EPA-R2-73-180
March 1973            Environmental Protection Technology Series
Evaluation of Measurement Methods
and Instrumentation
for Odorous Compounds
in  Stationary Sources
Volume II, Field Testing

                            Office of Research and Monitoring
                            U.S. Environmental Protection Agency
                            Washington. D.C. 20460

-------
                                              EPA-R2-73-180

Evaluation of Measurement  Methods

            and  Instrumentation

         for  Odorous  Compounds

           in  Stationary Sources

                Volume II, Field Testing
                           by

                        H. J. Hall

              ESSO Research and Engineering Company
                Government Research Laboratory
                  Linden, New Jersey 07036
                       GRU.2DJAB.73


                   Contract No. 68-02-0219
                  Program Element No. 1A1010
                EPA Project Officer: F. C. Jaye

                Chemistry and Physics Laboratory
              National Environmental Research Center
            Research Triangle Park, North Carolina 27711
                       Prepared for

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

                        March 1973

-------
This report has been reviewed by the Environmental Protection Agency
and approved for publication.  Approval does not signify that the
contents necessarily reflect the views and policies of the Agency,
nor does mention of trade names or commercial products constitute
endorsement or recommendation for us.
      NOTE:  Volume I of this report was issued as APTD-1180
                                 11

-------
                            TABLE OF CONTENTS
                                                                      Page
1.  INTRODUCTION .........................     1
2.  SELECTED APPROACH
    2.1  Selection of Instruments .................      /
         2.1.1  Instruments not Tested ..............      g
         2.1.2  Initial Adjustments to Instruments ........      g
    2.2  Sulfur Odorants .....................    10
         2.2.1  Odorants at Selected Field Sites .........    10
         2.2.2  Laboratory Gas Samples ..............    ,,
    2.3  Test Sample Requirements .................    ,r
         2.3.1  Gas Manifold ...................    15
         2.3.2  Field Sampling ..................    ig
         2.3.3  Effects of Water Vapor on
                Specific Instruments ...............    24
    2.4  Data Logging .................. .....    2g
         2.4.1  Instrument Output Limits .............    26
         2.4.2  Recording System .................    32
         2.4.3  Linearity of Scales. .... ...........    -jg
                2.4.3.1  Barton Calibration
                         Curves ..................    o
                2.4.3.2  Bendix ..................     42
                2.4.3.3  Houston Atlas and RAC ..........     45
    2.5  Equipment Van ......................     AC
3.  PERFORMANCE OF INSTRUMENTS ..................     53
    3.1  Response Time and Zeroing ................     53
         3.1.1  Barton ......................     co
         3.1.2  Bendix ......................     58
         3.1.3  Houston Atlas. . .  ................     eg
         3.1.4  RAC Sampler ....................     62
         3.1.5  Overall Comparison ................     ,^
                                  iii

-------
                       TABLE OF CONTENTS  (Cont'd)

                                                                      Page
    3.2  Performance with Specific Gases  .............     55
         3.2.1  Hydrogen Sulfide  .................     55
                3.2.1.1  Basic Parameters .............     55
                3.2.1.2  Changes  in Scale .............     71
                3.2.1.3  Tests for Accuracy ............     77
                3.2.1.4  Effect of Stack  Gas CO/CO ........     81
         3.2.2  Interference from SO  ...............     83
                3.2.2.1  Effect on H^S Response
                3.2.2.2  KAP Scrubber for SO
                         Removal  .................     gg
         3.2.3  Effects of COS  ..................     93
                3.2.3.1  Response and Interferences ........     93
                3.2.3.2  Effects of KAP Scrubber .........     96
                3.2.3.3  Alternate GC Packings  ..........    104
         3.2.4  CSH and Heavier Sulfides .............    107
         3.2.5  Stack Gas Results .................    112
                3.2.5.1  Refinery Glaus Plant ...........    112
                3.2.5.2  Kraft Mill Furnace ............    116
    3.3  Operating Limitations  ..................
         3.3.1  Instrument Advantages and
                Disadvantages ...................
         3.3.2  Data Logging Limitations .............    130
         3.3.3  Sampling System Limitations ............    132
4.  GENERAL CONCLUSIONS AND RECOMMENDATIONS ............    134
    4.1  General ..........................    134
    4.2  Barton Titrator .....................    135
    4.3  Bendix Environmental Chromatograph ............    137
    4.4  Lead Acetate Tape Systems ................    139
    4.5  Minimum Requirements for Future
         Instruments .......................    139

-------
                       TABLE OF CONTENTS (Cont'd)




                                                                      Page
BIBLIOGRAPHY ........... ................




APPENDIX I:  IBM PROGRAMS .....................    A-l




        II:  REGRESSION ANALYSIS  ...... ...........    A-9

-------
                             LIST OF FIGURES

No.                                                                   Page

 1        DESIGN OF PROBES AND GAS MANIFOLD	      17

 2        PUMP-TRAP SYSTEM 	      22

 3        LINEARITY TO H S, BENDIX VS BARTON	      28

 4        ENERGY LOSS IN FLAME PHOTOMETRIC
          DETECTOR	      30

 5        ESTERLINE ANGUS TAPE	      34

 6        DAILY PRINT-OUT	      36

 7        BARTON CONVERSION FACTORS	      40

 8        DISCONTINUITY IN H S FACTORS	      43

 9        ODORANTS VAN LAYOUT..	      47

10        VAN INTERIOR:  FRONT	      48

11        VAN INTERIOR:  REAR	      49

12        VAN ON SITE	      52

13        BARTON CHART TRACES	      56

14        EFFECT OF ALTERNATING AIR BLOW
          IN SAMPLE	      61

15        EFFECT OF CO- ON BENDIX RESPONSE	      82

16        BENDIX LINEARITY TO COS	      94

17        CSH RESPONSE	     109

18        TAIL GAS FROM REFINERY  CLAUS PLANT	     113

19        KRAFT MILL STACK H S/COS/CSH	     117

20        KRAFT MILL STACK H2S/S02/CSH 	     118
                                   vi

-------
                             LIST OF TABLES

No.                                                                   Page

 1        INSTRUMENTS SELECTED ..................      4

 2        MANUFACTURERS1 CLAIMS (INSTRUMENTS AS
          OF 1971) ........................      5
 3        WORKING GAS BLENDS
 4        BARTON TITRATOK, H_S CONVERSION FACTORS
 5        NORMALIZED CONVERSION FACTORS                                4
          FOR BARTON .......................     44

 6        BARTON RECORDER ZERO LEVEL .RANGE
          PERCENT OF FULL SCALE ..................
 7        PERFORMANCE CHARACTERISTICS FIELD
          TEST RESULTS, ZEROING AND RESPONSE ...... .....     64

 8        LINEARITY TO H S CONCENTRATION,
          PARALLEL TESTS SYNTHETIC BLENDS,
          NOMINAL 3-12 ppm ....................     66

 9        LINEARITY AND PRECISION (MV AT
          3, 1.5, 0.5 ppm) .....................    68

10        DIFFERENCES IN STABILITY (H2S BLENDS) ..........     69

11        BARTON DEVIATIONS WITH CHANGES IN
          GAS RATE ......... ...............     70

12        LINEARITY TO CHANGES IN SCALE ..............     72

13        LINEARITY AND SPEED OF RESPONSE
          (MV READINGS) CELL //1 ..................     74

14        SLOW RESPONSE ON BARTON xO.3 SCALE ...........     76

15        BENDIX PHOTOMULTIPLIER EFFECTS .............     78

16        TESTS FOR ACCURACY ...................     80

17        EFFECT OF S0_ ON H S RESPONSE ..............     85

18        EFFECT OF S02 CYCLES ON H2S RESPONSE ..........     86

19        EFFECT OF FRESH SOLUTION IN KAP
          SCRUBBER ........................     89
                                  vii

-------
                         LIST OF TABLES (Cont'd)

No.                                                                   Page

20        KAP SCRUBBER:  COMPENSATING EFFECTS
          BENDIX/BARTON READINGS FOR H2S/COS/S02 	     91

21        RESPONSE TIME WITH S02 SCRUBBER IN
          LINE	     92

22        EFFECT OF COS IN USED KAP SCRUBBER	     97

23        COS + SO  IN KAP SCRUBBER	    100

24        CONTINUOUS FLOW KAP SCRUBBER	    103

25        SCREENING TESTS ON ALTERNATE PACKINGS	    106

26        RESPONSE TO CSH/CSC/CSSC BLENDS	    HI

27        CORRELATION OF COS AND MILL
          OPERATING LOG	    119

28        EVALUATION OF INSTRUMENTS	122
                                   viii

-------
                                  - 1 -
                            1.   INTRODUCTION






          The project covered herein relates to the evaluation of instruments




commercially available for the measurement of H S and reduced-sulfur odorants




in stack emissions, under field conditions.  The project started with a




state-of-the-art review to determine basic requirements, covered in Volume I




of this report.  Instruments selected on the basis of this review were ac-




quired on loan from the manufacturers or from EPA, and tested first in the




laboratory and then in the field.




          The nature of the problem changed significantly as soon as the




test program began, when it became apparent that almost none of the instru-




ments available had in fact been field tested in the stack emissions range.




A single instrument, the Barton Titrator, had been used extensively in kraft




mills only, and the others were either ambient instruments or prototypes.




Further work was based on simple modifications of these instruments to adapt




them to measurements in the stack emissions range.  The following meanings are




used for basic terms in this discussion.




      commercial availability - an instrument which can be purchased




      on the open market and used for the stated purpose,  as de-




      livered.   It should not require research work other  than fol-




      lowing instructions provided, either to put it on stream or to




      keep it in operation.






      field use - the instrument can be operated with routine mainte-




      nance, in the absence  of  laboratory facilities (such as running




      water) , by properly trained non-technical personnel  following

-------
                                  - 2 -
     standard instructions.  The amount of periodic replenishment


     needed for consumable supplies such as compressed gases or rea-


     gents should be predictable, and infrequent.



     sulfur-'containing odorants - the instrument should measure both


     H-S and organic sulfur odorants, preferably as separate readings,


     free of interference from common stack gases.  SO  can either be


     excluded or measured separately.  Satisfactory alternates are to


     measure either  h_S  and total reauced  sulfur,  or  seieccea  major


     individual  S  compounds present.   These are primarily H2S,  SO,,,  COS


     (carbonyl sulfide)  and CSH (abbreviation used herein for  methyl-



     mercaptan).



     stack emissions - the instrument and its sampling system must be


     able to operate on a gas which is hot, wet, and dirty, containing


     percentage  concentrations of water vapor, C0», CO and possibly


     SO , plus NO .   The typical range of odorant emissions is about 1 to
       £.         X


     30 ppm, or  more broadly, 0.1 to several 100 ppm.  Problems of in-


     terferences and sampling are very different in this range than


     they are in the ambient range which is more dilute, or the pro-


     cess gas range which is more concentrated.  The ratio of


     potential interferants to odorants can easily be 1000:1, or more.


          None of the instruments available proved capable of making the


routine measurements desired, in the refinery and paper mill stacks


selected for field testing.  Accordingly, the procedure for their quanti-


tative evaluation was modified (1) to achieve operability, then (2) to


develop their relative advantages, disadvantages and problems encountered.


The selection of procedures and  instruments will first be considered in


more detail.

-------
                                  -  3  -
          Acknowledgment is made of the effective assistance of




Mr. Leigh Gove as research technician throughout this project, and




Messrs. G. J. Piegari and A. C. Hall as engineers, F. W. Church and




R. S. Brief as staff advisors, and P. K. Starnes and P. B. Gerhardt as




analytical consultants.  Invaluable cooperation and hospitability were




received from the Exxon Company, U.S.A. (former Humble Oil and Refining




Company) in providing a site for refinery field tests, and from the




S. D. Warren Mill of the Scott Paper Company at Westbrook, Maine for




paper mill field tests under the direction of Messrs. R. T. Labreque




and S. T. Broaddus.




          Loans of equipment were arranged through the courtesy of




Messrs. Robert K. Stevens and Frederick C. Jaye of the Environmental




Protection Agency,  John Robison of ITT Barton, Oliver Cano of Bendix




Process Instruments, Charles Kimbell of Houston Atlas, John Rex of




Research Appliances Corp., Frank Kabot of Philips Electronic Instruments,




R. J. Joyce of Dohrmann Envirotech, and Troy Todd of Tracor, Inc.

-------
                         2.  SELECTED APPROACH
2.1  Selection of Instruments

          Three principles of measurement were selected for study:

coulometry, flame photometry, and lead acetate tape sensing.  Eight instru-

ments using these principles were acquired on loan, or on nominal rental.

These are listed in Table 1, and manufacturer's claims for these instru-

ments are summarized in Table 2.  Four varieties of coulometers were

included, two flame photometric detector systems (FPD) combined with gas

chromatography (GC), and two lead acetate tape sensors (PbAc2>:
                                 Table 1
    Principle and Instruments

   Coulometry

     Barton Titrator

     Dohrmann Microcoulometer
       Oxidative Cell
       Reductive Cell

     Philips SO- Monitor

   GC + FPD

     Bendix

     Tracer

   PbAc2 Tape Sensors

     Houston Atlas

     RAC
          Auxiliary Equipment
    bubbler, chemical absorption system
Pyrolysis furnace, independent GC
Direct injection or sample valve

Dry absorption filters (ambient)
None; integral unit

Separate function selector"



Pre-pyrolysis + catalytic reduction

Dynamic dilution

-------
TABLE 2
Ranges:
Instrument + Method ppm H2S
(7)
Barton 400 .010-1000
Br/HBr Cell
Bendix 8700 (11)
GC/FPD
Houston-Atlas (3)
825 .010- 20
PbAc2 Tape
855 Same
+ red. to H2S
RAC 5000 .001 to .15
PbAc2 tape
Dohrmann
oxidative MC 301 . 1 at 10 ul
furnace, I2 Cell to
reductive MC 401 200 at 2 pi
Ag Cell
Philips PW 9700 .010 to 5
MANUFACTURERS
Measures
Total S Compounds

Yes all together
Yes TS/H2S/S02

No H2S (low S02)
(Ag tape, + S02)
Yes all together
No H2S (low S02)
Yes GC effluent
No H2S, CSH direct
No H2S or S02
CLAIMS (INSTRUMENTS AS OF 1971)
Zero drift, Accuracy Unattended
COS Time Cycle 24 hrs. + % FS Repeatability Operation

No Continuous reset 4% 4% 7 days
(1 hr. for timed at 10 ppm
filter system)
as H S 5 min. 1% 2% 1% 7 days
(2% 3 days)
i
No Continuous (2%) — 5% Yes ,
Yes Continuous 1% — 5% Yes
(linearity)
No 5 min. (2%) 15-30% Yes
multiples
Yes pulsed injection reset 10% 5% No
(at 10 ppm)
Yes pulsed or reset 5% 8% No
continuous (at 10 ppm)
No Continuous .01 ppm .01 ppm 2% 90 days

-------
                                  - 6 -
          The principles of operation of these instruments and the basis




of their selection are discussed more fully in Volume I of this report with




literature references which should be consulted for further information.




Auxiliary equipment required for each instrument is listed separtely,




because it had an important bearing on the performance of the instrument




in every case.  The failure or inadequacy of this equipment for field use




ruled out at an early stage the Philips, Dohrmann oxidative system and




Tracor instruments as originally acquired, leaving five instruments for




the bulk of the program.  These were primarily the Barton, Bendix,




Houston Atlas (combination unit) and RAG, plus the Dohrmann reductive




cell.




     2.1.1  Instruments not Tested




          The Philips S02 Monitor enjoys an excellent reputation for good




engineering and reliable field performance in the measurement of ambient




concentrations.  Its instruction manual is the best and easiest to use




of all the instruments received, and the laboratory data obtained were




in line with manufacturer's claims.  It was dropped before field evaluation




for four main reasons:




     (a) The maximum range to which it could be adjusted by the manu-




         facturer for direct measurement of H~S was below 5 ppm.   This




         is barely above the ambient range, and too low to be useful




         for monitoring even a well-controlled stack without dilution.



     (b) The barium acetate filter system provided to remove S02 when




         measuring H2S is designed only for ambient use.  It is rapidly




         exhausted when the sample contains a large excess of SO^,

-------
                                   -  7 -










         which is characteristic of many stack emissions.  Also,




         no provision is available for switching filters so that




         the same unit can be used for either H S or S09.




     (c) Safety features built into the instrument limit its use to




         ambient pressure.  It could not accept even a 1 mm (Hg)




         temporary pressure fluctuation, on starting gas flow from




         the manifold used to distribute gases to the instruments




         on test.  This could probably have been corrected.




     (d) Very poor service was received from the manufacturer in attempts




         to adjust the filter system and the instrument's safety




         pressure controls for our use.   Service was good on routine




         maintenance, but anything not routine was referred to the home




         office in Holland, which took six months or longer for a reply.




          The Dohrmann Microcoulometer has many operating requirements




which are difficult to achieve in the field.  First is the necessity for




daily calibration, changing the cell electrolyte, and constant skilled




attention to keep the instrument in adjustment.  The oxidative cell system




converts all sulfur compounds present to S0_ using a high-temperature




pyrolysis furnace, which has a large demand for cooling water.  A portable




system for cooling water was tried out but this proved to be an undesir-




able nuisance, and tests on the oxidative cell were discontinued.  A




limited amount of further testing was carried out on the Dohrmann




reductive silver cell, which has the advantage that is responds to




either H~S or mercaptan but is not sensitive to S0_.  This unit could

-------
                                  -  8 -
be operated in the field, and gave somewhat smoother results in continuous

operation than on spot samples.  This was of strictly limited interest,

however, because the electrode was consumed as well as the electrolyte.

          A Tracer Model 250H unit combining GC/FPD was loaned to the  project

by EPA,  as a possible  alternate  to the  Bendix unit using  the same approach.

The Tracer system was essentially a  prototype  with separate boxes housing

a"function selector" for electronic controls, and the equipment for separat-

ing and sensing.  The function selector box was inoperable as received, due

to damage in shipment.  Reasons for this could not be diagnosed with the

incomplete operating manual supplied, and the control box was returned to

the manufacturer at his request.  It was apparent that the GC oven and

other parts of this equipment were riot as far along in engineering as the

Bendix unit, and no further tests on it were made.  An improved Model

270H has been announced during 1972, combining the function selector

and other equipment in a single box.

     2.1.2  Initial Adjustments
            to Test Instruments

          The Barton Titrator was ready to go on stream as received.

Two modifications were made after initial laboratory tests.  A one-liter

surge pot was removed from the sample line, to give a much more rapid in-

strument response.  This surge pot has several functions which are discussed

further below, one of which is to decrease the sensitivity of the instru-

ment to short term variations in sample composition.  With this more rapid

response, the recorder chart speed was set at 30" per hour instead of the

-------
                                   -  9  -
usual 2" per hour, to simplify the diagnosis of cell behavior.  A one-




liter chemical bubbler is also provided to scrub S0? out of the sample




gas.  This proved to be a limiting factor in operability, and all tests




were made without this unit in line except where its presence is




specified.




          The Bendix unit was operable as received, after tightening




slide valve tensions, but limited in range to about 2 ppm of H^S/SO .




This range was increased tenfold by cutting back on the photomultiplier




voltage, and another fivefold by cutting the GC sample loop size from




5 cc to 1 cc.  The FPD sensor at this setting can detect up to 100 ppm




of most S compounds, or 30 to 50 ppm of COS,  with some loss of linearity




at the upper concentrations.




          The Houston Atlas System was tested first as the Model 825




tape sampler only.  This was replaced early in the program by a



Model 855 combination, in which a pyrolytic chamber is followed by




catalytic reduction to convert all S compounds present to t^S, for




the tape sampler.  The 825 was operable as received, and the 855 com-




bination was set up by the manufacturer, with no complications.  A more




accurate feed control valve was added.  The instrument is direct reading




to 25 ppm, and this can be increased tenfold by a sample timing sector.




Higher dilutions using a dual dilution option gave erratic results and




this system was not employed.   No other changes were required.

-------
                                  - 10 -
          The RAG is limited strictly to the ambient range.  An effective




dilution ratio as high as  2000:1 was achieved simply by cutting the




cumulative sample time   to 5 minutes and using the sample feed pump to




aspirate a dilution gas  into the system, while cutting back at a con-




trolled rate on sample flow.



          The Dohrmann Microcoulometer which ordinarily operates with a




syringe for sample injection was provided with a 4-port Chromatronix valve.



This was arranged for continuous sample flow, or timed injection from a





sample loop.  This system was operable but not much used.   Smoother




results were obtained on continuous flow but this was of limited interest,




because the electrode was  consumed by the continuous reaction.



2.2  Sulfur Odorants




     2.2.1  Odorants at  Selected Field Sites




          The selection  of odorants considered in this program is outlined




in Volume I.  This was based on previous studies of the sulfur odorants




emitted in stack gases by  petroleum refineries and kraft paper mills.  These




two sources have many features in common.  Hydrogen sulfide is the most




important compound in both.  It is important at very low concentrations up



to 10 ppm as an odorant, and in slightly larger amounts as a potential




toxic hazard.  It is almost always present in larger amounts than the mer-




captans or other higher  S  compounds.  When H2S is present other sulfur




odorants may be, but if  it is absent, they are unlikely.  Its amount can,




therefore, be used, particularly in differential measurements from a single




source, as a semi-quantitative indicator of odor intensity.  It must be




emphasized that this does  not measure odor itself, which is a subjective




human  response.   This program  is  concerned  only  with  the measurement of




odorants.

-------
                                 - 11 -
          Sulfur dioxide, and the small amount of sulfur trioxide which




usually accompanies it in combustion gases, is important as a noxious and




corrosive gas, entirely apart from its odor.  It is the one gas most fre-




quently measured and controlled as a harmful air pollutant, and instrumenta-




tion for the purpose is well developed.  It is not nearly as odorous as  the




reduced sulfur compounds, and it is not considered primarily as an odorant.




It is important in odorant measurements, however, as a component which must




be separated or removed to measure other sulfur compounds.




          There are major differences in the S0_/odorant ratios in stack




emissions from kraft mills and from petroleum refineries.  In the paper in-




dustry sulfur is an expensive raw material, which must be purchased and




conserved for recyling to the system.  The amount of S0» released is kept




to a minimum, and it is of ten  less than the  H2S emitted.  Exactly the



opposite is true for most petroleum refineries.  Crude oils contain excess




sulfur compounds which must be removed or destroyed for corrosion control,




as well as for odor.  Sour crude extracts which are essentially unmarketable




are burned in the refinery to recover their heat content.  Refinery stack




emissions may contain hundreds or thousands of times as much SO- as H_S, and




still be objectionable as to odor for the small amount of reduced sulfur




compounds they do contain.




          Total sulfur content can be a confusing concept because of the




very different effects due to SO  and reduced S compounds present,  which may




come from quite different sources in the plant.   Total reduced sulfur which




correlates better with odor is usually determined by first removing S02»




then measuring what is left.   These measurements are complicated by the fact

-------
                                   - 12 -
 that the sensing device often has a different response factor for dif-




 ferent sulfur compounds.  The results must be interpreted with caution,




 since different compounds give a markedly different odor response for




 the same sulfur content.  The usual plant practice has been to report all




 S compounds "as ILS".





          Total sulfur  (TS) or  total  reduced  sulfur  (TRS) are useful  pri-




marily as differential measurements,  which can be  related empirically to




changes  in air quality  from a given source.   The response from FPD systems




(following separation by GC) has  an advantage  for  either TS or TRS in





being proportional to volumetric  concentration  (ppm).  This is much less




confusing than the response from  coulometric systems which varies from




one sulfur compound to another, depending on its state of oxidation.






          Methyl mercaptan  (CSH)  is an  important odorant.   It  plays a major




role  in  wood pulping, as a reagent  in the  delignification process, and its




quantitative recovery for recycling is  important.  At  the same time,  it




is easily converted either to H S by  cracking,  or  to SC>2 by burning.   Its




amount is decreased rapidly by common methods of pollution control, and




little or none may be found in stack  emissions  from  a well-controlled  plant.




It will normally appear only under conditions where  substantial amounts of




H S are being released.  CSH is also  the simplest  of the alkyl mercaptans




present  in crude oil, or a raw refinery fuel  gas.  Here again  these are




mostly converted to H~S by thermal cracking.



          Carbonyl sulfide (COS)  is an  odorless gas which is much more of a




problem in sulfur odorant analyses than is commonly  realized.  It is much




less  toxic than H2S, but is mistakenly  identified  as HZS by many methods of

-------
                                   -  13  -
 analysis.   It is  present  in many  refinery emissions,  particularly those




 produced under reducing conditions.   In some cases, such as  the stack gas




 from a Glaus  plant  after-burner,  it  may be found in much larger amounts




 than the residual concentrations  of  H S.   COS can also occur in the stack




 gases  from a  kraft mill recovery furnace,  particularly  under overload con-




 ditions.  Unfortunately, it  creates  severe analytical problems with  the




 potassium acid phthalate scrubber system  used by Barton  to separate  SO  ,




 if the sample is  a COS-containing gas.  This had not previously been de-




 scribed.  cs  in  iarRe amounts is also  an  interferent In the Barton.




          Heavier sulfur compounds in the  kraft mill effluents are chiefly




 dimethyl sulfide  (CSC) which is the  thio-ether of methyl mercaptan,  and di-




 methyl disulfide  (CSSC), an oxidation product.  These compounds may  occur




 in relatively high concentrations in process gas streams within the  pulping




 process, but  their amount in kraft mill effluents is normally less than the




 CSH from which they are derived and much less than the amount of H2S.  They




can  appear, however, and make a significant contribution to total odor ef-




 fects, whenever emissions are uncontrolled or temporarily out of control.



 A somewhat analogous situation applies  to  the varying amounts of carbon




 disulfide which may be emitted from a Glaus plant burner stack.  The




 total amount of CSH, CSC, CSSC and heavier sulfides present is included




 in any measurement of TRS, or the difference between TRS and H_S.  The




 control procedure usually preferred  is  to  collect all emissions con-




 taining collectible amounts of any of these odorants, and incinerate to

-------
                                   - 14  -
      2.2.2   Laboratory Gas  Samples



          The  gases  primarily  considered  in  this  study were H~S, SO  ,
                                                              ^     £,


 COS,  and  CSH.   Working gas  samples  of  each of  these were prepared as



 blends  in nitrogen,  as listed  in  Table 3, and  stored as compressed gases



 in No.  1A cylinders.   These cylinders were each aged with the gas



blend before filling and initial analysis by the manufacturer.  This



gave good results except for E^S.  No  satisfactory method of preventing



the decay of H2S in steel cylinders was available, and an aluminum cylinder



was no better.  A Teflon-lined cylinder gave even poorer results, since



the concentration of H_S which it delivered went down and up inversely



to the temperature to which the cylinder was being exposed.  In steel.



cylinders the decay rate tended to fall off with time,  so the gas


could ultimately be used at a known lower concentration.




                                  Table 3
                            Working  Gas  Blends
Sample Gas
                   Blend  in
                                                         Changes in Use
                Nominal   125  ppm



                          200  ppm



                          74 ppm



                          51 ppm



                Nominal   670  ppm



                          22,700 ppm



                Nominal   2460/255 ppm


                          Diluted to 246/26 ppm



                Nominal   205/195 ppm



 CSH/CSC/CSSC   Nominal   500/500/500


                          Diluted to 50/50/50
                                               (Decayed  gradually  to  40  ppm)


                                               (Used mostly at  180 ppm)


                                               (Used mostly at  30  ppm)


                                               (Teflon-lined, mostly  70-85  ppm)
SO
COS/CS,
H2S/CSH

-------
                                  - 15 -
          Concentrations of H S in these working blends were calibrated




against permeation tubes in a constant temperature bath, as recom-




mended by Stevens (1) and 0'Keefe(2).




