United States        Air and Radiation        EPA420-R-02-030
           Environmental Protection                  October 2002
           Agency
vxEPA     On-Road Emissions
           Testing of 18 Tier 1
           Passenger Cars and
           17 Diesel-Powered Public
           Transport Busses
                                   > Printed on Recycled Paper

-------
                                                                   EPA420-R-02-030
                                                                        October 2002
                                                             1
              17
                         Assessment and Standards Division
                       Office of Transportation and Air Quality
                       U.S. Environmental Protection Agency
                               Prepared for EPA by
                             Carl Ensfield Sensors, Inc.
                        EPA Contract No. QT-MI-01-000659
                                     NOTICE

   This technical report does not necessarily represent final EPA decisions or positions.
It is intended to present technical, analysis of issues using data that are currently available.
        The purpose in the release of such reports is to facilitate the exchange of
     technical information and to inform the public of technical developments which
       may form the basis for a final EPA  decision, position, or regulatory action.

-------
USEPA STATEMENT OF WORK
Reference# QT-MI-01-000659
 ON-ROAD EMISSIONS TESTING OF 18 TIER 1 PASSENGER CARS AND 17 DIESEL
                  POWERED PUBLIC TRANSPORT BUSSES
                               Final Report
                                 V1.4
Carl Ensfield
Sensors, Inc.
October 22, 2002
Copywright 2002, Sensors, Inc.

-------
1   INTRODUCTION	4

2   TEST EQUIPMENT	5

  2.1    SEMTECH-D	5
    2.1.1    Heated Flame lonization Detector and Sample System	5
    2.1.2    CO and CO2 Analyzers	6
    2.1.3    NO and NO2 Analyzers	7
    2.1.4    Vehicle Interface (VI) Modules	7
    2.1.5    GPS	8
    2.1.6    Weather Probe	8
    2.1.7    Datalogger	8
  2.2    SEMTECH-G	10
    2.2.1    HC Measurement.	11
    2.2.2    Vehicle Interface	11
    2.2.3    Other components	12

3   DIESEL BUS TESTING	13

  3.1    BUSSES AND ROUTES	13
  3.2    ANALYZER SETUP AND OPERATION	14
    3.2.1    SEMTECH-D Vehicle Interface Setup	16
  3.3    FUEL ANALYSIS	17

4   GASOLINE PASSENGER VEHICLE TESTING	18
  4.1    VEHICLES AND OPERATORS	18
  4.2    ANALYZER SETUP AND OPERATION	19
  4.3    VEHICLE INTERFACE SETUP	21
    4.3.1    Load Factor	22
    4.3.2    MAP	23
    4.3.3    AC on/off Indicators	24
    4.3.4    Fuel Trim	24
    4.3.5    Lambda	24
    4.3.6    APT (All/Part Throttle)	24
  4.4    FTP CORRELATION TESTING PROTOCOL	24

5   DATA PROCESSING	25
  5.1    DATA SYNCHRONIZATION	25
  5.2    DIESEL ENGINE EMISSIONS CALCULATIONS	26
    5.2.1    Fuel-specific emissions	26
    5.2.2    Mass Emissions (grams/second) by Fuel Flow Method	26
    5.2.3    Fuel Mass Flow Rate and Fuel Economy	27
    5.2.4    Exhaust Flow Computation	27
    5.2.5    Engine Torque	28
    5.2.6    Brake Specific Emissions (grams/BHP-hr)	29
    5.2.7    Not to Exceed Zone	29
    5.2.8    Road  Grade	31
  5.3    GASOLINE ENGINE EMISSIONS CALCULATIONS	31
    5.3.1    Engine Airflow	31
    5.3.2    Exhaust Air/Fuel Ratio	32
    5.3.3    Fuel Flowrate and Fuel Economy	32
    5.3.4    Mass Emissions (grams/second)	32
    5.3.5    Distance Specific Mass Emissions (grams/Mile)	32
  5.4    QUALITY ASSURANCE	32
    5.4.1    Erroneous ECM data	33
    5.4.2    Vehicle Speed Validation	35
    5.4.3    GPSDropouts	36
    5.4.4    Fuel flow QA	37
Copywright 2002, Sensors, Inc.	

-------
    5.4.5   Erroneous zero Calibration	39
    5.4.6   Erroneous NO results	39
    5.4.7   FID flame outs	39
    5.4.8   Erroneous Throttle Position Data	40

6   BUS EMISSIONS SUMMARY	41

7   HEAVY DUTY DIESEL CORRELATION TESTING	43

  7.1    STEADY-STATE CORRELATION TESTING	43
  7.2    TRANSIENT FTP COMPARISONS	47
  7.3    ECM VALIDATION DATA	53
  7.4    OVERALL SYSTEM ACCURACY	54

8   GASOLINE VEHICLE CORRELATION TESTING	56

9   APPENDIX	59
Copywright 2002, Sensors, Inc.

-------
 1  Introduction
Sensors, Inc. has supplied gas analyzers and portable emissions testing systems worldwide for
over three decades. They currently supply over 75% of the worldwide market for OEM gas
analyzers for the Inspection & Maintenance (I/M) industry. In the past four years, Sensors,
Inc. has devoted considerable effort in the development of gasoline and diesel emissions
testing systems for in-use (on-road) applications. This new direction was greatly accelerated
when Ford Motor Company selected Sensors, Inc. as their development partner for
PREVIEW, an on-vehicle emissions tester for gasoline vehicles. This development spawned
three generations of equipment for Ford alone. Subsequently, Sensors, Inc. built on this
experience and its expertise to develop variants to the Ford gasoline system.  Current products
are named SEMTECH-G for gasoline powered vehicles, and SEMTECH-D for diesel
powered vehicles, and are commercially available.

In XXX, 2001, Sensors, Inc. was awarded a contract by the USEPA to conduct real-world,
on-road emissions measurements of 15 gasoline powered passenger cars and 15 heavy-duty
diesel vehicles.  The purpose of the contract was to generate on-road gaseous emissions data
to facilitate the development and evaluation of EPA's mobile emissions models.  The
passenger cars used in the study were primarily recruited from EPA personnel, although some
rentals were also used. The cars were  subject to Tier 1 emissions standards.  The Ann Arbor
Transit Authority generously provided city transit busses for the heavy-duty diesel on-road
data collection.

This report describes the test vehicles,  test equipment, validation testing, on-road test
procedures, and data reduction procedures. Analysis of the on-road data itself was not
performed by Sensors, Inc., and is not  covered in this document.

The SEMTECH systems used in this study rely on certain ECM data to compute mass
emissions results. Some of the exact equations and methodologies used to compute mass
emissions are considered proprietary to Sensors,  Inc., and are not disclosed in this document.
However, adequate data is provided to validate the calculated results.  Future models of
SEMTECH will include a direct exhaust measurement device as a redundant or primary
method of mass emissions determination.
Copywright 2002, Sensors, Inc.

-------
 2  Test Equipment
To support this study, two SEMTECH-G systems and one SEMTECH-D system were
produced. These were the first of the current generation of SEMTECH analyzers, which later
served as prototypes for commercial production systems. These systems are described in
detail below.

   2.1 SEMTECH-D
The SEMTECH-D prototype analyzer used in the study measures raw vehicle exhaust,
collects vehicle ECM data, and stores the data on an internal data logger. A post processing
utility computes real-time fuel-specific, distance specific, and brake-specific mass emissions.
                                            Dimensions:
                                            Weight
                                            Power:
                                            Power Consumption:
10"H x 19"W x 18"D
80 Ib
12 VDC
400W steady state
      Figure 2.1 SEMTECH-D Analyzer used in Study
The SEMTECH-D prototype analyzer consists of the following components:
       Heated FID Total HC analyzer
       195 C Heated Sample line, filter, and sample pump
       NDIR CO2 analyzer
       NDIR CO analyzer
       NDUV NO and NO2 analyzers
       Thermo-electric chiller and coalescing filters for sample conditioning
       Sample pressure control via proportional feedback control valve
       Vehicle Interface with J1587 (HD engine protocol) data acquisition.
       GarmanGPS
       PC104 Data logger with VenturCom ETS 10.0 RTOS
       Compact flash removable data storage
       2.1.1  Heated Flame lonization Detector and Sample System
The first consideration for proper sampling of any type of hydrocarbon is to minimize the loss
of hydrocarbons from the exhaust sample before they enter the cell or chamber where analysis
is performed. Some species of hydrocarbons characteristic to diesel exhaust have very high
condensation temperatures and will collect on the various surfaces of the sample system and
never make it to the analyzer. Some species of hydrocarbons found in diesel exhaust are also
soluble in water, so any condensation of water in the sample system will further remove HC
before it enters the analyzer. For these reasons, the standard practice for sampling of diesel
HC is to operate the entire system, including the analyzer, at 195 C.
Copywright 2002, Sensors, Inc.

