Study o
Mandatory Engine Maintenance
for Reducing Vehicle Exhaust Emissions
VOLUME Yl. A Comparison of Oxides of Nitrogen Measurements Made
With Chemi luminescent and Non-Dispersive Radiation
Analyzers
Year End Report
July 1972
In Support of:
APRAC Project Number CAPE-13-68
for
Coordinating Research Council. Inc.
Thirty Rockefeller Plaza
New York. New York 10020
TRW
SYSTIMS SHOUT
and
. Environmental Protection Agency
Air Pollution Control Office
5600 Fishers Lane
Rockville. Maryland 20852
SCOTT RESEARCH LABORATORIES, INC
P. O. BOX X4I«
AN BERNARDINO. CALIFORNIA B4O«
OH SP/ICE
CALIFORNIA
-------
A Study of
Mandatory Engine Maintenance
for Reducing Vehicle Exhaust Emissions
VOLUME 33 . A Comparison of Oxides of Nitrogen Measurements Made
With Chemi luminescent and Non-Dispersive Radiation
Analyzers
Year End Report
July 1972
In Support of:
APRAC Project Number CAPE-13-68
for
Coordinating Research Council. Inc.
Thirty Rockefeller Plaza
New York. New York 10020
TRW
SYITIUS OJHMJf
ONI SMCf M»* KCOONOO SftCM C»LlfO1"l» »?/«
and
Environmental Protection Agency
Air Pollution Control Office
5600 Fishers Lane
Rockville. Maryland 20852
SCOTT RESEARCH LABORATORIES, INC
P. O. »OK Ml*
AN CHNAHDINO. CAi.iroi.NlA »««o»
-------
PREFACE
This report, "A Study of Mandatory Engine Maintenance for Reducing
Vehicle Exhaust Emissions," consists of six volumes. The following are
the subtitles given for each volume:
t Executive Summary, Volume I
Mandatory Inspection/Maintenance Systems Study, Volume II
A Documentation Handbook for the Economic Effectiveness
Model, Volume III
Experimental Characterization of Vehicle Emissions and
Maintenance States, Volume IV
t Experimental Characterization of Service Organization
Maintenance Performance, Volume V
A Comparison of Oxides of Nitrogen Measurements Made With
Chemiluminescent and Non-Dispersive Radiation Analyzers,
Volume VI
The first volume summarizes the general objectives, approach and
results of the study. The second volume presents the results of the
mandatory inspection/maintenance system study conducted with a computer-
ized system model which is described in Volume III. The experimental
programs conducted to develop input data for the model are described in
Volume IV (Interim Report of 1971-72 Test Effort) and V. Volume VI
presents comparative measurements of NO and NO using chemiluminescence
/\
and NDIR/NDUV instruments and differences in these measurements are
examined.
The work presented herein is the product of a joint effort by TRW
Systems Group and its subcontractor, Scott Research Laboratories. TRW,
as the prime contractor, was responsible for overall program management,
experimental design, data management and analysis, and the economic
effectiveness study. Scott acquired and tested all of the study vehicles,
Scott also provided technical assistance in selecting emission test
procedures and in evaluating the test results.
-------
TABLE OF CONTENTS
Page
INTRODUCTION 1
TEST VEHICLES AND PROCEDURES 1
EXPERIMENTAL APPARATUS AND TECHNIQUES 2
EXPERIMENTAL RESULTS - PARALLEL CONFIGURATION 4
EXPERIMENTAL RESULTS - SPLIT CONFIGURATION 6
CONCLUSIONS 26
ii
-------
LIST OF FIGURES
Figure Title Page
1 System Schematics (Parallel & Split) 3
2 Parallel Configuration (CUB Vs. CL#A) 5
3 Parallel Configuration (NDIR/NDUV Vs. CL#A) 7
4 NDIR/NDUV Vs. CLA (NOX) 8
5 CLB (NOX) Vs. CLA (NOX) 10
6 CLB (NO) Vs. CLA (NO) 11
7 Removal Efficiency 14
8 Production Efficiency 16
9 NDIR (NO) Vs. CLA (NO) 17
10 NDIR (NO) Vs. CLB (NO) 19
11 NDIR (NO) Vs. CL#A (NO) & CL#B (NO) 20
12 Independent Evaluations of the Effect of 22
Drierite on NO Measurements
13 NDIR/NDUV (NOX) Vs. CLB (NOX) 23
14 Independent Evaluations of the Effects of 25
Drierite and Interference Gases on NOX
Measurements
iii
-------
A COMPARATIVE STUDY OF THE USE OF CHEMILUMINESCENT
AND NON-DISPERSIVE RADIATION ANALYZERS FOR THE
MEASUREMENT OF OXIDES OF NITROGEN IN DILUTE
SAMPLES OF AUTOMOTIVE EXHAUST
INTRODUCTION
At present, two distinctly different types of instrumentation are
widely used to measure automotive exhaust emissions of the oxides of
nitrogen (NO and N02). Until quite recently, virtually all such measure-
ments were made using non-dispersive infrared (NDIR) analyzers for NO
and non-dispersive ultraviolet (NDUV) analyzers for NO?- The use of
chemiluminescence (CL) analyzers has recently gained wide acceptance for
the measurement of both NO and N09 (NO ).