2.3  Test Sample Requirements




     2.3.1  Gas Manifold




          Initial laboratory tests indicated that all the instruments re-




spond at once and in direct proportion to any change in inlet pressure.




Special provisions are required whenever two or more instruments are con-




nected to the same manifold, to avoid a disturbance to all the instruments




when any one of them passes through a cycle which involves a change in flow.




Tests in parallel also require that the pressure within the sensing cell of




all instruments be the same.  This is complicated by the fact that each




instrument as received has its own pump or aspirator system,  with a dif-




ferent capacity or gas flow rate.   The  expedient of using an over-




sized gas manifold to minimize these effects was ruled out for this




program, because it involves time delays and gas mixing which is




undesirable with samples of continually changing composition.




          The effect of sample gas pressure on instrument reading is a




simple matter of physics.   It is recognized in the manuals, but still fre-




quently overlooked.   The effect of barometric pressure within the sensor,




for example, is almost never controlled,  and yet it alone can cause varia-




tions of 5% in readings on the same sample.  While the effect of ambient




temperature may be greater, as much as 10% or more, it is frequently con-




trolled.   A more frequent  source of error is to depend on a gas rotameter




to measure flow rates.   These instruments repond to changes both in gas




pressure and composition,  and it is only when these variables are held




exactly constant that the  necessary corrections for them can  be  ignored.






(1)   See  Bibliography.

-------
                                 - 16 -
          A pressure controlled manifold system was constructed to minimize




these problems, and used both in the laboratory and in the field.  It was




operated at a positive pressure of 2 psig at all times.  This




was accomplished by feeding into the manifold an amount of gas equal




to 30% or more above the total demand of all instruments on line, and




bleeding  the  excess gas  to vent  through  a simply but  accurate Moore pres-




sure  release  valve  (63  SDL).   The manifold  pressure was let down  into  the




inlet of  each instrument  through a solid block  needle valve, and  monitored




by an open-end mercury manometer which was  held at a  constant level against




the  instrument pump or other  feeding system.




           The design basis of this manifold is  shown  in Figure 1.  This




was  designed  to be compatible with a sampling system  for  field use to  provide




a constant supply of sample at all times during any purging or cleaning




cycle,  starting with dual  stack  probes and  sintered porous metal  filters.




Compressed and de-watered  sample gas was passed through a treated Teflon




line  (up  to 200 feet used).   The same gas manifold was used both  in the




laboratory and in the  field.   Excess gas was vented to the atmosphere  after




every gas  measurement  point,  for dynamic balance.  Bleed  lines just past




the  inlet  points for dilution air or calibration gases were used  for more




accurate  control of gas  flows, rather than  for  the control of pressure.




The  principal pressure  release point through the Moore valve was  placed




at the end of the manifold, past all the instrument feed  lines,  to




minimize  the  effect of valving changes  in any one  instrument on  gas flow to




the  others.

-------
             - 17 -
             Figure 1
DESIGN OF PROBES AND GAS MANIFOLD

-------
                                   - 18  -
          Flow controls for either dilution air or large volumes of




calibration gases were provided through batteries of 3 rotameters each, to




cover the full range from about 5 cc/min. to 3 liters/min.  Mass flow meters




for the more accurate measurement of small volumes of calibration gases were




mounted in parallel to the last bank of rotameters.  These were sensitive




to + 1 cc/min., in the range from 1 to 100 cc.  This sensitivity was




sufficient to detect very small differences in the flow of a given gas




when additional amounts of another were fed to the manifold; all rates were




controlled or readjusted as required for each change in composition.  The




permeation tubes used for H.S calibrations produced blends of about 3 ppm




to 9 ppm, depending on the tube size, bath temperatures, and air flows to




the manifold of 1 to 2 liters per minute.  Outlets were available for the




supply of sample gas directly to different instruments in parallel, before




or after the final gas dilution.




          The basic principles of this design were kept while details were




altered during the project, as shown in Section 2.3.2 below.  The original




design provided a separate pump in each probe line for a positive sample




gas pressure.  This was altered to provide a rapid dewatering of sample gas




by compression and re-expansion,  (see Figure 2) using the second pump for




rapid removal of water.  This displaced the gas refrigerator, which was




moved to the end of the sample line from the probes to the instrument van.




          With this system in use, total gas demand for the four instruments




normally on stream was about 600  cc/minute, and the usual air feed rate to




the manifold was just over a liter (1060 cc), plus sample or calibration




gases.

-------
                                   - 19  -
     2.3.2  Field Sampling




          The problems of getting a satisfactory gas sample  for  analysis




from a stack gas which is hot, wet and dirty are well known  and  chronic.




This is particularly  true of efforts  to get a  continuous  sample  gas  flow,




which have usually been abandoned in  favor of  taking a  spot  grab sample.




The situation is worse for odorants which must be measured at  low ppm or




ppb levels than it is for the usual combustion gas components, which occur




at higher concentrations.  The problem is partly mechanical, partly  physical,




and partly chemical.  Excess water vapor drops out as soon as  the hot gas




is cooled below its dew point, which  is of the order of 140-180°F.   This




water appears in the  lines as a mixture of slugs in gas.  This is dif-




ficult to pump,  and the whole length of the line can act as a gas




scrubber or chemical reaction zone.   The degree to which this changes




the composition of the final gas is  variable,  depending on time and




temperature and the possibility of interactions between the gas and




any particulates present.   All such  changes are undesirable,  and  they




can easily destroy the integrity of  the sample.



          These problems are aggravated if the gas is to be pumped through




the sample line at any pressure above atmospheric.   This increases the




solubility or reaction rate of gases, raises the dew point, and  thus makes




it more difficult to prevent condensation by using heated lines.   Once con-




densate has been formed in a line, the amount of heat which can be supplied




through ordinary heating is not enough to cause much re-evaporation.

-------
                                  - 20 -
          A common expedient to reduce this tendency to condensation is to




run the sample line under reduced pressure, using a pump or aspirator.




The problem is that air leaks into the system are difficult to avoid, easy




to ignore, and hard to detect.  "Quantitative" results reported with a




leaky sample line under these conditions may be analytically correct




but always too low, by an unknown amount.  The alternate expedient of di-




luting the stack gas promptly with air to below its dew point involves




large uncertainties in temperature/volume corrections, and is less desir-




able than direct measurements.




          A solution to these problems requires a method of rapidly removing




both particulates and excess water with a minimum of contact between sample




gas and any condensate, and keeping the sample above its dew point for




the largest possible fraction of the total time until it reaches the




analytical sensor.




          A field sampling system was designed and built to supply sample




gas under pressure, without hot dilution, with low hold-up, and with




minimum scrubbing effect from any condensate.  It consists of five elements:




     (a) Two probes in parallel, alternately on stream at all times




         while the other is "blown back with  air,  each probe includ-




         ing an integral filter inside the stack.






     (b) A combination of  three knock-out traps and two pumps in




         series placed next to the stack, which dewaters the sample




         gas rapidly at ambient temperature  and 10-20 psig, and




         feeds it to the sample line.






     (c) A heated line which keeps the sample gas above its dew




         point until it reaches the instruments.

-------
                                  - 21 -
      (d) A gas refrigeration unit with a centrifugal separator which




         chills the sample rapidly to about 40°F at the instrument




         end of the line, and removes any residual condensate without




         further contact with the sample.




      (e) A pressure-reducing valve on sample fed into the instrument




         manifold, either directly or plus diluent air if desired.




         Both the pressure reduction and any dilution gas help to pre-




         vent any further condensation.




          The system of two probes with alternating blow-back is shown in its




initial design in Figure 1.  This was built to provide a continuous gas




flow, from one probe or the other, without interruption for the necessary




intermittent blow-back by a blast of air.   Air blow-back operates satisfac-




torily with a single probe in the Barton system,  where it produces a blip




of air which is entrapped in the probe and transferred into the sample line




once in every 5-minute cycle.   The single system with intermittent air




flow cannot be used in the testing of a battery of instruments feeding from




the same manifold,  since varying amounts of air could be delivered to dif-




ferent instruments.   In the final design the timing of the two probes was




set by a repetitive self-energizing cycle, where the air blast in the first




started 30 seconds after the second came on stream, and the first came back




on stream 30 seconds after the air blast to it was shut off by the




timer.




          The pump-trap  combination  detailed in  Figure  2 was  primarily re-




sponsible for  the success  of this system.   It  was  added  to  the  sampler




after  the initial combination of dual probes,  pumps, heated lines and

-------
                           - 22 -
                           Figure 2

            PUMP-TRAP SYSTEM FOR RAPID WATER REMOVAL
  U
ALTERNATING
STACK PROBES
t_
                D
                 PUMP #2
                 SLOW
                  I
           H20
                                     PUMPft
                                     FAST
                                          BY PASS
                                                        TO VAN
                                                       SAFETY
                                                        VENT

-------
                                  - 23 -
refrigerator as shown in Figure 1 failed repeatedly in  the same service.




The failures were caused partly by the uneven swelling of gas-resistant




flapper valves in the diaphragm pump, which could handle either water or




gas but was not able to handle alternate slugs of water and gas.  The suc-




cessful combination adds a double gas recycle and low-volume water




traps (A, B, D).  The primary gas recycle returns compressed gas from the




second knock-out trap (B), after the compressor (C), at a variable ratio




from 1/1 to 10/1.  The water from this trap (B) passes  through valve (R),




to the third trap (D) to disengage and return entrained gas, along with  the




fresh feed.  This minimizes sample demand through the probes.  The com-




pressed gas which is recycled through valve (E) has been dewatered by




compression, while feeding a limited amount of fresh sample to the primary




trap (A).  This decreases the excess water load on the compressor (C).   The




rate of water removal pumped (I) from the third trap is adjusted to vent a




minimum amount of gas, usually about 500 cc/minute, with a fresh sample feed




of 2 to 4 liters/minute.  A minimum amount of fresh feed is desirable to




limit the amount of heat dissipation required in the line from the




probes to the Teflon air-blow-back valves, nearest the stack.




          Teflon lines were used throughout, and the use of Teflon valves




in the hot gas line at the stack was felt to be particularly important to




minimize sample degradation.  These valves were part of the standard con-




trol system provided with the Barton probe.  Existing equipment was used




here to simplify the design.  The pump diaphragm was Viton, which is said




by the manufacturer to show less tendency to swell than other elastomers in




S02~containing gases.  The probes, traps (pipe-T's),  pump body and

-------
                                   - 24 -
refrigerator were made of 316 stainless steel.  This use of metal parts

would not be permissible in a system designed for the ambient gas

range, below 1 ppm.  Laboratory tests at an early stage of the program

showed that stainless steel fittings cause a loss in H?S in air of a

fraction of one part per million, on contact with the metal.  This can

be severely limiting In analyses at the 1 ppm level, and catastrophic

below it.  The same loss has much less effect at ppm levels, and pro-

gressively less at normal stack concentrations.  Any effect that it

has in this range is comparable on all samples tested at the same time,

and can therefore be ignored for purposes of comparison.

          The water removed from the refrigerator trap was normally odorless

and showed no evidence of elemental sulfur.  Water removed from the initial

knock-out traps was also nearly odorless, and showed barely detectable

amounts of 4 to 16 microliters of H^S and SO  per liter on colorimetric

analysis.

     2.3.3  Effects of Water Vapor
            on Specific Instruments

          The amount of water vapor in the system is an important differ-

ence between laboratory and field conditions.  Excess water has a specific

effect on many instruments. Tape samplers are designed for a gas which is

moist, but not wet.  A certain amount of moisture is necessary for the

PbAc7 tape reaction, but too much can wet the tape and cause it to break.

Both tape samplers tested had a humidifier vessel, designed to trap excess

water or supply it if the gas was too dry.  This system was a source of

trouble with both instruments.  The RAG had a simple water bottle in which

-------
                                 - 25 -
the gas impinged on the surface.  The RAG could not be used in a plant




environment with ambient air or unanalyzed plant air as a diluent, con-




taining uncontrolled trace amounts of S, and  the use of dry compressed




air caused the humidifier bottle to evaporate in a little over a day.




The Houston Atlas humidifier uses a 5% aqueous acetic acid trap to




either add or remove water, and this does not disappear as fast with




normal gas flow.  The system gave problems with the carryover of spray




from the humidifier into the sensing head, where it wet the tape and




caused intermittent breakage.  Partial condensation or back-pressure




in the elevated vent line may have been a factor.  The 1972 model of




this instrument has been able to eliminate the bubbler for samples in




air, by balancing the amount of water vapor formed in the catalytic




reduction against that required for the tape.




          Stack gas samples are very wet.  The amount of water vapor com-




monly present in a kraft mill recovery furnace stack is 20 to 40%, and a




30% volume shrinkage is assumed in the usual routine calculation of the




concentration of sulfur compounds in the gas as emitted, based on gas as




analyzed.   Similar moisture contents appear in the stack gas from a Glaus




sulfur plant, which produces one mole of water for every mole of sulfur.




          In addition to mechanical problems in sampling, this excess mois-




ture limits the choice of GC packings which can be used for any gas separa-




tion desired.  It is not a problem with the polyphenyl ether packing used




in the Bendix as supplied, or with Poropak Q.   It is measured by Poropak R,




which is a limitation if H_0 is present in very large amounts.  It destroys




the activity of silica-base materials such as Deactigel,  which is one of

-------
                                  -  26 -
the packings recommended for separation of COS from H«S.  These effects




may be moderated or controlled by the use of a guard bed and backflushing,




but it is simpler if possible to select a packing which is not so affected.




          Sulfur dioxide has more of a tendency to hold up in sample lines




in the field than in the laboratory.  This reflects the influence of




moisture which tends to plate out in the lines with the SO , as sulfurous




acid.  Line adsorption of moist SO  can be critical, and significant




even in Teflon lines at highly variable concentrations.




2.4  Data Logging




     2.4.1  Instrument Output Limits




          The output signal from each of the five instruments on test was




designed or adjusted by the manufacturer to 0 - 100 millivolts, full




scale.  The Barton, Houston Atlas and Dohrmann give a continuous measure-




ment, and each has its own strip chart recorder.  The Bendix gives a




separate output  for each of three GC peak heights, on demand or on a




5-minute timed cycle, and does not need a recorder for  routine operation.




A continuous output to record the GC/FPD signal is available, if desired.




This is useful for the study of column behavior or the  analysis of




unknown samples.  The output from the RAC is a continuous timed reading




of cumulative spot density, fed normally to a meter.  Its cycle can




be set on any multiple of 5 minutes up to 4 hours.




          The Barton sensor gives a useful output well  beyond the upper




limit of its recorder scale.  Its ability to give approximately linear




results up  to 300% of scale is a practical advantage.   This may be used




to avoid the complete loss of data under conditions of  unexpected over-

-------
                                   - 27  -
load, if it is linked to an auxiliary system such as a meter or a




digital read-out.  The Barton electronic system gives an almost linear




but delayed response to a change in attenuation scale, and the capability




to operate over-scale can be used to advantage to avoid undesirable




changes in scale setting.  The instrument is most useful in the top 5




of its 7 scales, xl, x3, xlO, x30 and xlOO.




          The Bendix system responds immediately  and arithmetically to




changes in scale setting.  Its upper limiting  factor is not electronics,




but the capacity of the FPD sensor.  The instrument has 11 scales, from




xl to x2000, and a setting for automatic attenuation.  As used in  this




program, the signal was consistently overscale at about 115 MV on  the




x50 or xlOO scale used for most measurements, or at about 45 on x2000.




These limits were quite constant, subject to variations in the cor-




responding zero base.  The Bendix zero was normally constant within




0.5% of scale, and much the best of any of the instruments on test.




The specific model on test lost zero stability after 10 months on




stream; improvements to this sub-system were made by the manufacturer




during the year.




          The limiting value of H9S response for  the Bendix was about




120 ppm,as shown for the x2000 scale setting in Figure 3.  This is a




plot of field test results obtained in paper mill stack measurements,




at a time when Bendix data showed no SO. or CSH in the stack.  Under




these conditions parallel measurements in the  Bendix and the Barton




provide a reliable extended basis for correlations.  The Bendix MV




readings are fairly linear through 100 ppm, but flatten out rapidly

-------
                                                        FIGURE 3
                                          LINEARITY  TO H2S,  BENDIX VS BARTON
      50
      40
o
o
o
CN
30
•o

0)
03
      20
      10
                                                                                                                           t-O
                                                                                                                           OO
                                                                     Readings in stack gas,  at

                                                                     time of no SO- or CSH output.
                                                                               J_
               10     20     30      40    50ppm    60     70      80    90


                                              Barton PPM Observed (xlOO)
                                                                        lOOppm    110     120
130

-------
                                   -  29  -
above this.  All concentrations above 120 ppm are off scale and give




the same reading of 45-46 MV.  Other data show that the off-scale level




at xlOOO is 88-89 MV, or almost exactly linear (two times 45).  The




limit is correspondingly higher at lower attentuationst but with less




difference between scales, appearing at about 113 to 115 MV (above zero)




for the xlOO and x50 scales.




          The theoretical basis for this flattening which places an




upper limit on output has been developed recently by Greer and




Bydalek (3).  They find that the loss of sensitivity of the flame




photometric detector at higher sulfur concentrations is due to the




self-absorption of the energy emitted.  Their basic data curve is




reproduced in Figure 4.  The conversion scale from nanograms of sulfur




(by weight) to ppm is 75% for a 1 cc sample.  As the amount of sulfur




in the flame increases, the loss of energy increases at an increasing




rate.  The inflexion point is at 100 ng,at which 50% of the energy is




lost.  This corresponds to 75 ppm in the instrument under test.  The




curve flattens sharply above 80% loss, at 160 ng or 120 ppm.  The experi-




mental curve in Figure 3 checks this value exactly for its upper limit.




The volumetric relationship in ppm will be the same for other gases,




such as S0_ or CSH, which contain one gram atomic weight of sulfur




per mole.  A lower value was found for the upper limit for COS, which




may be attributed to a difference in flame chemistry for this compound.




          The Houston Atlas tape sensor operates at a low fraction of




the saturation capacity of the PbAc. tape.  As a result, instrument




output up to 25 ppm is nearly linear on scale, but overscale capacity

-------
                      - 30  -
                      FIGURE 4


     ENERGY LOSS IN FLAME PHOTOMETRIC DETECTOR^3'
                         LEGEND
                           •  HYORDGEN  5ULFIDE
                           i  SULFUR DIOXIDE
D.D
                                     BD
        2HD
                 Nanograms of Sulfur

                I           I	
               50         100

               Equivalent ppm (at  25°C)
150

-------
                                  - 31  -
is limited.  Sample dilutions for scale changes are made by a variable




sector on a 20 second timer.  This controls the percent of sample flow




which passes into the conversion chambers, or is vented.  The repro-




ducibility of the timer setting is about + 1% on full scale.  This




becomes + 10% of reading at a dilution of 10:1, however, and corres-




pondingly more at higher dilutions.  The small rotameter and valve




supplied with the instrument are not sensitive enough to cut back on




sample flow per unit time for higher concentrations, and a more accurate




control system was added for some tests.  Zero adjustment is set




manually as desired for the reflectance of each tape.  This is subject




to normal variations of + 2-3% FS at zero for a given tape, at all




levels of reading, with occasional spikes of -10% at a constant gas




input.




          The logarithmic output and narrow range of the RAC instrument




is a serious limitation, except at a very low concentration.  The undiluted




range of the sensor is 0.001 to 0.100 ppm.  This is multiplied by dilution,




but only the lower half of the scale is approximately linear.  The out-




put in terms of percent transmission was converted to percent absorption




for convenience in recording, by subtraction from a constant bucking



voltage.  If logarithmic data are to be used in the unconverted form,




the exact time of starting the absorption cycle is critical and not




well-suited to data logging.  A computer calculation to eliminate this




need is possible, but it requires recording enough data points to




determine differences in slope.  This means taking one minute readings,




which is a nuisance for the other instruments, if the RAC cycle modulus

-------
                                  - 32 -
is cut to the minimum of  5 minutes which was necessary for samples of




a concentration higher  than 10 ppm.  This calculation if made, however,




eliminates the necessity  of determining or assuming an accurate zero for




the tape spot in each cycle.  Similar advantages could be realized by




providing a linearizing circuit to convert the logarithmic output to a




direct measurement of slopes.




          The Dohrmann  cell can operate as a continuous sensor for TRS




within the limit of its capacity, or for the measurement of discrete




injected samples.  The  instrument can be used on stream or tied to a




GC or other pretreatment  system.  The normal output is the area under




a GC peak, which is calculated by a disc integrator but read manually to




interpret the results.  The problems inherent in the Dohrmann system do




not encourage repetitive  sampling of a variable stream.  It is necessary




 to make  critical adjustments  of  electronic bias,  gain, and  range  to  get




 a well-shaped  curve which starts  and  ends at  zero.  These adjustments




 establish a base line which may  vary with each sample, and without them,




 the results are  difficult to  interpret.  When the cell is used as a  con-




 tinuous  indicator  for TRS,  these  adjustments are not so critical.  The




 output  is proportional  to total  concentration of  H~S,  CSH,  and COS in  part




(10-15% response), but not S0_.  This is a useful combination for total




sulfur, but its capacity  in this service is strictly limited by the con-




sumption of the electrode as well as the electrolyte.




     2.4.2  Recording System




          A data acquistion system was designed and built to accept the




simultaneous input of all instruments, print out their MV readings in




parallel for comparison,  and store the data for the calculation of

-------
                                  - 33 -
 standard  deviations.   These  calculations  are based on entries selected




 by manual inspection of  the  printout.   The  recording  system comprised




 three basic  elements:




           a)  An  Esterline Angus  D-2020 Data Logger




           b)  A Tally  P-120  Paper Punch




           c)  An  IBM 1130 Tape Reader-Printer-Computer.




           The Esterline  Angus provides  its  own  printed tape output.




 Readings  on  any number of channels up to  20 are printed in  sequence




 in a single  column, and  simultaneously  digitized for  external output.




 A typical set of  printed entries  is reproduced  in  Figure 5.   The  first




 line in each entry gives the date (by last  digit of the month,  and day)




 and the hour (HR), on  the basis of a 24 hour day.   Succeeding lines




 give the  number of the channel recorded (Ch)  and its  reading in milli-




 volts  (MV).  The  instrument  as set on the 0 - 100  MV  scale  records the




 outputs received  to the  nearest 000.1 MV.   Any  channel can  also be set




 if desired to read to  the nearest 00.01 or  0.001 MV.   It can read and




 print about  one line per second,  or about seven entries per minute when




 reading seven channels on a  continuous  setting.  The  normal setting was



to read once every five minutes,  when operating properly.   The time cycle




of this setting was adjusted to start 2  to 5 seconds after the Bendix and




RAC instruments had completed their 5 minute cycle.  Readings every minute




were taken when it was desirable to follow more closely the output of the




Barton or RAC,  and every 20 minutes or every hour on most overnight runs.

-------
          - 34 -
         FIGURE 5
ESTERLINE ANGUS TAPE RECORD

-------
                                  -  35 -
           Except  for  persistent  and  annoying  mechanical  failures  in




 the  Esterline Angus unit,  this system worked  well  as  designed.  No




 problems whatever were  encountered with  the Tally  Paper  Punch.




           The IBM 1130  was programmed to read the  paper  tape and  feed




 it into disc storage.   Each tape normally represented one  day's operation;




 or a longer period such as a week-end with less  frequent entries.   Four




 computer programs were  set up for this project  (see Appendix I).




 Program 1  reads the tape and stores  up to 300 entries, for  any number




 of channels up to 12  per entry.  The number of channels  entered is




 kept  to a  minimum, depending on  the  number of instruments on stream.




 This  allows more  frequent  entries if desired.  For extra long tapes




 the  tape reader stops after  the  first 300  entries and waits  for




 printing and processing the  first 300 before  it  is started  again to




 continue reading.  The  disc  was  used for  temporary data  storage of  the




 daily record and  discharged  each time another set of  entries  was




 read.




           Program 1A reads  the disc and prints out for each  channel in




 parallel columns  the data  recorded in successive entries.  A  typical




 print-out  is reproduced in Figure 6.  The  first  column is an  arbitrary




 run number, assigned by the  IBM printer  to successive entries on the




 tape.  The next two columns are the date and  time.   Channel number 0




was used at first for the  Barton instrument and  later for a diagnostic




 entry of a zero or constant voltage, after the Esterline Angus troubles




 became chronic.   The next  three columns were  used for Bendix

-------
      - 36 -




     Figure 6




 DAILY PRINT-OUT





CHANNEL   NUMBERS

163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
170
179
180
101
102
183
1R4
185
136
187
168
189
190
191
192
193
194
195
196
197
19R
DAY
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
219
Htf
1545
1550
1555
1555
1555
16 0
16 5
1610
1615
1617
1617
1620
1625
1629
1629
1630
1635
1640
17 0
1720
1740
IB 0
1820
1840
19 0
1920
1940
20 0
2020
2040
21 0
2120
2140
22 0
2220
2240
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
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
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1
54.0
53.6
53. R


54.2
55.5
55.3
54.3


0.0
0.0


0.0
51.1
51.3
50.4
49.5
48 .4
47.6
47.6
47.2
47.1
47.3
46.8
46.2
47.2
47.7
47.7
47.2
47.5
47.4
46.0
^6.1
2
78.9
79.2
79.0


1.4
0.4
0.2
0.0


107.1
103.1


105.6
0.5
0.1
0.0
0.0
0.0
0.0
C.O
0.0
0.0
0.0
-0.1
0.0
0.0
0.0
0.0
-0.1
-0.1
0.0
-0.1
-0.1
3
8.7
R.3
8.4


7.6
7.3
7.2
7.3


0.3
0.2


0.3
3.1
-0.2
-0.5
-0.5
-0.3
-0.5
-0.5
-0.3
-0.7
-0.6
-0.5
-0.6
-0.8
-0.7
-0.6
-0.8
-0.9
-0.fi
-0.7
-0.7
4
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
0.0
0.0
0.0
0.0
o.o
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
5
0.1
0.1
0.1


0.1
0.1
0.1
0.0


0.1
0.1


0.0
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
6
24.1
2^.3
24.1


23.3
23.0
22.8
23.1


14.0
11.1


15.9
22.6
21.6
15.2
1^.2
10.5
10.4
10.7
9.4
9.8
9.1
7.9
7.7
0.1
7.3
6.1
7.2
6.5
4.9
4.2
3.0

-------
                                  - 37 -
channels 1, 2 and 3, which were set for various gases during the pro-




gram depending on the GC column in use.  Channel A was used for the




Houston Atlas, 5 for the RAG, 6 for the Barton, and other channels as




needed for the Dohrmann or auxiliary equipment.  Negative entries were




printed as read, usually as small values during zeroing.  A consistent




column of negative values appears for the RAG when it is set to record




tape absorbance by difference, rather than transmission.




          Program 3 computes the mean and standard deviation for any




number of columns and entries, up to 60 per column.  Instruction cards




for this computation are made out after inspection of the print-out




from Program 1A.  This calculation usually includes all entries between




those selected.  It can also specify taking every nth entry, to cover




a longer period of time, or to pick out readings for the RAG at a 10




minute or longer period while the other instruments are being read




every 5 minutes.




          Program 2 is a special computation of the RAG absorption curve,




which can reproduce the whole curve from a set of three to five points




at regularly timed intervals.  This computation was demonstrated but not




much used, since it requires taking and processing data at intervals




more frequent than those necessary for the other instruments on test.




It can be replaced by the linearizers available as standard equipment




in other units, such as the Tracer 250H.




          The printed tape on the Esterline Angus and the chart record




on the Barton are both convenient places to record changes in sample,




processing, or instrument scale setting.  The Houston Atlas recorder

-------
                                  - 38 -
chart was not adequate for  this purpose.  Records from the lab notebook


and charts were entered on  the daily print-out as shown in Figure 6,


during periods of satisfactory operation, for the interpretation of


data and selection of entries for computations.