-------
Through affiliations with several manufacturers of portable FIDs, Sensors has incorporated a
compact heated FID and sample system operating at 195 C in SEMTECH-D. This heated FID
has been used in commercially available portable FIDs designed for raw diesel exhaust
measurements for several years.  This same FID chamber is also used in laboratory
equipment. Once the sample is extracted from the exhaust stack, SEMTECH-D maintains the
diesel exhaust sample at elevated temperatures throughout the entire sample system and
analyzer with the use of special heated lines, pumps, and filters.

    •   Filtered  probe:  A  stainless steel sample probe integrated with a 2 micron stainless-
       steel mesh filter is  inserted into the exhaust stack of the vehicle. The filter removes
       the bulk of the particulates.
    •   Heated sample line: A 10 ft heated, insulated sample line operating at 195 C delivers
       the exhaust to the analyzer. The wetted surface of the line is teflon because of its high
       heat resistance and  low absorbing properties.  The teflon line is wrapped with a heater
       and molded inside a larger insulated flexible tube with a durable outer skin.
    •   Heated filter and pump: The heated sample line connects directly to a heated filter and
       pump operating at  191 C.  The heated filter is constructed of boro-silica  glass and
       removes  particulates as  low as 0.1  micron, and is accessible to the user  for easy
       replacement. The pump is specially designed with a 191 C heated head.
    •   Fittings and tubing: All fittings and tubing leading to the HC analyzer are maintained
       at  195 C and constructed of stainless steel.  Stainless steel has very low absorbing
       properties, so HC loss will be minimized.

       2.1.2 CO and C02 Analyzers
Like typical certification grade systems, CO and CO2 are measured on a dry basis with non-
dispersive infra-red (NDIR) technology. SEMTECH utilizes Sensors' second generation
Automotive Microbench (AMB-II) module for CO and CO2, which has been extremely
successful in the inspection and maintenance industry, with over 60,000 sold worldwide. For
SEMTECH products, the AMB-II NDIR analyzer was further enhanced in efforts to achieve
laboratory quality data in a portable application. After expanding the memory capacity and
processor speed, more sophisticated algorithms were  developed that accounted for second
order effects from cross interference of other gases and water vapor that were previously
ignored due to lack of processing power in earlier models.

The primary sample system concerns with NDIR are liquid water contamination or
interference from water vapor, and effects of operating temperature and pressure. To
eliminate the interference of water vapor, the sample is conditioned with a thermo-electric
chiller and two coalescing  filters after first passing through the heated filter to remove
particulates. To  eliminate  pressure effects, a feedback control to a proportional valve
maintains  a constant backpressure to the analyzers. Pressure fluctuations in the exhaust stack
or plugged filters will not alter the operating sample pressure. Finally, precise temperature
control is maintained for the NDIR detectors to eliminate ambient temperature effects.

One area of improvement needed for the CO analyzer in  diesel applications is the calibration
range.  The CO analyzer in the SEMTECH-D prototype unit for this study was designed for
gasoline exhaust, and is not yet optimized for low concentrations that are often found in diesel
exhaust. The analyzer is calibrated between 0.5%, or 5000 ppm and 8%.  However, the levels
measured on the Ann Arbor busses rarely approached the lower portion of this range.
Copywright 2002, Sensors, Inc.	

-------
Correlation testing of the SEMTECH-D prototype demonstrates a detection limit and
measurement uncertainty of 50 ppm (see Correlation Testing results).

       2.1.3  NO and N02 Analyzers
Chemiluminescence (CLD) is the most established technique for measurement of NO in
vehicle exhaust in stationary certification systems. It offers wide measuring ranges down to 1
ppm with a fast response time. However, there are several drawbacks for this technology that
makes it impractical for portable applications. The analyzer requires an ozone generator
(operating at high voltage), a vacuum pump, and a de-ozonizer to scrub out toxic ozone from
the analyzer vent gases. Because of these requirements, the analyzer is quite bulky and not
suitable for integration into a portable system. It is also likely that vibrations in a vehicle may
adversely affect performance of its photomultiplier tube. Since the CLD measures only NO
(and not NO2 ), it requires the use of a high-temperature catalytic converter to convert NO2 to
NO prior to measurement. The CLD  cannot measure NO and NO2 simultaneously.

Aqueous electro-chemical cells are one alternative commonly used for in-use testing, since
they are compact and very simple to operate. However, they suffer from high response times
(5-10 sees) and even higher recovery time.  They also provide relatively poor accuracy (± 25
ppm) especially in the low measurement range, and are known for poor reliability.  Like
chemiluminescence analyzers, electrochemical sensors measure NO and hence a catalytic
converter is necessary to monitor NOX.

In order to provide laboratory grade performance of NO and NO2 detection for in-use testing,
Sensors, Inc. developed a proprietary, non-dispersive ultra-violet (NDUV)  analyzer. Both NO
and NO2 have unique adsorption bands in the UV range, with little interference from other
gasses, and can be measured independently with this technology. Any interference (typically
from HC) is compensated with reference detectors. UV analyzers have not  commonly been
used for NOx measurements, primarily because  the UV lamps typically have limited life of
less than 1000 hours. However, Sensors has developed  a partnership with a company that has
developed an electrode-less UV lamp with a life of at least 20,000 hours, which makes it
practical for emissions analyzers.

Sensors, Inc.'s NDUV NO  and NO2 gas analyzer is unique because it offer simultaneous
monitoring of NO and NO2 in the exhaust gas. For SEMTECH-D, this analyzer has an
operating range of 0 to 3000 ppm for NO, and 0 - 500 ppm for NO2 with better than +/-10
ppm accuracy through these ranges.  The amplified signal from the NDUV optical  bench is in
the range of 1 mvolt/ ppm of NO with less than  1 ppm noise. This patent pending analyzer is
compact and easy to integrated into on-vehicle emission analyzers. Active temperature control
stabilizes the instrument at 60 C and provides immunity to changing ambient conditions. Its
steady state power consumption is less than 7 watt. As discussed in later sections,  correlation
testing of the NO and NO2 analyzers in the SEMTECH-D prototype analyzer shows
outstanding correlation to laboratory systems for both accuracy and response time.

       2.1.4  Vehicle  Interface (VI) Modules
SEMTECH-D utilizes a Nexiq Technologies (formerly MPSI)  Serial Data Module (SDM)
vehicle interface to retrieve real-time engine operating conditions during in-use testing. The
Copywright 2002, Sensors, Inc.	

-------
SDM module complies with SAE J1587 communications and hardware protocols for heavy-
duty diesel engines.  The data transfer rate is variable up to 10 Hz. The SDM module is
integrated into the SEMTECH system.  One end of a serial data cable connects through a
serial port on the SEMTECH, and the other end connects to the vehicle data port with
standard SAE J1587 connectors.

The SAE J1587 communications protocol defines the parameters that are available from the
ECM data link.  Unlike OBDII, the J1587 hardware continuously broadcasts all of the
available data. There is also much better consistency as to how the parameters are defined
than with OBDII. To retrieve the desired data, the SEMTECH PC 104  data logger/controller
communicates to the SDM module to tell it which parameters to retrieve. The SDM module
then intercepts the broadcast data, decodes it, and sends it back to the SEMTECH  data logger.
A complete list of parameters obtained from the diesel busses is provided in section 3.2.1.

       2.1.5 GPS
SEMTECH uses Garmin's Global Positioning System (GPS 25LP) to keep track of the route
taken by the vehicle under scrutiny. Garmin International is the industry leader in Global
Positioning System technology. This GPS module is suited for broad spectrum of OEM
system applications. Its compact design is ideal for applications with minimal space. It does
not require any user initialization. This module can simultaneously track up to twelve
satellites providing fast time-to-first-fix, one-second navigation updates and low power
consumption. It is also designed to withstand rugged operating conditions. SEMTECH
communicates with this module via RS-232 compatible bi-directional communication
channel.