C- s\
Many investigators, however, have noted that significant differences
'exist in the data obtained with both of these measurement techniques,
although little discussion of these differences is found in the litera-
ture. This latter fact is attributable in part to both the wide accept-
ance of CL techniques and the specification by the Environmental Protec-
tion Agency that CL instrumentation be used as the measurement technique
for the Federal NOV exhaust emissions standards for 1973 and subsequent
X
model-year light duty vehicles. It is the purpose of this report to
describe an experimental study which was undertaken to describe differ-
ences between CL and NDIR/NDUV measurements, as applied to dilute samples
of automotive exhaust gas.
TEST VEHICLES AND PROCEDURES
The test vehicles that were employed for this experimental program
were part of the experimental fleet of cars being tested on the CAPE
13-68 Engine Parameter Deterioration Program. This fleet originally
consisted of 450 privately owned 1960 through 1971 model-year automobiles.
The fleet was composed of all domestic makes of vehicles, in addition to
some foreign makes. These automobiles were selected at random.
The experimental NO measurements were conducted concurrently with
A
the regular Engine Parameter Deterioration tests. These tests consisted
of measuring the automobile exhaust emissions in accordance with the 1975
Federal Test Procedure and the Federal Short Cycle test procedure, as
1
-------
well as during selected diagnostic driving modes. The NO experimental
A
investigation was based on the constant volume sampler (CVS) bag emis-
sions: the three samples obtained for the 1975 Federal Test Procedure
and the single bag resulting from the Federal Short Cycle test procedure.
The "Drierite" brand of dessicant was renewed at the beginning of each
vehicle test and was not changed during measurement of the four successive
bag samples.
EXPERIMENTAL APPARATUS AND TECHNIQUES
The NDIR instrument used for these experiments was a Beckman Model
315A Non-Dispersive Infrared Analyzer and NDUV measurements were made with
a Beckman Model 255 Non-Dispersive Ultraviolet Analyzer. Two chemilum-
inescence instruments were used, both Thermo Electron Model 10A NO-NO.
X
Analyzers. All sample handling lines were 1/4 inch O.D. Teflon, wrapped
with black (opaque) shielding to prevent any photochemical reactions from
occurring in the sample lines. Dilute exhaust samples were collected in
Mylar CVS bags, also opaque to ultraviolet and visible radiation. Sample
flowrates through each of the NDIR and NDUV instruments were maintained
at 10 cubic feet per hour (cfh) with a positive displacement pump placed
upstream (a "push" pump) of each instrument. The sample flowrate through
each of the CL analyzers was maintained at cfh with a pump downstream
("pull" pump) of each analyzer. It should be noted that a Drierite
dessicator was positioned upstream of the NDIR unit to remove water vapor,
whereas dessicant is not normally included in any of the sample lines
leading to the other instruments. This point will become very important
in the analysis of the results, since several effects may be attributed
to the presence of Drierite in the sample handling system of the NDIR
analyzer.
Two instrument configurations were used in this investigation, and
they are schematically diagrammed in Figure 1. In the parallel configura-
tion, all four instruments were placed in parallel with one another, and
the sample line lengths to the instrument cabinets were equal in order to
eliminate any differences which might result from reactions in the sample
line. This configuration was used to establish both the degree of agree-
ment between the two CL analyzers and the amount of difference between
-------
FIGURE 1
SYSTEM SCHEMATICS
PARALLEL CONFIGURATION-
CL i
1 B
SPLIT CONFIGURATION:
CL i
* A
-------
either CL measurement and those of the NDIR/NDUV instruments. These
results will be discussed in the following section.