     2.4.3  Linearity of Scales


          The instruments on test did not give readings suitable for direct


output in concentration, or for computer conversion of the raw data from


MV observed to calculated ppm.  The principal reasons for this were the


unpredictable variations in zero values for the Barton, by day and by


attenuation scale, and for  the Houston Atlas and RAG for each new tape.


The RAG system to correct for tape zero required an excessive amount of


maintenance to keep it in operation at a high rate of tape feed.  When it


did work, the instrument span and the placing of the data on the logarithmic


output curve were changed by the automated zero shift.  The Bendix zero


was inherently more stable, but its linearity is limited at increasing


concentrations, and each change in GC packing or in gating for a new


set of GC peaks requires a  new calibration curve.


          2.4.3.1  Barton Calibration Curves


          The linearity of  response of the Barton Titrator over con-

                            4
centrations differing by 10 or more is a distinct advantage.   It was


used for this reason as an  intermediate reference in this program,  to


extend the range and volumes of calibration gas blends matched against


permeation tubes or known small synthetic samples.  Since we were


dealing with known blends or pure compounds, the variable response


of the Barton to different  compounds could be allowed for, and did

-------
                                  -  39 -
not constitute a problem in calibration.  The conversion factors


supplied with the instrument do not fall exactly on a  smooth curve,


however, and various efforts to provide a better set of values have been


reported in the literature.


          The handbook data are plotted in Figure  7.   Both  the attenuation


scale and the response scale are logarithmic, and  the  values reported


for five sulfur types are all close to a straight  line with a slope of


1.  Deviations from the line are greatest at the lower attenuations.


Experimental values as determined by various authors and modified  for


use in this report are given in Table 4.  The ranges given  by Cooper

           (4)
and Rossano    refer to different detection cells, which they note


should be calibrated individually for accurate results.  Their values


are not based on all the information available, however, since their


ranges do not include three of the seven factors in the Barton hand-


book.


          A more consistent set of factors can be generated by treating


the values for each compound at different attentuation scales as a


single set of data, and constructing a faired curve by standard


regression analysis.  Adjusted values are shown in Table 4  for the


H S factors of the Barton handbook, and Blosser   .  Details of the


Barton calculation are shown in Appendix II:   it is apparent that the


value of .023 at x.3 is suspect and should be discarded.   A modified


faired curve constructed without it leaves the values at xl and x.l


close to the unit level, for which the instrument was originally


designed.   Preferred values as used in this report are based on this


modified curve.

-------
                                                           Figure 7

                                                 BARTON CONVERSION FACTORS
100
                                                                                                                             o
                                                                                                                              I
                                                                                                                     100

-------
TABLE 4
BARTON TITRATOR
H2S Conversion Factors
Attenuation
Range
x .1
x .3
x 1
x 3
x 10
x 30
xlOO
Preferred
Values
.010
.029
.095
.28
.90
2.6
8.6
Barton
.010
.023
.10
.27
.86
2.6
9.0
Blosser
+ Cooper
.013
.033
.110
.25
.80
2.4
8.0
. (x MV Scale
Cooper +
Rossano
.008-. 013
.030-. 035
.09-. 12
.24-. 26
.7-. 9
1.5-2.5
4.5-9.0
Reading
= ppm)

Faired Curve
Barton
.009
.027
.091
.269
.890
2.66
8.80
Blosser
.012
.034
.103
.284
.867
2.40
7.32
Modified
.010
.029
.095
.278
.904
2.64
8.60

-------
          All  three  of  the sets of experimental  factors reported in




Table 4 show a discontinuity  in the values  reported above and below




the  x3 and  xl  range.  This  is illustrated in Figure  8:   a  line  drawn




between these  two  factors extrapolates  consistently below  the experi-




mental values  for  lower ranges, and above the experimental  line  for




higher ranges.   This corresponds  to a dividing point  in  the  performance




of the Barton  cell:   it equilibrates within 2 minutes or less at the




x3 scale and above,  and much  more slowly  at xl and below.   Directionally,




any  experimental reading which had not  been allowed to equilibrate




fully at the lowest  attenuations  could  be translated  into a  pseudo-




factor which is  too  high.  The uncertainties observed in all factors




were greatest  at the xl scale and below,  which are used only for ambient




measurements.




          The  factors given  for SO-, CSH, CSC and CSSC in  the Barton




handbook are accepted by Blosser. Faired curves calculated  for  these




show a consistent relationship for the data  as a whole.   The equations




for these curves are  given in Table 5,  with  original and adjusted




values for the factors  at each attenuation.  The low-level delayed  re-




sponse to CS_ encountered in  this project is not noted in these  references.





          2.4.3.2  Bendix




          Linearity  of  the Bendix instrument  for a given sample size




is related basically to the limits of the FPD  sensor  and to  the




chemistry of the sulfur compound  reactions  in  the hydrogen-rich




flame, as shown  above in Figures  3 and  4.   The plots  of Bendix readings




vs Barton values show linear  correlations within + 3% for H«S through




about 80 ppm.  Similar  data  are given in Section 3  below  for SO-  and




CSH  through 80 ppm,  and for COS through 30  ppm.

-------
100
                                                                          Figure 8
                                                               DISCONTINUITY IN H2S FACTORS
                                                              10                           JO
             • .BARTON HANDBOOK
             » COOPER -ROSSANO RANGE
             a BLOSSER,  AND COOPER
  /Ol (Barton)
.01 (Cooper-Rossano)
  .Oldlosser)
CONVERSION FACTOR (x Barton MV=ppm H2S)
                                                                                                                                                            OJ
                                                                                                                                                            I

-------
                                - 44 -
                               TABLE 5
NORMALIZED CONVERSION FACTORS FOR BARTON
Scale
xO.l
xO.3
xl
x3
xlO
x30
xIOO
Log-log equation
A
B
Std. Error
Correl. Coeff.
H?S*
.010
.029
.095
.28
.90
2.6
8.6
(A x log
1.022
1.0448
.0203*
.9993
CSH
.013
.039
.128
.38
1.25
3.7
12.1
S02
.021
.063
.21
.62
2.05
6.1
20.3
CSSC
.030
.087
.28
.81
2.6
7.6
24.6
CSC
.035
.107
.36
1.08
3.6
10.9
36.5
factor = log scale -B)
1.012
.9032
.0308
.9997
Barton Handbook Values
xO.l
xO.3
xl
x3
xlO
x30
xIOO
.01
.023
.10
.27
.86
2.6
9.0
.014
.04
.12
.35
1.3
3.5
13.0
1.005
.6861
.0427
.9993
(Unmodified)
.023
.053
.23
.62
2.0
6.0
20.7
1.029
.5692
.0230
.9998

.031
.09
.26
.8
2.5
8.0
25.0
.9954
.4446
.0272
.9997

.035
.1
.37
1.2
3.5
11.0
35.0
*  After discarding reported factor for H_S at xO.3 scale.

-------
                                 - 45 -
          2.4.3.3  Houston Atlas and RAC




          The linearity of the Houston Atlas sensor at + 3% is better




than the inaccuracies due to unevenness in the tape, at + 5 to 10%.  The




simple timing sector gives dilution values which are reliable as far as




10:1, but increasingly less so at values of 20:1 or higher.  An optional




system provided for dual dilution which sends an extra pulse of hydrogen




into the lines along with the timed sample pulse gives non-linear  results,




probably due to the interrupted flow.  This system was not used, and is




not recommended.   A microliter injection system is now available,




designed for liquid samples, which can be adapted for gaseous samples




at high dilution.  The 1973 model of the Houston Atlas tape sampler




has gone to a measurement of slopes instead of direct scale readings.




This should extend its useful range, as well as decreasing its




sensitivity to differences in the tape.  The logarithmic response




curve of the RAC sampler and the necessity to consult a conversion




chart to get even a qualitative result are distinct disadvantages




compared to the other instruments.




2.5  Equipment Van




          For the field test program all instruments, calibration systems




and gas manifold controls were mounted in a mobile air-conditioned




equipment van.  This was a standard production model 8' x 20' office-




type trailer, Model M20 from the Coastal Mobile and Modular Corporation.




The complete sampling console for the probes, filters, blow-back system,




knock-out traps,  pumps and accompanying valves was mounted afthe stack.




This was at the 50* level 100' horizontally from the van location in

-------
                                 - 46 -
the refinery, at a 100' level 30' from the van at the paper mill




recovery stack, and at a 20' level next to the  van at the lime kiln




stack.  A suitable length of electrically heated 3/8" Teflon line led




from the console to the refrigerator at the trailer van.  Power controls




for the heated line were placed  in the van.




          Satisfactory operation was achieved with gauge pressure in




the range of 10 to 15 psig  at the sampling compressor outlet, and a




total sample flow of two to four liters per minute.  The Hankinson




refrigerator unit had a cyclone  separator chamber to drop out any water




separated promptly, with a  minimum of subsequent gas contact.  The




amount of water separated out here at AO°F was normally about 30 cc



in 16 hours, overnight, or  0.1 mol per hour.  This is equivalent to




2 volume %  of water removed, at  a sample  flow of 2 liters per minute.




The rate of sample flow, the use of a heated line, and rapid chilling




and separation were sufficient  to avoid any evidence of sulfur formation




or dissolved S0? in the refrigerator trap.




          The  interior layout of the trailer van is shown in Figures 9,




10 and 11.   Gas from  the refrigerator mounted under the van passes up




into  the gas blending control box for the manifold system, diagrammed




above in Figure 1  (Section  2.3.1).  Either charcoal filtered plant air




or cylinder air may be used for  sample dilution, or a calibration gas




can be used for dilution if desired.  The sample and dilution gas




streams were each  supplied  by a  bank of three rotameters in parallel,




to cover the range from 10  cc to 4000 cc  per minute.  Smaller amounts




of calibration gases were fed directly to the gas manifold after the

-------
                                  _ 47 -
                                  Figure 9

                           ODORANTS VAN LAYOUT
i - $.c- o
3-
4-
5-
6-
7-
S>
                        10
                                 II

-------
                 - 48 -



               Figure 10






         VAN INTERIOR, FRONT




INSTRUMENTS, DATA RACKS, GAS CONTROLS

-------
                - 49 -




               Figure 11






         VAN INTERIOR, REAR




INSTRUMENTS AND ELECTRICAL CONTROLS

-------
                                  -  50  -
blender, through a bank of three mass flow meters.  These were calibrated




for accurate metering of 1 cc to 100 cc per minute of dilute gas blends




in nitrogen.  The total gas supply from the blender to the manifold




could be passed through a permeation tube cell in a water bath, or




the permeation cell could be put in the line from the manifold to




individual instruments to get a higher gas concentration.




          The gas manifold pressure was held at a constant 108 mm of




Hg (+ 0.5 mm), equal to 2.09 psig.  Gas from the manifold was throttled




down to atmospheric pressure by individual needle valves controlling




the sample feed to each instrument.  This was monitored by an open-end




Hg manometer controlled to + 1 mm at each instrument for feed to the Barton,




Bendix and Dohrmann cells.  The Houston Atlas pump which was disconnected




for these tests normally comes before the instrument and flow is controlled




by a rotameter, against the normal pressure drop through the instrument.




Sample flow was normally held at full scale of 0.3 CF/hr plus 0.3 CF/hr




of hydrogen, but it could be cut to e.g., 10%, 20% or 40% of F.S. for




samples of variable high concentration.  Flow through the RAC was




controlled by aspirating dilution air or cylinder gas at atmospheric




pressure through one end of an open T, while pulling a constant small




volume of sample through a rotameter into the other end of the T.




This dilution was controlled to 10%, 20%, 40% or 100% of the rotameter




scale of 0.9 CF/hr.  The RAC calibration scale covers the nominal range




of 0.001 to 0.1 ppm of H?S in a 15 cu. ft. sample, equal to 0.25 CF/inin




for one hour.  This gives a factor of 200 to 2000 for a 5 minute sample

-------
                                 -Sl-
at 10% to 100% of the rotameter setting, or an upper range of 2 to




200 ppm;  Lower ranges are covered by lengthening the cycle time to




10 minutes or longer multiples of 5 minutes, as desired.




          Instruments were arranged on benches around the interior of




the van, as shown in Figure 9.  The Esterline Angus and Tally Paper




Punch were rack-mounted on the records desk, next to the blending




control box.  An interior view of this end of the van with the RAC




and Barton instruments is shown in Figure 11.




          All vent lines from the blender, manifold, and individual




instruments were conducted inside a common stack to an outlet extend-




ing three feet above the roof of the van (top of Figure 10). Storage




cylinders for air, carrier gases, and calibration gases were chained




in racks mounted on the side of the van.  Auxiliary ventilation was




supplied by a window mounted fan in addition to the air conditioner.




A telephone was used for regular communications with the office or" a




plant foreman in a neighboring unit, for added safety precaution.  These




arrangements can be seen in Figure 12, an exterior view of the van on




site, at the lime kiln stack of the Warren Paper Mill.

-------
       Figure 12

EQUIPMENT VAN ON SITE
                                                          I
                                                          in

-------
                                  - 53 -
                      3.  PERFORMANCE OF  INSTRUMENTS






 3.1   Response  Times  and  Zeroing




      3.1.1   Barton




          The  Barton coulometer  shows an  initial  response  to  changes




 in sample concentration  within a few seconds  and  can  travel  full




 scale in 10  seconds  or less,  at  its higher  scale  ranges.   This  intrinsic



 speed of response is an  advantage which is  not  commonly  utilized in plant




 practice, however.   It accentuates a very short-time  variability of




 readings and zero levels which is related to  the  hydraulic properties




 of the cell.   This instability appears as a restricted fluctuation,




 between limits which are related to the rate  of generation of titrant




 bromine in the cell  and  the rate at which excess  bromine is  stripped




 out by the constant  flow of carrier gas.  The rate  of fluctuation is




 determined by  the circulation of electrolyte  within the  Barton  cell,




 which depends  on a gas lift effect of the sample  gas  and a "wiggler




 valve" between cell  chambers.  Coulometric  cells  in the  Philips SO




 Monitor and  the Dohrmann Microcoulometer  are  provided with a  more




 positive circulation of  electrolyte, and  they do  not  show  the same




 instability.   Laboratory experience with  Dohrmann cells  shows that




'improved stirring is directly related to  stability  of output.



          This short-term instability is  more of  a  hazard  in  electronic




 data  logging than it is  in a  visual record.   The  variations  concerned




 take  place within a  few  seconds,  and their  apparent effect is minimized




 by running  the recorder  at  a  low chart  speed.  This produces a  record




 in the form  of narrow band, instead of a  moving line. The width

-------
                                   - 54 -
of this band varies with  the hydraulic properties of  the cell, and  for

any cell its width is  inversely related  to the attenuation scale.   The

uncertainty which it  introduces appears  both  in  zeroing and  in sample

readings.  The average effect is given in the Barton manual; selected

cells do better than this, and a noisy cell does not do as well.

                                 TABLE 6

                     BARTON RECORDER ZERO LEVEL RANGE
                  %  OF FULL SCALE,  AT 250 cc/min. FLOW
Range Switch
Setting
.1
.3
1
3
10
30
100
Blank Level
(Instruction Manual)
20 + 8
12 + 4
6 + 3
3 + 1.5
2 + 0.5
1.5 + 0.8
0.7 + 0.4
Typical Averages
Observed

15.6
8.9
4.4*
2.2
1.3
1.2
       *  Daily averages  (and  standard  deviations)  at x3
          on  successive days =3.5  (+ 0.3),  4.1  (+  0.3),
          5.4 (+  0.4).

          The + values  in this table represent expected deviations  from

the mean  in any instantaneous  reading,  not  just  an  uncertainty  in

the average value.

          The Barton commercial unit allows for  this effect essentially

by visual averaging.  This also has another purpose. The compositions

of stack  gases from a paper mill recovery furnace and from other large

combustion units  are subject  to rapid fluctuations  in concentration,

within a  few  seconds, whose effect  is averaged as soon as they  reach

the atmosphere.   For most environmental studies, therefore,  the

analysis  is better  averaged than recorded in detail. The standard

-------
                                 - 55 -










Barton procedure recommends a mechanical averaging which masks the




intrinsic speed of response, by placing a one liter surge pot in the




sample line   .  This serves as a trap for any water and particulates




condensed in the line, but it also imposes a time delay of about 15 minutes




in full response.  The standard gas flow rate is 250 cc/minute, which




would give a plug-flow time of 4 minutes for one complete gas exchange.




The experimentally observed time interval of 12-15 minutes for a full




response with the surge cell in line corresponds to an exponential




decay curve, to allow for dilution and mixing.  This is a good check




with theory, which should require 3.3 times the plug-flow time to




achieve a 99% exchange of gases.  The instrument tests in this program




were run without the surge cell; this is also common practice in paper




mills when the Barton is being used for process control studies.




          For a better understanding and diagnosis of cell behavior and




instrument limitations, the recorder was run at a more rapid chart speed




of 30 inches per hour during initial laboratory evaluations, instead




of the normal 2 inches per hour used for field tests.  At the slow




chart speed, it became apparent that the variations in cell reading




are not in fact random variations about the mean, but are a bipolar




distribution within limits.  This appears in the statistical analysis of




data as a normal distribution of data at the 1 a level, but a much




higher fraction of data points than expected fall within the 2 o limit.




The effect is clearly illustrated by typical chart traces at different




attenuations, as shown in Figure 13.

-------
                                  - 56  -
C  O U O  G  O O O  O  O O G  -O  1.1 O O  O  O O (•  O  O O O  O  O G O  O Ci o C
                                             BARTON CHART TRACES
                                           AT DIFFERENT ATTINUATIONS
o no  n  n n q o  o o o o  o o o  o  n n o  G n o o  p n o o  o n o n  n

           ® ^ O  G O O O  iy O u O  O O O  O  O O O  O O O C  O O O O  O
S C »  O

-------
                                   -  57  -
          Figure 13 also shows the effect of scale on band width.




Plant  operators using  the Barton  instrument normally operate  it at




the xlO  scale  (or higher) which minimizes the visual effect of the




instability discussed  above, and  gives more nearly a straight line  on




the chart.  Electronic averaging  of  the signal  from the  cell  over a period




of 30  seconds  to a minute might be used to give a smoother output for




data logging.



          The  Barton cell zero also  shows a continuous short-term drift




over a period  of a day or less, which  requires  at least  daily re-determina-




tion.  This can be a serious disadvantage.  The zero blank,like all sample




readings, is a dynamic balance involving the small amount of  titrant




bromine  stripped out of  the electrolyte by carrier gas,  and the current




required to keep the concentration of free bromine available  in the  cell




at a constant  level.   This zero varies with changes in barometric pressure,




temperature, the condition of the electrolyte,  and above all  with small




changes in the "constant" flow rate  of total gas.  The uncertainties in




zero level which this  represents  fall within the same limits  as shown in




Table  6, since they are related   to  the same dynamic balance  within  the




cell, but in this case it is the average which  is shifting within the limits




shown.  The lack of an absolute zero means that all measurements are dif-




ferential values,  subject to an uncertainty of  1/2 to 3 units on scale.




This uncertainty varies with attenuation,  and the length of time since  the




last measurement of the zero blank.




          A cumulative anomalous effect of CS- on cell zero was observed




on continued exposure at ppm levels.   Apparently CS  is only partly oxidized




and builds up  slowly to poison the cell response, until it is stripped  out




by a clean carrier gas.

-------
                                  - 58 -
           It  should be noted that  these difficulties with zero level are




not unavoidable in coulometric titrations, at ambient levels.  Most of them




have been  corrected in the Philips S02 Monitor by using a thermostated cell,




with stirring and more  positive  flow control,  and an  internal  calibration




unit for  automatic  re-zeroing  on demand.




     3.1.2 Bendix




           The  GC/FPD  combination in  the Bendix Chromatograph is much




better  in  zero stability than  the other instruments on  field test, by an




order of magnitude.   Its normal  zero  variation was less than 0.5 scale




unit, with a  standard deviation  of 0.1 or 0.2.  This held stable for




the first  6 months on test.




           Stability was  not as good after shipment from New Jersey to




Maine in a trailer van without factory readjustment, after which the




standard deviation on zero level was  about 0.2 to 0.4.  This was still




superior performance, by comparison.  The instrument is provided with




two systems for adjusting zero level, either automatically each cycle




or on demand by a manual control.  Rebalancing the automatic zero when




it is out  of adjustment  requires circuit testing equipment, and it was




not attempted  in the  field.  The manufacturer has redesigned this




automatic  zero system during 1972  to  make it more stable and easier




to adjust.  The need  for this change  was further confirmed by the return




trip from Maine to New Jersey, when the automatic zero adjustment became




inoperative and the manual setting for electronic zeroing was actuated




daily,  or  as required.

-------
                                  - 59 -
          Response time in the Bendix is normally 98% in the first




or second 5 minute cycle, depending on exactly when in the GC cycle




the change in concentration to be measured occurs.  This is not affected




by large changes in concentration, as long as the gas components remain




the same.  Exceptions may occur when one component has been absent




(such as CSH) with a large excess of others (H-S and SCO, requiring a




third cycle to get full response.  This depends upon the GC packing.




          Response to changes in attenuation scale are immediate and




arithmetic, up to where the instrument goes off scale.  An automatic




attenuation subsystem is provided but it was found to give a different




zero level at different attenuations, and was not included in the lab-




oratory evaluation.  This sub-system was apparently no longer balanced




properly with the rest of the system after the GC columns were shortened




to 20% of their original length, which was done at the start of the




test program to increase the instrument range.  It operated satisfactorily




in field tests for differential measurements, and in making matched




synthetic blends to duplicate selected stack gas analyses.  Rebalancing




it is a factory operation, and the lack of the automatic feature did not




interfere with operations at a constant attentuation scale.




     3.1.3  Houston Atlas




          Zeroing in the Houston Atlas tape recorder is an approxima-




tion achieved by a simple screwdriver adjustment, to allow for the




average reflectance of the unexposed tape and subtract it mechanically

-------
                                   - 60 -
 from the recorded sample readings.   This average must be reset for




 each tape,  and it varies by + 3 to  5 percent for a given tape.




 Overnight runs regularly show occasional spikes of -10% or more in




 reflectance,  due to tape unevenness which is greater than average.




           The combination of the Model 825 tape recorder and the pyrolytic/




 catalytic converters in the Model 855 unit on test gave a critically




 slow response time, on the order of 25 minutes.  This interfered




 seriously with instrument evaluation.  Several factors were apparently




 involved:  one was the volumetric effect of the enlarged pyrolytic and




 catalytic conversion chambers,  which create the same type of gas  mixing




 and  dilution  delays as  the gas  surge pot in the Barton Titrator.   This




 effect was  confirmed at one point in the field tests when one of  the two




 Barton probes slipped its  timing adjustment and began to blow air back




 into the line every other  5 minutes on alternating with the second probe,




 which was providing stack  sample gas.   This gave a sawtooth reading on




 the  Barton,shown in Figure 14.   The minimum Barton readings were  from




 80 to 90% below the maximum,  approaching but not quite reaching the




 zero level  due to incomplete  air dilution between sample peaks.   The




 Houston  Atlas shows the same  sawtooth  effect but greatly damped,  with




 valleys  only  20% below  the peaks.   This  pattern indicates a substantial




 amount of gas mixing within the instrument.




          Sample adsorption on  unheated  metal lines  within  the  instrument




was  also  involved,  as a time  factor.   The  hold-up of SC^,  a highly  polar




gas,  and heavier  sulfur compounds such as CS? and CSSC was substantially




longer than H-S  or  COS, on the  order of  45 minutes or  more.   This  is

-------
                      Figure 14
EFFECT OF ALTERNATING AIR  BLOW  IN SAMPLE LINE
              HOUSTON ATLAS (damped cycles)
                BARTON MV (5 minute cycling)

-------
                                  - 62 -
attributed to gas adsorption and desorption effects  involving either




the contact  agent in  the  catalytic  reduction  chamber or  the  extensive




amount of stainless steel tubing used  to  connect various  reaction chambers




in the instrument.  The time lag for polar gases in  the  simple  825




recorder was from 5 to 10 minutes,  instead of  25 or  more.




          An improved model produced by Houston Atlas during 1972 is




said to cut  this response time lag  from 25 minutes to k or less by




greatly reducing the  free volume in the two conversion zones, and by




using heated lines for all connections between the reaction chambers.




No tests on  this improved model were made.




     3.1.4   RAG Sampler




          The RAC instrument is designed  with  an automatic zeroing




system which is intended  to measure the absorbance of each spot on




the tape before it is used, and reset  to  that  zero for the succeeding




measurement.  The system  did not work  reliably, and  the lack of any




manual of instructions made trouble-shooting difficult.  Contributing




factors may  have been a slow warm-up time each time  the instrument




was shut off to save  tape, and an excessive rate of  lint formation




when measurements were made regularly  at  the minimum time interval




of 5 minutes, for use at  high gas concentrations.  The lint production




requires frequent cleaning and adjustment of the optical sensor.  Both




of these factors would be much less noticeable if the instrument were




being used for ambient gas measurements,  for which it is designed.

-------
                                   -  63
In ambient use the sampler can be left on stream continuously to avoid




warm ups, and a time cycle of hours instead of minutes for cumulative




sampling produces much less paper lint, which has to be cleaned off of




optical surfaces.




          The cumulative response may be read at any time but it is




interpreted most conveniently against a fixed time scale, such as the




5 minutes modulus for which the instrument is set.   Better results with




a constant sample or a slow rate of concentration change might be




obtained by measuring the rate of change of absorption rather than its




absolute value, but this would not solve the problem of scale limits with




rapidly changing concentrations.




     3.1.5  Overall Comparison




          An overall comparison of zeroing and response characteristics




for the instruments on test is shown in Table 7.  Preliminary laboratory




results are included for the Philips Monitor which does well on this




basis and for the Dohrmann cell, which responds rapidly once it is




adjusted.

-------
                                        TABLE 7
Scales used

Response time,
  (minutes)

Cycle time
  (minutes)

Response limits
  ppm H2S
      COS

Precision (% FS)
S.D. (MV)

Stability of
  reading, % FS

Zero drift, % FS

  Automatic zero
    + MV

  Time to Failure
FIELD
Barton
1 to 100
2
Contin.
0.1-1000
nil
0.2-1.5
0.2-0.9
0.6-1.5
0.5-2
3
0.5
reset
daily
PERFORMANCE
TEST RESULTS,
Bendix
50 to 2000
5-10
5
0.05-120
0.05-50
0.2
0.0-0.3
0.2-1
0.5
0.2
10 months
CHARACTERISTICS
ZEROING AND RESPONSE
RAC
5 to 20 rain
5
5
.001-. 15
nil
1-3
1.7-3.1
2
5
1-2
2 weeks
Houston Dohrmann
Atlas Reductive Philips
200 yl
20-100 6 seconds 1-3
Contin. 5 Contin.
.5-25 .005-2 .01-5
.5-25 (10% of H2S) nil
2-10 5 2
1-7
2 0.5 1
5 reset 1
none none 1
none reset for
each sample

-------
                                  - 65 -
3.2  Performance with Specific Gases




     3.2.1  Hydrogen Sulfide




          3.2.1.1  Basic Parameters




          The basic parameters of linearity, accuracy, reproducibility,




and equilibration time for all instruments were measured on blends of




H2S in air.  Field test data on H2S up to 120 ppm in  the Bendix and




Barton are given above in Figure 3.  Typical data on  linearity for a




laboratory test run on the Barton, Bendix, RAC and Houston Atlas




(Model 825) are given in Table 8.  This shows the effects of concentra-




tion changes for synthetic blends of a nominal 3 to 12 ppm, and of




changes in attenuation scale with a given sample.