The Garmin GPS has a resolution of one meter and an absolute accuracy of 15 meters for
latitude, longitude and altitude. However, the repeatability has been shown to be significantly
better at approximately 1-2 meters.  Because grade is computed with  the GPS data,
instantaneous results can have significant uncertainty,  especially at low speeds.  The grade
calculations  are described further in section 5.2.

       2.1.6 Weather Probe
A combination probe provides ambient pressure, temperature, and humidity to the data logger
each second. This information is used to compute the NOx humidity correction factor, Kh. It
also provides useful information about the test conditions.  The probe can be placed remotely
from the SEMTECH analyzer. However, the systems used in this study had the probes
mounted internally within the SEMTECH. For the bus data, windows were kept open in order
to provide ambient conditions to the analyzer. The gasoline vehicles were tested with the
analyzers enclosed in the trunk, so the temperatures may differ from ambient.

       2.1.7 Datalogger
The electronics and software for data logging and control provides the  ability to acquire, store
and transmit data, and also controls the  operation of sample system and other modules. The
core component is a CPU card with two asynchronous serial ports, parallel interface  and
integrated RAM. Individual support cards include a power supply, flash memory, digital I/O,

Copywright 2002, Sensors, Inc.	

-------
A/D, eight port asynchronous serial expansion, signal conditioning driver, and an Ethernet
controller card. PC-104 was the standardized format selected for SEMTECH, which was
designed for industrial applications and it lends itself for ease of packaging.
             Figure 2.2  Semtech Data Logger

The following briefly describes performance capability and/or capacity of each of the
modules selected for the current SEMTECH prototypes.

CPU Card: Industrial grade AMD-586 running at 133 megahertz. The processor card also
includes support for the parallel port and two EIA-232 serial ports. These can also supports
RS485 communications. PC-AT system architecture leverages x86 platform's popularity and
widespread compatibility.

Power Supply: Rugged industrial grade supply.  Input power can change from 8 to 30 volts
while maintaining +5, -5, +12, and -12 volt output.

Flash Memory:  Support for both Compact Flash and disk-on-chip technologies.  IDE
interface ensures future compatibility as memory densities increase.  Type I Compact Flash
cards with capacities up to  192 MB are readily available.

Digital I/O: Provides simple logic output for heater and pump controls. Unused channels can
be used for application specific purposes. This card supports up to 48 digital  channels.

Analog to Digital Converter: Sixteen channels with 12-bit resolution.  Used to measure
analog signals from thermocouples, pressure transducers, humidity sensors or other auxiliary
modules.

Eight Port Serial:  Expansion board adds an additional eight EIA-232 communication ports to
the system. This board incorporates eight individual UARTS each with a 32 byte FIFO.  The
deep FIFO memory helps eliminate bottlenecks often associated with serial communication.
Also, this card includes EIA-485 and EIA-422 communication support.

Ethernet Controller: Network card provides a standard lOBase-T connection  conforming to
IEEE 802.3 standard. This provides high-speed  link to control analyzer and access files
stored on the system. This card can easily be changed to support the faster lOOMbps speed
networks or to a 10Base-2 connection.

Copywright 2002, Sensors, Inc.	

-------
Signal Conditioning/Driver: Provides the amplification and thermocouple junction
compensation that precedes the analog to digital conversion.  The driver circuitry buffers the
digital I/O from the pumps and heaters.

The software for the data logging and control function runs on the hardware platform
described above. The design of the software architecture accommodates easy addition and
removal of hardware. A VenturCom ETS 10.0 real-time operating system (RTOS) provides
the foundation for the software system. It provides superior reliability and reduces the
overhead.  The operating system implements a subset of the Win32 API providing the
programmer with a well documented and extensively used tool set.

The SEMTECH control software itself was developed using the standard Microsoft Visual
C++ 6.0 compiler. The C++ compiler supports object oriented design methodologies, which
facilitate extensible design and long term maintenance. The software has been designed to
allow for maximum flexibility. Each device that connects to PC 104 stack has a corresponding
object in the software.  As devices are added and removed from the system, the software
dynamically configures itself to add or remove them. Another feature of the software design is
the wide set of communication options it supports to send the logged data to a host computer.
Since the ETS operating system incorporates a TCP/IP stack, data can be sent seamlessly over
numerous physical links. The  TCP/IP stack also provides the mechanism needed to send
information across the Internet.

   2.2 SEMTECH-G
The two SEMTECH-G prototype analyzers used in the study measure raw vehicle exhaust,
collect vehicle ECM data, and  store the data on an internal data logger that is automated by
key-on / key-off events. A post-processing utility computes real-time fuel-specific and
distance specific mass emissions based on engine airflow computed from the ECM data.
                                                  Dimensions:
                                                  Weight
                                                  Power:
                                                  Power Consumption:
10"H x 19"W x 18"D
45 Ib
12 VDC
100W steady state
    Figure 2.3 SEMTECH-G System used in Study
The two SEMTECH-G prototypes systems used in the study consist of the following
components:
   •   NDIR CO2 analyzer
   •   NDIR CO analyzer
   •   NDIR HC analyzer

Copywright 2002, Sensors, Inc.

-------
       Thermo-electric chiller and coalescing filters for sample conditioning
       Sample pressure control via proportional feedback control valve
       NDUV NO analyzer
       J1850 (OBDII) ECM data acquisition.
       Carman GPS
       PC104 Data logger with VenturCom ETS 10.0 RTOS and compact flash removable data storage
       2.2.1   HC Measurement
The primary difference between the SEMTECH-G and the SEMTECH-D analyzers used in
this study is in the measurement of HC. Since gasoline engines emit hydrocarbons primarily
in the form of hexane, which has strong infra-red absorbing properties, SEMTECH-G can
measure HC with NDIR technology instead of a FID.  While the NDIR technology cannot
measure total HC like the FID, there is a reasonable correlation between NDIR and FID
measurements as described below.

In addition, NC>2 from gasoline engines is generally considered negligible, and was not
measured. This results in significant differences in the sampling system for SEMTECH-G vs
SEMTECH-D.  Since hexane hydrocarbons do not condense at ambient temperatures,
SEMTECH-G does not require a heated sample system like SEMTECH-D. SEMTECH-G
pulls the sample through an unheated teflon sample line, then conditions the sample through a
thermo-electric chiller prior to analysis. Without the heated sample system, SEMTECH-G
consumes less than 100 Watts, which is significantly less than SEMTECH-D. This allows
continuous operation of the instrument for over 8 hours on a deep cycle 12V battery.

There is still a considerable difference in performance between NDIR hexane measurements
compared to a total HC measurement with a FID, as the correlation data shows.  That is
because the hexane HC species  only accounts for approximately 60% of the total HC during
an FTP test. That percentage is not exact for every vehicle or all driving conditions, but
provides a reasonable estimate of total HC over the test cycle.

Current versions of SEMTECH-G are available with a heated FID, if desired.

       2.2.2  Vehicle Interface
SEMTECH-G utilizes a Vetronix Enhanced OBDII Interface Module, with custom software
to allow proprietary parameter (PID) collection.  It supports SAE J1850 variable pulse width
protocol (VPW) for GM and Chrysler vehicles, and pulse width modulation (PWM)  for Ford
vehicles.  It also supports ISO 9141 protocol that is  commonly found on foreign-made
engines.  With this combination, all OBDII equipped vehicles are supported.

The Vetronix Enhanced OBDII Interface Module is integrated into the SEMTECH system.
One end of a serial data cable connects through a serial port on the SEMTECH,  and the other
end connects to the vehicle data port with standard OBDII connectors. To retrieve the desired
data, the SEMTECH PC 104 data logger/controller communicates to the VI module to tell it
which parameters to retrieve. Using the appropriate protocol, the VI module then sends a
request to the vehicle electronic control module (ECM) requesting the specific data.  The VI
module intercepts the responding data, decodes it, and sends it back to the SEMTECH data
logger. Each parameter is requested individually, one at a time.

Copywright 2002, Sensors, Inc.	

-------
The scope of the Statement of Work required a large number of PIDs that are proprietary to
the vehicle manufacturers.  Sensors, Inc. contracted Vetronix to develop custom interfaces in
order to access large numbers of proprietary PIDs at 1 Hz sample rates. However, not all of
the requested PIDs were available on GM and Ford vehicles, and Chrysler does not publish
information on their proprietary PIDs. A summary of the PIDs accessed by each vehicle is
provided in Table  4.2.