It may also be seen in Figure 1 that the split configuration placed
one CL analyzer in parallel with the NDIR/NDUV cabinet (as in the parallel
configuration), and the other CL instrument was placed downstream in series
with the NDIR analyzer. It again should be noted that the sample is
passed through Drierite before entering the NDIR instrument. This con-
figuration was designed to serve four purposes. First, results from both
CL #A and the NDIR/NDUV cabinet could be compared with those of the pre-
vious experiments (using the parallel configuration) in order to ensure
that nothing had changed with time. Second, a comparison of NDIR data
with that obtained from CL #B in the NO mode would quantify differences
between NDIR and CL measurement techniques, with the Drierite effect
removed. Third, a comparison of CL data when both CL instruments were
operated in the NO analysis mode would quantify any Drierite effects on
NO concentrations. Fourth, a similar comparison of CL data when both were
operated in the NO mode would establish any Drierite effects on NO (on
X X
both NO, which would now be known, and on NOp). The results of this
series of experiments will be discussed in a later section.
EXPERIMENTAL RESULTS - PARALLEL CONFIGURATION
The experiments with the instrumentation in the parallel configura-
tion consisted of 51 dynamometer runs, each of which yielded 4 bags to be
analyzed. Thus 612 independent NO data points resulted from this series
X
of experiments. Measured NO concentrations ranged from approximately 20
X (
to 250 ppm.
In Figure 2 may be seen a plot of pairs of data from the two CL
analyzers. The line appearing in the plot is the 45° datum and it is
seen that the data fall very closely about this reference line. A linear
regression yielded a line obeying the equation
[NOX]C|_#B 0.0097 [NOx]a#A + 1.01867 ppm (1)
-------
FIGURE 2
PARALLEL CONFIGURATION
CL #B vs, CL #A
-------
This best fit of the data had a correlation coefficient of 0.99943, which
means that all but 0.11% of the data is explained by the above relation-
ship. The slope and intercept indicate that the best fit of the data lies
very close to a 45° line.
Figure 3 presents a plot of pairs of data from the NDIR/NDUV cabinet
and one of the CL instruments (the previous data indicate that virtually
either CL analyzer may have been selected for this comparison). Again the
line of unity slope is shown for reference and it is seen that the NDIR/
NDUV data are consistently higher than corresponding measurements from CL
#A and that this trend becomes more pronounced (diverges from the 45°
line) at higher NO concentrations. The statistical interpretation of
X
these data will be discussed in the next section. Reference to the paral-
lel configuration in the system schematics (Figure 1) will indicate that
the data differences appearing in Figure 3 may be possibly attributed to
both Drierite effects on the sample and interference gas effects which
differ between CL and NDIR/NDUV instrumentation.
EXPERIMENTAL RESULTS - SPLIT CONFIGURATION
The experiments with the instrumentation in the split configuration
consisted of 38 dynamometer runs, each of which yielded 4 bags to be
analyzed. Since both NO and NO determinations were made, a total of 912
X
independent data points resulted from this series of experiments. Measured
NO concentrations ranged from approximately 25 to 320 ppm and NO values
J\
were observed between approximately 20 and 250 ppm. It should be emphasized
at this point that all subsequent discussion will address the split con-
figuration only.
Figure 4 presents a plot of pairs of NOV data from the NDIR/NDUV
X
cabinet and CL #A, which were in parallel to one another. It should be
noted that there is a striking degree of similarity between these data
and analogous data obtained from the parallel configuration (Figure 3).
It may thus be concluded that there had not been a measurable change
between the two series of experiments. The equation which best describes
the data appearing in Figure 4 is given by
-------
FIGURE 3
PARALLEL CONFIGURATION
NDIR/NDUV vs, CL #A
-------
FIGURE 4
NDIR/NDUV (NOX) VS. CLA (NOX)
40,00 80.00
120.00
160.00 200.00 240.00
CLfl(NOX) , PPh
280.00
-------
[NOx]NDIR/NDUV = T'21890 tNOx]CL#A - 6.84340 ppm (2)
This best fit of the data yielded a correlation coefficient of 0.99181
(all but 1.63% of the data is explained by the above relationship). It
is again noted that the data lie above the 45° line and diverges at
higher NO concentrations.