          Considering first the Barton, the precision observed is + 4%,




well within the accuracy of the gas blends.  Blends of a nominal 3, 6




and 12 ppm gave a net reading on the xl scale of 30.3, 61.1, and 127.7 MV,




for observed analyses of 2.7, 5.5 and 11.5 ppm.  A change in scale from




xl to x3 gave an immediate shift in MV reading from 132.9 to 59.7 MV




(including zero) and a slower equilibration to 51.5,  corresponding to an




observed value of 12.2 ppm.  A check run with a blend of 4.5 ppm (nominal)




gave a net reading of 45.5 MV (xl) and an observed value of 4.1 ppm.







          The same blends were checked in the Bendix alone against




a permeation tube sample having a known H«S concentration of 3.17 ppm.




This gave a net H S reading of 19.5 MV at x500, compared to 19.4 MV




for the 6 ppm blend at xlOOO, and 18.9 MV for the 12 ppm blend at




x2000.   These are check readings, showing a direct arithmetic ratio

-------
                                                      TABLE  8
                                     LINEARITY TO I^S  CONCENTRATION,  PARALLEL TESTS
                                            SYNTHETIC BLENDS,  NOMINAL 3-12 PPM
   H2S        Barton, Cell #2
  in Air     Scale    MV  (ppm)*    Scale
 3 ppm      (xl)
          35.5 + 1.7
             (2.7)*
                                                Bendix
                                  //I, TS
                                    MV
                        2, H2S
                        MV
        RAC
Scale
MV
Houston Atlas
 (Time)  to MV
(x500)    18.6+1.4   17.2+0.4              38.7,  39.8,  38.7  (16)      6.9
                                 (20 min.)        (IJ)**      (80 min.) 8.6
                                                                      (0.5)
Air
(zero)

6 ppm
12 ppm
Air
4.5 ppm
Permeation
   Tube
= 3.17
  ppm
          5.2 + 1.5


(xl)       66.3+1.7   (xlOOO)  23.1+0.3  19.4+0.3             44.3, 40.9, 42.7   (5 min.)  15.7
             (5.5)                                      (10 min.)        41.6         (15 min.)  18.5 + 2
                                                                         (5.0)                 (2.9)

(xl,  x3)   132.9, 59.7                                              48.4, 47.9, 47.7   (5 min.)  33
             (11.5)                                                      48.4         (15 min.)  41+2
(x3)       51.5+1.7   (2000)   22.6+0.2  189+0.2   (5 min.)         (12.0)                (8.5)
             (12.2)
(x3)       4.5+1.5    (x500)   5 min. to
                                -0.9 + 0.1
                                                           0.0
                                            10 min. to 0.5    (10 min.) 12.3
                                                              (15 min.) 8.7 + 0.5
(xl.O)     50.0+2     (x500)   29.1+0.9  25.3+0.7  (5 min.)   12.2, 12.3, 12.0   (5 min.)  12.3
             (4.1)                            (4.11)                     12.8         (10 min.)  8.2 + 0.5
                                                                         (3.3)                  (2.5)
                       (x500)   24.0 + 0.5  19.5 + 0.5
 *  Calculated ppm values  based  on  net MV, above zero blank.

**  Add 10 MV  (+ 0.8  ppm)  to  RAC, as estimated correction for improper zeroing.

-------
                                 - 67 -
of scale.  The observed value for the nominal 3 ppm blend was 1070 low,




at 17.2.  The reading for the nominal A.5 ppm check blend was 25.3,




which gives a calculated value of 4.1 based on the 3.17 permeation




tube value.  This checks exactly the observed value of this blend in




the Barton.




          The RAC likewise gave approximately the same readings for





3 ppm at 20 minutes exposure, for 6 ppm at 10 minutes, and for 12 ppm




at 5 minutes.  Observed values calculated for these three are 1.7, 5.0



and 12.0 ppm, and 3.3 for the 4.5 ppm check blend.  A zero correction




of +10 MV for improper automatic zeroing of the instrument will add




about 0.8 ppm to these readings, and bring them in line with the Barton




and the Bendix.  The Houston Atlas (simple tape sampler) gave observed




values for the same four blends of 0.5, 2.9,  8.5 and 2.5 ppm.  These are




all low, and probably reflect insufficient time for full equilibration




in a run period of 10-15 minutes.



          A similar set of data with standard deviations for each




instrument during the first month on stream is shown in Table 9.  In




the range of 0.5 to 3 ppm, the usual standard deviation for the




Barton  (Cell #2) was about 0.9, the Bendix 0.1, the Houston Atlas




(Model 825) and the RAC both about 2.0.  The calculated values for




standard deviation were usually close to 2/3 of the range of values




estimated by visual inspection.




          Differences in the stability of the instruments with a con-




stant sample are shown as standard deviations in Table 10.   The first




four runs relate to a single blend from a cylinder originally contain-




ing 30 ppm, and the fifth is a one half blend of the same sample.

-------
                                                         TABLE 9

                                 LINEARITY AND PRECISION (MV AT 3. 1.5. 0.5 ppm)
   Nominal                       Barton         Bendix  fphotovolt at 200)   	H-A	 	RAG
Blend, in Air      Time     Scale  (Cell #2)  Scale   //I  (TS)    //2 (H?S)     Time     Reading     Time       Reading
   3 ppm       12:50-16:00  (x3)   7.5+1.5*  (x500) 27.0+0.320.5+0.3  20 min.  12.0+2.0   (5  min.)  86+3.5     11.9
                                                                                     14.0 a 2.0
               8:25-9:00    (x3)   1+0.5           -0.2       0.0         20 min.  -4.5+1.5             98+1.5
                 (air)             0.5 a 0.9         -0.3 a 0.1                       3.8 a 0.8             97.9 a  1.3    0


   3 ppm       9:15-10:30   (xl)   15.5+1.0        22.9       17.2        20 min.    10+3                87.5+2          ,
                                   15.4 a 0.8        22.6 a 0.2 17.2 a 0.1           9.0 a 2.1              87.8 a  1.4   10.1  ^
                                                                                                                             00
   1.5 ppm     10:45-12:15  (xl)    7.5+1.2        10.3       8.5         10 min.   4.5+2.7             94.5+2.5         '
                                    7.6 a 0.6        10.2 a 0.1 8.5 a 0.1             4.4 a 1.9             92.8 a  2.3    5.1

   0.5 ppm     12:25-12:45  (xl)   1.6+1.5   (xlOO)15.1       13.3        10 min.   0.5+1.0             9.55+1
                                   0.6 a 1.1         15.3a 0.2 13.2 a 0.1            0.0 a 0.7             95.7 a  0.7    2.2
*  + values are visual range by inspection, a are calculated standard deviations.

-------
                                      - 69 -
                                    TABLE 10
                  Scale + Blank

Time: Start
      Finish
      Readings

Barton (Cell //2)
  MV              xlO
                           (2  5")
  S.D.                     U0;

  % < la

  % < 2a

  net
  obs. ppm

  S.D. ppm


Bendix            X2000    , n ,^
                           (-0.4)
  //I MV

  S.D.

  #2 MV                    (0.0)
  S.D.


H/A               .3(100%)

  MV                       (8.0)

  SJ).

  % < la

  % < 2a

  ppm obs.
F STABILITY

Run //I
10:55
11:40
10
21.1
0.6
80
100
18.6
16.8
0.5
35.4
0.1
35.4
0.0
82.7
1.3
70
100
19.8
(H2S BLENDS)
Nominal
n
11:45
12:45
14
25.2
1.1
71
100
22.7
20.4
1.0
35.7
0.1
35.3
0.0
89.6
0.8
78
93
20.4
30 ppm
//3
13:10
13:13
48
24.7
0.9
71
98
22.2
20.0
0.8
35.7
0.1
35.3
0.1
91.0
0.6


20.8

//4
15:00
15:25
6
23.7
0.9
67
100
21.2
19.1
0.8
36.1
0.1
35.3
0.0
87.2
2.1
50
100
19.8
Norn.
15 ppm
95
14:00
14:45
10
12.1
1.3
70
100
9.6
8.7
1.1
19.7
0.2
21.5
0.3
49.8
1.4
50
100
10.4

-------
                                  - 70 -
 The  cylinder gas blend had lost H S to the level  of about  20  ppra at
 the  time  of this test.  Considering first  the  Barton,  the  differences
 in readings in  the first four periods  reflect  slight adjustments in
 sample  flow,  to which this instrument  is quite sensitive.   The  standard
 deviation of 0.6-1.3 MV averages 0.9,  or 4% of the  reading, which
corresponds to  observed values  of 20.0 + 0.8 ppm.   The percent of all
readings which  fall within 1 a  is normal, ranging from 67 to 80, but
it is unusual for  98-100% of all to fall within the 2 a limit.  This
was a consistent pattern for the Barton, and it is attributed to the
internal properties of the cell.
          The criticality of gas flow rate control in the Barton is
easily overlooked.  Typical data on this effect are shown in Table 11.
Deviations of + 25 cc  from the normal flow rate of 250 cc/min. give
an immediate inversely proportional change of 4- 10% in the analytical
value obtained.
                                TABLE 11

BARTON
DEVIATIONS WITH

CHANGES IN GAS RATE
Air
cc/min
150
225
250
275
300
MV Reading
Zero Reading
2.5 48
33
2.0
27
25
(x3)
Net Calc.
45.5 12.7
30.5 8.5

25 7.0
23 6.4
H2S, ppm
Equiv/250 cc
7.62
7.64

7.69
7.68

-------
                                  - 71 -
          In the comparative results presented in Table 10, as in Table 9,




the Bendix showed much better stability for the same periods on test,




ranging between 35.A and 36.1 MV.  It is relatively insensitive to




changes in sample flow rate, which must be critically controlled for the




Barton and other instruments.  The standard deviation of readings is far




superior, at 0.1 for most observations.  The one half blend deviations




are a little high for all instruments, suggesting that the composition




of this blend was less exactly controlled.




          The Houston Atlas is higher in standard deviations,  from




0.6 to 2.1, and somewhat less sensitive than  the Barton to changes  in




flow rate.  The observed values which it gives are  close to the Barton




and follow the same pattern, allowing for the characteristic time lag




of about 25 minutes (+ 5 minutes) in H_S sample readings.   Tests on the




RAC in other series showed a similar standard deviation range of about




1.5 to A, and a sensitivity to changes in flow rate comparable to the




Barton.





           3.2.1.2   Changes in Scale





           Linearity to  changes in attenuation scale from xO.3  to xlOO




with the  Barton is good, as  illustrated further in  Table 12.   The ppm




values observed for the  first blend  at xl, x3, xlO  and x30 fall within

-------
                                       -  72  -
                                     TABLE 12

Scale:
Cell #1
Blank (in air)
Start
Finish
Readings
Mean MV
S.D.
Sample
Start
Finish
Readings
Mean MV
S.D.
Net
Reading
ppm obs.
S.D. (ppm)
LINEARITY TO CHANGES IN SCALE
0.3x xl x3 .xlO x30 xlOO x3
Cell #2

14:07 14:36 14:54 15:09 15:23 15:40 9:55
14:11 14:40 14:57 15:13 :27 :48 10:47
25 25 15 26 24 51 53
15.5 9.0 4.4 2.1 1.3 1.2 3.5
0.33 0.31 0.12 0.21 0.1 0.17 1.1
% 300(1) 30.3 11.8 4.5 2.1 2.0 31.4
0.40 0.48 0.1 0.1 0.1 1.2
%0.9 2.0 2.1 2.1 2.1 6.9 7.53
0.04 0.13 0.08 0.03 0.9(2) 0.32

x3
(+ fresh
electrolyte)







13:18
14:17
60
30.3
1.0
58
100

26.1
7.05
0.27
(1)   Off scale
(2)   xlOO reading is meaningless, too close to blank and SD.

-------
                                 - 73 -
 the narrow range of 2.0 to 2.1 (2.02 to 2.13).   The xlOO reading is




 too low and meaningless for this sample, however, since it falls within




 1 a of the zero blank.   The reading at xO.3 allowed much too short a




 time for equilibration, but an interval between readings of about 20




 minutes was entirely adequate at the higher scale settings.




           The standard deviation of readings is consistently higher




 for both the instrument zero and sample readings at the lower scales,




 as indicated above in Table 6.  It also varies with the individual cell.




 Cell //I operated somewhat more smoothly than Cell //2 which gave the




 results  in  Table 10 and fche  second half  of  Table 12:  Its standard



  deviations were about  0.3  (from 0.1 to  0.5) compared to 1.0  (from 0.6 to




 1.3).  Table  12  also  shows  that when readings with  Cell  #2  had  became




 more erratic  as  the cell solution was used,  they were restored  to a




 normal range  by  supplying fresh electrolyte-(100% within 2a).



          More complete data on linearity to scale and speed of res-




 ponse are given in Table 13.  This is a  continuous set of readings in




 which  the scale was changed  from xl to x3 (at the end of column 1),




 from x3  to xlO (at the end of column 2), and so on.  The "zero-time"




 reading  in each column shows an immediate effect of the scale change,




within 5-10 seconds.  The response time  is consistently 2 minutes or




 less for 98% of the final reading change on going up scale, with the




 zero-time effect of overshooting the final reading on each change

-------
  - 74 -
TABLE 13
LINEARITY AND SPEED OF RESPONSE
(MV READINGS) CELL #1
Time
(minutes)











Mean
S.D.
0
2
4
6
8
10
12
14
16
18
20


xl
86.9
86.8
83.4
86.7
84.5
83.8
86.5
84.7
85.1
84.7
84.4
85.1
1.26
2 minutes to 98%
Immediate to 98%
3x and Ix.
Blank (air)
Net
Calc
S.D.
Reading
. ppm
(ppm)
8.1
77.0
7.7
0.13
x3
43.2
33.3
32.0
33.9
32.5
34.1
32.7
33.1
33.7
33.9
34.6
33.2
0.73
xlO
x30
36.1 13.6
12.6
12.0
11.8
11.7
11.1
11.6
11.9
11.5
11.8
12.2
11.8
0.40
of scale change
of scale change
4.2
29.0
7.8
0.20
2.5
9.3
8.0
0.34
5.2
4.9
.4.9
5.1
4.8





5.0
0.16
going
going
1.8
3.2
8.3
0.42
xlOO
2.4
1.7
2.0
1.9
1.6
1.9
1.9




1.8
0.15
x30
4.9
5.0
4.7
4.6
5.0
5.0
4.8
5.1
5.0
4.8
5.3
4.9
0.20
xlO
11.6
11.5
11.4
11.7
11.5
11.4
12.0
11.6
11.2
11.1
11.2
11.4
0.25
up scale; overshoots
down scale, as far as
1.0
0.8
7.2
1.05
1.8
3.1
8.1
0.52
2.5
8.9
7.7
0.22
x3 xl
28.8 75.7
29.2 75.2
29.9 77.2
77.1
74.2
79.3
81.5




29.3 77.1
0.56 2.5
at each change .
lOx; slow at
4.2 8.1
25.1 69.0
6.8 6.9
0.15 0.25

-------
                                 - 75 -
up-scale.  The reason for this is explained in Section 3.1.1 above, and




in Volume I.  For each increase in scale setting there is built into the



instrument an automatic increase in the rate of bromine generation,




to compensate for an expected increase in demand for titration reagent




at the higher range.  As a result, the signal overshoots each time




the scale setting is moved up, and undershoots to a slightly lesser




extent each time it is moved down.  The zero-time response is 98% of




final reading going down scale as far as xlO , but slower at x3 and xl.




          The effect is not great enough to be disturbing either up-scale




or down at the xl scale or higher.  This covers the range starting from




0.1 ppm,for a reading of 1 MV, or one scale division on the chart.




Readings at the higher scales place a strong leverage on the blank,




since the uncertainty at reading levels of 10 times the standard




deviation (zero or net) equals + 10% on the result.




          The caution necessary in using attenuations below the xl




scale is illustrated in Table 14.  This is a series of three blank




runs and three runs on a 2.80 ppm permeation tube blend at the xO.3




scale.  Starting at a previous sample reading of 88, the blank




reading on air took 15 minutes to reach 90% of final response.  It




had not reached a final at 10.5, at the end of 4 hours.  The visual




range of readings at this point was + 1.5 MV.  Two blank runs starting




at the other end of the scale with a zero sample reading came up  to




10 and 9 in 15 minutes.  These are within -1.5 MV of the 10.5 reached




on coming down scale.  Sample run //I starting from a "zero" level  of




11 was still climbing at 86.5 at the end of 4 hours, and run //2 reached

-------
                                 - 76 -
                               TABLE 14

Scale xO. 3
Time
Start
15 min.
1 hr.
2 hrs.
3 hrs.
4 hrs.
16 hrs.
Visual Range
SLOW RESPONSE ON BARTON xO.
(Permeation Tube: H2S at 2.
Blank
88 0 0
17 10 9
13
12
11
10.5

+1.5
3 SCALE
80 ppm)
Run
//I
11

79
83
85
86.5

+1
Sample ppm obs.
ppm (perm.
tube)


Run
n
10.5
85
89



93.5
+1.2
2.45
2.80

Run
03
17

88







Start-up Equilibration Time on xO.3 Scale = >15 min.
  (from 0.0 to blank for air)  xl   Scale =   1 min.
                               x3   Scale =  <1 min.

-------
                                  -  77  -
93.5 in 16 hours, overnight.  Run //3 gave about the same reading as




Run //2 at the end of one hour, but starting from a much higher "zero".




The 93.5 MV reached at the end of 16 hours corresponds to an observed




value of 2.45 ppm on the 2.80 ppm sample.




          The reproducibility between any of these readings short of




equilibrium is not reliable.  The Barton handbook recommends dumping




the inner cell electrolyte by hand each time the scale is changed, at




attenuations of xO.l or xO.3.  Even this is not enough to correct the




very slow response to changes in sample, or changes from sample to




blank to sample, at a constant low scale setting.  This very slow




equilibration is apparently related to the circulation of electrolyte




between the inner and outer cell chambers.  It should be greatly




improved by better circulation.




          Linearity to concentration in the Bendix is affected to some




extent by increases in photomultiplier voltage which expand the scale.




Data for low concentration blends at three settings are shown in




Table 15,for both the TS and H S channels of the instrument.  The




increase is directionally higher at the lower concentrations.  This is




in line with the observation that the more expanded scales tend to




become non-linear at a lower concentration, which was found in the




more detailed data presented below for COS (Figure 16).



          3.2.1.3  Tests for Accuracy




          Most of the data obtained in this program,  like those in




Tables 6 to 15, were tests on operability and precision,  not accuracy.




Tests for accuracy require a known sample with an instrument of

-------
               - 78 -
               TABLE 15
BENDIX PHOTOMULTIPLIER EFFECTS (TS/H2S)
           TS/H2S Readings, MV
   Ratios of
TS/H2S Readings
ppm H2S
(Nominal)
3
1.5
0.5
Attenuation
Scale
x500
x500
x500
xlOO
at Photomultiplier Settings:
200
23.4/17.3
11.0/8.9
3.1/2.5
16.0/13.2
50
14.4/10.5
6.6/5.1

0
12.2/8.8
5.5/4.1

at Different Settings
200:50 200:0
1.63/1.64 1.92/1.96
1.67/1.74 2.00/2.17
xlOO: x500
5.16/5.28

-------
                                 - 79 -
constant calibration, under fixed operating conditions.  This com-




bination was rare and such tests were run on a relatively few




occasions, since the project involved a continuing series of changes




to achieve operability with instruments not originally designed for




the stack gas range.  This was particularly true for the Bendix where




repeated changes were made in GC packing, gating, photomultiplier




voltage range, automatic zero imbalance and combinations thereof, and




each required essentially a new calibration.  Tests for accuracy




against permeation tube blends were therefore run (primarily) on the




Barton, which had its own calibration curves, and less often on the




Houston Atlas and RAC.




          Accuracies of about + 2% above calibration are shown in Table




16 for Barton Cell #1 and 4- 6% for Cell //2^ at known concentra-



tions of 4.36 and 9.17 ppm of H S.  A cylinder blend to match the




lower known sample was made,  and a calculated higher blend derived




from this by volumetric proportion.  The second blend,  run overnight,




gave a completely different value from its daytime setting.   Investi-




gation revealed that this blend was prepared from a Teflon lined




cylinder, which released more H_S when cool at night than when warm




during the day.  The ratio was 7.8 to 6.2 at a constant dilution,  or




16.6 to 13.2 at a higher flow rate.




          Cell //2 in these permeation tube tests shows  the same higher




standard deviation as before,  but Cell #1 had deteriorated slightly,

-------
TABLE 16
TESTS FOR ACCURACY
(Permeation Tube: H S)
Barton (Cell //I)
Blend: Start
Finish
Readings
Mean
S.D.
Sample: Start
Finish
Readings
Mean
S.D.
% <10
% <2C7
ppm obs.
ppm calc.
% Error obs.
Cyl.
13:35
14:15
10
5.1
0.2
4-28
11:20
13:25
26
20.1
0.5
77
92,
4.20
(4.31)
Perm.



42-49
14:30
15:05
8
20.5
0.9
75
100
4.32
4.36
-0.9
* H«S concentrations in Teflon-lined
night 7.8 to day 6.2; at different
Perm.
16:20
16:20
2
6.2
0.1
55-58
15:35
15:50
3
39.6
0.2

100
9.35
9.17
Cyl.



76-49
18:20
8:40*
44
29.3
0.5
77
98
6.46
(5.14)*
+2.0
cylinder-blend changes
flow rate day 13.2 to
Houston Atlas Barton (Cell
Perm. Perm. 8 ppm 9.1 ppm
13:40 13:05
17:30 14:45
11 11
4.5 7.5

42-49
14:30 15:15
15:05 15:25
8 3
30.6 43.8 36.3 42.8
7.0 3.0 1.3 1.5
62 67 73 63
100 100 100 100
7.14 10.5 8.9 9.89
4.36 9.17 9M7
+64 +15 +7.8
with temperature: daytime 6.2 ppm to
night 16.6.
#2)
4.3 ppm
11:15
11:30
3



23.9
0.6
67
100
4.59
4.36
5.3
                                                                oo
                                                                o

-------
                                  -  81  -
after 4 months in use, and is averaging about 0.5 (0.2 to 0.9) as




compared to its initial averages of about 0.3.  This is attributed to




slight differences created in cleaning the cells, during prolonged




experimental use.




          Houston Atlas tests on the same samples showed high results




and high standard deviations, approaching 25% of the mean observed.




This is higher than usual for the instrument, but not predictable,




since it depends in part on unevenesses in the paper tape.




          3.2.1.4  Effect of Stack Gas CO/C02




          The interferent effect of stack gases on analytical results




reflects their composition and also the very large or overwhelming




ratio they may have compared to the odorant gases to be measured.




This has been noted above with respect to water, and with the FPD




sensor it applies to both CO and CO .  Percentage amounts of either




gas in a combustion stack will completely mask the FPD response to




S compounds at 0-10 ppm.  Data on this effect are given in Figure 15.




The same result appears in blends containing hydrocarbons at percentage




levels or higher.  These ratios do not usually occur in combustion




stacks,  but they can occur in other vented gases.




          The amount of CO  it takes to suppress the FPD response is




roughly 4000 times the amount of S present.   Less than 3% of CO  has




no effect on the response to an H S blend of 7 ppm,  but amounts above




0.5% begin to show the effect with an H S blend of 0.7 ppm.   At this




ratio the background or noise level due to CO  becomes equal to the

-------
                                                        FIGURE 15


                                                EFFECT OF C02 ON BENDIX

                                       RESPONSE (PPE, xlOO) AT DIFFERENT  S LEVELS
   100
0)
0)

o
ex
in
0)
fcu
<*-(
o
    20 _
                                                                                         0.7 PPM  7  PPM   Measurement
H2S (by GC)


TS (open tube)
                                                                                                                           00
                                                                                                                           to
                                                    %  C02 in Sample Gas

-------
                                 - 83 -








signal, even though it can be ignored when the interferent and S



compound are present at an equal order of magnitude.  Qualitative



tests showed an equivalent effect for CO, C0? and butane  in air,



without the GC column.



          This interference has no effect on the Bendix response in



Channels 2 and 3 of the instrument as supplied.  These use the GC/FPD



combination to measure H S and SO .  It destroys the response to total



S in stack gases in Channel 1, which measures the signal from a portion



of the untreated original sample.  This places an important constraint



on the choice of GC packings which can used to obtain special separations.



They must be able to separate CO, CO  and hydrocarbons if present, from



the S compounds to be measured.  This is true for polyphenyl ether/H_PO



on Teflon as supplied with the instrument, but there are many other



packings to which it does not apply.



          A brief review of scrubbing procedures as a possible



alternate to GC separation indicated considerable difficulties in



obtaining a satisfactory separation of C0« from SO. and H«S, and even



less promise for a quantitative separation of CO.



     3.2.2  Interference from S02



          3.2.2.1  Effect on H2S Response



          SO  is not distinguished from H S in the Barton and gives an
            £.                            *-


additive response when both are present, as indicated above.  The con-



version factor for SO  is 2.3 times that of H2S (Table 5), so a small



scale reading represents a relatively large amount of S02-  The Bendix

-------
separates the two by GC and gives entirely independent measurements.




The simple Houston Atlas  (Model 825) and RAC recorders have a low




tolerance for SO , which  bleaches the PbS color on the tape.  Data




showing these effects are given in Table 17.




          The Barton values for concentration are taken as a reference




point, starting with an air blank on all instruments.  The first sample




was 10 ppm of H_S and SO  was added to this in three successive incre-




ments.  Results on the Bendix for H-S alone and the first increment of




45 ppm of S0_ are exactly additive for total S, as the sum of H S and




SO .  This amount of S0«  diminishes but does not totally suppress the




initial response of the two tape recorders to H^S, and has about the




same effect on both instruments.  A larger increment of SO  to 130 ppm




is nearly but not quite off scale for the Bendix, at 29.5 MV for SO-




vs 30.9 for total S.  This amount returns the Houston Atlas reading to




its zero level, but still leaves a detectable small response for the




RAC.  Continued exposure  at a slightly higher SO  level of 150 ppm is




off scale for the Bendix  and gives no response for the RAC.  At both




the 130 and 150 ppm levels the response of Bendix Channel //2 to H«S is




unaffected, even though both Channels //I for TS and #3 for SO  are




going off scale.  The upper response limit of 130-150 ppm shown here




for SO,, checks the 120 ppm limit for H S discussed earlier in Figures 3




and 4.  The MV level of the limit (and readings) is different, because




these data were obtained  at a photomultiplier setting of 50 instead of




the usual 100.  The response limit is the same for separations using the




PPE packing (Table 17) and Poropak Q (Table 3), since it depends only on




the gas analyzed and not  its method of separation.

-------
                                                    TABLE 17
Scale

Zero

H S ('vLO ppm)

   -  Vt5 ppm)
+SO
+SO   (-VL30 ppm)

+SO   (^150 ppm)   280
Barton
x3 xlO
3.5 3.0
38.7
24.0

78.5
280 100.1
H-A
F.S.
7
37
20

7
7
EFFECT OF S02 ON
TS
RAG Bx //I
5 min. x2000
-.8 -.3
-37.4 9.6
-13.0 27.4

-3.1 30.9
-.8 31.2
H2S RESPONSE
Bx //2
(50 t
0
9.8
9.8

9.9
10.0
so2
Bx //3
>hoto)
0
-.1
18.3

29.5
31.1
ppm Calculated
Ba H-A RAC Bx Response

9.9 7.5 6 //I = 2 + 3
(+20) 3.2 2.5 //I =2+3

(+129) 0 0.5 End of Scale
(+150, 172) 00 Off Scale



i
00
1
Conclusions:   Bx  is  almost off scale at 129 ppm.
               Bx  H_S is  not affected,as S0_ and TS go off scale.

