FINAL REPORT
An Investigation of Remote Sensing Devices
for Chemical Characterization
of Motor Vehicle Exhaust
By
Julian W. Jones, C. Ted Ripberger, Niranjan Vescio*
U.S. Environmental Protection Agency
National Risk Management Research Laboratory
Air Pollution Prevention and Control Division
Research Triangle Park, NC 27711
(*) Currently with Remote Sensing Technologies, Inc., Tucson, AZ
Prepared for:
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
ABSTRACT
The Air Pollution Prevention and Control Division (APPCD) of EPA's National Risk
Management Research Laboratory, located in Research Triangle Park (RTF), NC, has conducted
a series of tests to (1) evaluate the accuracy and precision of two different remote sensing devices
(RSDs) for measuring carbon monoxide (CO), hydrocarbons (HCs), and nitric oxide (NO), and
(2) evaluate the capabilities of three RSDs for characterizing fleet emissions of NO. This report
summarizes the results of these tests and lays the groundwork for analysis efforts that will be the
basis for future articles.
"Puff tests were conducted in which bursts of simulated motor vehicle exhaust were
repeatedly measured by the RSDs. The accuracy and precision of data from these measurements,
which show a generally linear response over a range of concentrations, were in the order CO >
HC > NO. Subsequently, three vehicles were driven at constant speed on a dynamometer and on
a test track. The average emissions data measured by three RSDs at the track, when compared to
the dynamometer emissions data (which were used as the "standard"), in general, showed just the
opposite result; i.e., NO > HC >CO. However, the track data show a considerable amount of
variation. Whether or not this was caused by the conditions at the site (e.g., only a slight grade,
which made maintaining constant vehicle speed difficult) or some other variable is a question for
possible future research.
The three RSDs were also tested for several hours on a freeway ramp in southwest
Raleigh, NC. Considerable differences in readings for each individual vehicle by the three RSDs
were noted. However, when the range of all emissions measurements for each RSD was grouped
into intervals, the distributions of the data were very similar. This supports the conclusion that if
enough data are collected, they should adequately indicate fleet emissions at a particular location.
ACKNOWLEDGMENTS
We wish to thank Kenneth T. Knapp of EPA's National Exposure Research Laboratory for
providing the dynamometer facilities for this study as well as the technical staff to operate the
laboratory equipment — without their assistance the dynamometer component of this study could
not have been completed. We would also like to acknowledge the assistance of both the technical
staff of Hughes Santa Barbara Research Center (Santa Barabara, CA) and Remote Sensing
Technologies, Inc. (Tucson, AZ). The valued assistance of Wojciech Kozlowski of ARCADIS
Geraghty & Miller, Research Triangle Park, NC, was supported by EPA Contract 68-D4-0005,
Work Assignments 1-046, 2-045, 2-054, and 3-041.
-------
Table of Contents
Section Page
Abstract ii
Acknowledgments ii
List of Tables iv
List of Figures v
List of Acronyms v
Introduction 1
Background 1
Experimental 4
Summary 4
Experiment 1: RSD-1000 "Puff Tests 4
Experiment 2: Smog Dog™ 3 "Puff Tests 5
Experiment 3: "Puff Tests with Three RSDs 6
Experiment 4: Dynamometer Tests on Three Vehicles 6
Experiment 5: Track Tests with Three RSDs 7
Experiment 6: On-Road Testing with Three RSDs 8
Results 9
Experiment 1 9
Experiment 2 15
Experiment 3 21
Experiment 4 24
Experiment 5 26
Experiment 6 29
Conclusions 33
References 34
Appendix A: Instrumented Vehicle Track Tests Al
in
-------
LIST OF TABLES
Tables
Page
1 Gas Mixtures Used to Calibrate and Puffin the RSD-1000 Precision/Accuracy Tests .... 5
2 Gas Mixtures Used to Calibrate and Puffin the Smog Dog™ 3 Precision/Accuracy Tests . 5
3 Gas Mixtures Used to Calibrate and Puffin the RSD NO Precision/Accuracy Tests 6
4 Gas Mixtures Used to Calibrate the RSDs in Experiments 5 and 6 8
5a Puff Test Results for the RSD-1000 Calibrated with Gas Mix 1 10
5b Puff Test Results for the RSD-1000 Calibrated with Gas Mix 3 12
5c Puff Test Results for the RSD-1000 Calibrated with Gas Mix 4 14
6a Puff Test Results for the Smog Dog™ 3 Calibrated with Gas Mix 1 16
6b Puff Test Results for the Smog Dog™ 3 Calibrated with Gas Mix 3 18
6c Puff Test Results for the Smog Dog™ 3 Calibrated with Gas Mix 4 20
7a RSD Puff Measurements - CO/CO2 Ratios 21
7b RSD Puff Measurements - HC/CO2 Ratios 22
7c RSD Puff Measurements - NO/CO2 Ratios 24
8 Summary of Second-by-Second Emissions During Dynamometer Testing 25
9a RSD Measurements at the Track - CO/CO2 Ratios 27
9b RSD Measurements at the Track - HC/CO2 Ratios 28
9c RSD Measurements at the Track - NO/CO2 Ratios 28
Al Instrumented Vehicle (IV) Runs Performed on Test Track Al
A2 RSD-1000 vs. Instrumented Vehicle in Track Tests A2
IV
-------
LIST OF FIGURES
Figi
1
9
3
4
5a
tires
Components of the remote sensing device system
Path length dynamics
Plotting the relative concentrations provides the ratio
Relative location of RSDs at NC Highway Patrol's test track in Raleigh, NC . .
Distribution of CO/CO2 emissions data for motor vehicles
measured bv RSDs at freewav entrance in Raleish. NC
Page
1
3
3
7
30
5b Distribution of HC/CO2 emissions data for motor vehicles
measured by RSDs at freeway entrance in Raleigh, NC 31
5c Distribution of NO/CO2 emissions data for motor vehicles
measured by RSDs at freeway entrance in Raleigh, NC 32
LIST OF ACRONYMS
APPCD Air Pollution Prevention and Control Division
FEAT Fuel Efficiency Automobile Test
HCs Hydrocarbons
I/M Inspection and Maintenance
IR Infrared
NCDEHNR North Carolina Department of the Environment, Health, and Natural
Resources
RSD Remote Sensing Device
RTF Research Triangle Park (North Carolina)
RSTi Remote Sensing Technologies, Inc.
UV Ultraviolet
-------
INTRODUCTION
Remote sensing of automobile emissions is a technique developed in the Iatel980s. The
remote sensing device (RSD) uses infrared (IR) and, in some cases, ultraviolet (UV) spectroscopy
to measure the concentrations of pollutants in exhaust emissions as the vehicle passes a sensor on
the roadway. RSDs have been used to develop a profile of the emission characteristics of the
overall fleet of motor vehicles in metropolitan areas and/or to identify those vehicles known as
"super emitters," which are responsible for much of automotive emissions.
The Air Pollution Prevention and Control Division (APPCD) of EPA's National Risk
Management Research Laboratory, located in Research Triangle Park (RTF), NC, has conducted a
series of tests to (1) evaluate the accuracy and precision of two different RSDs for measuring
carbon monoxide (CO), hydrocarbons (HCs), and nitric oxide (NO), and (2) evaluate the
capabilities of RSDs for characterizing fleet emissions of NO. NO measurements are based on a
newer technology than that for measuring CO and HCs.
This report summarizes the results of these tests and lays the groundwork for analysis
efforts that will be the basis for future articles.
BACKGROUND
Remote sensing enables the exhaust emissions of a motor vehicle to be measured as the
vehicle passes by on the road. With this technique, which was developed in the late 1980s, IR
spectroscopy, or a combination of IR and UV spectroscopies, is used to measure concentrations of
carbon dioxide (CO2), CO, HCs, and NO. A schematic of a typical RSD system appears in Figure
1. In addition to the source and detector, remote sensors may be equipped with meteorological
stations and speed/acceleration systems which are important in interpreting exhaust measurements
by the RSD.
EMISSIONS
DETECTOR
, CALIBRATION
GAS
IR/UV SOURCE
Figure 1. Components of the remote sensing device system.
1
-------
The RSD technology is designed to measure CO, HC, and NO exhaust emissions of many
thousands of vehicles per day and provides a practical approach for routinely characterizing on-road
vehicle emissions. As such, remote sensing has several potential uses: determining fleet average
emissions for inventory purposes,1'2 characterizing fleet emissions distributions to evaluate vehicle
inspection and maintenance (I/M) programs,3 and to compare with other fleets for bench marking
purposes.4 EPA has issued detailed guidance on the use of RSDs in I/M programs.5'6
Previous remote sensing studies have indicated that most of the measured on-road emissions
(over 50%) come from a disproportionately small percentage of the vehicles (approximately
10%).7'8 This has been shown to be true for CO, HCs, and most recently for NO.9 Since the
remote sensing signal can be integrated with a video image of the license plate of the passing
vehicle, RSDs can also be used to relate emissions data to the characteristics of specific classes of
vehicles and, in some cases, identify high emitters.10 With these potential applications, several states
are considering adopting remote sensing as a supplement to their air quality improvement
programs/7
Note that steady-state conditions do not represent vehicle emissions under transient
operation where the highest emissions occur (e.g., during cold engine starts or rapid accelerations
resulting in fuel enrichment). Thus, RSD results may in some cases over- or under-predict
emissions. Despite these potential errors, RSDs, properly used, are very useful tools.
Calibration of the instrument and the calculation of exhaust constituent concentrations will
be discussed briefly. The RSD functions much like a bench top spectrophotometer except that it
has been adapted for the roadway. Each of the principal species of vehicle exhaust can be identified
by its characteristic IR or UV spectrum. For example, in the RSD-1000 unit produced by Remote
Sensing Technologies, Inc. (RSTi), absorption peaks produced by CO, HCs, and CO2 in the exhaust
plume are isolated by separate detectors focused at 4.6, 3.4, and 4.3 jim in the IR region,
respectively. Donald Stedman's group at the University of Denver determined that using the
absorption peak at 227 nm in the UV spectral region would avoid the interference from exhaust
water vapor in the IR spectrum.7 RSTi selected this approach for their NO channel. A reference
table of measured absorption intensity at various constituent concentrations is created which enables
the concentrations of constituents in vehicle exhaust plumes to be determined. If the width of the
exhaust plume in the sensor beam (i.e., the path length) were known, concentrations of sample
constituents could be calculated directly. But the path length of the plume is not known.
Furthermore, the path length changes as the plume disperses and dilutes (Figure 2). By using the
reference table values without correcting for the plume path length dynamics, incorrect
concentrations would be produced because we would include the path length of the sample chamber
in the calculations.
The inability to determine the path length of exhaust plumes is addressed by measuring the
relative plume concentrations of exhaust constituents, not the absolute plume concentrations.