A
In order to quantify the effects of Direrite on NO concentrations,
A
a comparison of CL #A results and those obtained from CL #B, when both
were operated in the NO mode (N09 converter in operation) appears in
X c.
Figure 5. A linear regression of the data was performed and the result-
ant best fit was found to obey the equation
[NOXJCL#B = 0.87264 [NOx]a#A + 6.59542 ppm (3)
This description of the data yielded a correlation coefficient of 0.98284
(96.60% of the data is explained by the above relationship). There are
several important observations which may be made at this point. In
Figure 1 it was seen that the two CL analyzers agree very closely to one
another on NO measurements when they are placed in parallel. Figure 5,
A
however, shows little agreement with Figure 1, and this difference may
be attributed to the presence of Drierite in the sample handling system
which precedes CL #B. Furthermore, it is seen from Figure 5 that the
two CL instruments diverge at higher concentrations, with the instrument
downstream of the Drierite (CL #B) yielding lower measured NO values than
A
CL #A. This was not found to be the case with the results from the NDIR/
NDUV instruments, (see Figure 3) which yielded increasingly higher NO
A
readings than CL #A at higher NO concentrations. Based on the high
A
degree of agreement between the two parallel CL instruments (Figure 1),
it may thus be concluded that the presence of Drierite results in lower
NO concentrations, although measured NO concentrations obtained by NDIR
A
result in higher measured NOV values for the NDIR/NDUV system.
A
Figure 6 presents a plot of NO data which helps to explain the above
observed NO differences. It is seen in this plot of CL #B against CL #A,
A
when both were operated in the NO mode (converter bypassed), that the
instrument downstream of the Drierite measures consistently higher values
-------
FIGURE 5
CLB (NOX) VS. CLA (NOX)
iHi'^Bi^H^^^^n
[ffll rinfr fl 1 (lifrrnT H nrlTniTTTnf^
:- r -T-:-"-.-^ I - - -- -Tt --;;;-:-
1-1- ::jH"i: - ::- -^ ::-:-f :;-;--
i--ii>j? _:: :Tl :::::: :: :.. :::[": . .:.:--:- :::.::: --± :.::: :i ": : ::
---------- - 4_ ------- -, I 4 ....... T - - - j- - - - -
fe::--l:V-3.::|:.-::'. E:E :A: ^ ^::: |i !:: :? . |EE :EE:! -EEEEEEt
Bj-'-jlE, [S: f f!:E_E:'?f:f|:4:E:E"k"rr::EEE iffi::EEE:EE
;;;:;:;;? :::::;::-:: :r.E-; ;EE:;"E;;; t;:;E"'E:| ?-E::.VE"EEE:LEEE:EEE
|;:EM;;;;y:E;:^:^;i
:|;;EEE; E|EE|-:::;EE; ;;;:::f::;£: E|E;E;E: ::|;;E::;;EE| EE:;|;E:;"
_..._[..: s _. "Tl . - - - - --J- ----- .!£ ._. _.-.!.. . .-
- +- - - 1 ; ; - P - - - - - - - - - - __...-......
; . T"" T---" ' -- " ".I. . : """: " _: .". : i J:: :::: ::T : ::.::::;: ::::. r::;. ..".::.. :~t:
' " f ' ' £ * " -i " i "" ---- - t_ - - - - - ------- - - -
::?':'r?: :,.:::: ::'.:: '..i ::-::: ::::::::: :":::;:::.:::.:: ::.:::::-:::::-::.
; = i - , - - - - - -
"" . . - -
. . B - I ? - . - - , _ _ _ _ . . _..__._
-- --| - h---
" ' ' " ~ _..-.-.- . . . - -J-. .... . . _ . . _ , . . . . .
" " ~ _......- ..... - _ .. . . ..._... ^ - -f
^^^^^^^^^^^^^L ^|j||5j.|ij mxaflSp f -f-FfTf FFf ff
'--''1^i-' -' " -J~ - . .-5j.. .._^_...
;':; ' J - ''i-jj
. . ' !^ !.:..:.::::::.