-------
                                     - 86 -
                                      TABLE 18
EFFECT 0
Bendi
#2
Gas Cone. (H2S)
Air (blank) 1.1
H S (3.5 ppm)* 26.7*
(8.0 ppm) 58.8
(3.5 ppm)* 26.7*
H2S (3.5) + S02 (15)* 25.5*
H2S C\>6 ppm) 45.0
+ S02 (15) 45.3
+ SO- (Scrubbed, 40.2
old soln. )
F S02 CYCLES ON H?S RESPONSE
.x 1x1000)
#3 RAC Houston Atlas
(S02) MV (ppm) MV (ppm)
0.9 -1.0
0.8 26.1* (4) 31+9 (7.7)
(1.7)
0.7 54.6 (11) 44+3 (11)
0.7 23.7* (3.7)
57.0 13.9-6.5 (2_ - 1)
0.7 51.8 (10)
82.6 23.9-7.6 (3.7-1.2)
6.1-24.9 21.7-15.6 (3.5-2.5)

*  Check runs

-------
                                  - 87 -
          A second set of data showing the effect of several cycles of




increasing and decreasing concentration is shown in Table 18.  Check




runs on H.S of the same concentration with or without added S0? gave




check results in Bendix #2, within the accuracy of the sample.   Check




results with the RAG were less precise for H_S alone,and the addition of



S0» caused an immediate loss in response, with a further loss in the




next 5 minute cycle.   The addition of a Barton SO  scrubber partly




restored the RAC response, at first, but this scrubber contained a




used solution.   This  was soon exhausted, and as the SO  response in the




"scrubbed gas" went up in Bendix #3, the RAC response went down.




          The Houston Atlas Model 855 which is not shown in these data




converts all SO  and other S compounds to H S, and makes no distinction




between them.  The chief problem in this instrument was the time lag




of about 25 minutes with all compounds, due to mixing effects in the




conversion chambers and unheated lines.  This lag was as much as an




hour for samples containing moist S0~, or some of the heavier S com-




pounds.  When the concentration of sample containing one of these




gases was dropped there was a fairly prompt partial drop in reading,




but a long equilibration time to show the full effect.  This behavior




may be attributed in part to line adsorption and desorption inside




the instrument..

-------
                                  - 88 -
          3.2.2.2  KAP Scrubber for SO  Removal




          The potassium acid phthalate (KAP) scrubber for SC>2 removal




supplied as part of the Barton system uses a solution containing 30 g. of




KAP per quart, or 0.155 mols per liter.  The theoretical absorption capacity




of this scrubber is 10 g. of S0? per liter of solution, assuming a molar




reaction.  This is equivalent to 10  liters of gas at 3.74 ppm (at 20°C.).




At the normal Barton gas feed rate of 250 cc/minute, this corresponds to




24 hours of gas flow at 10,390 ppm.  This would represent a maximum useful




life of 1000 days at 10 ppm of S0_, in the ambient range, but only one day




or less at an average stack gas S0_ concentration of 1% or more.




          The selectivity of the KAP solution for passing H?S while retaining




SO  varies with the concentration.  It is reported as a 10-15% loss of H_S




at the 1 ppm level and a tenth of this at the 20 ppm level, in tests by




Blosser and Cooper using a continuous flow of fresh solution  (5).  Data




obtained in the present program confirm the loss of 10-15% on scrubbing at




low concentrations, in the range of 0-10 ppm.  A typical result is the last




entry in Table 18, which shows 12% of the H S lost on scrubbing (a decrease




from 45.3 to 40.2 MV in Channel #2), with a feed gas containing 6 ppm of




H  S and  15 ppm of  S0~.   These  data were obtained with a scrubber approaching




exhaustion:  similar results with a fresh solution are shown  in Table 19.




The H_S blends in this case were permeation tube samples having a calculated




concentration of 9.17 and 4.36 ppm, and the observed Barton values represent




a loss of 17-18% of H.S on scrubbing.




          Parallel tests between the Bendix and Barton did not confirm




literature indications that the KAP scrubber quantitatively removes S0»

-------
                                           - 89 -
                                           TABLE 19

EFFECT OF FRESH
SOLUTION IN KAP SCRUBBER
Scrubber Barton MV (x3) Minutes to Read
(ppm) in Air
Air
Air
S02 (16)
S02 (16)
Air
S02 (16)
S02 (16)
S02 (16), H2S (9)
S02 (8), H2S (4)
Same
Same
(+ 1 hr)
(3-5 hrs)
(5-13 hrs)
(13-15 hrs)
Action Initial
_ -
+ (old) 11.1
+ (old) 11.6
91.8
6.4
+ (old) 26.6
+ (new) 4.1
+ 129.8
+ 17.3
64.0
+ 13.8




Equil. Value 90% F.S.
3.5 2 5

26.9 30 40
5
4
26.0
3.7 1 5
31.2 5 10
16.2 5 10
64.0 5
16.0
16.2 (a 0.9)
16.5 (a 0.3)
16.5 (a 0.4)
16.3 (0 0.4)
                                                                                 Sample by
                                                                                 Perm. Tube
                                                                               Calc.
                     Obs.
                                                                               9.17

                                                                               4.36
                    7.62(83%)

                    3.56(82%)
"Old" scrubber gives false high value, but reproducible after equilibration.
Value comes down with new solution, reproducible at lower level.
partially absorbed (17-18%) at this level.
H2S is

-------
                                   -  90  -
while retaining a constant small volume of H^S which shows only at low




concentrations.  On the contrary, there is evidence that under the




conditions used here, the KAP solution passes some S0_ while retaining




some H?S; the two amounts are in approximate balance, so as to give an




approximately correct reading in the Barton which does not distinguish




between them.  Data on this effect are presented in Table 20, taken from the




"daily print-out" reproduced in Figure 5.  The Barton and Bendix in this




study were run in parallel, and the KAP scrubber when used was placed in




line with the gas manifold, feeding both instruments.  The addition of




16 ppm of S0« to 4.5 ppm of H_S shows separately in the Bendix, and




increases the Barton total response from 24 to 49 MV (including the zero




blank), without the scrubber.  Placing the scrubber in line decreases the




SO  reading from 98 to 14  (Bx //3) and the Barton total to 18.  Only on




shutting off the SO- does Bx //3 drop to  zero, and the Barton TS to 15.




These data show consistently 15% of the S02 as unabsorbed.  The action toward




H_S is erratic, depending apparently on the past history of the scrubber.




          The time required for equilibration after placing the 1-liter




scrubber in line is basically related to the time required for gas mixing,




as discussed above with respect to the Barton surge chamber (see 3.1.1).




The effect is illustrated  in Table 21:

-------
                                                   TABLE 20
     Gas Blend
Nominal ppm in Air
Air blank
H2S (2.5)
H2S + S02 (16)
H2S (4.5) + S02
H2S (4.5) + S02
H2S (4.5) only
H2S (4.5) only
Air
H2S (4.5)
H2S + S02 (9)
H2S + S02 (9)
H2S + S02 (9)
H2S + S02 + COS (30)
H2S + S02 + COS (30)
H2S + S02 only
COS (30) only
H2S (4.5) only
H S, overnight
KAP SCRUBBER: COMPENSATING EFFECTS^,
BENDIX/BARTON READINGS FOR H^S/COS/S00U;
KAP
Scrubber
-
-
-
-
+
+
-
-
-
-
+
-
-
+
+
-
+
+

ill

33
33
57
50
52
52
0
52
50
53
53
54
51
54
0
51
47
Bendix MV (xlOO)
(H^S) //2 (COS) I
.1 -.1
.5 0
.5 0
.7 0
.8 0
.5 0
0
-.2
0
0
0
0
77
65
0
103
•> 47 -.1
-.4
e.
n (so2>
-.2
0
99.5
98
14
0
0
-.4
0
46.5
6.5
47.5
52.5
9
7
0
-.6
-.7
Barton
MV (x3)
7
-
25
49
18
15
24
8
24
39
22
38
44
24
23
14
22 -»• 7
3
           Comments
Scrubber left 15% of S02  (Bx)
Scrubber left 16% of S02  (Ba)
SO  (Ba) reading restored
Check readings

Scrubbing left 15% of S02 (Bx)
Check readings
Initial ASO-, Bx and Ba
-20% COS, partial S02
More SO. removed
Ba reading is false
Ba reaches "zero" in 3 hrs
Ba cell poisoned
 i
VO
(1)  COS blends contain CS , 10/1 ratio of COS/CS-.

-------
                                   - 92 -
                                 TABLE 21
                   RESPONSE TIME WITH SO  SCRUBBER IN LINE
                              (BARTON, X 3 SCALE)

Air
12 I
7.3
3.6
Air
Sample ppm
(after high cone.)
I2S, 24 S02
v
V
(after low cone.)
Initial Response
Time, % of Final

5 min, 60%
2 min, 75%
2 min, 85%
2 min, 95%
Full Response
Time to + 2% of Final
15 min.
10-15 rain, 98%
10 min, 98%
10 min, 99%
5 min, 99%
          The time delay of response approaches 15 minutes at anything
other than low concentrations.  This curve has a significant result in the
automatic time cycle built into the Barton instrument for plant operation,
which cuts the KAP scrubber out of the line for 10 minutes out of every 2
hours.  This reads "reduced sulfur, as H2S" for 110 out of the 120 minutes,
but the 10 minutes off scrubbing is less time than is required for full
equilibration of the response.  Thus, the answer which this provides for
"SO  by difference" is an empirical reading which may be correlated with
plant operation, but it is not an absolute value.  To be realistic, not much
more would be gained in accuracy by allowing more equilibration time, since
the final correlations with plant operation will still be empirical.  If
more information is desired on actual measurements of SO- and H.S or other
components, a GC separation is much to be preferred.

-------
                                  - 93 -
     3.2.3  Effects of Carbonyl Sulfide




          3.2.3.1  Response and Interferences




          The original assumptions made as to the effect of COS on the




detection and measurement of other sulfur-containing gases changed signifi-




cantly during the course of this study.  The initial picture was that the




Barton and RAC do not respond, while the Bendix and Houston Atlas (catalytic




unit) give a full response.  The first complication is that COS itself is




slowly hydrolyzed to H-S in moist air.  This is a probable contributing




factor to conflicting reports in the literature on the odor of COS, which




is usually reported as odorless.  The Barton is very sensitive to H_S, and




a standard gas cylinder blend of COS which gave no response at the 250 ppm




level at the start of this program gave a response 10 months later




equivalent to 0.2 ppm of H2S.  This could be due to either a 0.1% hydrolysis




in the cyclinder, or a more critical evaluation of the data based on longer




experience.




          The response of the Bendix to COS is quantitative and equal to H^S




on an equal  volume (mole) basis, but only at low to moderate concentrations.




A calibration curve showing the linearity of response in the range of




0-35 ppm and the effect of photomultiplier voltage setting is shown in




Figure 16.  These data were obtained with the original GC packing of




polyphenyl ether (PPE) plus H-PO,, on Teflon.  The gas blends were made up




from COS in air.




          Linearity to COS is somewhat better at the lower voltage settings




of the photomultiplier potentiometer, although higher settings spread the




scale.   Linearity is almost identical at the 0 and 100 settings,  and the




standard setting at 100 was selected on this basis.

-------
                                                         FIGURE 16
                                            BENDIX LINEARITY TO COS,  0-35 PPM
    50
    40
•o

-------
                                   - 95 -
          The deviation of individual Bendix COS readings from the smoothed




curve is about +0.5 ppm, up to 35 ppm.  A straight line calibration is




within + 2% of the readings up to 20 ppm, and another straight line of




different slope can extend this range to about 40 ppm.  These data indicate




that the upper limit of linearity in the FPD sensor is lower for COS than




it is for H2S and S02 (Figures 3 and 4).  This relationship merits further




research, which is beyond the scope of the present project.  It is




presumably tied to the chemistry of COS in the hydrogen-rich flame.




          A complicating factor in the study of COS in the Barton, with or




without the scrubber, is the unusual behavior of the Barton cell on




exposure to carbon disulfide.  The incorrect assumption was made at the




start of the program that the Barton does not respond significantly either




to COS or to CS_, which is present regularly in small amounts in field




samples containing COS.  The standard cylinder for representative




synthetic gas blends was made up accordingly, to contain 250 ppm of COS




and 25 ppm of CS«.  Subsequent data indicate that the Barton does respond




to CS?, but slowly and in an anomalous manner which is not fully understood.




          CS~ in air at or below 1 ppm shows no response in the Barton cell




during an exposure of 100 minutes (see Table 22, below).  The effects




observed here might not be noted, therefore, in calibration tests or




measurements at ambient levels.  A higher concentration blend at 80 ppm does




respond, but only to the extent of a 3.5 MV increase in reading at the end




of 20 minutes.  This reading, on the x3 scale, would correspond to 1 ppm




of H S.  A shot of the straight calibration gas at 250 ppm of CS




(+ 2500 ppm of COS) gave a reading which continually increased,  with

-------
                                  - 96 -
no indication of a definite level of response.  More significantly,




however, the CS_ reading does not revert to zero when the sample gas is




replaced by air.  Several hours or more on air are required to re-establish




a blank reading, depending apparently on the cumulative exposure to CS_




(at 80 ppm).  These observations suggest that CS9 may form some sort of




molecular complex with free bromine instead of being oxidized in the




coulometric titration.  The net effect is to poison the cell.




          A still further indication is that the KAP scrubber may convert




COS in part to H2S and in part to CS_.  The loss in activity which is




discussed below for a 10:1 blend of COS/CS_ is apparently more rapid than




that obtained with the same small amount of CS_ alone.  This is a complex




interaction which invites further research.  The Barton, with or without




its scrubber, is of no value in the measurement of COS-, and it is poisoned




by COS + CS2-




          3.2.3.2  Effects of KAP Scrubber




          Comparative data for COS/H^S blends in four instruments are given




in Table 22, showing the effect of the Barton KAP scrubber on readings




for TS, H.S and SO^.  The feed concentrations given are nominal ppm in air,




calculated from cylinder gas calibrations.  The addition of 8 ppm or 6 ppm




of COS blend to 4 ppm of H^S shows no response or change in the Barton.




The Bendix shows full response in both TS (Channel 1) and "H-S" (Channel 2),




in direct proportion to the total of H~S plus COS.  The RAC shows no change,




and the Houston Atlas is qualitatively proportional to total concentration.

-------
                                                    TABLE  22
   Time
9:45-10:20
-11:10
-12:00
-12:30
-13:15
-15:00

-15:30
-15:55
-16:10
Gas Feed
(ppm) in Air
H2S (4)
H2S (4), COS (8)
H2S" (4), COS (6)
.Same
H2S (4), COS (8)
Same
Same
Same.
EFFECT OF COS IN USED KAP
Scrubber Barton
Ba Bx +RAC Obs.
18.2
18.2
18.3
+ + 87.5-39
36-23.4
+ 18.8-8.6,9.8
Cell //2
+ 22.7-21.2
+ + 35.2-33.8
23
SCRUBBER

(x3) Bendix (xlOOO)
Net ppm //I
14.1 3.8 20.7
60.0
50.2
39.0
51.6
48.5
49.0
55.9
//2
26.5
59.3
50.7
38.0
47.5
43.1
44.0
49.6
//3
0.3
2.0
0.3
0.4
0.4
0.4
0.4
0.4
RAC
10 Min.40%
31
27-33
29-30
22-32
35-39
35-39

H-A
Range
33-39
80-90
58-62
75-82
67-93 vo
i
77-87 :
75-94
74-97
COS/CS» blend causes sharp increase in Ba, then loss in activity.
New Ba cell shows activity restored, but pattern is repeated.

-------
                                   -  98  -
          The COS blends in this series were all made up from a cylinder




containing 250 ppm of COS plus 25 ppm of CS_, so that "8 ppm" of COS actually




means 8 ppm of COS and 0.8 ppm of CS_.  The Barton showed no response to




either constituent during 100 minutes, without the scrubber.  This was




considered at the time as evidence that the Barton does not respond to




CS?, as well as COS, and tests were continued using the calibration blend.




          The KAP scrubber caused an immediate drop in Bendix TS and H.S




(after time 12:00, Table 22) and an immediate increase in these readings




when it was subsequently removed  (after time 15:55).  The Barton, on the




contrary, showed a sharp increase on scrubbing, to over twice its previous




reading (time 12:00 to 12:30).  This can only be interpreted as converting




a material inert to the Barton (COS) to an active form, or stripping out




of the scrubber materials previously retained (SO- or CS-), or both.  On




further scrubbing, the Barton value fell and continued to decrease for




another hour.




          Two separate KAP scrubbers were employed for these tests, one




for the Barton and the other for the Bendix and RAC, to avoid overloading




the scrubbing action by a high gas rate.  At this point the Barton scrubber




was taken off line (13:15), but the Barton reading continued to decrease




even without the scrubber.  It was apparent that something in the COS/KAP




combination had poisoned the cell  (//I), and it was switched to a new cell




(//2) with fresh electrolyte.  The reading was restored promptly to 21 MV




for cell //2 (vs. 18 MS for cell //I), without scrubbing.  The Bendix at the




same time (15:00 to 15:30), on the same gas stream as the Barton, but with




continued scrubbing, showed a 20% increase in TS and H^S, corresponding to




a 20% increase in the total of H_S  (4 ppm) plus COS blend (6 ppm to 8 ppm).

-------
                                   -  99  -
          The same pattern was repeated with the new Barton cell and fresh




electrolyte.  Placing the KAP scrubber in line caused an immediate increase




on the Barton from 21 MV to 35, and this fell off gradually to 33 in 25




minutes (to 15:55).  At this time both KAP scrubbers were taken out:  the




Barton reading dropped promptly to 23, and the Bendix rose to 56/50 on the




first cycle.




          The RAC was running on the same gas stream and scrubber as the




Bendix during these changes.  Scrubbing caused an initial slight drop in




reading, which was not clearly outside the experimental error.  Continued




scrubbing showed directionally a slight increase in reading.  This suggested




that some of the increase in Barton might be due to COS absorbed in the




scrubber and converted to H.S.  The Bendix with PPE packing can not answer




this question, however, since it is equally sensitive to H?S and COS in




both channels //I and //2.  The Houston Atlas, with no scrubbing, showed




the expected response to total concentrations of 4, 10 and 12 ppm  but




its deviations between readings are as great as the difference in response




between the 10 and 12 ppm samples.




          The possibility that the COS blend was stripping something out




of a "used" KAP scrubber solution was considered next, in the tests shown




in Table 23.  The same pattern of Barton and Bendix results as before was




obtained on adding 8 ppm of COS blend to A ppra of H_S, without, with, and




without the KAP scrubber.  The used KAP solution in both scrubbers was




then replaced with fresh KAP.  The Bendix showed the same results (COS




absorbed) with fresh solution, but the Barton showed no corresponding




increase.

-------
                                                    TABLE 23
Scrubber
   + (used)
         	Gas	
         Air
         H2S (4 ppm)
         H2S + COS (8 ppra)
         H2S + COS
         H2S + COS
+ (new)  H S + COS
         H S + COS
         H2S + COS + SO,
+        H S + COS + SO,
           (10 min.)   '
           (30 rain.)
           (2 hr.)
           (4 hr.)
           (8 hr.)
           (13 hr.)
COS + SO,, IN KAP SCRUBBER
Barton
MV
5.0
23.0
23.0
33.8
23.5
23.0
23.0
46.8
23.5
26.8
28.1
31.7
32.7
36.1
(x3) (Cell //2)
ppm

4.7
4.7
7.8
4.8
4.7
4.7
obs.
14.8 (S02)
(=initial,
no S02)






//I (TS)

9
57.1
48.1
56.5
51.0
55.4
101.2
46.0
50.5
52.4
56.5
58.5
65.9
Bendix
//2 (H^S)
£.
22.7
52.7
43.7
51.2
45.7
51.1
51.0
39.5
42.4
41.4
43.7
44.1
45.0

C/3 (SOJ

0.4
0.4
0.4
0.4
2.8
0.4
55.8
3.9
7.1
15.0
19.6
21.4
23.6
     RAC Range

29.3 - 32.0
(37.2) 30.5 - 28.7
       22.5 - 30.2
       31.6 - 28.5
       26.3 - 25.5
       23.4 - 28.6
       18.8 - 12.9
                                                                                                                       o
                                                                                                                       o

-------
                                  - 101 -
          This suggested that the gain in Barton  reading was due  to a reaction




with or displacement of the S0? absorbed in previous use.  This theory was




tested by adding 14 ppm of S02 to the H2S/COS blend, with fresh KAP in both




scrubbers.  The SO- increment in the Barton readings was quantitatively




removed at first but the scrubber began to fail at once, with both SO- and




COS present.  The readings continuously increased with continued  exposure.




The Bendix reading for H-S (from COS plus H_S) was unchanged by the added




SO- before scrubbing, but it came down to a lower level on scrubbing than




it did on scrubbing without the added SO..  In other words, the interaction




of SO-, COS and KAP definitely caused more absorption of COS than with the




COS blend and KAP alone.  The presence of CS- has no effect on this




comparison, since it is separated in the Bendix GC and does not come out




with H S.




          On continued exposure overnight, the Bendix channel 3 results




(Table 23) indicate an increasing release of SO- through the scrubber, and




an apparent gradual increase in readings for H-S.  The KAP scrubber by this




time has almost completely failed, and all readings, both Barton  and Bendix,




have regained 40 to 50% of the initial decrease or increase observed on




scrubbing.  The RAC showed no response to this COS blend, with or without




scrubbing, but lost half its reading for H?S on adding SO- to the sample.




          The Bendix //3 readings for SO- in Tables 22 and 23 are  not




adequately explained.  The addition of COS apparently released S0» on two




occasions, both of which may be due to an initial purging of adsorbed gas




on first introducing the COS blend (S0_-free) to a gas manifold or scrubber




which has had previous contact with SO-.  The effect was erratic, if real,

-------
                                  - 102 -
and it is not known whether this was due to COS itself or possibly to the




effects of CS« in the blend.  The tests on CS2 alone discussed above




(Section 3.2.3.1) were run at the end of the project, and it is obvious




that the definition of its effects requires further research.




          A continuous KAP scrubber of reduced gas hold-up was recommended




by Blosser and Cooper (5).  This would also have the advantage of avoiding




continued contact with spent solution.  The time factor of gradual failure




noted in Table 23 above directed attention to continuous scrubbing as a




possible improvement.  Tests were conducted in a column made up from a




3 foot section of 1 inch tubing packed with irregular 1/2 x 3/4 inch glass




cylinders.  This had a 100 cc of liquid hold-up at a liquid flow rate of




180 cc/hr.  The results obtained are shown in Table 25.  The concentrations




given are nominal values, for synthetic gas blends.  At a gas rate of 1 liter




per minute, S0_ was completely removed at 200 ppm in air, but only partially




removed at 400 ppm or more.




          The addition of 15 ppm of COS blend (1.5 ppm of CS-) caused a




marked increase in Bendix SO  and Barton total S.  The poisoning effect of




the COS/CS- was decreased by the addition of 6 ppm of H_S, but not




eliminated since neither the Barton nor Bendix response is normal for this




amount of H_S.  A slight increase in S00 content from 200 to 280 ppm shows




that the COS blend has greatly decreased the capacity of the scrubber for




S0« removal.  Another increase in H S to 10 ppm directionally reduces the




poisoning effect of the COS blend, but neither the Barton nor Bendix




readings are normal.  Continuous scrubbing did not cure the problems of the




KAP reaction, and the results obtained can only be interpreted by further




research.

-------
                                       - 103 -
                                      TABLE 25

CONTINUOUS FLOW KAP
SCRUBBER
180 cc/hr. , 100 cc on wetted column
1^/min. total gas flow
Response ppm
Gas Blends
Nominal ppm in Air
so2



so2
so2
so2
(700)
(500)
(400)
(200)
(200) + COS (15)
, COS + H2S (6)
(280), COS (15),
H2S (6)
Barton
as H0S as S00
L .i
8
4.5
3
0
3
5
(6) +17
Bendix
TS
10
7
5
0
15
9

SO
i.
8
5
4
0
29
15
16
Comments
Partial removal


Limit of complete removal
COS releases SO
H2S reduces effect of COS
less S0_ released
COS decreases capacity fo
S0» removal
SO  (280), COS (15),
  Z HS (10)
(10)     +16
10
reduces effect of COS

-------
                                  - 104 -
          3.2.3.3  Alternate GC Packings




          The failure of the Barton cell in samples containing COS/CS2, with




or without the KAP scrubber, rules out this instrument for use in refinery




stacks such as the Glaus plant burner, cat cracking regenerator, or fluid




coke burner.  The failure of the KAP scrubber in gas streams containing




both S0? and COS likewise rules out its use to achieve operability for the




RAG tape recorder, or to distinguish between total sulfur and total reduced




sulfur when using the Houston Atlas catalytic conversion unit.




          Separation by GC seems a desirable alternate, but a packing




different from the original polyphenyl ether column in the Bendix is




required.  Five possibilities were considered:  a longer PPE column,




Triton X-305, Deactigel, Poropak R, and Poropak Q.  Screening tests on these




alternates led to the selection of Poropak Q for further field testing:




          a) Polyphenyl ether in the original Bendix packing shows a slight




separation between COS and H^S, as a shoulder on the side of the H S peak




when COS is added as a minor constituent.  The indivudual peaks for H~S




and COS lie entirely within the time gate for H_S alone.  A double length




column of PPE widened this separation significantly, after pre-conditioning




with a high concentration of COS, but not far enough to avoid overlapping




peaks.  The double length column also gave an improved separation of C0_/C0,




but a slower response to SO™.  The use of a still longer PPE column was




considered as a possibility, unless other alternates were more attractive.

-------
                                 - 105 -
          b)  Triton X-305 has been used regularly in these laboratories,




with the Dohrmann oxidative cell as a detector, to separate COS, H~S, and




various organic sulfides.  The Dohrmann cell is not subject to the C0?/C0




interference which prevents the use of the FPD sensor for sulfur compounds




in stack gas samples.  Several difficulties appeared with the Triton X-305.




New packings from the original supplier which duplicated exactly the




description of the original method of preparation did not give the same




separation as retained samples of the original supply, which have been in




use for two to three years.  After six new samples had been tested, from




two different suppliers, it developed that the new packing does in fact




give similar results, but at a different operating temperature which must




be very exactly controlled.  A good split between COS and H?S is obtained




at 45°C, but not at either 30° or 60°.  A more serious defect is that the




Triton X packing does not solve the problem of CO^/CO interference at high




concentrations, so that it cannot be used in combustion stack samples.




          c) Deactigel is recommended by Thornsberry and by Hartmann of




Varian Aerograph (7) for the separation of COS, H S, and other odorant




sulfides.  This is a silica-based material, partially deactivated and




double-washed with chromic acid and HC1 for improved selectivity.




Independent tests in these laboratories have indicated that Deactigel is




highly sensitive to water vapor, which acts as an irreversible poison under




ordinary conditions.  Consultation with the manufacturer confirmed previous




indications that a Deactigel packing which has been poisoned by water can




be regenerated only by prolonged baking at temperatures above 250°C., and




reactivation by an acid wash with HC1.  It is" theoretically possible to

-------
                                  - 106 -











avoid contact with samples high in water vapor.  This would not avoid a




cumulative effect at lower concentrations, however, with a packing so




easily poisoned, and other alternates seemed preferable.




          d) Poropak R and Poropak Q both give a sharp separation of CQ^/co




and a good separation of H S, COS, H_0, S02> CSH and heavier S odorants.




Poropak R was tested first, but given second rating because it is slow for




SO-.  It also gives an exceptionally strong response to H_0, which is not




harmful at low concentrations but strong enough to create a high background




for neighboring peaks if water vapor in the sample is high.  Both Poropak R




and Q give a good separation for CSH, with a peak which is strong enough




for stack gas measurements (1 ppm or more), although not as strong as might




be desired for ambient measurements at the ppb level.