Although the path length of the plume is unknown, it is assumed to be the same for all detectors at
any given instant in time. Since the relative concentration at each instant in time is a function of
concentration and path length, by dividing the concentration of one gas by another at each instant in
time, the path length is removed from the equation. The slope of the relative concentrations as the
-------
plume disperses and dilutes
provides the ratio of two gases
(Figure 3).
The RSD determines
the concentrations of CO,
HCs, and NO in the tailpipes
of passing vehicles by
measuring the ratios of these
gases to CO2 in the exhaust
plumes and by employing
knowledge of the
stoichiometry of gasoline
combustion. For example, for
the RSD-1000, the effective
tailpipe concentrations of CO,
HCs, NO, and CO2 are
computed through the
combustion equations by
assuming an empirical fuel formula and an ideal air/fuel ratio.12'13 Since ambient CO2 levels can be
significant (350 ppmv or 0.035%)14, a field calibration is conducted using a known gas mixture to
remove the effects of ambient CO2. During field calibration, a table of absorption versus
concentration is constructed using CO as the reference in this case since ambient CO levels are low
(0.1 ppmv).15 Absorption and concentration are
mathematically related by a polynomial equation
typically of third or fourth order. New coefficients
are derived for the terms in each constituent's
polynomial equation, with the exception of CO,
each time the RSD-1000 is field calibrated. The
equation for CO remains unchanged since CO is
used as the reference.
Source
ii i
ii i
1 1 1 Time=3
1"
Sample
chamber
111 Time=2^^i
1 1 1 Thm=L—r — -^ '
" '^A ^
iiftl ,
II 1
•
•
MTime=4
LI Changing
i path length
w
1 ' ' Exhaust
Ml nlnvMA
plume
^
\ 1 ^
1
Detector
Car Direction
Figure 2. Path length dynamics.
|
1
o
O
O
o
Slope = CO/CO2 ratio^i
• /
• /*
/•
B /•
9*
•/•
7
CO2 Concentration
Because the shape and contents of an
exhaust plume from a motor vehicle are affected by
a number of variables [e.g., the rate of acceleration,
the speed, the location of the tailpipe (rear vs. side),
the condition of the engine and its emission control
system], and because remote sensing is a relatively
new technology, there are concerns about the
accuracy and precision of RSD measurements. For
this reason, a series of laboratory and field
experiments were conducted to evaluate the performance of several RSDs. These experiments are
described below, after which summaries of the experimental data are presented. Note that detailed
analyses of the data are not included in this report; these will be the objectives of subsequent efforts.
Figure 3. Plotting the relative concentrations
provides the ratio.
-------
EXPERIMENTAL
Summary ~ A series of six experiments using RSDs was performed:
(1) "Puff tests in which simulated motor vehicle exhaust was repeatedly measured by
an RSTi RSD-1000 using five certified gases representing a range of concentrations
of CO2, CO, propane (C3H8), and NO in a nitrogen matrix;*
(2) Puff tests in which simulated motor vehicle exhaust was repeatedly measured by a
Hughes Environmental Systems, Inc. Smog Dog™ 3 using five certified gases in a
manner similar to the tests with the RSD-1000;
(3) Puff tests in which simulated motor vehicle exhaust was repeatedly measured by
three different RSDs [the RSD-1000 and Smog Dog™ 3 devices used previously, as
well as a University of Denver Fuel Efficiency Automobile Test (FEAT) #3008
device] using three certified gases, each containing, in a nitrogen matrix, the same
percentages of CO2, CO, and C3H8, but a different percentage of NO;
(4) Dynamometer tests in which the actual exhaust of each of three different vehicles (a
Dodge Minivan, Ford F-150 Pickup Truck, and Ford Mustang II) was measured by
laboratory equipment while the vehicles were "driven" at constant speed;
(5) Track tests in which the actual exhaust plumes of the three vehicles tested on the
dynamometer were repeatedly measured by three different RSDs (used previously in
puff tests) as the vehicles passed the devices at a constant speed; and
(6) Road tests in which the actual exhaust plumes of vehicles were measured by the
three different RSDs (used previously in puff tests) as the vehicles passed the devices
at a freeway entrance ramp.
* These puff tests were supplemented by track tests in which the actual exhaust plume of an instrumented vehicle (IV)
was repeatedly measured by the RSD-1000 as the IV passed the device. See Appendix A.
Experiment 1: RSD-1000 "Puff" Tests
This experiment involved repeated measurements of five known mixtures of simulated (dry)
motor vehicle exhaust. Each of the mixtures was puffed into the path of the sensor 100 times with
the system calibrated using one of three available calibration gas mixtures (only those mixtures
containing NO could be used to fully calibrate the instrument). Table 1 lists the gases that were
puffed, which include the three used to calibrate the sensor. Gas components at these relative levels
represent tailpipe concentrations ranging from 1 to 9% CO, 300 to 4100 ppm HCs as propane, and
1500 to 3600 ppm NO. Gases used in the experiment were Scott Master Certified® (±2%) gases.
Because the calibration procedure is a possible source of error, the experiment was designed
so that separate variance components for calibration error and gas puff error could be estimated.
The 100 puff measurements were obtained in four groups of 25, each group after a fresh calibration.
-------
Table 1. Gas Mixtures Used to Calibrate and Puffin the RSD-1000 Precis! on/Accuracy Tests
Gas Mixture
1 (Calibration 1)
2
3 (Calibration 3)
4 (Calibration 2)
5
CO, %
3.05
4.00
6.44
6.03
1.00
HC, ppm
3940
1200
6350
2795
300
NO,
ppm
3590
0
1500
1997.5
0
C02, %
12.90
12.00
10.54
5.98
14.00
C0/C02
ratio
0.236
0.333
0.611
1.008
0.071
HC/CO2
ratio
0.030
0.010
0.060
0.046
0.002
N0/C02
ratio
0.028
0
0.014
0.033
0
The objective of this experiment was not only to determine the accuracy and precision of the
RSD-1000 over a typical range of operation (0 - 1.0) for CO/CO2 ratio, (0 - 0.06) for HC/CO2
ratio, (0 -0.04) for NO/CO2 ratio, but to determine whether the instrument's response was linear.
This was accomplished by calibrating with gas mixtures that spanned the observable range of
automobile exhaust. The concentrations of components in each of the mixtures, except gas #4,
were selected using the dry gas combustion equation and, therefore, reflect proportions observed in
dry tailpipe exhaust. Gas #4 is the manufacturer's recommended calibration gas.
Experiment 2: Smog Dog™ 3 "Puff Tests
This experiment involved repeated measurements of five known mixtures of simulated (dry)
motor vehicle exhaust, and was, with a few exceptions, identical to experiment 1 with the RSD-
1000 device. Again, each of the mixtures was puffed into the path of the sensor 100 times with the
system calibrated using one of three available calibration gas mixtures. Table 2 lists the gases that
were puffed, which include the three used to calibrate the sensor. Gas #4 is the manufacturer's
recommended calibration gas. Note that gas #1 has almost the same composition as gas #1 in
experiment 1, while gases #2, 3, and 5 are identical to those used in that experiment with the RSD-
1000 device.
Table 2. Gas Mixtures Used to Calibrate and Puffin the Smog Dog 3 Precision/Accuracy Tests
Gas Mixture
1 (Calibration 1)
2
3 (Calibration 3)
4 (Calibration 2)
5
CO, %
3.01
4.00
6.44
15.11
1.00
HC,
ppm
4297
1200
6350
19900
300
NO,
ppm
4137
0
1500
3820
0
CO2, %
13.30
12.00
10.54
15.01
14.00
CO/CO2
ratio
0.226
0.333
0.611
1.007
0.071
HC/CO2
ratio
0.032
0.010
0.060
0.133
0.002
NO/CO2
ratio
0.031
0
0.014
0.025
0
-------
Experiment 3: "Puff" Tests with Three RSDs
This experiment, like experiments 1 and 2, involved repeated measurements of known
mixtures of simulated (dry) motor vehicle exhaust. However, this experiment was focused on
assessing the accuracy and precision of NO measurements, and involved three different RSDs. The
RSDs tested were the RSD-1000 and Smog Dog™ 3 devices used previously, as well as a
University of Denver FEAT #3008 device. Each of three gas mixtures was puffed into the path of
each sensor 100 times; each RSD system was calibrated using the same single calibration gas
mixture. Table 3 lists the gases that were puffed, which basically differed only in their
concentration of NO, and the gas used to calibrate the three different devices.
Table 3. Gas Mixtures Used to Calibrate and Puffin the RSD NO Precision/Accuracy Tests
Gas Mixture
1
2
3
4
(Calibration
Gas)
CO2, %
15.01
14.99
15.00
13.30
CO, %
0.300
0.300
0.299
3.01
HC, ppm
500.0
502.6
502.0
4297
NO, ppm
500.
999.
1496.
4137
NO/CO2 ratio
0.0033
0.0067
0.0100
0.0311
Note: Prior to the testing, the RSD manufacturers all agreed that gas mixture # 4, which was readily available, could be used for
calibrating their instruments in these tests. Although somewhat different from what they would recommend for fleet
characterization testing under road conditions, the gas was similar enough to be adequate for these tests.
Experiment 4: Dynamometer Tests on Three Vehicles
This experiment involved measurements by laboratory equipment of the actual exhaust of
each of the three vehicles used in experiment 5. Each vehicle was "driven" at constant speed (45
mph*) on a dynamometer at EPA's facilities in Research Triangle Park, NC. The results of this test
established the "standard" by which the track test emission measurements using the RSDs would be
evaluated.
Each vehicle was tested three times for 10 minutes (30 minutes total). Emissions from the
vehicles were diluted and analyzed on a second-by-second basis; "bag" samples were also taken and
analyzed to obtain a mean concentration value for the entire test. However, because some of the
NO in the bag samples was converted to NO2, the second-by-second data was adjusted and used as
the "standard." The adjustment used the "middle" 8 minutes of test data (the first and tenth minutes
of data were dropped to eliminate some of the effects of start-up and driver variability).
* Readers more familiar with the metric system may use the following factors to convert from nonmetric units in this report:
1 ft = 30.48 cm; 1 in. = 2.54 cm; 1 mi = 1.61 km; and 1 mph = 1.61 km/h.
-------
Experiment 5: Track Tests with Three RSDs
This experiment involved measurements of CO, HC, NO, and CO2 concentrations in the
automotive tailpipe exhaust of three different vehicles driven repeatedly past several different RSDs.
The RSDs used were the three devices tested in experiment 3, as well as a Smog Dog™ device
owned and operated by the North Carolina Department of Environment, Health, and Natural
Resources (NCDEHNR). Since the NCDEHNR's device did not measure NO, which, like
experiment 3, was emphasized in this testing, no results for that device are included here.
