane and THC channel of the analyzer should be calibrated
by use of mixtures of CH, in "air" (of the composition described above).
The calibration standard must be traceable to a National Bureau of Standards
methane in air Standard Reference Material (SRM 1658 or 1659).  A propane
in "air" calibration standard is required for determining the NMHC response
factor (see section 6,2) for the particular analyzer.  This standard must
be traceable to an NBS propane in air (SRM 1665 or 1666).   Procedures for
certifying the methane and propane cylinders  (working standards) against
the appropriate NBS traceable methane or propanes standards are given in
reference 4,  The cylinders should be rectified on a regular basis as
determined by the local quality control program.
6.0  PROCEDURE
     The analyzer should be installed on location preferably by the
manufacturer or his authorized representative.  If this is not done,
installation and preparation for operation should be carefully performed
by qualified instrumentation specialists following the analyzer manu-
facturer's instruction manual.
                                    168

-------
     6.1  Initial Setup
          The analyzer should be Installed on location preferably by the
manufacturer or his authorized representative.  If this is not done,
installation and preparation for operation should be carefully performed
by qualified instrumentation specialists following the analyzer manu-
facturer's instruction manual.
     6.2  Initial Calibration
          Following preparation of the analyzer for operation and after
stability of instrument temperature, fuel, carrier gas and support air
has been achieved, calibration can be started.  Because the ambient
air water vapor content (humidity) may have an effect upon instrument
response,  the calibration gases should be conditioned prior to intro-
duction into the instrument.  This can be accomplished by the use of a
bubbler containing distilled water kept at a constant temperature by
a thermostated bath.  The bath temperature should be maintained at a
temperature that provides an equilibrium partial pressure of water vapor
corresponding to the relative humidity anticipated at the measurement
site.  For many sites a wide range of relative humidity values is
encountered, varying seasonally and even diurnally. • In these cases the
calibration should be carried out at the upper and lower limits of water
vapor consent anticipated.  These calibration curves, along with measure-
ment of the actual water vapor content throughout the monitoring period,
will allow interpolation to determine the actual calibration corresponding
to the given water vapor content obtaining at the time of measurement.
     With the humidifying bubbler adjusted to give the desired water
vapor content, introduce the zero gas.  Allow the analyzer output signal
to stabilize and then adjust analyzer controls to obtain a zero reading
on the output indicator.   Depending upon instrument design and method of
data collection and recording, the output indicator may be an analog meter,
strip chart recorder, digital indicator or digital printout.  Following
setting of the zero level, which is performed for both instrument channels,
introduce the nominal 80% full-scale methane calibration gas and adjust the
analyzer CH, and THC span controls for the corresponding channels to obtain
                                    169

-------
an output indication corresponding to the certified analysis of this
Calibration gas.  Recheck zero and span until adjustments are no longer
necessary.  Then sequentially introduce intermediate value calibration
mixtures of 10, 20, 40 and 60% of full scale.  Plot the output values
corresponding to the cerfifed analysis concentration numbers and
construct a calibration curve.  The response should be linear within
+1% of full scale.  If the calibration points do not lie within this
limit, the calibration gases may need replacement or there may be some
equipment malfunctipn which should be isolated and corrected.
     Using the same humidification level as for the calibration of the
CH, and THC channel above, introduce a series of propane calibration
mixtures at levels of 10, 20, 40, 60 and 80% of full-scale response.
Plot the analyzer NMHC values corresponding to the certified analysis
concentration numbers (as equivalent methane) and construct a NMHC
calibration curve.  The slope of this curve determines the NMHC response
factor of the particular analyzer under calibration.  If more than one
humidity level is to be used, repeat the above procedures at each level.
     6.3  Routine Operation
          Following calibration, manufacturer's instructions should be
followed to commence routine analysis operation.  Performance should be
checked for several cycles to assure that automatic operation is reliable
before committing the instrument to unattended operation.  Subsequent
inspections at least every 72 hours and preferably every 24 hours should
be made to verify continued satisfactory operation.
     6.4  Routine Calibration
          Periodic checks and adjustments as necessary, at least every
72 hours and preferably every 24 hours, or zero and span should be made
to assure data quality.   Some instrument manufacturers may offer automatic
zero and span cycles as an option.  When these are not used, manual
calibration checks using span (nominal 80% full scale) and zero gases
can be scheduled to coincide with routine maintenance tasks such as
reagent replenishment.   An operational log should be maintained to aid
in identifying incipient problems with calibration stability,
                                    170

