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
------- |