       2.2.3 Other components
Aside from the HC measurement and vehicle interface, all other analytical equipment is
identical between  SEMTECH-G and SEMTECH-D, including the data logger/controller,
GPS, NDIR CO and CO2, NDUV NO, and the weather probe. The only remaining
differences are in the data processing and computation of mass emissions, which are
described below.
Copywright 2002, Sensors, Inc.

-------
 3  Diesel  Bus Testing

   3.1 Busses and Routes
EPA approached the Ann Arbor Transit Authority (AATA) as a possible source for test
vehicles in the study, and were greeted with enthusiasm and complete cooperation. Not only
did AATA allow us to test the busses on an aggressive schedule, they provided technicians as
needed and dedicated drivers who performed whatever routes we requested. In addition,
AATA provided busses for on-road emissions testing demonstrations at EPA in Ann Arbor
and the SAE World Congress in Detroit.

The busses that were tested include 15 New Flyer models with Detroit Diesel Series 50
engines. All of these busses were of model year 1995 or newer.  In addition, a model year
2000 Gillig bus was tested with a Detroit Diesel Series 40 engine. Also, a model year 1992
New Flyer was tested with a 6V92 two-stroke engine that had  recently been overhauled and
upgraded to a DDEC4 engine control and a new emissions system including a reduction
catalyst. AATA was interested to know how the upgraded two-stroke engine compared to the
newer engines.  A summary of the busses tested is shown below, and a more detailed table is
provided in the appendix.
Bus#
1
2
3
4
5
6
1
8
9
10
11
12
13
14
15
16
17
Bus ID
BUS380
BUS381
BUS382
BUS383
BUS384
BUS385
BUS386
BUS379
BUS377
BUS363
BUS361
BUS375
BUS360
BUS372
BUS364
BUS404
BUS352
Model yr
1996
1996
1996
1996
1996
1996
1996
1996
1996
1995
1995
1996
1995
1995
1995
2000
1992
Odometer
223471
200459
216502
199188
222245
209470
228770
260594
252253
283708
280484
211438
270476
216278
247379
60000
206443
Engine series
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
SERIES 50 8047 GK28
S40E8.7LTA
6V92TAC/JWAC
Displ.
liter
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.7
9.1
Peak
torque
890
890
890
890
890
890
890
890
890
890
890
890
890
890
890
900
766
Test date
10/23/01
10/22/01
10/17/01
10/19/01
10/17/01
10/18/01
10/19/01
10/23/01
10/24/01
10/24/01
10/25/01
10/25/01
10/25/01
10/26/01
10/24/01
11/1/01
11/1/01
                              Table 3.1 Buses used in Study

Sensor's employees preformed the instrument setup and data collection for all the tests. All
of the bus tests consisted of approximately 2 hours of data collection during standard Ann
Arbor bus routes.  The drivers displayed an "out of service" message on the busses, but
stopped at all regular stops as if they were performing real routes. The routes were mostly
different for each test, and were selected for a wide variety of driving conditions.
Copywright 2002, Sensors, Inc.

-------
   3.2 Analyzer Setup and Operation
The physical setup for the bus testing was relatively simple, and took approximately 30
minutes for all of the busses.  The instrument warmup time was an additional 30 minutes
minimum (60 minutes for a cold start).  The following steps were involved in setting up an
operating the equipment:
       •   Place SEMTECH-D analyzer in rear of bus with 12V deep cycle battery.
       •   Attach power cables from the SEMTECH battery to the bus 12V equalizer. The bus power was
           used to keep the SEMTECH battery charged during the test.
       •   Turn on power to the SEMTECH to begin warmup.
       •   Run heated sample line out rear window to stack. Install exhaust probe in stack and secure line
           with a rail clamp.
       •   Attach J1587 vehicle ECM interface to the connector in the rear engine compartment. Verify
           communication by turning the bus engine control to "on" (without starting). Secure wires to bus
           chassis with tape. ECM PIDs are all consistent for J1587 protocols, so no PID setup was required
           after the first test.
       •   Mount GPS antenna on roof of bus.
       •   Insure that adequate FID fuel pressure is available in the portable bottle.
       •   Ignite FID when testing is ready to begin.
       •   Attach laptop computer, and zero the instruments.
       •   Perform single point gas audit with portable gas bottle.
       •   Manually start and stop data collections at approximately 30 minute intervals. This is to keep file
           sizes manageable, since VI data was collected at up to 10 Hz.
       •   Check zero between data collections.

Unlike the passenger vehicles, the heated components of the SEMTECH-D draw too much
power to operate for more than two hours on a battery.  Fortunately, the power output on the
diesel busses is sufficient so that the SEMTECH did not place any noticeable load on the
engine.  The steady-state power consumption of the SEMTECH-D prototype system was
approximately 400 watts.

The following photographs illustrate the SEMTECH-D setup on the busses.
            Figure 3.1  SEMTECH-D in back of Bus


Copywright 2002, Sensors, Inc.	

-------
       Figure 3.2 Vehicle Interface Connection in Rear Engine Compartment
       Figure 3.3 Heated Sample Line
Copywright 2002, Sensors, Inc.

-------
       3.2.1   SEMTECH-D Vehicle Interface Setup
The vehicle interface setup for the heavy-duty diesel engines is very simple, because the SAE
J1587 communications specification provides broadcast data instead of two-way
communications as required for OBDII communications. Also, the parameter availability and
definitions are more universally applied for heavy-duty engines, and it is unnecessary to
experiment with the PIDs to see what is available or what definitions are used. One set of
PIDs seems to apply universally for ECM equipped heavy-duty diesel engines sold in the U.S.
The following is the list of parameters that were collected during the Ann Arbor bus testing
through the vehicle interface:
Parameter
Engine speed
Vehicle Speed
Throttle position
Engine Load
Fuel flowrate
Fuel economy
Oil temperature
Oil pressure
Coolant temperature
Warning lamp
Units
RPM
MPH
%
%
gal/sec
MPG
DegF
KPa
DegF
binary
         Table 3.2 Heavy-Duty Vehicle Diesel Interface Parameters

One issue that arose during the testing was the large amount of data that was collected from
the ECM. Because the J1587 data is broadcast continuously at rates as high as 10 Hz, the test
files became large rather quickly. Because of this, new data files were started every 30
minutes during the 2 hour test, so a total of four files was generated per bus. The four files for
each bus were later combined into one file after post-processing. The post-processing
procedure eliminates the extra data by interpolating and synchronizing all the data to 1 Hz as
described in section 5.1.  Current SEMTECH-D analyzers store the ECM data into a buffer
and computes a moving average at user-defined intervals, so the user can specify the data
collection rate.

Another issue that arose was a significant amount of data errors from the ECM, including
random, non-physical data spikes. The source for the errors is unknown; they could have
been generated from external sources or from the ECM itself. This erroneous data was
filtered using techniques described in section 5.4.

For HD diesel engines, the Load Factor is defined as the fraction of maximum engine torque
for the current RPM. When combined with the engine lug curve (maximum torque curve),
real-time engine torque can be computed as described in section 5.2.  Also, NTE zone
operation can be determined with this information, as well as brake-specific emission. These
calculations are described in section 5 in more detail. The fact that fuel volumetric flowrate is
provided directly from the ECM is very  convenient for emissions calculations. However, an
accurate specific gravity measurement of the fuel is required.
Copywright 2002, Sensors, Inc.

-------
   3.3 Fuel Analysis
In order to compute mass emissions accurately using the reported fuel flow, the specific
gravity of the fuel was required.  Other fuel analysis data was required by EPA in addition. A
single sampling of the fuel from the AATA depot was provided at the beginning of the study,
and was analyzed at the EPA laboratories.  AATA personnel believed that the fuel properties
would not change significantly during the course of the two-week study.  The results of the
fuel analysis are summarized below:
AATA Fuel Sample
Analysis Parameter
Specific Gravity
API Gravity
Density
Sulfer content, ppm
Cetane index
ibp
t10
150
t90
ep
Result
0.8814
42.7
7.355
150
44.7
348
391
432
482
504
Copywright 2002, Sensors, Inc.