-------
7.0  QUANTIFICATION OF AMBIENT AIR HYDROCARBON LEVELS
     7.1  Difference Methods
          For  the batch difference method  the output indication pairs
corresponding  to total hydrocarbons  (equivalent methane) and methane
for a given cycle are converted  to equivalent ppm methane in air
directly from  the calibration curve  for the respective channels.  The
methane value  is subtracted from the total hydrocarbon value and the
difference is  multiplied by the  NMHC response factor to obtain the
value of the methane equivalent  to the non-methane hydrocarbons in the
sample.
     For the continual difference method the outputs corresponding to
the methane channel and the NMHC channel are directly cpnverted to
equivalent CH, concentrations using  the respective calibration curves for
each channel.
     7.2  Backflush Method
          The  output indication  pairs corresponding to the methane
and non-methane hydrocarbons, expressed as equivalent methane, for a
given cycle are converted to ppm methane in air directly from the,
calibration curves for the respective channels.
     7.3  Equivalent Mass Concentration
          Conversion of the value of ppm methane in air obtained from
the procedures above to an equivalent mass concentration of carbon in
air is obtained by multiplying the ppm CH, by 0.654 to obtain equivalent
carbon (as CH,) in milligrams per cubic meter of air at 25°C and 760
ton.
8.0  SOURCE OF ERROR
     Sources of error in environmental monitoring instrumental measurements
are discussed  in detail in reference 4.  Some of these error sources
of particular  importance in hydrocarbon measurements are briefly pointed
out below to aid in operator recognition.
     8.1  Interferences
          Other gas and vapor phase  pollutants at the concentrations
which are likely to occur in ambient air,  with the exception of water
vapor, do not  interfere with this hydrocarbon measurement method.  Water
vapor can introduce measurement  errors by  altering flame ionization
                                     171

-------
detector response.  The magnitude of this effect can vary from instrument
to instrument, depending upon design.  Conflicting reports have been
published in the literature.  One report  has indicated that a change of
over 4% in response may pccur over the normal range from 40% RH to
100% RH at 25°C air temperature when the water vapor is injected directly
with the hydrocarbon sample, as is done in those analyzer designs that
use the difference method.
     8,2  Zero Drift
          Most commercial GC-type instruments incorporate electronic
circuitry which automatically corrects for baseline drift when the
instrument is working as designed.  Usually this correction is updated
once per measurement cycle when the instrument is, used in the bargraphic
mode.   In addition, there are usually adjustments to set the zero position
of the pen on the strip chart recorder.   Instrument manufacturers specify
the performance of these functions, usually expressed as a percentage
of full-scale indication over some definite time period-  The stability
of these adjustments can be affected by many factors.   Conformity with
these specifications should be routinely checked in order to prevent
gradual or catastrophic degradation of performance causing the generation
of spurious data.
     8.3  Span Drift
          Most commercial instruments incorporate controls for adjusting
the amplification factor of electronic amplifiers in order to obtain a
desired output signal level (full scale indication) qorresponding to a
given input concentration.  The stability of these adjustments,  once
made,  is usually specified by the manufacturer as a maximum deviation,
usually expressed as a percentage of full scale over some definite time
period.  This stability may be affected by many factors and should be
routinely checked in order to prevent gradual or catastrophic degradation
of performance causing the generation of spurious data.
     8.4  Calibration Gases
          The use of the calibration gases is required tp allow adjustment
of the analyzer zero and span controls in order to obtain a calibration
curve.   The specificatiqns of these calibration gases are given in Section 4.
                                    172

-------
The calibration procdure is given in Section 5.  Care should be exercised
to assure that the calibration gases conform to the specifications.  Gases
should be ordered with an analysis certification by the supplier traceable
to the National Bureau of Standards, Standard Reference Materials 1658,
1659 and 1660 as appropriate.
     Pressure regulators, gauges and connecting tubing used with the
calibration gases should be scrupulously clean to prevent the spurious
introduction of unknown and uncontrolled amounts of hydrocarbons into
the analyzer.
     8.5  Support Gases
          Specifications for fuel, carrier and combustion support gases
are given in Section 4.  Variable hydrocarbon content from tank to tank
can cause baseline shifts.  In analyzers that have automatic baseline
correction, these shifts can be compensated.
     8.6  Data Reduction
          In the difference method, where peak heights are determined by
measuring scale displacements and subtracting the methane value from
the THC value, the occurrence of large values of methane and relatively
small values of NMHC can introduce significant error in the determination
unless exceptional care is exercised in the graphical measurements.  Use
of digital technqiues can alleviate this problem to the point where
calibration stability of the FID and associated electronic signal
processing circuitry is the limit.
     The backflush method does not require a subtractive step; however,
it may suffer from problems inherent with the strip chart recorder
graphical reduction process.  If the NMHC concentration is small, the
corresponding pen displacement will be small and may not correspond to
the actual value due to dead-band and backlash effects in the recorder
pen servo-system.  This problem can be alleviated by digital, signal
processing techniques or by scale compression-expansion methods.
     The sample temperature and pressure should be known in order to
reduce the concentration to standard conditions.  Errors in the measure-
ment of these values can cause similar errors in the calculation of HC
concentrations.
                                    173

-------
9.0  REFERENCES

1.  	.  Air Quality Criteria for Hydrocarbons.  National Air
    Pollution Control Administration Publication No. AP-64.  March 1970.
    Chapter 3.