          The results of this comparison are summarized in Table 25:







                                TABLE 25




                  SCREENING TESTS ON ALTERNATE PACKINGS
Packing
PPE
Triton X
Deactigel
Poropak R
Poropak Q
CO /CO
2
Poor
Poor
Good
Good
Good
H00
A.
None
None
Poison
Strong
Good
Response to
COS/H^S
i.
Shoulder
Good
Good
Good
Good
SO
Good
Slow
Slow
Slow
Good
CSH
Slow
Slow
Good
Moderate
Moderate
Poropak Q was  selected  on  this  basis  for  further  tests with  reduced S




compounds, and in  field tests at  the  paper mill site.

-------
                                 - 107 -
          3.2.4  CSH and Heavier Sulfides




          Methyl mercaptan shows a quantitative response in the Bendix,




Barton, Houston Atlas and Dohnnann cells, with some qualification in each




case on the interpretation of the data obtained.  The same qualifications




apply to the use of these instruments with heavier sulfur compounds.  The




RCA does not respond to any sulfide heavier than H.S.




          The Bendix instrument as supplied, with PPE packing, shows a CSH




response which is just outside the 3 minutes time period allowed for sample




elution, before blow-back for the next cycle.  The peak can be easily




found by holding the valve timing on manual control, at this point in the




cycle.  The choice of GC packing and elution conditions can be controlled




to bring this peak within the 3 minute cycle, and this was achieved with




Poropak Q.  With this packing, the CSH response of the Bendix was linear




through 80-90 ppm, as shown by the data in Figure 17.  This is a plot of




parallel test results obtained in the Barton (x30) and Bendix (x200) on a




series of synthetic blends based on a gas sample of a nominal 1000 ppm.




The Barton gave observed values of 86.1, 86.1 and 91.8 ppm for a nominal




blend of 86 ppm.  The correlation curve is a straight line which extrapolates




to zero on the Bendix at 8 ppm.




          The Bendix in this configuration gave entirely satisfactory




results for CSH blends or samples above about 10 ppm, but no response below




about 5 ppm.  The intercept at a zero response indicates an adjustment




required in the elution conditions and instrument pre-column, which




apparently held back a constant small amount of CSH equivalent to 8 ppm in




a 1 cc. sample.  The CSH peak is not strong, and the initial laboratory




calibrations were made at higher concentrations.

-------
                                  - 108 -





The data in Figure 17 were obtained after field tests had begun, and no


changes in the pre-column were made at this time.  The open circles at the


lower end of this plot are taken from data obtained with an equimolar blend


of CSH/H?S, after back-calculating  the H S contribution out of  the Barton


values.
                               »

          A major problem of the Barton is the uncertainty of what conversion


factor to use in the presence of heavier sulfides, as discussed above in


Table 5.  Assuming the absence of  S02, either  in  the sample or  after a


satisfactory scrubbing,  the use of  the factor  for H»S  (at any scale,


say xlO) gives results which are 40%  too low if  the sample is all CSH.


If the sample is all S0_ but considered  as H_S,  the results are too low by


130%.  The ratio for heavier sulfides  is worse:   the error for  any CSSC


present is +190%, or for CSC +300%  of  the calculated value.


          The recommended plant procedure for  using the Barton  to monitor


a paper mill stack gives a value "as  H S" which  is at  least as  high as that


reported, but too low by an amount  proportional  to the relatively small


concentration of heavier sulfides  present.  The  accuracy of the reported


value usually improves at lower values,  under  conditions of better control,


because the control procedures applied remove  the heavier sulfides more


easily than they do H S.  When heavier sulfides  are suspected,  the selective


prefilter system used with KAP to  remove S0~ and leave H.S (and other


sulfides) can be provided with a CdSO, solution, buffered with  boric acid.


This removes both SO  and H  S, to  measure "total organic sulfides" by


difference.  The bubbler system requires about 15-20 minutes  to come  to


equilibrium, however, as it  does  with KAP, and it gives a moving value

-------
                                                        FIGURE 17
  100 _
   90 -
   80 -
   70 _
B
(X
a
   60-
a  so
4-1
a
co
o

a)
S
   40 _
   30 -
   20 -
   10 _
        - 25
        - 20
              03
        - 15
        - 10
        -  5
                                                 CSH RESPONSE, 0-100 PPM

                                  BARTON (x30), HOUSTON ATLAS (40%), VS. BENDIX  (x200)
                                                                                                          (86)
                                    •  Ba ppm CSH


                                    A  H-A ppm CSH


                                    O  Ba from H-S/CSH blend


                                    A  H-A from H2S/CSH blend
                                                                                                                            o
                                                                                                                            vo
                                                                                                          (86)
                                                                          ( ) Figures in parenthesis  are

                                                                               nominal sample ppm
                    I
                  10
                                I
                                20
 I
30
I
40
50
70
                                                        Bendix MV

-------
                                  - 110 -
averaged for this period of time.  The GC approach is considered preferable




when any distinction between compounds is desired.




          Houston Atlas data obtained at the same time as the Bendix are




shown by the dotted line in Figure 17.  The response is comparable up to




about 20 ppm, and linear above  this but at about 50% of the actual value.




The time lag for response to CSH in the Houston Atlas was the same as for




H_S, at about 20-25 minutes.  It is not clear whether the low response in




this series was due to line-adsorption at high concentrations or to the




operation of the sample timing  sector, which was set at 40%.  Further tests




should be made on the new model, with heated lines.




          The response of the Dohrmann cell to CSH is quantitative, on an




equimolar basis to H_S, but with a different electronic setting for each




compound.




          The response of these four instruments to CSH, CSC, and CSSC in




blends at several concentration levels is summarized in Table 26.  The




Bendix responds normally to the CSH and CSC in paper mill vent lines.  The




CSC peak appeared just beyond the normal GC cycle of 3 minutes elution time,




with the Poropak column used.   A further adjustment of column/pre-column




conditions is required to bring this on scale and to get a signal for CSSC,




which gives a normal response on direct injection but did not get through




the GC pre-column.  CS  came through the Poropak Q far ahead of CSSC, at




about 8 minutes elution time under the conditions of this test.





          The Barton responds to all three organic sulfides,  with no




evidence of an unusual time lag in the reaction.   The nominal concentrations




used for these tests are not definitive for accuracy.   The effect of differing

-------
                                                     TABLE 26
RESPONSE TO CSH/CSC/CSSC BLENDS
Nominal
Blend ppm
CSH
Bendix 3.1
(Poropak _
Q)
Barton 3.1
50
Houston 3.1
Atlas 50
Dohrmann
reductive
CSC
3.1
50
45
3.1
50
45
3.1
50
CSSC
3.1
50
(*)
3.1
50
3.1
50
Response (net)
Scale CSH CSC CSSC
xlO 0(22)
x50 65 + -
x5 +
x5 -
-irl ~!n
XI /U
von 	 	 i 7
XJU 4 /
-irT 01
XJ i J
10% 4 (10 to 30 min)
100% 84 (10 to 40 min)
+ +
ppm Scale Remarks
CSH erratic, at low concentration
second peak is beyond normal GC time
identifies second peak in "inlet gas"
no response in 45 minutes on GC
8.9 (as CSH) calc. 18 ppm as 3 comp. aliquot
173 (as CSH) calc. 348 as 3 comp. aliquot
25 (as CSC) pure compound blend.
20 (as H£S) Shows immediate partial response, 60-80
100 (as H2S) minutes to equilibrate, readings contini
after shut-off.
normal response to 2 compounds
*  Vapor above a drop of liquid in line, v.p. ca. 10 mm Hg.
                                         v.p. about 10,000 ppm.

-------
                                   - 112 -










 factors for different compounds is noted again in the remarks, since




 calculated values on the aliquot basis are twice those assuming CSH alone.




     3.2.5  Stack Gas Results




          A primary characteristic of stack gas emissions is their




highly erratic composition, both in terms of the ratio of components




present and their absolute amounts.  Field test  data on this  point




were obtained both at the refinery and at the kraft mill stacks.




          3.2.5.1  Refinery Claus Plant




          Tail gas from the Claus plant as fed to the burner stack was




analyzed after dilution 100:1 with air.  In the first run, composition




changed from 2 ppm to 6 in 20 minutes, with H S and SO  both present in




about equal amounts.  The sample line at the burner inlet plugged at




this time, due to a slug of molten sulfur  which  temporarily coated  the




walls of the unit.  The line was reamed clear of solidified sulfur and




analysis continued a day later.




          A second run starting at 100:1 dilution showed the results




plotted in Figure 18  for Bendix total S, H S and SO  ,  (plotted as MV)




and for Barton total ppm, calculated "as H S".  At the start both S02




and H S were present, and the Barton data curve paralleled closely the




Bendix TS  (Channel //I).  During the first 45 minutes  the Bendix TS




and H S (Channel //I and //2) moved closely together, but for most of  this




period the Bendix SO   (//3) went down close to zero.  After 13:10 (time)



sample dilution was changed from 100:1 to 30:1, to give a





higher reading level and better data.  A prompt  increase in Barton ppm




and in Bendix TS/H S was observed but no increase in  SO  readings, at




3 1/3 times the sample concentration.

-------
                                         - 113 -


                                        Figure 18


                     TAIL  GAS FROM REFINERY CLANS PLANT, DILUTED
                                13:00

                                 00 10
                                            14:00

                                      30   50
o
o
o
 X
CO

-------
                                  - 114 -










          Shortly thereafter, there was an abrupt change in sample




compositions: H_S disappeared (Bx  #2), SO  became the major component




(Bx  #3),  and the Barton curve and Bendix total S climbed smoothly




together.   The Barton reading "as H.S" reached 7.9 ppm at 14:50 on the




diluted sample, equivalent to 265 ppm in the stack.  RAC readings at the




same time, not shown in the figure, decreased to a blank.  This is a




qualitative indication of low H^S, or high S02, or of both, which was the




present situation.




          At 14:52 the Barton SO  scrubber was placed in line.  The




Barton reading for H S dropped within 3 minutes to zero.  It then




climbed gradually, while the Bendix H S (#2) stayed at zero at first




and then increased, a few minutes after the Barton, corresponding to




the time delay in GC separation.  Between 15:50 and 16:00 another abrupt




change in composition occured: the  Barton changed from 0.7 to 8.0 on




the diluted sample, representing 23 to 270 ppm in the stack.  Rapid




fluctuations continued, with readings off scale (over 300 ppm in the




stack) three times in the next hour.  The sample line plugged again




with a slug of molten sulfur after 10 hours on stream.




          Rapid fluctuations of this type are characteristic of plant




operation, during any change in operating cycle.  Time delays in the




analytical system are critical at such a time, and differential measure-




ments of different constituents are meaningless unless these samples are




taken at precisely the same instant.  This was true for the GC sample




separations, Bx //2 and //3 in the present analysis, and for no other pair




of measurements obtained.  The time difference was small for Bendix

-------
                                  - 115 -











total S (//I), less than a minute between sample injections into the




FPD sensor, but more for the Barton vs. the Bendix to allow for gas




mixing vs. GC separation.  It is not possible to measure Barton




"total S" (+S02) and "H2S" (scrubbed) closer than 3 minutes apart,




which is the minimum time for 95-98% response, and full equilibration




takes 10-15 minutes at higher concentrations.  Simultaneous measure-




ments of the separate constituents as provided by GC are a much more




powerful diagnostic tool for operating controls than averaged values,




although the averaged values may be adequate for monitoring alone.




            In a third run, the  sample  diluted  100:1  changed  from  a con-




centration  of 14 ppm by  the  Barton with  the  SO. scrubber  in  line,to 7,  20,




11,  14  and  10 during a period of 4 hours.  A continuous plot  of Bendix




GC peak shapes  during this run  showed  a  definite  shoulder  for COS on  the




side of the  H S peak, but not strong enough  (with  the polyphenylether GC




packing)  to  be  readily quantified.  At the end of  4  hours  with this gas




mixture SO   was coming through  the scrubber, and  the electrolyte  in the




Barton  cell  was apparently exhausted.  Later data  suggest  that this




exhaustion may  be due in part to small amounts of  CS_ coming  through  the




sample  at the same time as COS.




           The  refinery field tests concluded with this demonstration of




these disturbing effects.  The  original plan to determine  COS by  difference




between the  H2S readings of  the Barton which sees  H?S but  not COS and the




Bendix  which sees COS and H_S together was rendered  impossible by the




interactions of COS and S0_  in  the Barton phthalate  scrubber.  The study of




GC packings  was completed and Poropak Q placed in  the Bendix  column before




field tests were continued at the second test site.

-------
                                   -  116  -
          Smooth operability of the sampling system in the burner stack




or in the tail gas line feed to the burner was not achieved during the




refinery test period, and the sampling system was rebuilt based on this




experience.




          3.2.5.2  Kraft Mill Furnace




          A plot of typical data obtained during two weeks on stream at




the kraft mill recovery furnace stack is given in Figures 19 and 20.  In




Figure 19 the Bendix was measuring H S, COS and CSH and in Figure 20 it had




been reset to H_S, SO  and CSH.




          In the first two periods on Figure 19 the Barton, RAC, and Bendix




all show parallel readings, with the Barton and Bendix both off-scale at




the end.  The composition is mostly H.S, with some CSH, and a slight showing




of COS once every 24 hours.  In the third period the RAC reading goes to




zero while both Barton and Bendix are high, suggesting the presence of SO .




Readings are parallel again in periods 4 and 5, with sharp but small peaks




in COS which appear in each case just before a sharp decline in H«S.




          Several interesting correlations appeared on comparing the




analytical record of these COS peaks with the operating log of the paper




mill recovery furnace.  Significant points from this comparison are




summarized in Table 27.  These data were obtained during a period of




unattended operation for the instrument van, with hourly readings throughout,




and with no special readings or notice to the mill operators.  The COS




peaks were observed only once a day:  4 out of 5 times this was with the




same operator, from 2 to 4 hours into the shift.  For this operator, the




COS peak was preceded by a sharp rise in H_S (Barton TS) about 30 minutes

-------
                                                                                  Figure 19

                                                                    KRAFT MILL RECOVERY FURNACE STOCK

                                                                               H,S/COS/CSH
1
        300
        200
        100
          0
         90
         80
        40
          0
       -10
       120
        80
        40
        80
        40
         0
        40
                                             \/
                                      \/
                                               
                                                                    \

A
                                                                                                    \
                                                                                                  1x30)

                                                                                                                                    \
                                                        Period

                                                          (1) parallel readings, mostly HjS, & CSH - Box off scale.

                                                          (2) RAC Is good, mostly HjS, Box off scale.
                                                          (3) RAC Is killed, SOj present, HjS high, CSH present.
                                                          (4) parallel readings, COS appears before sharp -  H-S.
                                                          (5) Sharp COS and   CSH, before- HjS.
              16171819202122
                         H       [
4  567 8  9 10 1112
      	 9/22
                                                                1819202122
                                                                                 45678
                171819202122
             9/23  	1
789 101112
|	9/24  	1

-------
    300
m
x i/i
** CM
I!
<   100
      0
    300
    200
    100
      0
    150
  i 100
  I
S    50
J    50
     40

     30

     20


     10
     10
                                                                             Figure 20

                                                                KRAFT MILL RECOVERY  FURNACE STOCK

                                                                           H2S/S02/CSH
                                   \/
                                                                                                                                                                     OO

•^•~^»«i
-w^










•»>
	 '







• •
17 17:30 18 21 21:30 22 22:30 5 5:30 8:30 9 10:30 18 19 20 21 2 3
9/26 9/27 9/28

-------
                                                      TABLE 27
CORRELATION OF COS AND MILL OPERATING LOG
Date
9/21
9/22
9/23
9/24
9/25
COS
Peak
Time Height
17:00 8.6
12:00 12.8
19:00 13.1
9:00 51.1
10:00 11.9
H2S Peak History
Barton: TS
(Continuous)
Sharp rise
@ 16:30
Sharp peak
@ 11:30
Sharp rise
@ 20:15
Slight rise
@ 8:30
Sharp rise
@ 9:00
Bendix: H2S
ppm (1 hr. Sampling)
50-160 level
>180 up at peak
50-135 up before,
down after
45-70 down after
54-210 off scale
Recovery Furnace Log: Hourly Readings
Black Liq. BL % Steam Hours After
Flow Rate Solids Rate Shift Change Operator
low si. low 2 A
low low low 4 A
low low max 3 C
low low si. low 2 A
low 3 A
                                                                                                                       VO
COS peaks once a day, 4 out of 5 are same operator (on different shifts)
                      4 out of 5 black liquor flow off, % solids off, or both
          Operator A: sharp rise in TS before COS peak, steam rate off
          Operator C: steam rate at max, TS low after COS peak.

-------
                                  - 120 -











before the COS appeared, and this pattern moved with the operator when he




changed shifts.  The other COS peak with a different operator followed a




different pattern, with a sharp rise in H_S after the COS instead of before.




For both operators the black liquor flow rate and % solids tended to be low




at the time the COS appeared, and the accompanying H2S peak was usually the




highest reading for the day.  The meaning of those data in terms of mill




operation are not known, and much more information might have been obtained




by taking COS readings or log data more often than once an hour.  The




potential value of such component analyses in the study of emission controls




is obvious.




          During the period plotted in Figure 20 the Houston Atlas




reductive combination shows a close parallel to the Barton curve, based on




total sulfur.  This plot is for a week-end of unattended operation, and




for most of this time the Bendix curve for H_S is off-scale.  A higher




dilution ratio should have been set for unattended measurements.  At the




end of this period H~S comes down on scale and S0_ which has been moderate




shows a sharp peak of about 4 hours duration.  This information is available




only from the Bendix GC/FPD plot.  It could not be read at all from the




Houston Atlas, or as reliably from the Barton plus phthalate scrubber.




          Operation continued at the recovery furnace stack for a period




of three weeks, including three week-ends unattended.  At the end of this




period the timing disc in one of the Barton probe control boxes slipped




out of adjustment, apparently because it had not been adequately tightened




on setting.  The result was to allow air to blow back into one of the two




probes as noted in Figure 14 above, creating a sharp sawtooth effect on

-------
                                  - 121 -




the Barton chart and a greatly damped sawtooth on the Houston Atlas.

Readings on the Bendix during this upset were erratic and useless, because

of the unpredictable effect of partial air dilution in the line at the

exact moment of GC sampling.

          The instrument van was then moved to the lime kiln stack for

further tests.  This gave essentially nul point readings on all instruments,

corresponding to previous data and current results in parallel obtained by

a Barton unit at the plant.  This stack runs very wet, with a constant

rain of water condensing inside the stack and running off at the bottom.

It is apparent that the lime kiln in this plant was not being overloaded

and could be run at a higher loading of injected gas, if this were

desirable for pollution control.  Two samples of diluted inlet gas as

fed from a gas accumulator to the lime kiln showed the presence of GC

peaks for about 130 ppm of CSC and 35 ppm of CSH, in addition to CSSC

and H S.


3.3  Operating Limitations

     3.3.1  Instrument Advantages
            and Disadvantages	

          The eight instruments from six manufacturers which were

examined in this study each had advantages and disadvantages.  These

are tabulated for convenience in Table 28, in five pages for different

instruments:  the Barton, Bendix, Houston Atlas  (pyrolytic unit), RAC

(and simple Houston Atlas), Dohrmann (two cells) and Philips.  The

-------
                                                                   TABLE 28
                                                      EVALUATION OF INSTRUMENTS  (BARTON)
              Advantages
Barton (Coulometer)

  1.  Simple operation, adequate In-
        structions.

  2.  Field tested, long unattended
        runs, long cell life, slow chart
        speed.

  3.  Wide ranges of total sulfur, re-
        sponds to S of 5 types; fast re-
        sponse:  0.3 to 1000 ppm, slow
        response:  0.01 to 10 ppm.
        Operates satisfactorily over-
        range .
                    Intrinsic Disadvantages
  4.  Little attention required, solu-
        tions last up to 30 days.

  5.  Control box can be remote from
        cell, electrical connections
        only.  Aspirator system on cell
        vent is good (Br_ corrosion no
        problem).

  6.  Electronic stability good, two
        cells well matched (0.98 and
        0.99 of theoretical output).

  7.  Field-tested sampling probe.
No improvements announced.
 1.   Response coefficient differs  greatly for different
       odorants,  difference changes  with  range.
       H.S gives  approx.  4x response of RSR for  same ppm,
       H2S/S02 approx.  2.3x,  l^S/mercaptan 1.4x.
       S02 scrubber solution  is  inoperative with S02 + COS.

 2.   Cell is poisoned by  CS-, in high amounts or cumulative.

 3.   One-liter surge tank used to knock out water in sample
       introduces a 15-20 minute delay in response.

 4.   Flowmeter provided is not nearly adequate to measure ac-
       tual flow, (reading-dependent); used extra rotameter,
       manometer and pump.

 5.   Flow rate to aspirator is sensitive  to very small pres-
       sure changes; "auto cycle" S02 scrubber can cause a
       10% drop in lUS readings.

 6-   No clear indication  when electrolyte in cell is de-
       pleted, varies with loadings; shows up in cell blank,
       which must be adjusted after each refilling.

 7-   Zero and readings change with ambient temperature.

 8.   Slow circulation of  cell electrolyte causes rapid cycl-
       ing of cell readings,  between limits which are pro-
       gressively further apart  at lower  ranges; cell equi-
       libration time becomes 20 minutes  at xl,  2 hrs at
       x.3, 8-16 hrs at x.l.

 9.   Fritted distributor  discs in cell or scrubbers get
       clogged with sulfur formed in sample lines.

10.   Wet chemical bubblers supplied to separate  5 gases  re-
       quire at least 20  minutes to equilibrate:  still a
       research-type measurement,  not quantitative under
       field conditions.
  Operating/Maintenance Problems	

1.   Probe operation can be improved
      by longer blow-back, alternate
      operation of two probes is
      better.

2.   Correction for water in stack
      samples  is a source of error,
      usually assumed constant or
      ignored.

3.   Cleaning cell or bubbler when
      frits are clogged requires lab-
      oratory facilities, not con-
      venient  for the field.

4.   Normal maintenance schedule
      must be changed if unusual
      samples  appear.

5.   Importance of cell temperature
      control increases for lower
      readings, requires air con-
      ditioning for accuracy.
i
i-1
hJ

I

-------
                 Advantages
Bendix (GC/FPD)
  1.  Measures separate components directly.

  2.  No wet chemistry.

  3.  Readings held in memory circuits, read-
        out cycle 15 minutes for 3 gases  (4
        possible), or on demand.

  4.  GC exceptionally reproducible, versatile
        as to sample size, gating; timing
        holds well for months.

  5.  Designed for easy adjustment, repairs,
        or substitution of GC columns.

  6.  Teflon lines and valves.

  7.  Safety features for flame-out.

  8.  Good field service when required, good
        diagnosis on phone calls.

  9.  Field tested (for ambient only).

Improvements Claimed During 1972

  1.  Automatic  flow control for reignition.

  2.  Positive sample  flow control.

  3.  Fiber optics in  FPD replaced by  insu-
        lated direct mounting.

  4.  Rapid automatic  zeroing.

  5.  Readable external potentiometers.

  6.  New valve  diaphragms, longer life.

  7.  Improved oven temperature control.
    EVALUATION  OF  INSTRUMENTS  (BENDIX)
	Intrinsic Disadvantages	  	

1.  FPD interference from CO/CO  in stack gas   1.
      amounts, or hydrocarbon in process ef-
      fluents, prevents total sulfur reading
      in these streams; requires different GC   2.
      column for this use.
                                                3.
2.  Rotameter not accurate enough for sample
      flow control; flow rate changes with gas
      temperature, limiting on reproduci-       4.
      bility.

3.  H- flow is critical, to hold flame; rota-   5.
      meter not reproducible.  Needs a sepa-
      rate shut-off valve.
                                                6.
4.  Electronics optimized for ambient range,
      needs different balancing for emissions
      range.                                    7.
                                           a
5.  Need readable potentiometers to adjust
      gating, time cycle, oven, attenuation,
      etc.                                      8.

6.  Zero stability good short term but high
      long-term drift; adjustment very sensi-   9.
      tive, not easily controlled.
   Operating/Maintenance Problems	

Not yet ready for an unskilled attend-
  ant.

Needs a better instruction handbook.

Manual zeroing better than automatic;
  adjustments not adequately explained.

Needs flow controller, or needle valve
  plus manometer.

Mechanical failures in valve diaphragms,
  needle valves, one soldered joint.

Oven temperature not easily adjustable;
  fiber optics vulnerable on runaway.

Automatic attenuation gave variable zero
  levels, needs internal balancing with
  rates and times.                         <
                                          M
                                          ho
Reignition should not require change in   u>
  flow settings.                           i

Needs indicator on valve-actuating pres-
  sure line.

-------
                                                                   TABLE 28

                                                 EVALUATION OF  INSTRUMENTS  (HOUSTON ATLAS 855)
                 Advantages
        Intrinsic Disadvantages
Houston Atlas 855 (conversion to H.S, Pb tape)   1.

  1.  Pyrolysis + catalytic reduction con-
        verts all S compounds to measurable
        H S.
         2                                       2.
  2.  Only field instrument which gave a true
        total S reading.

  3.  Simple controls, few adjustments of any    3.
        type, good flow valves.

  A.  No electronic problems.
                                                 A.
  5.  Sample preconversion system could be
        used with better sensors.

  6.  Ag tape available, not sensitive to S02<
Improvements Claimed During 1972

  1.  Delay in response time cut (4 minutes
        modulus), by redesign of conversion
        chambers.

  2.  Accurate sample control valve and meter-
        ing, better recorder.

  3.  Spray carryover to tape from humidifier
        corrected.

  4.  All process flow lines heated (200"F);
        greatly reduces hold-up of polar con-
        stituents.
Excessive time lag in response, average 30   1.
  minutes, attributed to gas mixing in con-
  version chambers; concentration pulses
  dampened when they do appear.              2.

Small diameter stainless lines selectively
  retain and later release SOj or heavy S
  cpds (RSSR).
                                             3.
Flowmeters inadequate, range adjustment
  poor, time setting for dilution not too
  accurate.

Two-stage dilution with single timer (op-
  tional), can cause pulse flow and exag-    4.
  gerated artificial swings in concentra-
  tion observed.

Problems inherent in PbAc- tape recorder:
  humidity control to get reaction, too
  much water wets tape and causes break-
  age; moisture on windows changes zero
  levels; variable zero on tape reflect-"
  ance +2% average, but spikes of  -10%;
  changes +3-5% from one tape to another.

No indication when tape has run out or
  broken, could use an optical signal or
  tension switch.

Significant ^-consumption,  some  is vented;
  running without H~ harms catalyst acti-
  vity, needs an automatic shut-down.
  Operating/Maintenance Problems	

Prompt field service and supplies, but
  no instructions or handbook available.

Photocell fatigue and lint from moving
  tape build up imbalance, beyond bridge
  adjustment; requires periodic cleaning
  and readjustment.

Dilute acetic acid bubbler may be either
  depleted or flooded by radical changes
  in sample humidity; tape breakage can
  result from an unexpected increase in
  sample water content.

Recorder unsatisfactory; hard to read or
  write on, needle broke in service and
  gave double readings.
  5.  Tape zero problem avoided by new sensing
        system: measures rate of reaction
        (slope), feeds continuous signal to
        digital storage for timed average.

-------
               Advantages
 RAG  (Tape Recorder)

   1.   Simple  system,  few adjustments,
         for H.S only  (ambient range).

   2.   Simple  procedure for dilution:
         cumulative reading affected only
         by flow of sample and not by ex-
         act control  of dilution gas.