--T ' ---' -- >[ | 'Sj - T-
' ,
, , " ' ..--. . ,,.,--
^'.00 40-00 80.00 120.0
^^BHII^B^^^^BH
-E^illEliEiiEi-iiT-Ei-EEJI-EEiiiEEiiETE-iliiEEEiiEEEiiiiiiE-EE '- : : :
------ '- . - . - - - - - T - - ._.... 0. J. , _
---- --- --- }- - -1- --. -T- - -
-- f- £- - ^T -
I T--- X --' " "
.:::::::;: ::.:-: ::::-:::;::::: IE: ::T l.r ... .. I . .. r\ 4' : : -.
- - - - ----- - - - - T - - - -
:-:::::::^::i:::::r :±:::lf "I:::|::::-:":i:r ~::::i::r !:::::: : "::
^ . ' " ~ 1 1 .1 - ' L '. ~ ~. ~. . '1 .'. ~ ~. . '. II '- '. '. ~ ' - '. '.'''. '~ '. . . - " ~ " Z ," . ' ' ~ ~ ~ - '.'. - '.. . '.'''.'..'I'. i I I I
: ; : ~ : : : - :.:":::.::::. i j± .: : : .1 -:.: + :":.: : : : : : ' : J. -u : : : : . t : : : : : :
. .
- - - -- J - - - - -T }- - - - - - - J .
'. '. ~ ~ '. '. _ 1 \ ~ '. ' '. .' " ". I " " " ~~ ' _ . ~~ '. _ ~ T ~ I ","-"'. '. . '. - - 11 1 ". ^ - . - '. - - - . - - . - - - - - -
.::::::::: ± ::::: :.:I ::.:::.. I:::::,: ;;-: + :;.:-:.::::.-:::::.: : . :
- - T S - - - * - t - - - - - - -t - _ ;
EiEE|EEE:;:E;E;;EEE:;"ES:-:;EEEE:EEi:'EEE::;LEi|EEEEE" EE;:;;E; :
- - T ----- i! --?'- - -t - - - -
::|j:j|i: :::::::::; .:.::::::.:::::::::::::: :: ::
^^^^^HPM^^^fi^^^K^H
......_ -. .: - :;;::-. .;.:-:;-.:-.--.:. .I.:-::.: ::;-::.::::::..::: :.. :_ .
0 160.00 200.00 240.00 280.00
CLfl(NOX) , PPM
10
-------
FIGURE 6
CLB (NO) VS. CLA (NO)
.00
40.00
80.00
120.00
11
160.00 200.00 240.QC
CLR(NO) , PPM
-------
of NO than does the instrument which does not include Drierite in its
sample handling system. The best fit of the data was found to obey the
equation
[NO]a#B = 1.04808 [NO]CL#A + 3.97905 ppm (4)
with a correlation coefficient of 0.99315 (all but 1.36% of the data is
explained by the above relationship).
A comparison of Figure 5 and 6 allows several important conclusions
to be drawn about the chemical effects of Drierite on dilute samples of
automotive exhaust gases. Figure 6 indicates that the presence of Dri-
erite in the sample handling system causes NO concentrations to increase.
Figure 5 indicates that Drierite causes NOV concentrations to decrease.
X
Since NO (NO and N09) decreases while NO has been seen to increase due
X ^
to the presence of Drierite, the conclusion is that Drierite removes N02
from the sample gas and partially converts some N02 to NO. This may be
explained by the following (conceptual) reaction on Drierite.
Dri eri te -,
N02 - » NO + ^ 02 (5)
That this reaction, or one similar to it, does not go to completion in all
cases may be seen by Figure 5. If Drierite converted all N02 to NO, or
converted all N09 that it removed into NO, the NO data would have fallen
c. X
along the line of unity slope (as was the case of the data in Figure 2).
It is appropriate at this time to reexamine the data in order to
quantify the efficiencies of Drierite in removing N02 and producing NO
in samples of dilute automotive exhaust gas. The efficiency of Drierite
in removing N02 may be expressed as
N02 removed by passage through Drierite
nN02 removal = No2 present before passage through Drierite
[NOX]CL#A- [NO]CL#A-
- LNOJa#A
12
-------
Since this psuedo-equilibrium constant could be a function of the NC^ con-
centration in the sample, the data was reexamined to yield a plot of nNC^
removal as a function of initial NCL concentration, as seen in Figure 7.