2.  	.  Environmental Protection Agency, Title 40, Code of
    Federal Regulations.   Part 50-National Primary and Secondary Ambient
    Air Quality Standards.  Federal Register, 36, 8186.  April 30, 1971.

3.  	_.  Environmental Protection Agency, Title 40, Code of
    Federal Regulations.   Part 50-Reference Method for Determination of
    Hydrocarbons Corrected for Methane.   Federal Register,  36,  8198.
    April 30,  1971.

4.  	.   Technical Assistance Document for the Measurement
    of Non-Methane Hydrocarbons in Ambient Air.   In draft  form  from the
    Quality Assurance Branch of the Environmental Monitoring and Support
    Laboratory.

5.  	.   Guidelines for Determining Perfqrmance Characteristics
    of Automated Methods  for Measuring Nitrogen pioxide and  Hydrocarbons
    Corrected  for Methane In Ambient Air.   Environmental Protection Agency,
    Research Triangle Park,  N.  C.   EPA Publication No.  EPA-650/4-74-018.
    November 1974.

6.  Harrison,  J. W.,  M. L. Timmons, R. B.  Denyszyn and C. E.  Decker.
    Evaluation of the EPA Reference Method for the Measurement  of Non-
    Methane Hydrocarbons  - Final Report.   EPA-600/4-77-033.   Research
    Triangle Institute, Research Triangle  Park,  N.  C.   June  1977.

-------
                                   TECHNICAL REPORT DATA
                            (Please read Instructions on the reverse before completing)
 1. REPORT NO.
                              2.
 4. TITLE AND SUBTITLE
    Evaluation of the EPA Reference  Method for the
    Measurement of Non-Methu ne Hydrocarbons - Final
    Report
                                                           3. RECIPIENT'S ACCESSIOI*NO.
              5. REPORT DATE
                 May 1976
              6. PERFORMING ORGANIZATION CODE
 7. AUTHOR(S)
   W.J.  Harrison, M.L. Timmons, R.B.  Denyszyn and
   C.E.  Decker
              8. PERFORMING ORGANIZATION REPORT NO.
 9. PERFORMING ORGANIZATION NAME AND ADDRESS
    Research  Triangle Institute
    Research  Triangle Park, North Carolina
27709
              10. PROGRAM ELEMENT NO.

                  1HD621
11. CONTRACT/GRANT NO.
                                                              EPA  68-02-1600
 12. SPONSORING AGENCY NAME AND ADDRESS
    Environmental  Monitoring and Support  Laboratory
    Office  of Research and Development
    U.S.  Environmental Protection Agency
    Research  Triangle Park. North Carolina   27711
                                                           13. TYPE OF REPORT AND PERIOD COVERED
              14. SPONSORING AGENCY CODE
                  EPA/600/08
 15. SUPPLEMENTARY NOTES
 16. ABSTRACT
        Many problems have been reported with  the  method and instruments presently
   used to monitor ambient air hydrocarbons.   The  study reported here was carried  out
   to determine,  if possible, the sources of error inherent with the present  technique
   and to make  recommendations as to what modifications can be made to eliminate or
   minimize these errors.   A flame ionization  detector and gas chromatographic
   instrument were designed, built and evaluated to determine the effects of  operating
   parameters on  hydrocarbon response and the  differences in response to various
   non-methane  hydrocarbon (NMHC) species.  This instrument was then used in  a  com-
   parative evaluation with six commercial  instruments.  The evaluation included
   determinations of calibration stability  and response to various NMHC species.
   Following measurements  with gases from cylinders from commercial sources,  the
   commercial instruments  were further compared in a three day test using ambient
   air.  Calibration stability was found to be reasonable, with span shifts of  greater
   than 5% the  biggest problem.  There were wide differences in responses to  different
   NMHC species.   These differences were somewhat  reduced by using propane responses
   as the basis of calibration rather than  methane.  When ambient air was analyzed
   there were large discrepancies between analyzer readings which appeared to be
   related to atmospheric  water vapor content.   Recommendations are made for  changes  in
   technique to minimize analyzer discrepancies.   	      	
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
   Hydrocarbons
   Air Pollution
   Flame Ionization  Detector
   Gas Chromatography
 b.IDENTIFIERS/OPEN ENDED TERMS  C.  COSATl Field/Group
    Air Pollution
    Quality Assurance
                  13B
 8. DISTRIBUTION STATEMENT

   RELEASE TO PUBLIC
 19. SECURITY CLASS (ThisReport)
    Unclassified
                            21. NO. OF PAGES
                                              20. SECURITY CLASS (Thispage)

                                                 Unclassified
                                                                         22. PRICE
EPA Form 2220-1 (9-73)

-------