The tests were conducted on the NC Highway Patrol's 1.13-mile oval test track in Raleigh,
NC. To ensure that the vehicles passed under constant-applied throttle, the RSDs were positioned
near the crest of a long and gradual 0.3% uphill slope. The equipment vans which supported the
RSDs were positioned beside the track so that the beams which emanated from the devices were as
close together as possible (Figure 4). Specifically, the distance between the FEAT #3008 and Smog
Dog™ 3 was about 9 ft., 4 in. (the NCDEHNR device was located between them), while the
distance between the Smog Dog™ 3 and the RSD-1000 beams was about 3 ft., 7 in. In addition,
the IR/UV beams were approximately the same height (16-19 in.) above and parallel to the road
surface.
Traffic Flow >
Video Video Video Van A Van C
Camera A Camera B Camera C Van B
Note: A = FEAT #3008; B =Smog Dog™ 3; C = RSD-1000
Figure 4. Relative location of RSDs at NC Highway Patrol's test track in Raleigh, NC.
The vehicles made approximately 100 laps around the track (in four groups of about 25),
passing through all of the beams during each lap. The participating devices were calibrated with
puff gases prior to testing and were recalibrated after each group of 25 laps. With the exception of
the RSD-1000, the calibration puff gases used were those recommended by the manufacturers prior
to the testing described in experiment 3. The actual gases used are described in Table 4.
-------
Table 4. Gas Mixtures Used to Calibrate the RSDs in Experiments 5 and 6
Cal Gas for:
FEAT #3008
RSD-1000
Smog Dog™ 3
CO, %
6.01
3.01
15.01
HC,
ppm
2005
4297
19900
NO,
ppm
1003
4137
3820
CO2, %
6.02
13.01
15.10
CO/CO2
ratio
0.998
0.231
0.994
HC/CO2
ratio
0.0333
0.0330
0.132
NO/CO2
ratio
0.0167
0.0318
0.0253
Experiment 6: On-Road Testing with Three RSDs
This experiment involved measurements of CO, HC, NO, and CO2 concentrations in the
automotive tailpipe exhaust of moving vehicles on two freeway on-ramps in the RTF, NC, area.
The RSDs used were the same four devices used in experiment 4. Again, of particular interest were
the capabilities of the devices to accurately detect NO emissions, so the NCDEHNR data are not
included here. The tests were conducted at two on-ramps to Interstate Highway 40 (1-40), one in
southwest Raleigh (Wake County), the other in the southern part of Durham County.
Unfortunately, the computer on board the RSD-1000 van failed early in the testing at the Durham
County site, so all three devices can be compared only through data collected at the Raleigh site.
The equipment vans, which supported the RSDs, were positioned on the shoulder of the
road in a manner similar to that shown in Figure 4. However, the order of the RSDs was different,
as was the distance between the IR/UV beams. Specifically, the order was:
S Gorman St. Entrance to 1-40 W in Raleigh, NC -
FEAT #3008, RSD-1000, and Smog Dog™ 3
NC-147 N Entrance to 1-40 E in Durham County, NC -
Smog Dog™ 3, RSD-1000, and FEAT #3008
In this experiment, the beams were about 3 ft. apart. Again, the beams were approximately the
same height (16-19 in.) above and parallel to the road surface.
-------
RESULTS
Experiment 1
A summary of the RSD-1000 puff test results is shown in Tables 5a-5c. The Group number
designates a series of puffs that were initiated after the device was calibrated; N indicates the
number of puffs in each Group. The mean and standard deviation (Std Dev) are shown for each
series of puffs.
Gas Mix 1 as the Calibration Gas (Table 5a) — In this set of tests, the CO/CO2 group mean
values were best for the calibration gas. For higher concentrations of CO, the CO/CO2 group mean
values were underestimations of the expected values; e.g., those for Gas Mix 4 represented about a
15 % underestimation. For Gas Mix 5, which had a lower concentration of CO, the group mean
values overestimated the expected value of CO/CO2 by about 9 to 20 %. The precision of this
series of tests was very good; the worst was for Gas Mix 5, but the standard deviations for these
groups were less than 6 % of the means. For all the other gases, the standard deviations were less
than 2 % of the means. A linear fit of all the CO/CO2 group means gave an r2 of 0.998.
The accuracy of the HC/CO2 measurements (Table 5a) was better overall than those for
CO/CO2, but the precision was not quite as good. Although the group means, in general, were
below the expected values, they were on occasion above; the largest difference was for one of the
Gas Mix 2 groups, where the mean was about 11 % lower than the expected value. The standard
deviations for the Gas Mix 5 groups ranged from about 12 to 21 % of the means, but this was for a
gas containing a very low concentration of HCs (HC/CO2 = 0.002143). For all the other gases, the
standard deviations were less than 4 % of the means. A linear fit of all the HC/CO2 group means
gave an r2 of 0.998.
Also in Table 5a, for Gas Mixes 1 (the calibration gas) and 3, which had similar expected
NO/CO2 values (0.027829 and 0.033403, respectively), the NO/CO2 group means covered similar
ranges, with values both higher and lower than the expected values. The maximum difference was
less than 8 %. However, for Gas Mix 4, the group means were all higher than the expected value;
they averaged about 13 % higher, with a maximum difference of about 21 %. This pattern was
similar to the results for CO, where the higher concentration calibration gas gave an overestimation
of a lower concentration gas. The precision for these measurements was worse overall than those
for CO and HCs. The standard deviations for data groups of the three gas mixes containing NO
ranged from about 9 to 21 % of the means. Despite this, a linear fit of all the group means gave an
r2 of 0.990.
-------
Table 5a. Puffiest Results for the RSD-1000 Calibrated with Gas Mix 1
Gas
Mix
1
2
3
4
5
Group
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
N
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
CO/CO2 Ratio
Expected
0.236434
0.236434
0.236434
0.236434
0.333333
0.333333
0.333333
0.333333
0.611006
0.611006
0.611006
0.611006
1.008361
1.008361
1.008361
1.008361
0.071429
0.071429
0.071429
0.071429
Mean
0.23629
0.236484
0.238174
0.234844
0.318579
0.316387
0.316362
0.314568
0.561189
0.548256
0.548804
0.537615
0.860694
0.851913
0.842408
0.853434
0.078752
0.077888
0.08543
0.080662
Std Dev
0.002993
0.003297
0.003225
0.002812
0.00281
0.004042
0.004797
0.003557
0.006228
0.007603
0.010791
0.008714
0.010364
0.014053
0.015118
0.008759
0.001433
0.001605
0.004696
0.00353
HC/CO2 Ratio
Expected
0.030543
0.030543
0.030543
0.030543
0.0100
0.0100
0.0100
0.0100
0.060247
0.060247
0.060247
0.060247
0.046739
0.046739
0.046739
0.046739
0.002143
0.002143
0.002143
0.002143
Mean
0.030641
0.030566
0.030446
0.030796
0.009466
0.009996
0.009516
0.008921
0.059668
0.057299
0.05773
0.058482
0.043349
0.044791
0.044944
0.042296
0.00219
0.0021
0.002351
0.002195
Std Dev
0.000635
0.000518
0.000586
0.000384
0.000343
0.000382
0.000316
0.000325
0.000933
0.001331
0.001591
0.001146
0.001123
0.001084
0.001138
0.001002
0.000386
0.000245
0.000491
0.000327
NO/CO2 Ratio
Expected
0.027829
0.027829
0.027829
0.027829
0
0
0
0
0.014232
0.014232
0.014232
0.014232
0.033403
0.033403
0.033403
0.033403
0
0
0
0
Mean
0.024753
0.028684
0.029632
0.027366
-0.00039
-0.00089
-0.00015
-0.00066
0.017194
0.014286
0.016596
0.01621
0.03484
0.035889
0.034157
0.032119
0.000068
0.0001
-0.00036
0.00002
Std Dev
0.003838
0.002978
0.00502
0.002509
0.002742
0.003109
0.002769
0.002128
0.003013
0.003056
0.002984
0.003179
0.005286
0.004349
0.003938
0.004446
0.002365
0.002835
0.002956
0.002307
-------
Gas Mix 3 as the Calibration Gas (Table 5b) — In this set of tests, the CO/CO2 group mean
values were best for the calibration gas. For higher concentrations of CO, the CO/CO2 group mean
values were underestimations of the expected values; i.e., those for Gas Mix 4 represented about a
5 to 7 % underestimation. For Gas Mixes 1, 2, and 5, which had a lower concentration of CO, the
group mean values overestimatedthe expected value of CO/CO2 by about 5 to 18 %. Again, this
fits the pattern of the test results using Gas Mix 1 as the calibration gas. The precision of this series
of tests was excellent; all of the standard deviations for the groups were less than 4 % of the means.
A linear fit of all the CO/CO2 group means gave an r2 of 0.997.
As with the tests using Gas Mix 1 as the calibration gas, the accuracy of the HC/CO2
measurements (Table 5b) was better overall than those for CO/CO2. However, with the exception
of the tests with Gas Mix 5, the precision was similar. The group means were above and below the
expected values; the differences between group means and expected values for all the gas mixes
were all less than 9 %. The standard deviations for the Gas Mix 5 groups ranged from about 12 to
19 % of the means, but this was for a gas containing a very low concentration of HCs (HC/CO2 =
0.002143). For all the other gases, the standard deviations were less than 4 % of the means. A
linear fit of all the HC/CO2 group means gave an r2 of 0.998.
The results for NO using Gas Mix 3 as the calibration gas (Table 5b) were inconsistent. The
tests with Gas Mix 1 resulted in NO/CO2 group means that were all underestimates of the expected
values. Since the NO concentration in Gas Mix 3 was lower than that of Gas Mix 1, this appeared
to fit the pattern seen with CO. However, the tests with Gas Mix 4, which had just a slightly higher
concentration than that of Gas Mix 1, resulted in group means that were both higher and lower than
the actual value. Even more surprising was the fact that the largest differences between the
NO/CO2 group means and the expected value occurred in tests with Gas Mix 3, the calibration gas.
In these tests, the differences ranged from about -4 to +22 % of the expected value. Differences
for Gas Mixes 1 and 4 were a maximum of-16 and -11%, respectively. The precision of the data
followed a similar pattern; the best was with Gas Mix 1 (14 to 17 % of the means), while the worst
was with Gas Mix 3 (13 to 32 % of the means). Overall, the precision for these measurements was
worse than those for CO and HCs. Despite this, a linear fit of all the group means gave an r2 of
0.984; however, a cubic equation fit gave an r2 of 0.991.