The solid curve drawn through the data in Figure 7 was found to best
describe the relationship between the observed NCL removal efficiency by
Drierite and the amount of NCL present before passage through Drierite and
obeys the equation
nN02 removal = 2.3994 [NCL]'0'3420 (7)
available
This decrease in removal efficiency with increasing concentrations of in-
coming N02 may be qualitatively interpreted as a saturation effect and
suggests the presence of active sites for reaction on the solid surface
of the Drierite material. Furthermore, close scrutiny of the analogous
curves for each successive bag sample analysis (three for the 1975 cycle
and one for an EPA short cycle) indicate that the NCL removal efficiency
of Drierite deteriorates with time; i.e., with the amount of exhaust
sample which has been passed through it. This might well be a direct
function of the amount of water vapor which has been absorbed. The fact
that observed Drierite NCL removal efficiencies often have values which
are greater than unity at the lower available NCL concentrations is at
this time unexplained. It is believed, however, that this results from
instrument errors which, although small in absolute value, are large com-
pared to NCL differences of only a few ppm.
In a similar manner, the efficiency of Drierite in converting NCL to
NO may be expressed as
Mn n nA,, +-; - NO produced by Drierite
nNO production - ^removed by Drierite
[NO]CL#B -
LNOXJCL#A - CL#A - tNox]CL#B + FNOTCL#B
[NO]CUB - [NO]a#A
nN02 removed ([NOX]CL#A - [NO]C|_#A)
13
-------
4.QO
-00 12,00 16.00 20.00 24-00 28-00
32-00 36.00 40-00 44.00 48-00
CLFKNOX 1-CLFKNO J , PPM
52.00 56.0'Q 60.00 64.00
-------
Since this efficiency could be dependent on the amount of NO- removed by
Drierite (the amount of NCL available for conversion to NO), the above
efficiency was plotted against the amount of NCL which was removed, as
seen in Figure 8.
These data are seen to have considerably less scatter than those of
Figure 7 and the solid curve was found to best describe these data accord-
ing to the equation
nNO production = 3.7888 [NOp]"0'7295 (9)
removed by Drierite
Again a possible Drierite saturation effect is observed, and this hypo-
thesis is also supported upon examining the data from each of the four
sample bags. Again efficiencies are seen to exceed unity at lower values
of the x-axis and this phenomena is believed to be a function of instru-
ment variability, but is otherwise unexplained. The important factor in
these efficiency analyses, however, is that both values are clearly not
zero, and that the hypothesized reaction scheme is effective in explaining
the observed effects on NO and N02 concentrations.
Now that the effect of Drierite on the chemical composition of dilute
automotive exhaust gas samples has been described, it is appropriate to
investigate the different response characteristics of NDIR, NDUV and CL
instrumentation. It has been well established that NDIR/NDUV instruments
respond strongly to water vapor and that the subsequent inclusion of Dri-
erite (and other dessicants) in the sample handling system has effectively
eliminated this problem. Also, it has been long known that NDIR/NDUV
systems respond positively to several interference gases that are also
found in automotive exhaust, such as propane, carbon dioxide and carbon
monoxide.
In order to investigate the magnitude of these interference gas
responses, data from the split configuration were again scrutinized.
Figure 9 is a NO plot of pairs of data from the NDIR and CL#A when it was
operated in the NO mode. As was the case for the analogous NO curve
X
(Figure 4), it is seen that the NDIR instrument yields higher measured
values than the CL instrument in parallel with it and that this tendency
15
-------
§
a
I -n
ii i«
O CT5
en f ; ; f:
oo
o
o^ao
a.GO
12-00
16.00 20.00 24,00
28.00 32.00 -36.00 40,00 44.00 48.00 52.00
CLFKNGXl-CLFHNO J-CLBC NOX J+CLBC NO) , PPH
56.00 SQ..OO
-------
FIGURE 9
NDIR (NO) VS. CLA (NO)
40.00
80.00
120.00
160.00 200.00 240.00
CLfl(NO). PPM
-------
becomes more pronounced at higher NO concentrations. A linear regression
of these data yielded a best-fit equation given by
[NO]NDIR = 1.12184 [NO]a#A + 8.36667 ppm (10)
with a correlation coefficient of 0.98788 (all but 2.41% of the data is
explained by the above relationship). The difference between the two
instruments exhibited by this plot is interpreted to result from both
the effects of interference gases and Drierite since there was no dessi-
cant upstream of CUA.