-------
 4  Gasoline Passenger Vehicle Testing

   4.1 Vehicles and Operators
Each passenger car was recruited and selected by EPA for the study.  Some of the vehicles
were owned by Sensors employees, some by EPA employees, and some were rented.  All of
the vehicles were of model year 1996 or newer, and certified to Tier 1 emissions standards
with the exception of one vehicle which was certified to California LEV standards.  A
complete table of the test vehicles is listed below. More detailed specifications for each test
vehicle is found in the appendix.
Vehicle #
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
model_yr
1998
1997
1996
1997
1998
1999
1999
1999
2000
1999
1998
1997
1998
1998
1996
1998
1998
1996
make
CHEVROLET
FORD
MERCURY
CHRYSLER
SATURN
CHEVROLET
SATURN
FORD
FORD
GEO
FORD
FORD
MERCURY
FORD
CHEVROLET
CHEVROLET
MERCURY
FORD
model_name
LUMINA LS
TAURUS GL
SABLE LS
CIRRUS LXI
SATURN
MALIBU LS
SATURN
ESCORT
ESCORT
PRIZM
TAURUS SE
ESCORT
SABLE GS
TAURUS SE
CAVLIER
CAVLIER
MYSTIQUE
SPORT
TAURUS GL
dispjiter
3.1
3
3
2.5
1.9
3.1
1.9
2
2
1.8
3
2
3
3
2.2
2.2
2
3
test_date
8/15/01
8/17/01
8/21/01
8/24/01
8/24/01
8/28/01
8/28/01
8/31/01
8/30/01
9/5/01
9/5/01
9/7/01
9/7/01
9/12/01
9/12/01
9/14/01
9/14/01
9/19/01
      Table 4.1 Gasoline Passenger Vehicles used in Study

During the study, the vehicles were all operated by their owners or renters with a SEMTECH-
G analyzer installed in their trunks. They were encouraged to operate their vehicles in a
normal fashion and on normal routes, in order to provide real-world emissions. The only
requirement was that they operate the vehicle for at least one hour with the SEMTECH-G
analyzer collecting data.  The operators were given a log sheet to record the time and distance
traveled for each trip. A trip was defined by an engine start and stop cycle.

The operators were given brief instructions and a test protocol handout, which is included in
the appendix. They were not required to perform any other duties, except to plug in the
automatic charger for the SEMTECH-G battery at night to keep it charged. The analyzer was
set up to operate continuously and automatically with no user intervention, as described
below.
Copywright 2002, Sensors, Inc.

-------
   4.2 Analyzer Setup and Operation

SEMTECH-G analyzers were installed in the vehicles by Sensors' employees. This
installation procedure took approximately 15 minutes, followed by a one hour warm up
period. There were two SEMTECH-G analyzers used in the study, so that two vehicles could
complete on-road testing simultaneously.  The physical setup included:

       •  Place SEMTECH-G analyzer in trunk of vehicle with battery and automatic charger
       •  Attach exhaust probe and sample hose
       •  Attach OBDII interface to vehicle connector
       •  Verify OBDII communication
       •  Set-up desired PIDs (paramenter IDs) specific to the vehicle.
       •  Allow one hour of warmup before data collection.

While the physical setup was very easy, setting up the desired PIDs proved to be difficult in
some cases because the definitions were not  always available from the vehicle manufacturer.
This is discussed in detail in section 4.3.

Because this study was conducted with the purpose of obtaining real-world emissions data,
the equipment was set up to operate automatically,  without any user intervention.  This
required the following:

       •  Continuous power and operation
       •  Automated zero calibrations with LED indicator
       •  Automated data collections with LED indicator

Power for the SEMTECH-G analyzers could have been supplied by the vehicle's generator
without adding significant load to the engine, but then the analyzer would have to be shut
down if the vehicle was off for long. For this study, the SEMTECH-G analyzers were
powered from external batteries with an automatic  charger attached. This way, the operator
could plug the charger in and keep the battery charged when the vehicle was not in use.
Typically, the operators would drive their vehicle home, plug in the charger, and let the
analyzer operate overnight.  The next day, they can simply unplug the charger, turn the key on
to initiate data collection, and then start the vehicle and drive away without having to warm-
up the instrument.

Like most instruments, the analytical instruments in the SEMTECH-G analyzers will drift
from their zero setting over time. This is corrected with a zero calibration, which was
initiated automatically by the SEMTECH data logger/controller every 30 minutes when the
instrument is not in  a data collection mode.  Ambient air is pulled through a sample port on
the instrument for the zero calibration. It first passes through a charcoal filter to remove trace
organic contamination before reaching the analyzers. However, this may not be sufficient if
there is  significant ambient hydrocarbon contamination. A green LED on the instrument and
also on a remote cable placed near the driver would illuminate steadily if a zero is in progress.
The process takes approximately one minute. If the instrument happened to be in a zero
calibration mode when the operator was ready to use the vehicle,  they were instructed to wait
for the zero process to complete before starting the engine.

Data logging was also automated through a key-on signal from the vehicle interface. When
the SEMTECH-G data logger/controller receives the key-on signal, it engages a solenoid that
switches the analyzer from the Ambient port to the Sample port, where the vehicle exhaust

Copywright 2002, Sensors, Inc.	

-------
sample hose is attached.  At the same time, it initiates a new data file on the data logger
which will capture all data from the analytical instruments, GPS, and vehicle interface. To
insure the data collection was initiated properly, the LED indicators would flash at 1 Hz.  The
operators were instructed to turn the key to the "on" position, wait for the LED indicator to
flash, then start the engine.  When the key is turned to the "off position, the data collection
would cease after an additional 15 seconds.  This additional time allows for the transport and
analysis of the exhaust sample at the time the key-off signal is  given.  It also allows for brief
interruptions in the ECM data, which can occur on some vehicles.
Copywright 2002, Sensors, Inc.

-------
   4.3 Vehicle Interface Setup
A significant amount of vehicle interface data was requested by EPA for the study, but not
everything was available. Many of the desired PIDs were proprietary to the vehicle
manufacturer, which caused additional difficulties because only GM and Ford provide the
required information to access these parameters. A list of parameters that was obtained for
each vehicle is shown below.
Parameter ID

Vehicle #
Used in Study

RFM

Vehicle speed
Load Factor type 1
Load Factor type 2
Load Factor type 3
MAF
Coolant Temp.
AC clutch on/off
AC high pressure
Inlet Air Temp.
Inlet Manifold Press.
Fuel Trim
Lambda
Spark Advance
Throttle position
All/Part throttle
MIL on/off
98 Lumina

1


X

X

X

X
X

X

X
X

X
X

X
w
I
H
f-
Os

2
X

X

X


X
X
X
X

X


X
X

X
X

"c3
CO
^0
OS

3


X

X


X
X
X
X

X


X
X

X
X
w
o
f-
Os

4


X

X

X





X
X
X


X

X
6
ts
CO
oo
OS

5
X

X

X
X



X
X

X
X
X

X
X

X
x>
1
OS
OS

6
X

X

X

X

X
X


X

X

X
X

X
6
ts
CO
oo
OS

7
X

X

X
X



X
X

X
X
X

X
X

X
t!
o
&
w
OS
OS

8


X

X


X
X
X
X




X
X

X
X
tt
o
&
w
o
o

9


X

X


X
X
X
X




X
X

X
X
N
PH
OS
OS

10


X

X
X






X
X
X




X
w
C3
H
oo
OS

11
X

X

X


X
X
X
X

X


X
X

X
X
t!
o
&
W
r--
OS

12
X

X

X


X
X
X
X




X
X

X
X

-------
       4.3.1   Load Factor
Engine Load Factor is critical to the computation of mass emissions, since it is used to
compute engine airflow. This method was first developed for specific Ford Motor Company
vehicles with the assistance of Ford personnel who provided information for their proprietary
engine load factor (referred to as type 3).

For other vehicles where proprietary ECM PID definitions were not identified, engine airflow
was computed using SAE J1850 Load Factor. The general SAE J1850 definition of Load
Factor is

                 ,„      Current Engine Airflow  ,_
              Load Factor =	xlOO
                           Reference A irflow

The reference airflow is loosely defined in SAE specifications as the "maximum engine
airflow". This appears to have been interpreted by vehicle manufacturers in two different
ways depending on vehicle Make and Model. Once the reference airflow is identified, the
current engine airflow can be computed by multiplying the reference airflow by the Load
Factor.

The variations of Load Factor are described below.  Sensors, Inc. has arbitrarily chosen to
refer to the various Load Factors as Type 1, Type 2, and Type 3.