11
-------
Table 5b. Puff Test Results for the RSD-1000 Calibrated with Gas Mix
Gas
Mix
1
2
3
4
5
Group
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
N
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
CO/CO2 Ratio
Expected
0.236434
0.236434
0.236434
0.236434
0.333333
0.333333
0.333333
0.333333
0.611006
0.611006
0.611006
0.611006
1.008361
1.008361
1.008361
1.008361
0.071429
0.071429
0.071429
0.071429
Mean
0.266558
0.259557
0.260142
0.257566
0.349382
0.357014
0.358065
0.363864
0.612471
0.61547
0.611711
0.610452
0.95066
0.94411
0.960901
0.940967
0.084074
0.082501
0.083149
0.080153
Std Dev
0.004823
0.004666
0.004075
0.00409
0.005195
0.005693
0.006496
0.00538
0.011556
0.010182
0.011846
0.010404
0.012189
0.011711
0.027607
0.017383
0.002469
0.002027
0.002886
0.002969
HC/CO2 Ratio
Expected
0.030543
0.030543
0.030543
0.030543
0.0100
0.0100
0.0100
0.0100
0.060247
0.060247
0.060247
0.060247
0.046739
0.046739
0.046739
0.046739
0.002143
0.002143
0.002143
0.002143
Mean
0.031405
0.032384
0.03152
0.033049
0.010036
0.010415
0.010197
0.010195
0.059904
0.059317
0.060423
0.060686
0.04516
0.04434
0.045191
0.045762
0.002165
0.002155
0.002038
0.002011
Std Dev
0.000612
0.000618
0.000504
0.000604
0.000229
0.000339
0.000211
0.000356
0.001603
0.001836
0.001411
0.001394
0.001013
0.001108
0.001256
0.000925
0.00042
0.000253
0.000277
0.000339
NO/CO2 Ratio
Expected
0.027829
0.027829
0.027829
0.027829
0
0
0
0
0.014232
0.014232
0.014232
0.014232
0.033403
0.033403
0.033403
0.033403
0
0
0
0
Mean
0.024676
0.026739
0.026092
0.023291
0.000048
0.000199
-0.00031
-0.00103
0.01732
0.015282
0.013625
0.015738
0.029663
0.033336
0.033483
0.035135
-0.00023
0.00087
0.000415
-0.00014
Std Dev
0.004077
0.004163
0.003579
0.003534
0.00256
0.002603
0.002706
0.002849
0.004539
0.003417
0.004295
0.002056
0.004663
0.004204
0.007898
0.004436
0.003241
0.003607
0.003684
0.00447
-------
Gas Mix 4 as the Calibration Gas (Table 5c) — In this set of tests, the CO/CO2 group mean
values were best for the calibration gas. Gas Mix 4 had the highest concentration of CO, so,
following the pattern of previous testing, the CO/CO2 group mean values for tests with the
remaining four gas mixes were overestimations of the expected values. The overestimation
increased (as a %age of the true value) as the CO concentration decreased. For example, Gas Mix
3 gave means about 3 to 4 % above the expected value, while Gas Mix 5 gave means 28 to 32 %
above the expected value. However, precision was good for all of the tests; the standard deviations
for all the CO/CO2 groups were 2 % or less, even for Gas Mix 5. As a result, a linear fit of all the
group means gave an r2 of 0.999.
As with the tests using Gas Mixes 1 and 3 as the calibration gases, the accuracy of the
HC/CO2 measurements (Table 6c) was better overall than those for CO/CO2. However, the
precision overall was not quite as good. The group means were above and below the expected
values. The differences between group means and actual values for Gas Mixes 1-4 were all less
than 5 %; for Gas Mix 5, they ranged from 7 to 11 %. The standard deviations for the Gas Mix 5
groups ranged from about 7 to 10 % of the means; for all the other gases, the standard deviations
were less than 5 % of the means. A linear fit of all the HC/CO2 group means gave an r2 of 0.999.
The results for NO using Gas Mix 4 as the calibration gas (Table 5c) were inconsistent. The
tests with Gas Mix 1 resulted in NO/CO2 group means that were all underestimates of the expected
value. In fact, these tests showed the largest differences between the group means and the expected
value (-5 to -24 %). The results with Gas Mixes 3 and 4 were similar, with group means above and
below the expected values; for Gas Mix 3, they ranged from -7 to +8 % and for Gas Mix 4 (the
calibration gas), from -5 to +8 %. The precision of the data was not particularly good, ranging
from about 11 to 26 %. Despite this, a linear fit of all the group means gave an r2 of 0.981;
however, a cubic equation was required to give a fit comparable to those for the CO and HC data
(an r2 of 0.992).
In summary, it appears that when the RSD-1000 was calibrated with a mixture containing
CO at a high ratio to CO2, the measurement of gas mixtures containing CO at a lower ratio to CO2
was increasingly overestimated, on average, by the device. The reverse was true when a low
CO/CO2 ratio mixture was used for calibration and a high ratio gas was puffed; in this case the
higher ratio gases were underestimated. This pattern did not appear to hold for HCs and NO. This
problem has reportedly been noted by the manufacturer and dealt with in more recent models of the
device.
The accuracy of the RSD-1000 HC measurements was better overall than those for CO,
while the accuracy of the NO measurements was not as good overall as those for CO and HCs. The
precision of the measurements was best for CO, not quite as good for HCs, and worst for NO.
However, the linear fit of the group means of all the data, at its worst., resulted in an r2 of 0.981.
13
-------
Table 5c. Puff Test Results for the RSD-1000 Calibrated with Gas Mix 4
Gas
Mix
1
2
3
4
5
Group
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
N
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
CO/CO2 Ratio
Expected
0.236434
0.236434
0.236434
0.236434
0.333333
0.333333
0.333333
0.333333
0.611006
0.611006
0.611006
0.611006
1.008361
1.008361
1.008361
1.008361
0.071429
0.071429
0.071429
0.071429
Mean
0.26391
0.265342
0.262416
0.260852
0.361436
0.355815
0.347082
0.349268
0.635164
0.630512
0.636317
0.634238
0.996811
1.011076
1.008984
1.020835
0.093625
0.094111
0.093876
0.091476
Std Dev
0.004664
0.005236
0.004273
0.00498
0.006292
0.006486
0.006802
0.006301
0.009284
0.012866
0.009167
0.006539
0.011986
0.012473
0.012262
0.012262
0.001868
0.001405
0.001497
0.001354
HC/CO2 Ratio
Expected
0.030543
0.030543
0.030543
0.030543
0.0100
0.0100
0.0100
0.0100
0.060247
0.060247
0.060247
0.060247
0.046739
0.046739
0.046739
0.046739
0.002143
0.002143
0.002143
0.002143
Mean
0.031738
0.030525
0.032
0.031887
0.009847
0.010265
0.009732
0.009661
0.059549
0.059909
0.061169
0.059488
0.046565
0.046808
0.046445
0.04688
0.002358
0.002303
0.002293
0.002335
Std Dev
0.000917
0.000587
0.000753
0.000473
0.00028
0.000433
0.000453
0.000367
0.001011
0.000889
0.001138
0.001127
0.000692
0.000975
0.000944
0.00121
0.000228
0.000258
0.000199
0.000171
NO/CO2 Ratio
Expected
0.027829
0.027829
0.027829
0.027829
0
0
0
0
0.014232
0.014232
0.014232
0.014232
0.033403
0.033403
0.033403
0.033403
0
0
0
0
Mean
0.026332
0.025214
0.021273
0.024172
-0.00076
-0.00076
0.000348
-0.00013
0.01539
0.013267
0.013775
0.014122
0.034958
0.033314
0.031805
0.036079
0.000167
-0.0003
0.000766
-0.00076
Std Dev
0.004403
0.004449
0.003709
0.003522
0.002032
0.002369
0.004074
0.003269
0.003743
0.003467
0.003625
0.003576
0.00532
0.003683
0.006459
0.00571
0.002706
0.002767
0.002413
0.003151
-------
Experiment 2
A summary of the Smog Dog™ 3 puff test results is shown in Tables 6a-6c. The Group
number designates a series of puffs that were initiated after the device was calibrated, and N
indicates the number of puffs in each Group. The means and standard deviations (Std Dev) of the
CO/CO2, HC/CO2, and NO/CO2 ratios are shown for each series of puffs.
Gas Mix 1 as the Calibration Gas (Table 6a) — In this set of tests, the CO/CO2 group means
were fairly close to the expected value in all cases. The worst discrepancy was in the test with Gas
Mix 4, where the group means ranged from about 4 to II % higher than the expected value. (Note
that Gas Mix 4 had a higher concentration that Gas Mix 1, so the bias noted in experiment 1 with
the RSD-1000 did not occur with this device.) The tests with the other gases gave group means
higher and lower than the expected value, but the differences were all less than 5 % of the expected
value. The precision was fairly good; the largest standard deviation for the Gas Mix 5 groups was
about 5 % of the mean, while those for the remaining gas mixes were all less than 3 % of their
means. A linear fit of the means gave an r2 of 0.997.
On the other hand, the situation was much worse for the HC/CO2 group means (Table 6a).
The HC/CO2 group means for Gas Mix 3 were about 13 to 15 % higher than expected value, while
the group means for Gas Mix 4 were about 93 to 143 % higher than the expected value. On the
other hand, the HC/CO2 group means for Gas Mix 3 were about 3 to 13 % lower than the expected
value. The precision for the HC measurements was also not as good as that for CO. Although the
standard deviations for data groups for Gas Mixes 1-3 ranged from about 2 to 6 % of their means,
those for Gas Mix 4 ranged from about 7 to 11 %. The standard deviations for data group for Gas
Mix 5 ranged from about 10 to 29 % of their means, but the concentration of HC in this gas mix
was very low (less than 10 % of the HC concentration of Gas Mix 1). Despite this, a linear fit of
the means gave an r2 of 0.94; however, a second order polynomial fit gave an r2 of 0.988.
The NO/CO2 group means (Table 6a) were inconsistent. The group means for tests with the
calibration gas (Gas Mix 1) were about 9 to 13 % higher than the expected value. For Gas Mixes
3 and 4, however, the NO/CO2 group means were about 32 to 35 % and about 8 to 12 % lower
than the expected values, respectively. The precision for the NO measurements was not quite as
good as that for HCs. The standard deviations for Gas Mixes 1 and 4 ranged from about 4 to 9 %
of the group means, while for Gas Mix 3, the standard deviations ranged from about 11 to 23 % of
the means. Despite this, a linear fit of the means gave an r2 of 0.965; however, a second order
polynomial gave an r2 of 0.993.