In order to remove all effects of Drierite, a similar plot of pairs
of NO readings from the NDIR and CUB is seen in Figure 10. Since both
instruments were downstream of the Drierite, the differences may be
attributed to interference gas effects. A linear regression resulted in
a best fit of the data given by
[NO]NDIR = 1.06970 [NO]CL#B + 4.17953 ppm (11)
with a regression coefficient of 0.99405 (explaining all but 1.19% of the
data). It is seen that these data lie below those of Figure 9, given by
Equation (10), which is to be expected since it has been shown that Dri-
erite increases NO concentrations. This is summarized by Figure 11.
Since Drierite effects are additive to any interference gas effects,
the difference between the regressions of Figure 9 and 10, when referenced
again to the line of unity slope, would yield the expression for Drierite
effects only (here it is assumed that the two CL instruments respond
identically, as was illustrated by Figure 2). Thus, for Drierite effects
only, Equations (10) and (11) yield
[NO]NDIR = 1.05214 [NO]CL + 4.18715 ppm (12)
This expression is seen to be extremely similar when compared to
Equation (4), which resulted from pairs of CL#A and CUB NO data (see
Figure 6):
18
-------
FIGURE 10
NDIR (NO) VS. CLB (NO)
.00
40.00
80.00
120.00 160.00 200.00 240
CLB(NO ) , PPM
.00
19
-------
FIGURE 11
NDIR (NO) VS. CL#A (NO) AND CL#B (NO)
EFFECTS OTHER THAN DRIERITE
EFFECTS OF DRIERITE
0
80 120 160
CL(NO), ppm
200
20
-------
[NO]CL#B = 1.04808 [NO]C|_#A + 3.97905 ppm (4)
For convenience of further comparison, these two best-fit equations are
replotted in Figure 12. It is seen that the two regression lines lie
extremely close to one another. It is thus concluded that the effects of
Direrite on NO are accurately measurable by either NDIR or CL techniques,
once the effects of interference gases on the NDIR have been eliminated.
In order to investigate whether similar conclusions can be drawn for
NO response, let us reexamine Figures 4 and 5. Figure 4 presented NO
X X
data for NDIR/NDUV plotted against corresponding data from CL#A which was
in parallel. Since no Drierite was upstream of CUA, Figure 4 represents
differences which result from all effects (both those of interference gas
responses and those due to the presence of Drierite). Figure 5 presented
NO data for CUB (downstream of Drierite) plotted against corresponding
X
data from CUA. Thus, this figure represents the effects of Drierite on
NO concentrations. It is noted that the presence of Drierite results in
X
lower NO values, i.e., negative effect on measured NO concentrations.
X X
Taking the difference between the regression lines of Figures 4 and 5,
given by Equations (2) and (3), the resultant equation, when referenced to
the original 45° line, is given by
[NOxJNDIR/NDUV = ]-34626 [NOx]CL ' 13-43882 PPm
This equation should represent only the effects of interference gas and
the Drierite depletion of N02 ("counting N02 twice") on NDIR/NDUV measure
ment of NO , with the effect of Drierite increasing NO concentrations
X
effectively cancelled out.
This result may be compared directly with Figure 13, which presents
a plot of corresponding pairs of NOV data from the NDIR/NDUV and CUB.
J\
Since CUB was downstream in series with the NDIR (and therefore the
Drierite), Figure 13 directly represents all effects other than those of
Drierite on the NDIR measurement. Due to the nonlinear nature of the
data, the best-fit was found to be given by
21
-------
FIGURE 12
INDEPENDENT EVALUATIONS OF THE EFFECT OF DRIERITE ON NO MEASUREMENTS
280 r-
240
200
0-
uT
h^
C£
UJ
5
Q
u_
o
to
I
U
160
IE 120
O
z
80
40
DIFFERENCE BETWEEN*
FIGURES 9 AND 10
(NDIRANDCL)
FIGURE 6 (CL)
0
40
80 120 160 200
NO WITHOUT EFFECTS OF DRIERITE, ppm
240
22
-------
FIGURE 13
NDIR/NDUV (NOX) VS. CLB (NOX)
_ '
H
t
40vOQ 80.00 120.00 160-00 200.00 240.00 280.00
CLB(NOX), PPh
23
-------
with a correlation coefficient of 0.99465 (explaining all but 1.07% of the
data). In order to facilitate comparison of Equations (13) and (14), Figure
14 presents a plot of both equations. It is seen that the agreement between
the two approaches is quite strong, especially at higher NO concentration,
X
and it is thus concluded that interference gas effects on NO response by
A
NDIR/NDUV instrumentation, as well as the effects of Drierite, can be
quite accurately measured by either NDIR/NDUV or CL techniques. A more
detailed discussion of this result will be found in the final section of
this report.