   3.   Enclosed model  recommended, if
         odorants present other than H-S.

   4.   Case well designed for routine
         maintenance,  parts accessible.
No  improvements  announced
        EVALUATION OF INSTRUMENTS (RAC,  HOUSTON ATLAS  825)

	Intrinsic Disadvantages	

1.  PbAc.  tape cannot be used  for most  stack gases; re-
       action  killed  by  S02, must be  scrubbed.   Narrow
       response range is too limited  for widely varying
       samples.

2.  Uses only lower  half of logarithmic scale;  + 15-30%
       accuracy claimed,depends on concentration.

3.  Stack  concentrations require high dilution with a
       carrier gas  (not ambient air,  or plant air, if  it
       contains any S).

4.  System of exhausting treated gas into case not recom-
       mended, except for ambient use.

5.  Variable  concentrations handled  only by change in
       flow or cycle  time; high levels and short times use
       up more tape.

6.  Humidifier chamber must be watched, too little = no
       reaction,  too  much = overflow.  HjO supply lasts
       1-2  days with  dry dilution air.

7.  Tape zero changes rapidly during warm-up:   stability
       poor when  pushing limits,  may  be more stable in
       routine operation, for limited range in  known en-
       vironment.
	Operating/Maintenance Problems	

1.  No handbook, instructions incomplete:
      considered a standard A1SI unit
      using PbAc, tape.

2.  Tape zero  system limits already nar-
      row range, frequent cleaning to
      remove lint from optics:  gave con-
      tinual trouble using samples of
      widely varying range, with rapid
    . tape consumption.

3.  Mechanical operation otherwise very
      good.

4.  Slow supplies, service average.
                                            I
                                           M
                                           to

                                            I
Houston Atlas 825 (Tape only)

  1.  Does not see S compounds other
        than H2S and SO .

  2.  Acetic acid bubbler can toler-
        ate limited S02 exposure
        (reversible).
1.   Usual problems of PbAc. tape (see above).
1.   Instructions needed,  both for normal
      operations and for  adjustments.

-------
                                                                   TABLE 28
               Advantages
Dohrmann (Microcoulometer)

  1.  High accuracy, if calibrated by
        matched samples (differential
        analysis).

  2.  A basic sensor device, adaptable
        to different systems of sample
        preparation.

  3.  Reductive cell (Ag) does not see
        SO. and responds only to H.S,
        CSH, COS (in absence of chlor-
        ide or cyanide).

  4.  Oxidative cell procedure converts
        all S compounds to SO. and
        measures this.
       EVALUATION OF  INSTRUMENTS  (DOHRMANN, PHILIPS)

	Intrinsic Disadvantages	

1.  Has to be checked every day for standard, using same gas
       and similar concentrations, and restandardize for each
       major change in sample.

2.  Range and standardization change with sample size, more
       critical at higher readings.

3.  Critical adjustments of electronic bias, gain, and range
       required for each different sample.

4.  Readings affected by cell temperature, stirring rate.

5.  Laboratory facilities required, in the field:  distilled
       water and sink, drain and flush cell; standard solu-
       tions renewed  daily; gas blending system for calibra-
       tion samples;  "standard sample" gas handling proce-
       dures; daily handling-of fragile glass cell.

6.  Oxidative system furnace requires cooling water.

 7.            system consumes  the electrode;  slowly  for  spot
       samples, more  rapidly at high concentrations,  critical
       in  continuous  service.
                                                              Operating/Maintenance  Problems

                                                           1.   Instructions  provide  nothing on
                                                                  general  trouble-shooting:
                                                                  limited  to  research papers de-
                                                                  scribing specific uses  and
                                                                  precautions.

                                                           2.   Requires skilled  supervision.

                                                           3.   Set-up  time an  hour each  day.

                                                           4.   Service fair  in emergencies,
                                                                  normal supplies slow.
                                                                                                                                                 I
                                                                                                                                                 h-1
                                                                                                                                                 to

                                                                                                                                                 I
 Philips  (S02  Monitor)

   1.   Excellently  engineered (for
         ambient  range).

   2.   Adaptation to  H.S  (ambient)  can  be
         extended to  max  3-5  ppm.
 Improvements  Claimed for 1972

   1.   Separate modules  for SO,  or H.S;
         same  limits  on  maximum  concen-
         tration.
 1.



 2.

 3.
Chemical filter supplied to scrub SO- out of H.S (BaAc.)
  has limited capacity, not adequate for stack concen-
  trations.

Max H2S range is too low for stack emissions.

Dilution has to be accurate, if used, since reading
  is of concentration and not cumulative amount.
 4.   Unit has low tolerance for samples under pressure.
1.   Excellent manual of instructions,
      informative and easy to use.

2.   Field service extremely slow on
      anything except routine
      servicing, which is good.

-------
                                  - 127 -









 first column for each instrument lists advantages observed in use.




 Intrinsic disadvantages are considered inherent in the present design




 and construction of the equipment.  Operating and maintenance problems




 encountered reflect the quality of the manufacturers'  service and



 operating instructions.




           It should be emphasized that the statements  made are based




 largely  on the  study of a  single instrument,  which was on loan from




 the manufacturer and not purchased.  They  might or might  not  hold  true




 in  exactly the  same form for  another instrument.   These statements




 were  discussed  with the manufacturers  from time  to time throughout the




 program  and confirmed  by them as valid general observations.   Improve-




 ments  to correct or eliminate many of  the  problems noted  were made




 during the year by  Bendix  and Houston  Atlas,  and  to some  extent by




 Philips.   These are  listed  in the  first  column on  each  page as a second




 heading,  under  Advantages.  Certain major  points can be noted from the




 Table for  each  instrument.




           The Barton Titrator is  simple  to operate, has good  instructions,




 a wide linear range, and can  be  used for rapid direct measurements  in




 stack gases  from 0.1 to  1000  ppm.  The lowest two  attentuation scales



 (x.3 and  x.l) are very slow to come to equilibrium, unless the inner cell




 electrolyte  is  dumped out by  hand  for  forced circulation, and measurements




below 0.1 ppm should not be considered routine.   The Barton is good for




either H2S or total S (as H2S), in the absence of significant amounts of COS




 or CS^.   It  can also be used with chemical bubblers on a research basis




to measure individual components, but this is not suitable for routine

-------
                                 - 128 -
use in the field.  Even  the simplest of  these chemical separations,



SO_/H S, is only  85%  quantitative  in the 1-10 ppm  range:   it  cannot be
  ^  £,


relied upon at this level to indicate the absence of S0_ in H^S or vice



versa.   This SO™ bubbler must be avoided when COS is present.



          The  Bendix  unit analyzes  for  individual  compounds,  without



chemical separation.   It  is stable  and  reproducible over long periods



of time in GC  sampling,  timing, and in  zeroing as compared with the



other instruments.  The  design and  construction of the instrument is



being actively improved, with better instructions, and is  further



along than the similar instruments  manufactured by Tracer  or  Varian



Aerograph.  The original GC packing of  polyphenyl ether cannot be used



with the FPD sensor on stack emissions  that  contain percentage amounts



of CO, C02 or  hydrocarbons.  Good  results were obtained with  Poropak Q, and



GC packings can be changed  in the  field if desired.  This  is  better .done .at



the factory, however,  since it is  likely to  require a careful rebalancing



of electronic  controls to get proper operation of such features as automatic



zeroing or automatic.  attentnatJ.on.



          The  Houston Atlas pyrolytic/catalytic reduction  unit is



basically sound,  and  gives a true reading of total S compounds.  The



very long time lag in the unit under test prevented a good evaluation,



but this was apparently  due to the fact that this  was a prototype.  The



instrument can use better controls, auxiliary equipment and instructions.



It is considered  promising  for further  development, and many  improvements



along these lines have been made during the  year.

-------
                                 - 129 -
          The RAC and simple Houston Atlas PbAc_ tape recorders measure




H_S only, and cannot be used for more than a very limited time in the pre-




sence of S0_.  The RAC has less tolerance for SO- than the H-A, which




was not evaluated in as much detail.  Both units need operating




instructions:  this lack was particularly troublesome in trying to




adapt what is essentially an ambient range procedure for use  in stack




gas measurements.  The logarithmic reading scale and narrow range at a




given setting of  the RAC are undesirable.  Somewhat better  results




might have been obtained by running it at a much higher dilution,




which means using more carrier gas and fewer samples analyzed per hour.




This should cut down the high  frequency  of cleaning and adjustments




required to stay in operation with samples above 1 ppm.




          The two Dohrmann cells are essentially research type instru-




ments.  They can be used for an absolute calibration procedure in the




laboratory, in electrochemical equivalents.  The reductive cell has




a further advantage in not responding to SO-, while it measures H S and




mercaptan.  Requirements for constant recalibration and adjustments are





far too complex for routine operation, however, and not really suitable




for use in the field.   The absence of interfering chloride and cyanide




in small amounts is also not to be taken for granted in petroleum refinery




stacks.




          The Philips Monitor is a good instrument, but its. upper limit




of adjustment at 6-10 ppm of S0_ corresponds to only 3 ppm or at the most




5 ppm of ^S.   It also requires chemical filters for any distinction




between sulfur compounds.  This concentration range is of.little value




for direct measurement of stack gas samples.

-------
                                  - 130 -











     3.3.2  Data Logging Limitations




          The Esterline Angus system was plagued by mechanical problems




which resulted in some 30 service  calls during the year, and were never




fully diagnosed in the field.  These were largely in the electronic




system, and not in the replaceable printed cards.  Some of them at least




were due to troubles with internal grounding, and others were due to




cold soldered joints which gave  intermittent open circuits.




          Three different E/A units were finally supplied before sat-




isfactory operation was achieved.  The  first one gave occasional false




readings in specific channels, which were traced to grounding.  It




subsequently began "machine gunning" a  series of overprinted characters,




all on the same line.  Tapes containing such a record stopped the




IBM 1130 as illegal characters, which could not be read.  This defect




developed gradually over a period  of weeks, and finally became chronic.




It was not repairable in the field by putting in either new circuit




boards for the printer or a new printer, and the unit was returned to




the factory.





          The second unit behaved  normally at  first, then began printing




occasional  zeros  only  instead  of the proper  reading.  This malfunction




began on a  day when  the  trailer  van was hot, because the air conditioner




was off.   It  stopped when  the  air  conditioner  came on temporarily, and




it was  then  found that a small cooling fan directed at  the E/A unit




stopped  the  zeros.   The  same pattern repeated  on  later  days when the




zeros appeared at normal van temperature,  and  the cooling  fan was kept




in use,  directed  at  the  power  pack inside  the  case.

-------
                                 -  131  -










          The unit next developed a failure in the timing circuit, which




reset itself to a lower value each time it came to 22:40 on the clock.




This was corrected by finding loose soldered joints.  The unit then




developed a pattern of showing an occasional extra 100 or 200 digit




in the first nixie display, for  less than a second, on switching to a




new channel.  The momentary reading could be caught and recorded if




the printer happened to hit that specific second in its cycle.  This




began to occur about once in a thousand readings.




          The original E/A unit was finally reworked completely by the




factory, and run in parallel with the second one during the paper mill




field test, as a double check against each other in case .of unexplained




variations.  This third unit was placed on line alone for the last eight




weeks of the program and showed no mechanical failures during that time.




          Two minor recommendations were made for improvements in the




design of this equipment.  The first of these, which was incorporated




in the next production models, is a control switch which makes it




possible to select or not select individual channels which are printed




on the tape, instead of printing all channels up through any one selected.




A second desirable modification would provide the capability to print



 an index character on demand, to mark the end of data or some special




 comment in the operating notes.




          The Tally punch gave no problems, other than printing an




illegal character when it received a false signal from the E/A.  It




was recognized, after the fact, that the IBM might have been programmed

-------
                                  -  132 -
either  to  accept  and  by-pass  these  signals,  or  at  least  to be able to




restart  the  tape  and  continue at  the next  reading  after  it was stopped,




instead  of losing the whole print-out or manual  reading  of the tape.




      3.3.3   Sampling  System Limitations




           The mechanical  problems which had  to  be  overcome in the sample




probes  and pumping system were primarily related to  the  large amount




of water present  in the stack gases selected for testing.  Excess




water,  in  variable amounts, is no longer a limitation  in the system




as finally built.   The original design had separate  timers operating




in parallel, which are not easy to  keep in phase with  each other.




The  final  design  with a single timer is much simpler to  adjust and




operate.   The Barton  stainless steel probes  with integral glass fiber




filters  proved  suitable for severe  service with  air  blow back.




Simpler probes may be enough under less severe conditions, depending




chiefly on the nature and amount of the particulates present.




          Operating the gas sample  manifold  to maintain  a constant gas




pressure with variable demand requires a slight  overpressure, to avoid




short-term changes in line pressure with changing  demand.  A manifold




gas supply of 650  cc/minute when  using 500 cc or less was enough to




maintain the manifold pressure  constant at 108 +0.5 mm.  A simple




but sensitive pressure relief  valve is adequate.  The principal




operating problems were the failure of gas pumps and needle valves in




individual instruments, as noted  in Table 28  .  The use of good valves




and calibrated flow meters in  the blending system is essential to




successful operation.   The three mass flow meters in parallel used to

-------
                                 -  133 -
 feed calibration gases into the manifold are very sensitive to line




 pressure,  and a change of 5% or more in setting for any one of them




 required a minor adjustment in the others to maintain a constant flow




 rate.




           Operating the gas manifold under pressure creates a problem




 in feeding synthetic gas blends, which are conveniently made up using




 calibrated syringes and known volumes of gas.   A suitable system for



 this was devised using hydrostatic pressure, in a 5 gallon paint can




 containing a Mylar gas bag.   With water emptied out of the can, the gas




 bladder was evacuated and then inflated at atmospheric pressure with the




 desired blend,  made up in a  10-liter Houston Atlas acrylic gas-blending




 cylinder.   The  can surrounding the bag was then vented to the air,  and




 refilled with water under a  15 foot head.   This created a pressure  drive



 of 8 psig,  under which the sample was fed  through a mass flow meter




 into the manifold.




          Operation of the permeation tube system in a water  bath at




 30°C  was not  easy to  maintain  in  the  trailer van,  without  a ready




 supply  of cooling water.   The  use  of  a themostatically controlled air




bath  at  35°C or 40°C  is recommended as a preferred procedure.  Satis-




 factory  results were obtained  in  the  laboratory using  air bath units of




 limited  capacity  for H2S at 35°C and  SO  at  40°C.  Improved equipment




of this  type is now available  from Analytical Instrument Development,




Bendix, and several other manufacturers.  A  carbonyl sulfide permeation




tube was also tested, but its  useful  life was limited  to less than  3




weeks which is too short for convenience.

-------
                                 - 134 -
               4.  GENERAL CONCLUSIONS AND RECOMMENDATIONS







4.1  General




          None of  the instruments commercially available was able to




provide a routine  analysis at emission levels for H-S/TRS, or  for




individual S compounds, in both the refinery and paper mill test.  The




differences between ambient monitoring and stack measurements  for




odorants had not been adequately considered by the manufacturers as of




the start of this  program in 1971, or recognized in  their manuals




of instructions.




          The lack of adequate instructions is clearly significant,




in any program for instrument evaluation.  It suggests either  limited




resources, inexperience, or a lack of understanding  on the part of




the manufacturer.  The company may be too new, or too small to afford




the expense.  A poor manual, however, must be recognized as a  signal




that the user is on his own, and that the manufacturer may not really




know yet how to keep the instrument working when it  runs into  trouble.




          The problem of what to measure in odorant  emissions has




changed in the last few years, with the advent of simple pollution




controls in the plants most subject to complaints.   There is less of




the odorous mercaptans and higher alkyl sulfides, which means H,,S is




more useful as an  indication of total reduced sulfur.  At the same time,




the ratio of other inorganic sulfur compounds to H_S may increase




markedly in the remaining gas, so that there is 10 to 1000 times as




much SO  (or COS)  as the odorant H~S left to be measured.

-------
                                 - 135 -








          The choice of which instrument is best depends in part on what



source is to be analyzed, and in part on how well improvements made by



the manufacturers during 1972 work out in actual practice.  Specific



recommendations are made for possible further developments on the three



instrumental approaches which were found most promising.
   «


          The preferred approach where individual components are of



interest is the combination of GC/FPD.  Coulometric titration, however,



is the only method which has been extensively tested in the field.  The



chemical filters on which this method depends for component analyses



have limitations which depend upon the source to be measured.



4.2  Barton Titrator



          The Barton coulometer has given satisfactory field service



for years in a number of kraft paper mills.  They can use its results



as differential measurements for control.purposes, without having to



determine the composition of the odorant mixture.  This success is due



in part to the fact that the odorants present have a fairly constant



composition and consist predominantly of reduced S compounds,  with



H«S in excess, a variable small amount of S0_, and little or no COS/CS-.



The lack of an exact zero means that  the Barton  cannot give absolute



values.  Readings are also directly proportional to sample flow rate,



which is not accurately and continuously measured in present plant  •



practice.  The addition of a flow rate controller is recommended.



          The dependence of the Titrator on chemical bubblers is a



distinct disadvantage.   The phthalate S0« bubbler absorbs a small amount



of   S, amounting to 10-20% of the sample in the range of 1-10 pp, and it

-------
                                 - 136 -











gives no  indication when it  has  become  exhausted.   This  becomes




critically  important when the  stack  gas  composition is swinging  from




an excess of one  gas,  through  zero,  over to  an  excess of the  other.




The system  cannot be depended  on to  define the  amount or absence of




either H2S  or S02 at 1 ppm or  less in the presence  of 10 ppm  of  the




other.  This type of change which is rare in kraft  mill  stacks is com-




mon in the  gases  from  other  sources, such as a  Glaus plant  afterburner.




The KAP bubbler fails  completely in  this mixture, which  contains




significant amounts of COS and CS  .  The failure cannot  be  ignored, since




the interaction of COS and the phthalate solution releases  SO ,  hydro-




drolyzes some COS to H2S  or CS_, and exhausts the activity of the electrolyte




in the Barton cell.  The  Barton  does not respond to COS  as such but it is




slowly poisoned by CS«,  in an  anomalous  reaction which appears to continue




after the sample flow is  stopped.




          Research on  the chemistry  involved in the COS/SO  /KAP  inter-




actions is  recommended.   The reasons for these  effects are not entirely




clear, and  the results might lead to improved scrubbing  procedures.




Until they  are available,  the  Barton system cannot  be used  to measure




either total S or total  reduced  S compounds correctly in the  presence




of SO- and COS or CS .  There  is no assurance that  any system better than KAP




can be found,  however, and improvements  in its  use may be more promising




than the search for alternates.




          The Barton can,  of course, be  used with other  methods  of




separation such as GC, or  pretreatment to convert other  S compounds to S0»




or H S.   The usual procedure to  use the  factor  for  H S in measurements




and report the results as  H«S  is a useful compromise, but it  gives no




absolute values.

-------
                                 - 137 -











          Research on the effect of improved circulation of the electrolyte




in the sensor cell is also recommended.  This could lead to a more stable




zero, more accurate measurements at low scale readings, and more rapid




equilibration at the lower attenuation scales.  The very short term




variations in cell readings attributed to uneven circulation make the




Barton much less suitable for electrical recording than for reading




on a visual chart.  Electronic integration or averaging is a possible




alternate approach for this particular problem.






           There is a coulometric instrument available which combines the




 wide range of the Barton, the stability of the Philips, and the precision




 of the Dohrmann cells.  Specific suggestions have been made along these




 lines which remain to be implemented.






 4-3   Bendix  Environmental Chromotograph




          The Bendix GC/FPD combination has a more stable  zero than  the




 Barton, and  the  response  of the  two  instruments  to gases  to which both




 are sensitive is similar  in accuracy,  stability  and linearity up  to




 about 80 ppm.  The Bendix is superior  in stability, but limited in




 linearity by the intrinsic response  curve of the FPD sensor.  Data




obtained in this program indicate that with a 1 cc sample size the




upper limit of linearity varies with the compound at  30-50 ppm for COS




compared to 80-120 ppm for HZS,  S02 and CSH  (methyl mercaptan).




          The 80-30 ppm range is high enough for direct measurement of




stack gas concentrations at the Ie8al limits of 10 ppm or 17.5 ppm




set by recent legislation in several states, and higher concentrations




can be measured  by appropriate dilution.

-------
                                 - 138 -
          The separations obtained in this instrument depend upon the



GC component:  the polyphenyl ether/H PO, on Teflon supplied by the



manufacturer cannot be used to measure ppm quantities of S compounds



in stack gases containing percentage amounts of CO and CO^.  Good



results are obtained using a packing of Poropak Q, which separates



H_S, COS, SO  , CSH, CSC, and CS  and is not swamped by CO and CO .  The
 £»          £                  £                                £•


system must be internally rebalanced for proper operation of automatic



zeroing and automatic attenuation when any change in GC packing is made.



          The instrument model tested in this program requires skilled



maintenance, and it would need rebalancing every six months for



optimum operation.  Numerous improvements made by Bendix and other manu-



facturers of similar instruments during the year are intended to give



better long-term stability.  This needs to be confirmed by further



testing.



          Additional development work is recommended on modifications to



permit the analysis of more concentrated samples, up to about 0.1% of



H-S/SO-.  This might be accomplished by using one of the present two



valved sample loops for the purpose.  The use of smaller sample size



by-micro-injection or an equivalent procedure such as a GC sample



splitter is recommended as a desirable alternate.  This or a built-rin



dilution system could extend the useful range of the instrument by a



factor of 10  to 20 or more with no significant loss in accuracy, to cover



the entire range of interest for S odorant emissions.

-------
                                 - 139 -



4.4  Lead Acetate Tape Systems

          The  Houston Atlas  system for sample pyrolysis and catalytic

 reduction to H2S  is  promising for  further development,  and suitable for

 use  with a  simple tape  sensor.   The instrument tested here was a

 prototype which suffered  from a  25 minute time lag,  attributed to

 oversized conversion chambers and  small  unheated  lines.   This  interfered

 with its evaluation, but  operation was basically  satisfactory.   Many

 improvements were made during the  year,  which are still to be  tested

 in  the  field.  One improvement which may help to  eliminate the problems

 of  zero drift  is  a new method of sensing which measures the rate of

 color change in the  tape,  rather than  its absolute value.

          The  Houston Atlas  pyrolytic/catalytic system  could be used

 equally well with other sensors, such  as a re-engineered coulometric

 cell.   Present indications are that for gases free of massive  amounts of

S02  or COS,  coulometric  titration after converting all S compounds to  H S

should provide  a  good monitoring  system.

 4.5  Minimum Requirements  for
     Future Instruments	

          The  nature of the  problem of monitoring sulfur containing

 odorants in stack emissions  has  been redefined, herein.  It now appears

 that an instrument for  this  purpose should be capable of monitoring

 total reduced  sulfur compounds,  or H-S alone as an alternate,  in the

 presence of SO ,  COS/CS-,  CSH, heavier S odorants and indeterminate.

 amounts of  CO., CO,  and water vapor.  Minimum performance requirements

 can be  set  for such  an  instrument:

-------
                       -  140 -
1.  Linearity, stability and reproducibility  of +  2-5% or better,




   which  are  adequate  in  the  present  Bendix and Barton instru-




   ments.




2.  Conversion from MV to ppm should be linear to + 3%




    of scale.   It can use different factors for different




    parts of scale but this should be minimized.




3.  Response time needed of 90% in 5 minutes or less, 98%




    in 15 minutes.




4.  Direct reading capability for the concentration range




    .1-30 ppm is of primary interest for odorant pollution




    control.




5.  Reading directly or with built-in dilution for the usual




    emissions range of 0-300 ppm or more, with a precision




    of + 2-5% of scale.




6.  A positive distinction between H~S and S0~ at .1 to




    30 ppm is required for monitoring odorous  combustion




    stacks.




7.  Significant advantages are recognized for an instrument




    using GC or an equivalent separation procedure which




    will permit the simultaneous analysis of 3 or 4 selected




    S compounds, including S02 and reduced S compounds, as




    odorants or as key components for odor control.

-------
           8.   Satisfactory freedom from interference  (e.g.,  + 10%  of




               measured  values)  must  be  demonstrated,  for  the sensor




               and  for auxiliary filters or  separators required,  at high




               ratios of the  interferent gases  (including  SO-,  CO , CO,




               and  H20)  up  to  10,000-fold and COS  up to  10-fold the amount




               of H2S or other odorant to be measured.




           9.   Construction and  engineering rugged enough  for field use,




               by skilled non-professional personnel with  occasional




               technical  supervision.




         10.  A manual of instructions  is required with an outline




              of trouble-shooting procedures which will enable the




              user to keep the  instrument in operation  for a prescribed




              warranty period (at least  6 months to a year), without




              special maintenance by manufacturers' representatives.




          For reduced systems where the  amount of H.S/TRS is high




relative to S02 or S03>  the ability to  give either a simple  total  S




or total reduced S measurement with high reliability and  consistency




would appear to be prime goals.   The non-uniformity of response  to




different S compounds which is a major  drawback of coulometric titra-




tion could be minimized by some sort of  pyrolysis/oxidation/reduction




unit.  The reductive unit developed by Houston Atlas appears more




rugged at this point than the earlier oxidative unit offered by




Dohrmann.  Either approach can convert all sulfur to H~S  (or to  SO )




for coulometric analysis.  Further developments are recommended along




the line of a Barton or  Dohrmann cell of improved stability,  or a




Philips cell of much wider range.

-------
                                 - 142 -
           Similarly,  for  total  S,  the  improvements undertaken  over  the




past  year  by  Houston  Atlas  in the  detector  system of  their  lead  acetate




paper tape sampler  may result  in an economical measurement  tool  of  improved




 reliability.   Further evaluation of the improved models is  recommended.






           A major emphasis  must  be placed on  the evaluation of sampling




systems which will  operate  in stack gases containing  high amounts of




water  and  particulates.   The rapid separation  method  developed in this




project is  one approach,  and several systems using the  diffusion of




water  vapor through permeable membranes have become available  since




1971.  This problem is particularly important  for  the measurement of




odorous I^S at ppm levels in the presence of percentage amounts of




moist  SO-,  to avoid the rapid interaction of H2S with SO- in adsorbed




or liquid water.   On  this basis, both approaches appear to work.

-------
                                 - 143 -
                              BIBLIOGRAPHY






1.  Stevens, R. K., O'Keeffe, A.E., and Ortmann, G. C., "Absolute




    Calibration of a Flame Photometric Detector to Volatile Sulfur




    Compounds at Sub-Part-per-Million Levels", Environ. Science Tech.




    1652-655 (1969); Stevens, Mulik, J.D., O'Keeffe,  and Krost, K. J.,




    "Gas Chromatography of Reactive Sulfur Gases in Air at the Parts-




    per-Billion Levels", Analytical Chemistry 4J^ 827-831  (1971).




2.  O'Keeffe, A.E., and Ortmann, G.C., "Primary Standards for Trace




    Gas Analysis", Analytical Chemistry 38, 760 (1966).




3.  Greer, D.G., and Bydalek, T.J., "Response Characteristics of the




    Melpar Flame Photometric Detector for Hydrogen Sulfide and Sulfur




    Dioxide", Environ. Science & Tech. _7 153-155 (1973).




4.  Cooper, H.B.H., Jr., and Rossano, A., Jr., "Continuous Monitoring




    of Sulfur Compounds in the Pulp and Paper Industry", presented at




    12th Air Pollution Methods Conference, California  State Dept.




    Public Health, Los Angeles,  April 1971.




5.  Blosser, R.O., and Cooper, H.B.H., Jr., "Compendium of Methods for




    Measuring Ambient Air Quality and Process Emissions", NCASI




    Technical Bulletin No.  38, National Council of the Paper Industry




    for Air and Stream Improvement, New York, Dec.  1968.




6.  Thoen,  G.N., DeHaas, G.G., and Austin, R.R.,  "Continuous Measurement




    of Sulfur Compounds and Their Relationship to Operating Kraft Mill




    Black Liquor Furnaces", TAPPI, 5_2^ 1485-87 (1969).