15
-------
Table 6a. Puffiest Results for the Smog Dog™ 3 Calibrated with Gas Mix 1
Gas
Mix
1
2
3
4
5
Group
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
N
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
CO/CO2 Ratio
Expected
0.226316
0.226316
0.226316
0.226316
0.333333
0.333333
0.333333
0.333333
0.611006
0.611006
0.611006
0.611006
1.006662
1.006662
1.006662
1.006662
0.071429
0.071429
0.071429
0.071429
Mean
0.227204
0.230684
0.224996
0.230312
0.327344
0.330488
0.327304
0.327652
0.622044
0.602744
0.616624
0.617844
1.11258
1.04726
1.09566
1.08156
0.074884
0.0725
0.070348
0.075188
Std Dev
0.004275
0.005077
0.004013
0.003112
0.005553
0.004391
0.00335
0.004622
0.010288
0.010624
0.01389
0.016115
0.025994
0.031056
0.018841
0.017945
0.002148
0.003124
0.003543
0.003613
HC/CO2 Ratio
Expected
0.032308
0.032308
0.032308
0.032308
0.0100
0.0100
0.0100
0.0100
0.060246
0.060246
0.060246
0.060246
0.132578
0.132578
0.132578
0.132578
0.002143
0.002143
0.002143
0.002143
Mean
0.032768
0.032268
0.032628
0.032352
0.010664
0.010376
0.010572
0.010348
0.068392
0.069304
0.06862
0.067832
0.25602
0.267396
0.322783
0.284868
0.002068
0.001872
0.00194
0.00228
Std Dev
0.000515
0.000775
0.000982
0.000601
0.000673
0.000296
0.000349
0.000318
0.002436
0.003194
0.002735
0.002192
0.024482
0.023786
0.023619
0.030145
0.000214
0.000786
0.000456
0.000665
NO/CO2 Ratio
Expected
0.031105
0.031105
0.031105
0.031105
0
0
0
0
0.014232
0.014232
0.014232
0.014232
0.02545
0.02545
0.02545
0.02545
0
0
0
0
Mean
0.033932
0.034388
0.033888
0.035256
-0.00224
-0.00254
-0.00309
-0.00276
0.009356
0.009332
0.009284
0.009676
0.023472
0.022324
0.023268
0.022796
0.000328
0.000384
0.000116
-0.00014
Std Dev
0.00134
0.00318
0.002901
0.002472
0.000868
0.001058
0.000828
0.001008
0.00135
0.001046
0.00211
0.00180
0.001274
0.001303
0.001677
0.001502
0.001327
0.001562
0.002286
0.001783
-------
Gas Mix 3 as the Calibration Gas (Table 6b) — In this set of tests, the CO/CO2 group means
were fairly close to the expected value in all cases; the maximum difference was a maximum of 6 %.
The precision was also good; all standard deviations were less than 5 % of their means. Not
surprisingly, a linear fit of the means gave an excellent r2 of 0.9995.
For gas mixes with HC concentrations lower than that of Gas Mix 3, the HC/CO2 group
means were all underestimates of the expected value. The lower the HC concentration relative to
Gas Mix 3, the greater the difference between the group means and the expected value. For
example, the group means for Gas Mix 1 ranged from about 9 to 11 % lower than the expected
value, while the means for Gas Mix 5 ranged from about 20 to 24 % lower. For Gas Mix 4, which
had a higher HC concentration than that of Gas Mix 3, the group means were about 1 to 8 % higher
than the expected value. Note that this result is the reverse of what occurred with CO
measurements using the RSD-1000 unit in Experiment 1, above. The precision for these
measurements was fairly good. For Gas Mixes 1-4, the group standard deviations were no more
than about 5 % of the means; for Gas Mix 5, about 5 to 10 %. A linear fit of the means gave an
excellent r2 of 0.998.
Again, the NO/CO2 group means (Table 6b) were inconsistent. The measured values for the
calibration gas (Gas Mix 3) were about 30 to 32 % lower than the expected value. This was the
worst discrepancy; the worst of the remaining discrepancies were in the test with Gas Mix 4, in
which the NO/CO2 group means were about 13 to 20 % lower than the expected value. On the
other hand, in the test with Gas Mix 1, the NO/CO2 group means were somewhat (about 3 to 11 %)
higher than the expected value. The precision again was not as good as the measurements with
HCs. The standard deviations for the Gas Mix 1 data ranged from about 5 to 8 % of the group
means, while those for the calibration gas (Gas Mix 3) ranged from about 8 to 16 % of the means.
Despite all this, a linear fit of the means gave an r2 of 0.969; however, a third order polynomial was
required to give a fit comparable to the CO and HC linear fits (an r2 of 0.992).
17
-------
Table 6b. Puffiest Results for the Smog Dog™ 3 Calibrated with Gas Mix 3
Gas
Mix
1
2
3
4
5
Group
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
N
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
CO/CO2 Ratio
Expected
0.226316
0.226316
0.226316
0.226316
0.333333
0.333333
0.333333
0.333333
0.611006
0.611006
0.611006
0.611006
1.006662
1.006662
1.006662
1.006662
0.071429
0.071429
0.071429
0.071429
Mean
0.220328
0.21486
0.225448
0.221532
0.320456
0.314056
0.319396
0.320948
0.600412
0.607348
0.608776
0.615264
1.0025
1.005116
1.010736
0.993236
0.072904
0.075628
0.075728
0.074008
Std Dev
0.010525
0.004770
0.008027
0.007640
0.002172
0.001543
0.002263
0.001754
0.004593
0.005276
0.004466
0.009277
0.038798
0.020561
0.019138
0.040287
0.001233
0.001774
0.001875
0.001786
HC/CO2 Ratio
Expected
0.032308
0.032308
0.032308
0.032308
0.0100
0.0100
0.0100
0.0100
0.060246
0.060246
0.060246
0.060246
0.132578
0.132578
0.132578
0.132578
0.002143
0.002143
0.002143
0.002143
Mean
0.028736
0.028876
0.02936
0.029252
0.008956
0.008364
0.00868
0.00882
0.059468
0.0603
0.060268
0.060628
0.1342
0.139104
0.135936
0.142988
0.001632
0.001704
0.001712
0.001636
Std Dev
0.000838
0.000459
0.000802
0.000468
0.000126
0.000104
0.000096
0.0001
0.000309
0.000509
0.000348
0.000644
0.006391
0.001752
0.004575
0.001932
0.000080
0.000093
0.000174
0.000104
NO/CO2 Ratio
Expected
0.031105
0.031105
0.031105
0.031105
0
0
0
0
0.014232
0.014232
0.014232
0.014232
0.02545
0.02545
0.02545
0.02545
0
0
0
0
Mean
0.033328
0.032112
0.034448
0.032364
-0.0043
-0.00425
-0.00422
-0.0043
0.009916
0.009672
0.010032
0.00994
0.022084
0.021132
0.021572
0.020252
-0.00061
-0.00054
-0.00079
-0.00070
Std Dev
0.002242
0.002688
0.002577
0.001614
0.000473
0.00048
0.000347
0.000459
0.000994
0.00099
0.000787
0.001629
0.001574
0.00118
0.001451
0.00212
0.00055
0.000535
0.001008
0.000614
oo
-------
Gas Mix 4 as the Calibration Gas (Table 6c) — In this set of tests, the CO/CO2 group means
were all underestimates of the expected value, although they were all less than 9 % low. The
standard deviations were about 1 to 2 % of the group means for Gas Mixes 1-4, and less than 5 %
of the means for Gas Mix 5. Not surprisingly, a linear fit of the means gave an excellent r2 of 0.999.
On the other hand, with the exception of the test with Gas Mix 4 (the calibration gas), the
HC/CO2 group means were considerably lower than the expected values. The worst discrepancy
was for Gas Mix 5, where the measured mean values were 3 6 to 47 % lower than the expected
value. (Note that the HC concentration in the calibration gas was higher than that of Gas Mix 5, so
this result is again the reverse of what occurred with CO measurements using the RSD-1000 unit in
Experiment 1, above.) However, the precision of the data was quite good; the standard deviations
for the data groups were all less than 3 % of their means. A linear fit of the means gave an excellent
r2 of 0.995.
The NO/CO2 group means were consistently lower than the expected values. The means for
the calibration gas (Gas Mix 4) were about 22 to 24 % lower than the expected value. However,
the worst discrepancy was in the test with Gas Mix 3, in which the NO/CO2 group means were
about 28 to 33 % lower than the expected value. Again, the precision of the data was not as good
as that for CO and HCs. The standard deviations were less than 5 % of the means for Gas Mixes 1
and 4, but were about 12 to 18 % of the means for Gas Mix 3. A linear fit of the means gave an r2
of 0.977; however, a second order polynomial was required to give a fit comparable to those for the
CO and HC data (an r2 of 0.99).
19
-------
Table 6c. Puffiest Results for the Smog Dog™ 3 Calibrated with Gas Mix 4
Gas
Mix
1
2
3
4
5
Group
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
N
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
25
CO/CO2 Ratio
Expected
0.226316
0.226316
0.226316
0.226316
0.333333
0.333333
0.333333
0.333333
0.611006
0.611006
0.611006
0.611006
1.006662
1.006662
1.006662
1.006662
0.071429
0.071429
0.071429
0.071429
Mean
0.212072
0.211192
0.216196
0.211892
0.31372
0.308068
0.31432
0.312668
0.575516
0.584508
0.572292
0.577436
0.994116
1.002756
0.99802
0.99568
0.065104
0.067464
0.068244
0.068636
Std Dev
0.002554
0.004265
0.003094
0.002862
0.00367
0.005505
0.00421
0.003242
0.009051
0.007877
0.00635
0.009468
0.005975
0.004405
0.006271
0.004791
0.001491
0.002041
0.003176
0.001835
HC/CO2 Ratio
Expected
0.032308
0.032308
0.032308
0.032308
0.01000
0.01000
0.01000
0.01000
0.060247
0.060247
0.060247
0.060247
0.132578
0.132578
0.132578
0.132578
0.002143
0.002143
0.002143
0.002143
Mean
0.02446
0.023632
0.024436
0.024552
0.006432
0.006388
0.006636
0.006456
0.051328
0.053952
0.051964
0.051076
0.133064
0.133572
0.132788
0.133164
0.001144
0.001236
0.001188
0.00138
Std Dev
0.00059
0.0001029
0.000603
0.000552
0.000107
0.000088
0.000091
0.000082
0.000699
0.000667
0.000797
0.001131
0.000402
0.000629
0.000651
0.000476
0.000058
0.000182
0.000322
0.000168
NO/CO2 Ratio
Expected
0.031105
0.031105
0.031105
0.031105
0
0
0
0
0.014232
0.014232
0.014232
0.014232
0.02545
0.02545
0.02545
0.02545
0
0
0
0
Mean
0.02958
0.029392
0.02938
0.02892
-0.00253
-0.00276
-0.00256
-0.00256
0.010012
0.010236
0.00958
0.009484
0.0198
0.019688
0.019844
0.019416
0.000192
0.000064
0.000072
-0.00023
Std Dev
0.001031
0.001322
0.000737
0.000794
0.000628
0.000586
0.000477
0.000286
0.001219
0.001268
0.001157
0.001672
0.000806
0.000655
0.000601
0.000581
0.00055
0.000535
0.001008
0.000614
to
o
-------
Experiment 3
A summary of the puff test results for the three RSDs is shown in Tables 7a-7c, below. The
Group number designates a series of puffs that were initiated after each device was calibrated, and
N indicates the number of puffs in each Group. The means and standard deviations (Std Dev) of
the CO/CO2,HC/CO2,and NO/CO2 ratios are shown for each series of puffs and for each device.