   Type 1 Load Factor
   With a Type 1 Load Factor, the reference airflow is defined based on:

      •  Current RPM
      •  Reference inlet temperature
      •  Wide-open throttle
      •  Engine displacement
      •  Current volumetric efficiency

   This definition appears to be the most common, and is used on Ford (J1850) and some
   GM vehicles.  However, Ford also uses a proprietary Type 3 Load Factor as described
   below. There were five non-Ford vehicles in the study that used the Type 1 Load Factor.
   All were GM vehicles. You can identify type 1 load factor by its range under idle
   conditions, which is around 15 - 25% for the vehicles tested. The vehicles had to be
   tested first in order to make this determination.

   The only uncertainty with this Load Factor definition is the unknown volumetric
   efficiency. Volumetric efficiency can typically vary between 0.80 and 0.95 depending on
   the engine speed and other variables.  Because this information was not available, a fixed
   "average" volumetric efficiency had to be assumed. The value selected was 0.85 based on
   limited information.  The simultaneous FTP correlation tests confirmed that this value was
   appropriate.

   Type 2 Load Factor
   With Type 2 Load Factors, the reference airflow is defined based on:

      •  Reference RPM (fixed)
      •  Reference inlet temperature
      •  Wide-open throttle

Copywright 2002, Sensors, Inc.	

-------
      •  Engine displacement
      •  Volumetric efficiency at the reference RPM

   This definition seems less common, but was found on the single Chrysler vehicle and two
   of the GM vehicles.  Type 2 Load Factor typically have a range of 3 - 8% under idle
   conditions, which is significantly lower than type 1 and easy to identify (after a test is
   performed). Unfortunately, there were now two unknown parameters with these vehicles:
   the reference RPM and the volumetric efficiency at the reference RPM.  Since specific
   manufacturer definitions were available, these parameters had to be determined
   empirically based on the FTP correlation data.

   First, it was decided to again use an average volumetric efficiency of 0.85. The reference
   RPM was then determined empirically for each of the three vehicles from the FTP
   correlation testing. From these experiments, the reference RPM appears to be the "rated"
   RPM that provides the maximum engine power.

   Type 3 Load Factor (Ford Proprietary)
   In addition to the standard J1850 Type 1 Load Factor, Ford provides this third
   (proprietary) variation of engine load factor. It is similar to type 1 load factor, except that
   no assumption of volumetric efficiency is required. This means that there are no
   assumptions or "calibrations' that need to be performed before collecting data and
   computing results. There are probably similar parameters that exist for the other engine
   manufacturers, but they are unknown at this time. For all the Ford vehicles, the Type 3
   proprietary load factor was used for engine airflow computations.
As noted above, the Load Factor type had to be determined "after the fact" for GM and
Chrysler vehicles. Furthermore, for the three vehicles with Type 2 load factors, the reference
engine speed had to be determined empirically based on the FTP correlation data. Since some
of the SEMTECH-G calculations were first "calibrated" using the FTP data in this manner, it
may not be appropriate to consider the final comparison to FTP true correlation data, which is
typically collected and analyzed "blind". However, the purpose of this FTP testing was to
provide validation that the SEMTECH-G on-road data is accurate, especially since GM and
Chrysler vehicles had never before been tested by Sensors,  Inc.

Until all the engine manufacturers provide a definition of Load Factor type by make and
model, any vehicles that have not previously undergone correlation testing against a
laboratory reference should be validated in this manner. Because the Load Factor is a standard
OBDII parameter required by EPA for engine diagnostics, it seems very plausible that the
vehicle manufacturers could provide information as to which definition applies to each make
and model. If type 2 is used, then they should provide the reference RPM. With this
information, the VI method of computing engine airflow and mass emissions could be applied
without having to perform verification testing.

       4.3.2 MAP
Mass airflow sensors are used for fuel engine control strategies on many vehicles, along with
the lambda sensors.  Those that do not have MAF sensors rely on intake manifold pressure.
The MAF data was not used for the mass calculations in this study, but can be used as
redundant engine airflow information for comparison.  It should further be noted that units for
Copywright 2002, Sensors, Inc.	

-------
MAP may not be reported as specified in SAE J1850. It has been reported that some
manufacturers report voltage, rather than mass flowrate units.

       4.3.3 AC on/off Indicators
All the Ford vehicles and the two Saturns have an "AC clutch on/off parameter, that
indicates when the AC compressor is actually engaged. It does not simply indicate that the
AC dashboard switch is on or off. The Lumina and Cavaliers have an "AC High Side
Pressure" parameter that also indicates when the AC compressor is engaged.

       4.3.4 Fuel  Trim
The short-term fuel trim parameter is a J1850 parameter that gives an indication of
stoichiometry. The engine control always attempts to maintain a lambda of 1, so that there is
exactly the correct amount of fuel for complete combustion without any excess air. When the
short-term fuel trim is positive (say - .15), it indicates that the ECM is calling for 15% less
fuel in order to achieve a lambda of 1. From this, you can infer that the current lamda is 0.85.
This PID is not required if lambda is available directly from another parameter.

       4.3.5 Lambda
Ford provides direct, real-time lambda readings from the on-board sensors,  so the fuel-trim
parameter is not required.

       4.3.6 APT  (All/Part Throttle)
When Ford parameters are  retrieved using proprietary protocols, the J1850 "% throttle
position" PID is not  available.  Instead, the APT (all/part throttle) PID is available, which has
three states:  Fully closed, partly open, and fully open.

   4.4  FTP Correlation  Testing Protocol
In order to validate the test data, each vehicle underwent FTP correlation testing at the EPA
facility in Ann Arbor.  Full FTP tests were conducted with SEMTECH-G collecting data
simultaneously.

Each vehicle was prepared for full FTP testing at EPA's facility in Ann Arbor. The vehicles
were cold soaked overnight in a temperature controlled environment,  and the fuel was
exchanged with certification grade fuel.

SEMTECH-G data collections during FTP testing were conducted in the same manner as the
on-road data collection. The analyzer was allowed to warmup and stabilize before the test. A
zero calibration and  gas audit were also performed prior to FTP testing for each vehicle.

For FTP correlation  testing, additional HC data was collected with a portable FID from
Sensors, Inc.  This is the same FID model that is integrated  into the SEMTECH-D system,
and is available for SEMTECH-G as well.  It was desirable  to see how much improvement is
possible with a FID versus  the NDIR measurements.

Complete correlation test results are provided in section 8.


Copywright 2002, Sensors, Inc.	

-------
 5  Data Processing

   5.1 Data Synchronization
The first step in post processing is to precisely synchronize the raw data. Each of the
analytical  instruments, vehicle interface, and GPS equipment report data to the SEMTECH
data logger asynchronously and at differing rates, but with a timestamp at millisecond
precision.  The data is first interpolated to a common 1 Hz time interval. The first data point
reported by any of the instruments defines the first interval.

With all the raw data synchronized to the same data rate, it is then time aligned so that engine
data corresponds to emissions data in real time.  Data from each of the analytical instruments
is time shifted according to its delay in the system relative to the engine data from the ECM.
The delay for NDIR CO2 and CO is typically 6 seconds, accounting for transport time of the
gasses and response time of the analyzer. The delay for total HC from the FID and NDUV
NO and NO2 is typically 5 seconds.

For the diesel bus tests, time alignment between engine data and emissions data were verified
on most tests by plotting CO2 and fuel flow rate in real-time. A typical time-aligned data set
is shown below for one of the busses.

                          CO2 - Fuel Flow Time Alignment: Bus 17
                                        400
                                    Elapsed Time
                     Figure 5.1 Time Alignment of VI and Emissions Data

For the gasoline vehicle tests, this method of time alignment is only possible during the
engine start transients, since CO2 is relatively constant thereafter. However, NOx typically
correlates well to engine load and fuel flow rate, which provided additional evidence of
proper time alignment.  This method of time alignment is somewhat subjective and time
consuming, and not appropriate for large-scale data collections. More sophisticated
techniques should be developed as part of the development of in-use testing equipment where
proper time alignment is essential.
Copywright 2002, Sensors, Inc.	

-------
For diesel engine testing, the fuel-flow vs CO2 alignment strategy would be plausible through
software that can identify the best correlation between the two parameters over a short test
cycle by iteratively shifting the data until the best fit regression is found. For gasoline
vehicles, it may be plausible to inject a fixed amount of propane into the air inlet of the engine
while performing several snap accelerations. The propane would be diluted by the engine
airflow during the snap accelerations, so the computed engine airflow could be correlated to
the diluted propane measured from the exhaust. In order for this to work, the propane levels
would have to be significantly higher than the exhaust HC, and high enough so that catalyst
break-through would occur.  The data could then be aligned with software through a best-fit
regression analysis described above.