7.  Thornsberry, W.L.,  Jr., "Isothermal Gas Chromatographic Separation




    of C02, COS, H2S,  CS2,  and S02",  Anal. Chem.  4_3 452-53 (1971);




    H. Hartmann, "Improved  Chroraatographic Techniques  for Sulfur




    Techniques for Sulfur Pollutants,  AIAA Joint  Conference for Sensing




    of Environmental  Pollutants,  Palo Alto,  Calif., Nov. 1971.

-------
PAGE  1  APP.ENDIX I   PROGRAM 1:  TAKE READING

// JOB

LOG DRIVE   CART SPEC   CART AVAIL  PHY DRIVE
  0000        0010        0010        0000

V2 M10   ACTUAL 16K  CONFIG 16K

// FOR
»LIST ALL
»ONE WORD INTEGERS
MOCSIPAPER TAPE.DISK)
»IOCS(1132 PRINTER)
»NAME ARGL
      INTEGER 0(11)
      REAL IMV(21>
      DIMENSION NZER(5)
      DEFINE FILE 1  ( 300 .45 iU» JREC )
      DATA 0/11»0/
      DATA IMV/21«0.0/
      DATA NZER/999«4«1/
      JREC='l
      DO 1 1=1,300
      WRITE!1'JREC) NZER
    1 CONTINUE
      JREC=1
      CALL SAVE
      IROW=0
   10 CONTINUE
      I ROW=I ROW*1
      IFUROW-301) 12,100,100
   12 READI4.200) IDAYiIHR•IMIN
      JCOL=0
      DO 80 J=l,21
      READU.210) 0
      IF(0!l>-7744)  15,90,15
   15 JCOL=JCOL+1
      IF(QU>-24640) 20,50,20
   20 CALL TRRED
      READI4.220I D1,IMV(J),02
      GOTO 80
   50 CALL TRRED
      READU.250) 01 • IMV ( J ) ,D2
   80 CONTINUE
   90 CONTINUE
      WRITE!1'JREC) I ROW,I DAY,IHR,IMIN,JCOL,(IMV(J),J=1,JCOLI
      GOTO 10
  100 CONTINUE
      WRITE(3,270)
  ?00 FORMAT!13,12,IX,I?,?X)
  210 FORMATI11A1)
  2?0 FORMAT(/\3,F5.1,A2)
  250 FORMAT(A3,F6,1,A2)
  270 FORMAT(T5,'THE NUMBER OF ROWS HAS EXCEEDED 300,EXECUTE  THE  PRINT
     ?PROGRAM A.MD RESTART THE READ PROGRAM')
      CALL EXIT
      END

-------
PAGE
            APPENDIX I  PROGRAM 1:  TAKE READING (Cont'd)
VARIABLE ALLOCATIONS
   IMVIR 1=0030-0008
   '  III )"0047
     J(I )=004D
         DKR )°0032
       IROWd 1=0048
             D2IR 1=0034
           IDAYII 1=0049
                      NZERII )=003A-0036
                       IHRU )=004A
                                                                      0(1  1=0045-0038  JRECU  >=0046
                IMINII >=004B
                                                                                                       JCOLII  )=004C
STATEMENT ALLOCATIONS
 200  =0059  210  =005F  220
 50   =0121  80   =0131  90

FEATURES SUPPORTED
 ONE WORD INTEGERS
 IOCS

CALLED SUBPROGRAMS
 SAVE    TRRED   FLD     FSTO
 SDFIO   SDWRT   SDCOM   SDAI

INTEGER CONSTANTS
     1=0050    300=0051

CORE REQUIREMENTS FOR ARG1
 COMMON      0  VARIABLES

END OF COMPILATION

// DUP
     ARG1
DB ADDR  343A
                              = 0062
                               013A
                      250
                      100
           = 0066
           = 0150
         270  =006A  1
=OOC3  10   =0006  12   =OOE2  15   =0101  20
                                                                                                                   = 010F
                                 PRNTZ
                                 SDFX
                          PAPU
                          SDI
                  SRED
                 SWRT
                                                                  SCOMP    SFIO
                                                                  SIOAI
                                                          SIOFX   SIOF
                                                                                                          SIOI
                                                                         SUBSC
                             0=0052     301=0053


                               80   PROGRAM     274
                                                      4=0054
                                                 21=0055   7744=0056   24640=0057
                                                                                                      3=0058
                                                                                                                          T
                                                                                                                          to
CART ID 0010
»STOPE      WS  UA  ARG1
CART  10 0010   DB ADDR   3B29
DB CNT


DR CNT
0017


0017

-------
PAGE 1 APPENDIX I  PROGRAM 1A:  PRINT.-OUT

// JOB

LOG DRIVE   CART SPEC   CART AVAIL  PHY DRIVE
  oooo        0010        noio        oooo

V2 M10   ACTUAL 16K.  CONFIG 16K

// FOR
«LIST ALL
#ONE WORD INTEGERS
»IOCS(DISK«1132BRINTER)
• NAME ,j4fi£i£
      REAL I.MVI21)
      DIMENSION NARAYI20)
      DEFINE FILE 1  I 300.45.U.JREC>
      DO 5 I 1 = 1.13
    5 NARAYI II) = I1-1
      DO 40 Mali 3
      JREC=1
      WRITEI3.330)
      WRITE(3.320)
      WRITEn.310)
      WRITE I 3,270)  (NARAYI 111.11 = 1.13)
      WRITE I 3.230)
      DO 30 1=1.300
      READII•JREO  iROW,iDAY, IHR,IMIN,JCOL,
-------
PAGF
            APPENDIX I PROGRAM 1A:  PRINT-OUT  (Cont'd)
   85
   90
  100
  120
  230
  240
  270
  2flO
  310
  320
  330
  340
  350
  360
  370
  380
  440
      IF(K)  90.85*90
      WRITEI3.320)
      wRiTE(3i3flO) IROW.IDAY.IHR.IMIN,(IMVIji
      CONTINUE
      CONTINUE
      FORMAT(IX)
      FORMAT<1X.I3.2X.I3.2X.2I2.13(2X,F6.1 ) I
      FORMAT(T7.'DAY1.T13,'HR'»13I8)
      FORMAT(T6.1	  	'.T18.13I'  	
      FORMATIT20.101('-'))
      FORMAT
                                             .J=14»JCOL)
      FORMAT
-------
PAGE   1
            APPENDIX I  PROGRAM 2:  SEMl"-i,OG CONVERSION  (RAC)
                                                                    PAGE   2  APPENDIX I  PROGRAM 2:  SEMI-LOG CONVERSION (RAC) (Cont'd)
// JOB

LOG DRIVE
  0000
CART SPEC
  0010
CART AVAIL  PHY DRIVE
  0010        0000
V2 M10   ACTUAL 16K.  CONFIG 16*

// FOR
•LIST ALL
•ONE WORD INTEGERS
«IOCS(CARD,1132 PRINTER)
•IOCSIDISK)
                                                                       60
                                                                       65
      FNTEGER CHANOUOI
      REAL IMVI10)
      DIMENSION JARAYI3) ,X(150»10)
      DIMENSION JROWI 150) »JO/ '(150) i JHRI 150). JM INI 150)
     2.TIME(150),YI150»10),SL(10I»2(10).R(10)»EI10),AA(10I
     3.YYI150.10) .YYY(150)iJJROW(150).TTIME(150)
      DEFINE FILE  1 ( 300,45 ,U ,JREC )
      EQUIVALENCE  ( I ROWS. JAR AY ( 1)1.1 IROWF . JARAY ( 2 ) ) , ( NCHAN , JARAY ( 3 ) )
      DATA X/1500*0.0/
      DATA Y/1500»0.0/
      DATA TIME/150»0.0/
      DATA YY/1500«0.0/.
      DATA YYY/150»0.0/
      DATA JJROW/150»0/
      DATA TTIKE/150»0.0/
      READI2.300)  IA
      00 190  JJ=1,IA
     . WRITE (3. 22 5)
      CALL NOFI (3. JARAY, 2)
      CALL NOFI 
                                                                      120
                                                          122
                                              123
                                              130
CONTINUE
WRITE(3,370)
TTIMEI 11=0
00 65 K=l» NCHAN
YY(1,K)=ALOG(ABS(X(1,K) )  I
CONTINUE
JJROWd )=JROW(1 )
12 = 1
DO 130 M=?,NROW
TIME (M) = <60*JHR(M)+JMIN(M) )-(60»JHRI 11+JMINI 1 I )
DO 120 K = l, NCHAN
Y(M,K)=ALOG(ABS(X(M,K) I )
CONTINUE
IFITIMEIM-1 I-TIME(M) )   122,130,122
IZ=IZ+1
N=I2
J JROWI N)=JROW(M)
                                              160
                                              170
                                              180
                                              190
                                              210
                                              225
                                              230
                                              240
                                              250
                                              260
                                              270
                                                              DO 123 K=l, NCHAN
                                                              YY(N,K)=Y(M,K)
                                                              CONTINUE
                                                              CONTINUE
                                                              DO 150 M=l, IZ
                                                                                              NCHAN)
                                                              K=MOD(L,6)
                                                              IFIKI 140,135,140
                                                          135 WRITE(3,340)
                                                          140 WRITF.13,350) JJROW(N) ,TTIME(N) ,(YY(N,K) ,K
                                                          150 CONTINUE
                                                              WRITE (3. 370 I
                                                              DO 170 K=l, NCHAN
                                                              DO 160 N=1,IZ
                                                              YYY(N)=YY(N,K)
                                                              CONTINUE
                                                              CALL FITSL   ( 12, TTIME.YYY, SLOPE, TINT. CORCO, SEE!
                                                              SL(K)=SLOPE
                                                              Z(K)»TINT
                                                              R(K)=CORCO
                                                              E(K)=SEE
                                                              AA(K)=EXP(Z(K))
                                                              CONTINUE
                                                              DO 180 K"li NCHAN
                                                              WRITE I 3, 360) CHANO(K) ,SL(K),ZIK),R(K),E«),AA(K)
                                                              CONTINUE
                                                              CONTINUE
                                                              FORMAT(1X,I3,2X,13,2X,2I2,10(2X,F6.1) )
                                                              FORMAT ( 1H1 )
                                                              FORMATI1X,' INITIAL ROW NUMBER" ' , I 3/ )
                                                              FORMAT11X, 'FINAL ROW NUMBER" ' t I3/ )
                                                              FORMATdX, 'NUMBER OF CHANNELS" ' I 2///// I
                                                              FORMATdX, 'CHANNEL  NUMBERS ' ,3X , I 2 ,9 ( 6X , I 2 ) // )
                                                              FORMATIT7, 'DAY' ,T13, 'HR' )
                                                         . 280  FORMATdX)
                                                          300  FORMAT(113)
                                                          340  FORMATdX)
                                                          350  FORMATdX, I 3,3X,F5.0,10(3X,F8.5) I
                                                          360  FORMATdX,I3,3X,F8,4,3X,F7.4,3X,F6.3,3X,F7.3,3X,F7,3)

-------
PAGE 3 APPENDIX I PROGRAM 2:
370 FORMAT!//////)
CALL EXIT
FND
VARIABLE ALLOCATIONS
IROWSII nQ009 JARAY1I
Y(R =18A4-OCEE SL(R
YYIR =24CO-190A YYYIR
CORCOIR =2732 SEEIR
JJROWII =2A25-2990 CHANOII
NROWI I = 2A34 Ml I
JCOLII )=2A3A JII
Nil )=2A40
STATEMENT ALLOCATIONS
210 = 2A4C 225 =2A58 230
350 =2AAC 360 =2AB5 370
122 =2C98 123 =2CDA 130
SEMI-LOG CONVERSION (RAC) (Cont'd)




1=0009-0007
)=lBB8-lflA6
>=25EC-24C2
1=2734
)=2A2F-2A26
)=2A35
)=2A3B


=2A5B 240
=2AC2 10
=2CE3 135




I ROWF I I
ZIR
T T I ME ( R
JROWI I
JRECI I
I ROW I I
K( I


= 2A6A
= 2B98
= 2000




1=0008
)=18CC-18BA
=2718-25EE
=27CO-2738
= 2A30
= 2A36
= 2A3C


250 =2A78
50 =2BA1
140 =2004




NCHANI I
R(R
IMVIR
JOAYI I
IAI I
IDAYI I
JCHANI I






1=0007





)=18EO-16CE




X(R
EIR
>=272C-271A SLOPEIR
)=2863-27CE
)=2A31
)=2A37
)=2A3D


260 =2A8B 270
52 =2BBE 55
150 =2029 160





= 2A9E
= 2BC2
= 2052
JHRI I
JJII
IHRI I
LI I




=OBCO-COOA
=18F4-18E2
= 272E

=28F9-2864
= 2A32
= 2A38
= 2A3E


280 =2AA6
60 =2RE8
170 =2090





300
65
180




T I ME I R
AAIR
TINTIR
JMINI I
K I
I M I N I I
IZI I
=OCEC-OBC2
=1908-18F6
= 2730
=298F-28FA
= 2A33
= 2A39
= 2A3F


=2AA8 340 =2AAA
=2C1B 120 =2C7A
=2087 190 =2DCO
FEATURES SUPPORTED
 ONE WORD INTEGERS
 IOCS

CALLED SUBPROGRAMS
 KOFI    MOO     FALOG
 SWRT    SCOMP   SFIO

INTEGER CONSTANTS
     2=2A46      1=2A47
                         FABS
                         SIOFX
                  FITSL
                  SIOIX
FFXP
SIOI
FSUOX
SUBSC
FLO
SDFIO
                             3=2A48
                                         6=2A49
                                                     Q=2A4A
FLOX
SORF.O
                                                                60=2A48
FSTO
SDFX
FSTOX
SDI
FLOAT
CARDZ
                PRNT2
                        SRED
                                                                                                                          T
CORE REQUIREMENTS FOR ARG2
 COMMON      0  VARIABLES  1082?  PROGRAM

FND OF COMPILATION

// DUP
                              900
"OELF.TF.
CART ID 0010
     ARG2
OB AOOR  3884
                               OB CNT   0275
»STORE      WS  UA  ARG2
CART 10 0010   OB AOHR  3SC4   OB CNT   0275

-------
PAGE 1  APPENDIX I PROGRAM 3: MEAN AND_STANDARD DEVIATION

// JOB
LOG DRIVE
  0000
CART SPEC
  0010
                        CART AVAIL
                          0010
PHY DRIVE
  0000
PAGE   2  APPENDIX I  PROGRAM 3:  MEAN AND STANDARD DEVIATION (Cont'd)

      DO 100 M=1»NROW
      Z=X(M,K)
      ZSO=Z«»2
      TZ=TZ+Z
V2 M10   ACTUAL 16K  CONFIG  16K
// FOR                                                                 70
«LIST ALL
»ONE WORD INTEGERS
»IOCS(CARD.1132 PRINTER)
»IOCS(DISK)
#NAME ARG3 ,
      INTEGER CHANO(IO)
      REAL IMV(IO)
      DIMENSION JARAYI3) .XI150.10)                                    100
      DIMENSION JROW!150).JDAY!150)»JHR(150).JMINI150)                110
      DEFINE FILE'1 (300.45.U.JREC)                                    150
      EQUIVALENCE (I ROWS.JARAY I 1)) . (IROWF . JARAYI 2) ) . (NCHAN .JARAYI 3 I)  210
      READ!?.3001 IA                                                  220
      DO 150 JJ=1 .IA                                                  225
      WRITEI3.225)                                                    230
      CALL NOFI(3.JARAY,2)                                            240
      CALL NOFI(NCHAN,CHANO.2)                                        250
      WRITE(3»21!0)  JARAY(l)                                           260
      WRITEI3.240)  JARAYI2)                                           270
      WRITE(3,250)  JARAYI3)                                           280
      WRITE(3,260)  (CHANO!I).1=1,NCHAN)                               290
      WRITE I 3,270)                                                    300
      JREC=IROWS                                                      310
      NROWaIROWF-IROWS+1
      DO 50 M=1»NROW
      READ(l'JREC)  I ROW,I DAYi
      JROW(M)=IROW
      JDAY(M)=IDAY
      JHR(M)=IHR
      JMIN(M)=IM1N
      DO 10 K=liNCHAN
      JCHAN = CHA,NO(K) + 1
      XIM,K)=IMV.11X.F3.0)
                                                              FORMAT I1H1)
                                                              FORMAT! IX.' INITIAL  ROW  NUMBER"'»I3/I
                                                              FORMAT!IX.'FINAL ROW NUMBER='.I3/I
                                                              FORKATIlXf'NUMBER OF CHANNELS"'I2/////I
                                                              FORMATdX.'CHANNEL  NUMBERS ' »3X , I 2 .9 <6X . I 2 1 // )
                                                              FORMATIT7.'DAY',T13.'HR')
                                                              FORMATI1X)
                                                              FORMAT I IX.'CHANO',12X.'MEAN'.12X.'STDEV.12Xt'DF'
                                                              FORMAT(113)
                                                              FORMAT!//////)
                                                              CALL EXIT
                                                              END
                              IHR.IMIN.JCOL.IIMV(J).J=liJCOL)

-------
PAGE 3  APPENDIX I PROGRAM 3:  MEAN AND STANDARD DEVIATION (Cont'd)
VARIABLE
I ROWS! I
Z(R
SD21R
JHRl I
I I I
IKINII
ALLOCATIONS
1=0009
)=ORD6
)=ORE2
)=ODAD-OD18
)=OE51
)=OE57

JARAYl I
ZSOIR
SD1R
JM I N ( I
NROWt I
JCOLd

)=OC09-0007
)=ORD8
)=OBE4
)=OE43-ODAE
)=OE52
)=OE58

I ROWF (
TZ(
AVGl
CHANOI
Ml
J(

I
R
R
I
I
I

1=0008
)=ORDA
)=ORE6
1 =OE4D-OE<
)=OE53
)=OE59

NCHANl I
TZSOIR
DF(R
»4 JRECU
I ROW ( I
MI

1=0007
I = OPCC
)=ORE8
)=OE4E
>=OE54
)=OE5A
STATFMFNT ALLOCATIONS
 210  =OE69  220  =OF75
 300  =OEEO  310  =OEE2
 150  =10C4

FEATURES SUPPORTED
 ONE WORD INTEGERS
 IOCS

CALLED SUBPROGRAMS
 NOFI    MOD     FSORT
 SRED    SWRT    SCOMP

REAL CONSTANTS
  .500000E-01=OE62
225  =OE7E  230  =OE81
10   =OFB8  50   =OFC1
          FSIGN
          SFIO
                                 FADD
                                 SIOFX
INTEGER CONSTANTS
  1=OE65
                             3=OE66
                FSUR
                SIOIX
                6=OE67
                                                  240   =OE90   250   =OE9E
                                                  52    =OFDE   55    =OFE2
                                                          260   =OEB1
                                                          60
X(R
SZSOIR
JROWl I
IAI I
IDAYl I
JCHA.Ml I
Bl 270
08 70
)=OBCO-OOOA
1=OBDE

)=OC81-OBEC
)=OE4F
)=OE55
)=OE5R
= OEC4
= 1063



280
100
I MV ( R
SDKR
JDAYl I
JJl I
I HR ( I
L(I
= OECC
= 10B2
)=OBD4-OBC2
)=OBEO
)=OD17-OC82
)=OE50
)=OE56
)=OE5C
290 =OECE
110 =10BB
         FLD
         SIOF
FLDX
SIOI
FSTO
SUBSC
FSTOX
SDFIO
FDVR
SORED
FAX I
SDFX
               OE6>8
FLOAT
SDI
CARDZ   PRNTZ
                                                                                                                            00
CORE REQUIREMENTS FOR ARG3
 COMMON      0  VARIABLES    3682   PROGRAM

FND OF COMPILATION

// DUP
                              620
*DELETE
CART  ID 0010
     ARG3
DB ADDR  3B09
*STORE      WS  UA   ARG3
CART  ID 0010    DB ADDR   3R09
      DB CNT


      DB CNT
002B


C02B

-------
      APPENDIX  II  REGRESSION ANALYSIS FOR FAIRED CURVE

             f?        L.«R
           30.000
          100.000
                      -1.000
                      -0.523
                       0.0
                       0.177
                       1.000
                         177
                       2.000
COL/ROW
   1
   2

   1
   5
   6
   7

CODE
?« 0

ANALYSIS
?* MU

NEW DATA
?« NO
SPECIFY THE DEPENDENT VARIABLE
?" 2

NO. INDEP VAR
?- 1
1.000
0.273
0.0
0.228
1.000
2.182
1.000
 1
0.010
0.023
0.100
0.270
0.860
2.600
9.000
  0.951
 6
1.000
  681
  000
  323
  001
0.172
0.911
SPECIFY THESE VARIABLES
? 5
VARIABLE
     5
           REG.COEF.
              1.00530
                       STD.ERROR COEF.
                             0.01656
       COMPUTED T
         60.71373
INTERCEPT                 1.05075
MULTIPLE CORRELATION      0.99932
STD. ERROR OF ESTIMATE    0.01355
                                    (ADJUSTED R  =
                                    (ADJUSTED SE=
BETA COEF.
   0.99932
                                                     0.99932)
                                                     0.01355)
                 ANALYSIS OF VARIANCE FOR THE  REGRESSION
    SOURCE OF VARIATION        D.F.   SUM OF  SQ.      MEAN  SQ.
ATTRIBUTABLE TO REGRESSION      1        6.991        6.991
DEVIATION FROM REGRESSION       5        0.009        0.002
     TOTAL                      b        7.001

PRINT RESIDUALS
J" YES
CASE NO   Y OBSERVED   Y ESTIMATED

               1.000
               0.523
               0.0
               0.177
               1.000
                 177
                               960
                               596
                             0.015
                             0.179
                             0.985
                             1.168
               2.000
                             2.010
RESIDUAL
-0.010
0.073
-0.015
-0.002
0.015
0.009
-0.010
STD.RESID.
-0.922
1.681
-1.011
-0.015
0.317
0.211
-0.231
TEST OF EXTREME RESIDUALS
  RATIO OF RANGES FOR THE SMALLEST RESIDUAL..
  RATIO OF RANGES FOR THE LARGEST RESIDUAL...
  CRITICAL VALUE OF THE RATIO AT ALPHA =  .10

SAVE ESTIMATES
?« NO
                                                     0.015
                                                     0.190
                                                     0.13t
                                                                F VALUE
                                                              3686.157
                                                                          c/e/r/c/K twu$

-------
      APPENDIX II  RE-ANALYSIS AFTER DISCARDING ONE POINT
COL/ROW
   1
   2
   3
   1

   6

CODE
?!t 0

ANALYSIS
?« MU

NEW DATA
?" NO
  0.100
  1.000
  3.000
 10.000
 30.000
100.000
-1
  000
0.0
0.177
1.000
1.477
 2.000
                                 fl^fff    AS
 .000
0.0
0.228
1.000
2.1B2
1.000
0.010
0.100
0.270
0.860
2.600
9.000
-2.000
-1.000
-0.569
-0.066
 0.115
 0.951
 6
1.000
1.000
0.323
0.001
0.172
0.911
SPECIFY THE DEPENDENT VARIABLE
?« 2

NO. INDEP VAR
SPECIFY THESE VARIABLES
? 5
VARIABLE
     5
 REG.COEF.
    1.02191
 STD.ERROR COEF.
       0.00860
                  COMPUTED T
                   118.80806
INTERCEPT                 1.01181
MULTIPLE CORRELATJON      0.999b6   (ADJUSTED R =
STD. ERROR OF ESTIMATE    0.02027   (ADJUSTED SE=
                    BETA COEF.
                       0.99986
                                           0.99986)
                                           0.02027)
                                                                                                                            >
                                                                                                                            i
                                                                                  cc  =
                 ANALYSIS OF VARIANCE FOR THE REGRESSION
    SOURCE OF VARIATION        D.F.  SUM OF SQ.     MEAN SQ.    F VALUE
ATTRIBUTABLE TO REGRESSION      1        5.802        5.802   11115.318
DEVIATION FROM REGRESSION       1        0.002        0.000
     TOTAL                      5        5.801

PRINT RESIDUALS
?» YES
                                                                                              Otit,
CASE NO   Y OBSERVED   Y ESTIMATED   RESIDUAL   STD.RESID.
                 000
                 0
               0.177
                 000
                 177
               2.000
                     999
                     023
                   0.161
                   0.978
                   1.169
                   2.020
                   0.001
                   0.023
                   0.013
                   0.022
                   0.008
                   0.020
                            -0.016
                            -1.128
                             0.662
                             1.092
                               106
                            -0.986
TEST OF EXTREME RESIDUALS
  RATIO OF RANGES FOR THE SMALLEST RESIDUAL..
  RATIO OF RANGES FOR THE LARGEST RESIDUAI	
  CRITICAL VALUE OF THE RATIO AT ALPHA =  .10

SAVE ESTIMATES
?* NO
                                           0.061
                                           0.193
                                           0.182

-------
  BIBLIOGRAPHIC DATA
  SHEET
1. Report No.
   EPA-R2-73-180
 4. Title and Subtitle
       Evaluation  of  Measurement Methods and Instrumentation
       for Odorous  Compounds in Stationary Sources
       Vol.  II:  Field  Testing
3. Recipient's Accession No.
                                                5. Report Date
                                                Prepared March 1973
                                                6.
 7. Author(s)
        Homer  J.  Hall
                                                  Performing Organization Rcpt.
                                                  No" GRU.2DJAB.73
 9. Performing Organization Name and Address
       Esso Research  and  Engineering Company
       Government Research  Laboratory
       P.O. Box 8
       Linden, New Jersey 07036
                                                10. Proiect/Task/U'ork Unit No.
                                                   Program Element
                                                   No. A1010
                                                11. Contract/Grant No.

                                                68-02-0219
 12. Sponsoring Organization Name and Address
       Office of Research and Monitoring
       U.S. Environmental Protection Agency
       Washington, D.C.   20A60
                                                13. Type of Report & Period
                                                   Covered „ .
                                                         Final
                                                30 Jun 71  - 31 Dec  72
                                                14.
 IS. Supplementary Notes
                  Presents  evaluation of  instruments selected as outlined  in Volume  I  -
       State  of the Art  (Publication No.   APTD-1180).
 16. Abstracts
      Three  types of commercially available  equipment  for  the  analysis  of H2S and
      other  odorous sulfides  have been evaluated for performance and  reliability at
      stack  emission levels of 0.1 to 100  ppm in air.  These included coulometers
      (3  models) flame photometric detectors plus gas  chromatography  (2  models)
      and  tape sensors with or without a preliminary gas 'converter (3 models).  None
      of  these instruments is capable of analyzing for H?S  in this range in the presence
      of  large amounts of S02,  C02,  CO, COS  and CS?, which  may characterize stack
      emissions from a Kraft  paper mill or a petroleum refinery Claus plant.   Reasons
      for  these failures are  examined, and  modifications  of presently available equip-
      ment are recommended for this purpose.
 17. Key Words and Document Analysis. 17a. Descriptors
       (air pollution,  measuring  instruments,  hydrogen sulfide)
       (field  tests,  performance,  reliability)
       (odors,  paper  pulp mills,  petroleum refinery)
      coulometers
      gas chromatography
      instrument  characteristics
 17b. Identifiers/Open-Ended Terms
      commercially  available instruments, tests  for  H2S/sulfides,  at  stack emission
      concentration
17e. COSATI Field/Group  14B:   Methods and  Equipment:  Test  Equipment
 18. Availability Statement

       Release unlimited
                                    19. Security Class (This
                                      Report)
                                    	UNCLASSIFIED
                                                         20. Security Class (This
                                                            Page
                                                         	UNCLASSIFIED
          21. No. of Pages
             159
                                                         22. Price
FORM NTIS-33 IREV. 3-721
                                                                               USCOMM-OC 14032-P72

-------