Note that prior to these tests, the RSD-1000 software had been upgraded, an improvement over
that used in Experiment 1, above.
Although the main objective of this series of puff tests was to evaluate the capability of
RSDs to measure NO emissions (the CO and HC concentrations in all three test gases were
essentially unvaried), it is interesting to note the response of the instruments to CO and HCs.
As shown in Table 7a, the CO/CO2 group means for the FEAT #3008 were both higher and
lower than the expected values for the three test gases; the standard deviations for the data groups
ranged from about 7 to about 20 % of the mean values. The CO/CO2 group means for the RSD-
1000 were always slightly more than the expected values; the standard deviations for the data
groups were less than 10 % of the mean values. [Note that this was despite the fact that the
calibration gas values were about 10 times the test gas values, indicating that the bias shown in the
earlier puff tests with the RSD-1000 (Experiment 1, above) had been corrected by the software
upgrade.] However, the CO/CO2 group means for the Smog Dog™ 3 were always considerably
greater than the expected values; i.e., about 60-90 % greater. The standard deviations for the Smog
Dog™ 3 data groups were similar (relative to the mean values) to those of the FEAT #3008.
Table 7a. RSD Puff Measurements - CO/CO, Ratios
Gas
Mix
1
2
3
Group
1
2
3
4
1
2
3
4
1
2
3
4
Expected
0.0199867
0.0199867
0.0199867
0.0199867
0.0200133
0.0200133
0.0200133
0.0200133
0.0199933
0.0199933
0.0199933
0.0199933
FEAT #3008
N
25
25
25
25
25
25
25
25
25
25
25
25
Mean
0.018311
0.021552
0.019603
0.022407
0.020026
0.020834
0.020429
0.021846
0.022123
0.0198
0.019829
0.021353
Std Dev
0.001421
0.002491
0.002264
0.004033
0.002608
0.00193
0.00262
0.002132
0.002294
0.003006
0.00208
0.002013
RSD-1000
N
25
25
25
25
25
25
25
25
25
25
25
25
Mean
0.018737
0.018748
0.018843
0.018709
0.019246
0.018696
0.018934
0.018907
0.019771
0.018684
0.019844
0.019657
Std Dev
0.000934
0.00091
0.000858
0.000869
0.001041
0.000756
0.000947
0.00104
0.001451
0.001425
0.001459
0.000857
Smog Dog™ 3
N
24
27
28
28
25
24
26
29
26
30
27
27
Mean
0.032188
0.033856
0.032929
0.033693
0.032108
0.032654
0.038365
0.031797
0.033385
0.034437
0.033874
0.034415
Std Dev
0.001556
0.001618
0.001736
0.002381
0.002329
0.002348
0.005379
0.006439
0.001451
0.002271
0.001993
0.001104
The pattern for the HC/CO2 group means (Table 7b) was similar to those for the
21
-------
CO/CO2. The HC/CO2 group means for the FEAT #3008 were both higher and lower than the
expected values for the three test gases, although over a considerably greater range than for
CO/CO2. Specifically, they ranged from about 8 % less to about 25 % more than the expected
values. The standard deviations for the data groups ranged from about 11 to 22 % of the mean
values.
The HC/CO2 group means for the RSD-1000 were consistently higher (11 to 22 % higher)
than the expected values for the three test gases. The standard deviations for the RSD-1000 data
groups ranged from about 5 to 15 % of the mean values. The HC/CO2 group means for the Smog
Dog™ 3 were also consistently higher than the expected values for the three test gases, but by a
wider margin; specifically, about 19 to 45 % higher. The standard deviations for the Smog Dog™ 3
data groups ranged from about 8 to 22 % of the mean values, again, similar to the FEAT #3008
results.
Table 7b. RSD Puff Measurements - HC/CO, Ratios
Gas
Mix
1
2
3
Group
1
2
o
J
4
1
2
o
J
4
1
2
o
J
4
Expected
0.003331
0.003331
0.003331
0.003331
0.003509
0.003509
0.003509
0.003509
0.003347
0.003347
0.003347
0.003347
FEAT #3008
N
25
25
25
25
25
25
25
25
25
25
25
25
Mean
0.003062
0.003592
0.0036
0.00426
0.004156
0.003852
0.004036
0.003994
0.003789
0.004038
0.003889
0.003698
Std Dev
0.000523
0.000541
0.000668
0.000952
0.000682
0.000411
0.000548
0.000671
0.000871
0.000685
0.000651
0.000832
RSD-1000
N
25
25
25
25
25
25
25
25
25
25
25
25
Mean
0.003702
0.003703
0.003829
0.003798
0.003932
0.003841
0.00396
0.004051
0.003843
0.00371
0.003962
0.003891
Std Dev
0.000259
0.000261
0.000262
0.000211
0.000263
0.000299
0.000264
0.000277
0.000562
0.000351
0.000398
0.000229
Smog Dog™ 3
N
24
27
28
28
25
24
26
29
26
30
27
27
Mean
0.003971
0.004211
0.004204
0.004214
0.004052
0.004179
0.004815
0.004169
0.004219
0.004037
0.004337
0.0042
Std Dev
0.0006
0.000404
0.00042
0.000619
0.000418
0.000551
0.00069
0.000919
0.000343
0.000487
0.000463
0.000348
FEAT #3008 NO/CO2 Values - As shown in Table 7c, the NO/CO2 group means for the
FEAT #3008 were both higher and lower than the expected values for the three test gases, but over
a higher range than for HC/CO2. Relative to the expected values, the group means were:
for Gas Mix 1, - 26 to + 30 %;
for Gas Mix 2, - 11 to + 22 %; and
for Gas Mix 3, - 19 to+ 6%.
The standard deviations for the FEAT #3008 data groups were considerably higher than those for
both CO/CO2 and HC/CO2 Relative to the means, they were:
for Gas Mix 1, 51 to 140%;
for Gas Mix 2, 31 to 49 %; and
for Gas Mix 3, 29 to 41 %.
22
-------
As might be expected, as the expected NO/CO2 value increased, the deviation of the means of the
data group from the expected value decreased, and the variability of the data decreased. A linear fit
of a plot of the NO/CO2 means resulted in an r2 of 0.883.
RSD-1000 NO/CO2 Values - The NO/CO2 group means for the RSD-1000 were both
higher and lower than the expected values for the gas mixes 1 and 3, but were always higher for gas
mix 2. Although not as large a range of differences as was exhibited by the FEAT #3008, the range
was greater than that exhibited by the RSD-1000 for HC/CO2. Specifically, relative to the expected
values, the group means were:
for gas mix 1, - 2 to + 13 %;
for gas mix 2, + 17 to + 22 %; and
for gas mix 3, - 2 to + 45 %.
The standard deviations for the RSD-1000 data groups were higher than those for both CO/CO2
and HC/CO2 Relative to the means, they were:
for gas mix 1,21 to 29 %;
for gas mix 2, 20 to 37 %; and
for gas mix 3, 16 to 20 %.
Unlike the results for the FEAT #3008, as the actual (expected) NO/CO2 value increased, the
deviation of the means of the data group from the actual value actually increased. In addition, the
variability of the data actually increased in the tests for gas mix 2, when compared to the results of
the tests for gas mix 1. However, for gas mix 3, the variability of the data decreased. A linear fit of
a plot of the NO/CO2 group means resulted in an r2 of 0.926, somewhat better than the fit for the
FEAT #3008 data.
Smog Dog™ 3 NO/CO2 Values - The NO/CO2 group means for the Smog Dog™ 3 were
consistently higher than the expected values for all three gas mixes. Although the differences from
the expected values were somewhat higher than those for HC/CO2, they were lower than those for
CO/CO2. The differences were, overall, larger than those for both the FEAT #3008 and the RSD-
1000. Specifically, relative to the expected values, the group means were:
for gas mix 1, 37 to 64 %;
for gas mix 2, 31 to 52 %; and
for gas mix 3, 26 to 35 %.
The standard deviations for the Smog Dog™ 3 data groups, relative to the means, were
considerably higher than those for both CO/CO2 and HC/CO2 for gas mixes 1 and 2, but lower for
gas mix 3. Relative to the means, they were:
for gas mix 1, 25 to 57 %;
for gas mix 2, 20 to 41 %; and
for gas mix 3, 9 to 11 %.
As with the FEAT #3008, as the expected NO/CO2 value increased, the deviation of the means of
the Smog Dog™ 3 data group from the actual value decreased, and the variability of the data
decreased. A linear fit of a plot of the NO/CO2 group means resulted in an r2 of 0.981, which was a
better fit than data from either of the other two devices.
23
-------
In summary, with group means both above and below the actual values, the mean of all the
FEAT #3008 NO/CO2 data (100 measurements for each gas mix) was the most accurate. Except for
the RSD-1000 test with gas mix 1, the means of the data for the RSD-1000 and the Smog Dog™ 3
were overestimates of the actual values. However, with the exception of the RSD-1000 test with
gas mix 3, the precision of the FEAT #3008 was not as good as the other two devices. On the
other hand, while the Smog Dog™ 3 exhibited consistent precision, its overall accuracy was not as
good in these NO puff tests relative to the other two devices. Nevertheless, because all three
devices responded in essentially a linear manner to the increases in NO concentration, they appeared
to be adequate for this application. Therefore, dynamometer, track, and road tests were
subsequently conducted.