   5.2 Diesel Engine Emissions Calculations
The next step in the data processing is the computation of emissions results. The following
sections describe the various calculations performed for diesel emissions.

       5.2.1  Fuel-specific emissions
Fuel specific emissions are the mass fractions  of each pollutant to the fuel in the combusted
air/fuel mixture.  This fraction is easily computed directly from concentrations of the
measured exhaust constituents. No additional measurements are required.

 For example, to express NO fuel specific emissions, you start by computing the mole fraction
of NO to fuel burned.  This is simply the ratio of the measured concentration of NO to the
sum of the CO, HC, and CO2 concentrations in the exhaust, which reflect the number of moles
of fuel that is consumed per mole of exhaust.  The ambient CO2 concentration must be zeroed
on the instrument or subtracted from the exhaust measurement.  Ambient CO and HC are not
subtracted from raw exhaust concentrations because it is assumed these are destroyed in the
combustion process.

For example, the mass fraction of NO to fuel burned is then computed by multiplying the
mole fraction by the ratio of the molecular weights of NO to the molecular weight of the fuel.
This mass fraction is expressed as an equation below.  Fuel specific emissions for all other
species are computed in a similar manner.

             NO,
                *
                                 J+[C02J-[C02Jambieat   MWfud

There are several advantages of computing fuel specific emissions, particularly with diesel
engines. First, you do not need any additional measurements such as torque, speed, exhaust
flowrate, or fuel flowrate. You can completely characterize a vehicle operating under various
driving and loading conditions, and later compute mass emissions for a drive cycle by
applying an estimated or measured specific fuel consumption along with the fuel specific
emissions for various segments of the cycle.

       5.2.2  Mass Emissions  (grams/second) by Fuel Flow Method
Because today's diesel engines that are equipped with an ECU generally provide real-time
fuel flow information, the fuel flow method for mass emissions calculations is preferred. This
was the case with all of the Detroit Diesel engines  on the Ann Arbor busses.

Copywright 2002, Sensors, Inc. _

-------
With access to real-timea second-by-second fuel flowrate, transient mass emissions is
computed by multiplying these by the real-time fuel-specific emissions. This method has
been commonly used for steady-state emissions testing, when exhaust flow measurements are
not available.  CFR40 Part 86.345-79 describes the fuel flow method as an alternative for
mass emissions computations for diesel engine dynamometer testing.  Using NO for example,

             NO(g / sec) = NOfs (^^-Jx Fuelflow(g / sec)

Real-time mass emissions were computed for other species in a similar manner.

As discussed in section 5.4, the fuel flowrate was found to be unreliable at engine speeds
under 1000 RPM.  Under these conditions, two other methods were explored to compute mass
emissions.

       1.  Compute engine airflow using speed-displacement method. This seems reasonable
          at lower RPMs when there is no turbo boost or sharp transients.  Compute fuel
          flowrate using air/fuel ratio and engine airflow.
       2.  Estimate engine fuel flowrate using computed engine torque (see below), engine
          speed, and BSFC curve provided by the manufacturer.
These methods were compared, and found to yield reasonable similar results. Method 2 was
selected for mass computations below 1100 RPM. This is further explained in section 5.4.

       5.2.3  Fuel Mass Flow Rate and Fuel Economy
The mass flow rate of the fuel is critical to the calculation of mass emissions, as described
above. For FID diesel engines, the SAE J1587 protocol provides volumetric fuel rate data
(gallons/second) directly based on the fuel injector pulse width. To convert to a mass flow
rate, the fuel specific gravity is required.  Any error in the fuel specific gravity (SG) will  have
a matching impact on the mass emissions results. The fuel SG was measured at EPA
laboratories for the bus testing.

Fuel economy is easily computed for a test period by summing the fuel consumed and
dividing by the distance traveled. These results are provided as a 30 second moving average,
and for the entire test duration.

       5.2.4  Exhaust Flow Computation
Exhaust flowrate was not used to compute mass emissions in this study, but is still computed
for both diesel and gasoline engines. For  diesel engines, exhaust flowrate was back-computed
from the mass emissions generated with the fuel flow method. This is because engine airflow
cannot be computed from the diesel engine ECM data provided. Unlike gasoline engines,
there is no parameter that leads to  engine airflow.

This back-calculation also served as a useful tool for checking the time alignment of the data.
If the time alignment is off by any significant amount (2 seconds or more), then back-
calculated exhaust flowrates can be larger than possible during transients. Maximum engine
airflow for a given RPM is easily computed assuming maximum boost pressure for the
engine.
Copywright 2002, Sensors, Inc.

-------
       5.2.5  Engine Torque
Engine torque is computed on diesel engines by applying the Percent Load parameter with an
engine lug curve (maximum torque curve).  It is required for the computation of brake-
specific emissions as described below.  The Percent Load parameter is defined as
              %Load =
         Current _Engine _ torque
          Max _ Engine _ torque
where the maximum engine torque is defined at the current engine RPM.  The engine lug
curve, supplied by the engine manufacturer, defines the maximum engine torque for all engine
speeds.  The following chart and data table shows the lug curves used in the study for the
Detroit Diesel Series 50, 6V92, and Series 40 engines. Data at highest and lowest engines
speeds were not available, and had to be extrapolated.  These data points were necessary to
compute the NTE zone boundaries as described below. The original data sheets supplied from
the engine manufacturer is included in the appendix.

The accuracy of the torque computed from Percent Load and the engine lug curve is not well
established. Engine manufacturers cannot provide precise accuracy specifications for this
measurement, but generally agree it is within 10 percent of actual values.  Limited correlation
testing performed by Sensors supports this claim for torque levels above 30% of maximum
(NTE zone operation), but significant errors were observed at lower torque levels near idle
conditions. There  was no opportunity to perform correlation testing on the Detroit Diesel
Series 50 engines tested in this study. However, the small variance of the brake-specific CC>2
                          Detroit Diesel Engine Maximum Load Curves
    1000 i
    900 —
    800
    700
    600
    500
    400
    300
    200
    100
"Detroit Diesel 6V92
"Detroit Diesel series 40
"Detroit Diesel series 50
                         //I
                                                "V.
      400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800
                                      Engine RPM

emissions indicates good consistency of the computed torque for all of the 15 buses.
                         Figure 5.2  Engine Lug Curves used in Study
Copywright 2002, Sensors, Inc.

-------
Detroit Deisel Series 50
RPM
500
800
1200
1350
1500
1650
1800
1950
2100
2500
2700
Torque
335*
578
890
888
847
824
802
741
688
500*
300*
Power
31.9
88.1
203
228
242
259
275
275
275
238
154
Detroit Deisel Series 40
RPM
500
800
1200
1300
1400
1600
1800
2000
2200
2500
2700
Torque
244*
420
820
900
900
893
811
735
630
467*
250*
Power
23.2
64.0
187
223
240
272
278
280
264
222
129
Detroit Deisel Series 6V92
RPM
500
850
1200
1400
1500
1600
1700
1900
2100
2500
2700
Torque
195*
400
766
755
740
720
700
670
633
400*
200*
Power
18.6
64.8
175
201
211
219
227
242
253
190
103
                             Table 5.1 Engine Lug Curve Data
       * Extrapolated data
       5.2.6  Brake Specific Emissions (grams/BHP-hr)
To compute brake specific emissions, engine torque is first computed from ECM data and the
engine lug curve (maximum torque curve). Engine torque is then converted to engine
horsepower using RPM from the ECM. Work (BHP-hr) is computed for each second of the
test, and then summed over the desired interval. Brake specific emissions are reported as the
sum of the grams of pollutant emitted over the interval divided by the total work.

       5.2.7  Not to Exceed Zone
To determine emissions compliance of heavy-duty diesel vehicles, the "Not to Exceed Zone"
was established for each engine model tested.  The Not to Exceed Zone is  a subset of the
engine lug curve (maximum torque curve), bounded by minimum and maximum RPM levels
and minimum torque levels at each RPM. When operating in this zone for a minimum of 30
consecutive seconds, the average emissions are not to exceed 1.5 times the certification
standard according to the Consent Decree. The NTE zone is illustrated in  the figure below.
The upper RPM limit (UL) is defined as Nhigh, the point at which only 70% of the maximum
rated power is generated.  The lower RPM limit is called the 15% ESC speed, which is a
function of NIOW, the point at which 50% of the maximum power is generated, and the ESC
speed, which is the difference between NM and NIOW.