As the last step of this experimental investigation, an attempt was
made to qualitatively explain the observed effects of interference gases
on NO measurements by NDIR. It is well known that NDIR instrumentation
responds positively to several interference gases, such as carbon dioxide,
carbon monoxide and propane (all of which are found in automotive exhaust
samples). With flame ionization detectors, as normally used in CVS test-
ing, it is impossible to conveniently determine the concentration of any
one species of hydrocarbons, since many are found in typical exhaust
samples and are analyzed as a group on the basis of carbon concentrations.
A gas chromatograph was not available during these experiments and no
quantitative attempt was made to investigate the NDIR response to typical
(but unknown) concentrations of propane in dilute samples of automotive
exhaust gases.
It was possible, however, to qualitatively examine the effects of
typical (measured) concentrations of carbon dioxide and carbon monoxide on
the NO NDIR response characteristics. A series of experiments was first
conducted to quantify the NDIR response to C0~ concentrations ranging from
1.25 to 2.50% in nitrogen. It was found that the NDIR responded to these
gases by recording as if 3 to 6 ppm NO had been introduced. Similarly, it
was qualitatively determined that the response by the NDIR to CO was about
40% as strong as that for CO^. These data cannot be quantified (and are
therefore not reported), since the signal to noise ratio observed during
these experiments was too low to reliably examine in detail. It is
24
-------
FIGURE 14
INDEPENDENT EVALUATIONS OF THE EFFECTS OF DRIERITE AND
INTERFERENCE GASES ON NOX MEASUREMENTS
320 r-
O
z
Q_
.Q.
Q
Q
Z
DIFFERENCE BETWEEN
FIGURES 4 AND 5
80 120
CL, ppm NO
25
-------
evident, however, that the NDIR responds positively to several gases which
are present in samples of automotive exhaust. Throughout this interference
gas response investigation, no CL response was observed. This observed
property of the NDIR to respond to interference gases helps in large part
to explain differences, between its measured values and those of the CL
analyzer, which are not directly attributable to the effects of Drierite.
CONCLUSIONS
Based upon the results of this investigation, several conclusions can
be drawn. First, using two independent CL instruments placed in parallel
with one another, it has been shown that NO measurements are with high
X
precision using identical CL instruments. Also, the results indicate that
NO values measured by NDIR/NDUV techniques (with Drierite placed in the
A
sample handling system of the NDIR) are significantly larger than those of
correspbnding CL determinations. This effect becomes more pronounced at
higher NO concentrations.
A
It has been shown that higher NDIR/NDUV readings for NO , relative to
A
measured CL values, result from both the effect of Drierite in the NDIR
sample handling system and the response of the NDIR to interference gases
found in dilute samples of automotive exhaust gas. The action of Drierite
on the sample gas is such that N02 concentrations are decreased, and NO
concentrations are increased. This conversion of N02 to NO is only a
partial one, and thus a net decrease in NO concentration results. Since,
A
in the NDIR/NDUV cabinet, Drierite is present upstream of the NDIR only,
higher NO readings result than would be the case if no Drierite was pre-
sent. In effect, then, N02 is measured independently by NDUV analysis, and
most N02 is then removed and partially converted to NO before a measurement
of NO by NDIR analysis is made. This is the primary reason that NDIR/NDUV
instrumentation consistently yields higher measured NO values than those
A
of CL instruments.
The remaining difference between NDIR/NDUV and CL determinations of
NO are explained in large part by the positive response characteristics
A
of the NDIR to interference gases which are present in automotive exhaust.
NDIR instruments are known to respond positively to carbon dioxide, carbon
26
-------
monoxide, and propane, and during this investigation it was found that the
response to CCL alone explains virtually all differences between NDIR/NDUV
and CL measurements of NO which are not directly attributable to the
/\
presence of Drierite.
27
------- |