Table 7c. RSD Puff Measurements - NO/CO9 Ratios
Gas
Mix
1
2
3
Group
1
2
o
J
4
1
2
o
J
4
1
2
o
J
4
Expected
0.003331
0.003331
0.003331
0.003331
0.006664
0.006664
0.006664
0.006664
0.009973
0.009973
0.009973
0.009973
FEAT #3008
N
25
25
25
25
25
25
25
25
25
25
25
25
Mean
0.004028
0.004341
0.00247
0.003184
0.007281
0.005921
0.007742
0.008104
0.010558
0.00803
0.01062
0.009715
Std Dev
0.002048
0.005763
0.003461
0.003794
0.003573
0.002982
0.002519
0.002503
0.003434
0.003277
0.004126
0.002794
RSD-1000
N
25
25
25
25
25
25
25
25
25
25
25
25
Mean
0.003254
0.003756
0.003588
0.003562
0.008565
0.007779
0.00805
0.008305
0.014437
0.012475
0.009809
0.011741
Std Dev
0.000773
0.000787
0.001043
0.001042
0.001713
0.001751
0.001735
0.00309
0.002438
0.002455
0.001537
0.002084
Smog Dog™ 3
N
24
27
28
28
25
24
26
29
26
30
27
27
Mean
0.005475
0.005389
0.005164
0.004564
0.00876
0.009654
0.010154
0.009314
0.012569
0.0134
0.012896
0.013444
Std Dev
0.001992
0.001534
0.00128
0.002619
0.001767
0.002353
0.004176
0.003399
0.001102
0.001509
0.001342
0.001225
Experiment 4
The results of the dynamometer testing are summarized in Table 8. The means and standard
deviations (Std Dev) of the CO/CO2 HC/CO2 and NO/CO2 ratios are shown for each vehicle test.
The intent was to conduct a series of three tests with each vehicle. However, because of problems
in the first series of tests with the Mustang, it was subjected to a second series of tests to obtain NO
data that could be used in the comparisons to the track test data. As stated earlier, the results of the
dynamometer testing were assumed to be the actual emissions from the vehicles, or the "standards"
by which the track results would be measured.
24
-------
Table 8 . Summary of Second-by-Second Emissions During Dynamometer Testing
Vehicle
Ford Mustang II
Ford Mustang II
Ford Mustang II
Ford Mustang II
Ford Mustang II
Ford Mustang II
Ford F- 150 Truck
Ford F- 150 Truck
Ford F- 150 Truck
Dodge Caravan
Dodge Caravan
Dodge Caravan
Test
#
1
2
3
4
5
6
1
2
3
1
2
3
Time
(sec)
481
481
481
481
440
481
481
481
481
481
481
481
C0/C02
Mean
0.000118
0.000018
0.000075
NA
NA
NA
0.000632
0.000364
0.000517
0.012875
0.010857
0.011414
Std Dev
0.000061
0.000049
0.000369
NA
NA
NA
0.000430
0.000587
0.000439
0.005070
0.004651
0.007296
HC/C02
Mean
0.000579
0.000543
0.000588
NA
NA
NA
0.000830
0.000862
0.000980
0.001290
0.000976
0.000995
Std Dev
0.000089
0.000061
0.000119
NA
NA
NA
0.000031
0.000050
0.000060
0.000517
0.000401
0.000547
NO/CO2
Mean
NA
NA
NA
0.015359
0.014501
0.014341
0.003517
0.003705
0.003683
0.001055
0.001015
0.001087
Std Dev
NA
NA
NA
0.000325
0.001808
0.001442
0.000306
0.000212
0.000228
0.000721
0.000702
0.000702
Note: NA indicates that data were unavailable from second-by-second testing.
Again, the main objective of this series of tests was to ultimately evaluate the capability of
RSDs to measure NO emissions; however, it is interesting to note all the relative emissions of the
three vehicles. The average CO/CO2 ratio was the highest for the Caravan (~10"2), almost two
orders of magnitude higher than that of the F-150 (~2xlO"4), which was higher than that of the
Mustang. However, the mean CO/CO2 ratios for the Mustang were also highly variable (~2xlO"5 to
10"4). The same relative order held for the HC/CO2 ratios (Mustang lowest, Caravan highest),
although the range of the means was much less (~6xlO"4 to 10"3). For the NO/CO2 ratios, the
relative order was reversed (Caravan lowest, Mustang highest).
The relative precision of the data was also interesting. For the CO/CO2 ratios, the lowest
standard deviation relative to the mean was for the Caravan at 39 % [(0.005070/0.0128750x100],
while the highest was for the Mustang at 492 % [(0.000369/0.000075)xlOO]. Note, however, that
the CO/CO2 means for the Mustang were lower than those for the F-150, and much lower than
those for the Caravan. Note also that, for the HC/CO2 and NO/CO2 ratios, relative to the means,
the standard deviations were lower overall when compared with the CO/CO2 data. These ranged
from about 2 % - one of the NO/CO2 ratios for the Mustang - to 69 % - one of the NO/CO2
ratios for the Caravan. In fact, the relative precision of the Caravan HC and NO emission data was
considerably worse than that of the other two vehicles. In addition, while these data for the
Mustang and F-150 were considerably more precise than their CO data, all of the Caravan data had
about the same precision. Relative to the means, they were:
for CO/CO2, 39 to 64 %;
for HC/CO2, 40 to 55 %; and
forNO/CO2, 65 to 69%.
25
-------
Note that the CO/CO2 ratios for the Caravan were higher than those of the other two
vehicles, and these data represented the best precision. On the other hand, the NO/CO2 ratios for
the Caravan were lower than those of the other two vehicles, and these data represented the worst
precision. Assuming a proportional bias in the instrumentation at the dynamometer facility, this
would possibly explain the numbers above. However, the fact is that the HC/CO2 ratios for the
Caravan were similar to those of the other two vehicles (~10"3), but these data also represented the
worst precision. Although there may have been differences in the behavior of the Caravan on the
dynamometer or in the driver's handling of the vehicle, there is no clear reason for this anomaly.
For the other two vehicles, the standard deviations for the NO/CO2 data relative to the
means were:
for the Mustang, 2 to 13 %; and
for the F-150, 6 to 9 %.
Note, however, that because the NO/CO2 means for the Mustang were about four times as high as
those for the F-150, and an order of magnitude higher than those for the Caravan, the actual
variation in the NO/CO2 data was higher for the Mustang than for either of the other two vehicles.
It should be noted that the Mustang was tested twice. During the first series of tests (1-3),
the NO measurements exceeded the range for which the instrumentation had been set. For this
reason, a separate series of tests (4-6) was conducted, during which only the NO was measured.
Because of probable changes in fuel - as well as changes in the condition (e.g., tuning) and
operation of the vehicle — the combination of CO and HC measurements may or may not be
comparable to the NO measurements.
Experiment 5
The results of the track testing are summarized in Tables 9a-9c. The Group number
designates a series of laps that were made by the vehicles around the track, and N indicates the
number of laps in each group. After each group of laps, all three instruments were recalibrated.
The means and standard deviations (Std Dev) of the CO/CO2 HC/CO2 and NO/CO2 ratios are
shown for each group.
The average of the means of the four groups of CO/CO2 track data for the Mustang and F-
150 from all three RSDs (Table 9a) was about two orders of magnitude greater than the average of
the three groups of CO/CO2 dynamometer data (Table 8). However, the CO/CO2 track data for the
Caravan from the FEAT #3008 and the RSD-1000 were very similar to the corresponding
dynamometer data. The Smog Dog™ 3 CO/CO2 track data for the Caravan were all lower than
the corresponding dynamometer data.
26
-------
Table 9a. RSD Measurements at the Track -- CO/CO9 Ratios
Vehicle
Ford
Mustang
II
Ford
F-150
Truck
Dodge
Caravan
Group
1
2
3
4
1
2
3
4
1
2
3
4
FEAT #3008
N
27
30
27
27
27
30
27
27
27
30
27
27
Mean
0.007098
0.006296
0.006295
0.006849
0.005942
0.045064
0.025625
0.005687
0.016204
0.016648
0.016406
0.021827
Std Dev
0.013453
0.012869
0.011272
0.014061
0.009736
0.029773
0.030056
0.008443
0.018283
0.017414
0.027429
0.016425
RSD-1000
N
27
30
27
27
27
30
27
27
27
30
27
27
Mean
0.011
0.003967
0.005
0.006185
0.004556
0.030433
0.018667
0.00463
0.012296
0.011733
0.013481
0.012593
Std Dev
0.024404
0.003409
0.011923
0.008458
0.006154
0.023242
0.02301
0.004508
0.011864
0.00762
0.012119
0.008824
Smog Dog™ 3
N
22
24
25
27
23
25
27
26
19
24
21
21
Mean
0.009682
0.00145
0.002388
0.003804
0.000404
0.03448
0.011722
-0.00045
-0.00547
-0.0031
-0.01006
-0.01026
Std Dev
0.023507
0.005705
0.007591
0.007228
0.014203
0.036575
0.029716
0.01301
0.017234
0.01561
0.017086
0.015637
The average of the means of the four groups of the FEAT #3008 HC/CO2 track data for the
Mustang (Table 9b) was about an order of magnitude greater than the average of the three groups
of CO/CO2 dynamometer data (see Table 8), while the RSD-1000 track data for the Mustang
averaged about twice the corresponding dynamometer data. On the other hand, the Smog Dog™ 3
HC/CO2 track data for the Mustang averaged about a third (over 60 % lower) of the corresponding
dynamometer data. For the F-150, the FEAT #3008 HC/CO2 track data averaged about four times
the corresponding dynamometer data, while data from the other two RSDs averaged lower than the
same data. Specifically, the RSD-1000 data averaged about 42 % lower, while the Smog Dog™ 3
data averaged more than 300 % lower (all of the group means were less than zero; Table 9b). A
similar pattern occurred for the Caravan data; the FEAT #3008 data averaged about seven times the
corresponding dynamometer data, while data from the RSD-1000 and Smog Dog™ 3 averaged
almost 200 % and almost 400 % lower, respectively (both averages less than zero).
27
-------
Table 9b. RSD Measurements at the Track -- HC/CO, Ratios
Vehicle
Ford
Mustang
II
Ford
F-150
Truck
Dodge
Caravan
Group
1
2
3
4
1
2
3
4
1
2
3
4
FEAT #3008
N
27
30
27
27
27
30
27
27
27
30
27
27
Mean
0.003606
0.003202
0.005687
0.006818
0.003684
0.001341
0.004815
0.004509
0.005563
0.005596
0.008643
0.00824
Std Dev
0.002441
0.005508
0.003333
0.004629
0.001562
0.015595
0.003408
0.002034
0.002558
0.014159
0.006292
0.004577
RSD-1000
N
27
30
27
27
27
30
27
27
27
30
27
27
Mean
0.001481
0.0001
0.000481
0.002074
0.000815
-0.00017
0.000593
0.000815
0.000296
-0.00427
-0.00074
0.00063
Std Dev
0.001451
0.003021
0.003043
0.003802
0.000879
0.00278
0.001366
0.000736
0.001514
0.009512
0.002566
0.001597
Smog Dog™ 3
N
22
24
25
27
23
25
27
26
19
24
21
21
Mean
0.000155
-0.0004
0.000224
0.000693
-0.00145
-0.00208
-0.00177
-0.00217
-0.00291
-0.00256
-0.00395
-0.00285
Std Dev
0.001451
0.001422
0.001556
0.002769
0.002847
0.015996
0.00273
0.001839
0.003091
0.00825
0.004102
0.003018
The average of the means of the four groups of NO/CO2 track data for the Mustang and F-
150 from all three RSDs (Table 9c) was the same order of magnitude of the average of the three
groups of NO/CO2 dynamometer data (see Table 8). However, the data for the Mustang tended to
be higher (e.g., the Smog Dog™ 3 data averaged over 40 % higher than the corresponding
dynamometer data), while the data for the F-150 tended to be lower (e.g., the RSD-1000 data
averaged over 56 % lower than the corresponding dynamometer data). For the Caravan, the
pattern that occurred in the HC/CO2 data reoccurred. Specifically, the FEAT #3008 data averaged
about six times the corresponding dynamometer data, while data from the RSD-1000 and Smog
Dog™ 3 averaged over 150 % and over 350 % lower, respectively (both averages less than zero).
Table 9c. RSD Measurements at the Track - NO/CO9 Ratios
Vehicle
Ford
Mustang
II
Ford
F-150
Truck
Dodge
Caravan
Group
1
2
3
4
1
2
3
4
1
2
3
4
FEAT #3008
N
27
30
27
27
27
30
27
27
27
30
27
27
Mean
0.0144
0.020002
0.019366
0.017844
0.001477
0.003879
0.002939
0.000691
0.00357
0.021506
0.00162
-0.00188
Std Dev
0.018402
0.020642
0.015082
0.01694
0.015551
0.018353
0.015819
0.007877
0.01366
0.026348
0.025439
0.018392
RSD-1000
N
27
30
27
27
27
30
27
27
27
30
27
27
Mean
0.020481
0.014833
0.012556
0.010222
0.003111
0.001633
0.001963
-0.00033
-0.00033
-0.0032
-0.00252
0.003037
Std Dev
0.015875
0.014264
0.019025
0.016073
0.008345
0.016407
0.007793
0.00826
0.010058
0.017327
0.01458
0.013523
Smog Dog™ 3
N
22
24
25
27
23
25
27
26
19
24
21
21
Mean
0.019168
0.023333
0.021152
0.021189
0.004091
0.000056
0.000148
0.003135
-0.00351
-0.00082
-0.00084
-0.00597
Std Dev
0.007188
0.008344
0.005972
0.008319
0.013821
0.013567
0.012636
0.009825
0.008331
0.015213
0.016217
0.014484
28
-------
The inconsistency in all these data cannot be readily explained, but there were a number of
factors to take into consideration; e.g.,
(a) The grade of the track was very slight (almost flat), making it very difficult to have
some pressure on the accelerator without exceeding the 45 mph speed target. The F-
150 was the only vehicle with cruise control, but it was noted that the engine
revolutions per minute (RPM) were at a variety of levels as the vehicle passed the
RSDs.
(b) The height of the exhaust above the road, and the location of the exhaust (behind or to
the side) varied from vehicle to vehicle. Specifically, the F-150 exhaust was the
highest above the road, while the Caravan exhaust was located on the right rear of the
vehicle. The RSDs had to be adjusted several times to detect all three vehicles.
Experiment 6
Because all three of the RSDs did not always "see" a specific vehicle, the data had to be
screened to identify those vehicles for which emission data were available for all three RSDs. In
addition, because of uncertainty in data for each individual vehicle and RSD combination, a "one-
on-one" comparison between the data from, say, RSD X for Vehicle A to that from RSD Y for the
same vehicle is not meaningful. So, it was decided that the range of measurements by the three
RSDs of the "fleet," or entire sample of vehicles would be grouped in intervals, or "binned,"and the
distribution of the measurements displayed. The results of the road testing, then, are summarized
in Figures 5a-5c. The figures show the fraction of the fleet (on a logarithmic scale) which falls in
each interval of CO/CO2, HC/CO2, and NO/CO2, respectively.
The CO/CO2, HC/CO2, and NO/CO2 data each appear to have quite different distributions.
However, within a particular figure, the distributions of data for each RSD appear to be similar to
one another. The only data which appear to be distributed in a near-normal manner are those for
NO/CO2. Note that most of the fleet emissions, for both HC/CO2 and NO/CO2, are near zero.
29
-------
1:00
0.10
c
O
'•4—•
O
CO
CD
0
0:01
0:001
FEAT #3008
IRSD-IOOO
Smog Dog™ 3
< 0.01250 0.08750 0.16250 0.23750 0.31250 0.38750 0.46250 0.53750 0.61250 0.68750 > 0.76250
CO/CO2 Ratio [Interval Grouping]
Figure 5a. Distribution of CO/CO, emissions data for motor vehicles measured by RSDs at freeway entrance in Raleigh, NC.
-------
1.0000
0.1000
c
o
"o
03
0.02375
HC/CO2 Ratio [Interval Grouping]
Figure 5b. Distribution of HC/CO2 emissions data for motor vehicles measured by RSDs at freeway entrance in Raleigh, NC.
-------
0.1
o
CD
CD
CD
0.01
0.001
FEAT #3008
| RSD-1000
Smog Dog™ 3
< -0.045 -0.035 -0.025 -0.015 -0.005 0.005 0.015 0.025 0.035 0.045 > 0.055
NO/CO2 Ratio [Interval Grouping]
Figure 5c. Distribution of NO/CO2 emissions data for motor vehicles measured by RSDs at freeway entrance in Raleigh, NC.
-------
CONCLUSIONS
RSDs represent a method for collecting data on emissions from a large number of motor
vehicles in a relatively short period of time. The tests described in this report represent an attempt
to better understand the capabilities of RSDs and the quality of the data that they can provide.
The accuracy and precision of RSD measurements made in the puff tests (Experiments 1-3),
although showing a generally linear response over a range of CO, HC, and NO concentrations, did
not necessarily indicate how the RSDs would perform in the field. For example, the puff tests
showed that the precision and accuracy of the RSD data were in the following order:
CO/CO2 > HC/CO2 > NO/CO2
However, the average track data for each of the three vehicles (Experiment 5) when compared to
the corresponding dynamometer data (Experiment 4), show quite a different result. With the
exception of the CO/CO2 data for the Caravan, the accuracy of the RSD track data, using the
dynamometer data as the "standard," showed just the opposite result; i.e.,
NO/CO2 > HC/CO2 > CO/CO2
On the other hand, the track data showed a considerable amount of variation (Tables 9a-c).
Whether or not this was caused by the conditions at the site (e.g., the slight grade) or some other
variable is a question for possible future research. This could include testing of a small number of
vehicles at another location (on the same track or elsewhere) with a steeper grade, where a constant
speed could be more easily maintained. Unfortunately, what is unclear at this writing is the effect, if
any, that differences between individual sites have on the quality of the data collected.
Because of the experience at the track, the differences in readings by the three RSDs for
each individual vehicle in the road tests (Experiment 6) — where each vehicle was "seen" only once
by the equipment - were not surprising. This uncertainty in each individual measurement was the
reason for presenting the data the way that they are in Figures 5-7, above. Nevertheless, as noted in
the discussion for the road tests, the distributions of data for each RSD are similar. This supports
the conclusion that, if enough data are collected, they should serve as an adequate indication of the
fleet emissions at a particular location. In addition, if testing is conducted at enough sites in a
metropolitan area, the data should provide an adequate indication of that area's fleet emissions.
Finally, the results in this report represent an initial set of analyses. More detailed statistical
analyses of these data should provide additional understanding.
33
-------
REFERENCES
1. Stephens, R.D.; Cadle, S.H.; Qian, T.Z. "Analysis of Remote Sensing Errors of Omissions and
Commission Under FTP Conditions," J. Air & Waste Manage. Assoc. 1996, 46, 510.
2. Stephens, R.D.; Cadle, S.H. "Remote Sensing Measurements of Carbon Monoxide Emissions
from On-road Vehicles," J. Air & Waste Manage. Assoc. 1991, 41, 39.
3. Davis, L.D.; DuBose, R.S.; Walden, T.; Rodgers, M.O. "Estimating I/M Program
Effectiveness Through Analysis of I/M Records and Remote Sensing Results in Atlanta," The
Emission Inventory: Programs and Progress, Air & Waste Manage. Association, Raleigh, NC,
October 1995.
4. Fogelson, M.S.; Canon, M.B.; Grodzinsky, G.; DuBose, R.S.; Klochko, P.; Barker, T.;
Rodgers, M.O. "Comparison of Remote Sensing Results from Different Cities with Centralized
and Decentralized I/M Programs," In Proceedings of the AWMA/EPA Conference on Emissions
Inventory: Durham, NC, October 1995.
5. User Guide and Description for Interim Remote Sensing Program Credit and Utility, Office of
Mobile Sources, Ann Arbor, MI, EPA-420/R-96-004, September 1996.
6. Description and Documentation for Interim Vehicle Clean Screening, Office of Mobile Sources,
Ann Arbor, MI, EPA-420/P-98-008, May 1998.
7. Stedman, D.H.; Bishop, G.A.; Zhang, Y.; Guenther, P.L. "Remote sensing of automobile
emissions," in Traffic Technology International '94, UK & International Press, 1994, 194-198.
8. Peterson, I.E.; Stedman, D.H. "Find and fix the polluters, " Chemtek 1992, January 47.
9. Zhang, Y.; Stedman, D.H.; Bishop, G.A.; Beaton S.P.; Guenther, P.L.; McVey IF.
"Enhancement of Remote Sensing for Mobile Source Nitric Oxide," J. Air & Waste Manage.
Assoc. 1996, 46, 25-29.
10. Lawson, D.R.; Groblicki, P.J.; Stedman, D.H.; Bishop, G.A.; Guenther, P.L. "Emissions
from In-use Motor Vehicles in Los Angeles: A Pilot Study of Remote Sensing and the Inspection
and Maintenance Program," J. Air & Waste Manage. Assoc. 1990, 40, 1096.
11. Woods, M. "Remote Emissions Testing Hits the Road," Environmental Science and
Technology, Vol. 30, No. 6, 1996.
12. Hey wood, J.B. Internal Combustion Engine Fundamentals. McGraw-Hill: New York, NY
1988, 62-99, 567-65.
34
-------
13. Stedman, D.H.; Bishop, G.A.; Pitchford, M.L. Evaluation of a Remote Sensor for Mobile
Source CO Emissions; Advanced Monitoring Division, Environmental Monitoring Systems
Laboratory, Las Vegas, NV; EPA- 600/4-90-032 (NTIS PB91-148320), January 1991.
14. Energy and Climate Change, Report of the DOE Multi-Laboratory Climate Change
Committee. Lewis Publishers: New York, NY, 1990, p.2.
15. Energy and Climate Change, Report of the DOE Multi-Lab oratory Climate Change
Committee. Lewis Publishers: New York, NY, 1990, p.32.
35
------- |