                       1 5% ESC Speed = NIOW + . 1 5(Nhi - NIOW),
Finally, the minimum torque is defined by 30% of the maximum rated torque, or by 30% of
the maximum power when the torque level at that condition is greater. A graphical depiction
of the NTE zone is shown below.
Copywright 2002, Sensors, Inc.

-------
                                  % ESC Speed
           a
           o
           H
                        'lo
                                                            70% Max Power
                                                            50% Max Power
                                                              30% Max Torque
                                                                30% Max Power
                               Engine Speed

                             Figure 5.3 NTE Zone Boudaries
The following tables show the NTE zone boundaries for the engines tested in this study.

30% Max Torque =
30% Max Power =
50% Max Power =
70% Max Power =

Nhi RPM =
Nlo RPM =
15% ESC Speed
Series 50
267
82.6
138
193

2607
973
1218
6V92
267
82.6
138
193

2607
973
1218
                                Table 5.2 NTE Boundaries
One important consideration is the curb-idle torque specification, which varies among
manufacturers.  Some report zero torque at the curb-idle condition, while other report actual
(estimated) torque. It is obvious which is the case during a test, but it is still important to
obtain the curb-idle torque specifications from manufacturers that omit it from the lug curve.
Sensors, Inc has already developed the software platform to enter a lug curve for an engine
and automatically indicate NTE zone operation during a test.

All of the Detroit Diesel engines report a Percent Load at curb idle, indicating that they
account for all loads on the engine, including alternator, A/C , hydraulics etc. The torque
computed using the above method is actually the torque at the crankshaft of the engine, rather
than the flywheel.
Copywright 2002, Sensors, Inc.	

-------
       5.2.8  Road Grade
Initially, the grade was to be measured with a liquid-level inclinometer. However, the vehicle
acceleration introduced too much error for this method to be used for in-use testing, even
when averaging over 10 seconds. Instead, the GPS data was used for grade calculation.
Velocity is provided as one of the GPS data strings in addition to positional data. Combining
the velocity at time t with the difference in altitude between time t and t-1 second, the
instantaneous grade is computed as shown below.  For this study, grade was computed on a 1
Hz basis without any time averaging or filtering.

                     _   ,    velocity•( ft / sec)
                     Grade, = -
                            altitude t — altitude^
Obviously, there was significant instantaneous error in this calculation given the uncertainty
in the GPS position. While the rated accuracy is 15 meters absolute, the repeatability in
altitude was found to be much better at 1 - 2 meters based on limited testing. Still, this can
generate significant errors, particularly at low speeds where there is less of a differential in
distance traveled over the one-second interval.

It was left to the modelers to use this data in a manner that they deem appropriate. It was
recommended that at least a 10 second moving average be applied to the raw data, after some
basic filter function is  applied to eliminate  singularities.  Future grade measurements can be
enhanced by more accurate GPS devices. Sensors, Inc. has already incorporated alternate
GPS products with significantly better accuracy in current production SEMTECH products.

   5.3 Gasoline Engine Emissions Calculations
Gasoline emissions calculations are computed somewhat differently than with diesel engines,
although many of the steps are similar.  The following sections describe the various
calculations performed for gasoline emissions.

       5.3.1  Engine Airflow
Since fuel flowrate is not directly available from the OBDII vehicle interface on gasoline
vehicles, it must be computed using engine airflow and air/fuel ratio. Section 4.3 describes
the parameters obtained from the vehicle interface. The engine Load Factor (Type 1, 2 or 3)
is used to compute real-time engine air flowrate in SCFM.  The SAE J1850 general definition
for Load Factor is
                          Current Engine Airflow
              Load Factor =	xlOO
                            Reference A irflow

As discussed in section 4.3, the reference airflow is defined as  the "maximum engine
airflow". This has been interpreted in a variety of ways, resulting in three types of load factor
identified.  See section 4.3 for further details. Complete equations for computing engine
airflow are proprietary for Sensors, Inc. and are not provided in this report.
Copywright 2002, Sensors, Inc.

-------
       5.3.2  Exhaust Air/Fuel Ratio
Engine air/fuel ratio can be computed in a number of ways. The vehicle interface provides
short-term Fuel Trim as a standard OBDII parameter, which can be used to compute lambda
as described in section 4.3 above. Also, Lambda is directly provided from certain proprietary
parameters based on the on-board sensors. However, there are limitations to this method.
Fuel-rich, open-loop conditions resulting from aggressive accelerations or engine starting are
often not reflected accurately with these parameters.  Also, lambda from the vehicle interface
does not account for mixing with the air volume in the tailpipe.  Actual exhaust air/fuel ratios
are much leaner at start conditions (due to mixing with the stagnant air) than indicated by the
ECM.  In fact, all transient exhaust air/fuel ratios are dampened by the mixing that takes place
in the exhaust system.

For these reasons, the exhaust air/fuel ratio is computed using exhaust analysis techniques. It
should be pointed out that the exhaust air/fuel ratio differs from the engine air/fuel ratio
during such transients for the reasons described. However, the engines operate at
stoichiometry for the vast majority of the time, when differences in lambda are negligible.
The exact equations for exhaust air/fuel ratio computation are not provided in this document,
but are widely available.

       5.3.3  Fuel Flowrate and Fuel Economy
For gasoline engines, fuel flowrate is generally not provided directly. Some manufacturers
provide pulse width of the fuel injectors, which could be use along with engine RPM to
compute fuel rate. In this study, gasoline fuel rate was determined from the computed engine
airflow and exhaust air/fuel ratio. Since the gasoline vehicles typically operated at
stoichiometry, the variable most important in the calculation is the engine airflow.

       5.3.4  Mass Emissions (grams/second)
Fuel-specific emissions and mass emissions are computed in the same manner as with diesel
engines.  The fuel flowrate is multiplied by the fuel-specific emissions.

       5.3.5  Distance Specific Mass Emissions (grams/Mile)
Distance-specific emissions  are computed for gasoline vehicles over a 30 second moving
window and cumulative over the entire test. The mass emissions are summed over the test
period and divided by the distance driven. The distance driven is computed from the vehicles
speed data from the vehicle interface.

   5.4 Quality Assurance
Because the SEMTECH products used in this study were prototypes, most of the data quality
assurance was performed manually. There were cases of improper data, primarily from the
ECM on some gasoline vehicles and many of the busses.  There were also some cases where
improper exhaust concentrations were recorded for reasons explained below.  All of the
anomalies were corrected if possible, or omitted from the data set.
Copywright 2002, Sensors, Inc.

-------
       5.4.1  Erroneous ECM data
There were many cases where certain engine parameters were well outside of physical limits,
particularly engine RPM and fuel flowrate on the busses. Because the errors were so great,
they were easy to filter out with the post-processor. The following filter limits were imposed
on the rate of change of RPM, fuel flow, and vehicle speed data:

       •  Rate of change limit for RPM = 10,000 (RPM)/sec  (both gasoline and diesel)
       •  Rate of change limit for Fuel flow =  .003 (gal/sec)/sec (diesel only)
       •  Rate of change limit for Vehicle speed =  21 (mph)/sec (both gasoline and diesel)

These filters will remove the data outside the defined limits. The SEMTECH post processor
automatically interpolates between the remaining data, and produces results at  1 Hz as before.
The reason the filters are applied to the rate of change of the parameters is because negative
spikes can also occur. Data could be erroneous because of unreasonable engine acceleration
or decelerations, and still be within reasonable absolute limits.

For example, Figure 5.4 shows erroneous engine speed data on bus 11,  where positive spikes
occur above 5000 and even 9000 RPM. Negative spikes also occur as pointed  out with the
arrows,  and RPM data is below the curb idle levels. Figure 5.5 shows the same test with the
RPM filters  applied. Both the positive and negative spikes are removed by the filters, and
RPM data is consistent and within physical limits.

Similar  results are achieved by applying the filters to the vehicle speed  and fuel flow data.
Most of the data was screened by manually plotting the ECM parameters and computed mass
results. However, there was no means to verify that all erroneous data was eliminated by these
filter settings for all the busses.  If any remaining data spikes exist in the data, then Sensors
can modify further optimize the post processing software and re-process the data.
Copywright 2002, Sensors, Inc.

-------
                                       Bus 11 Engine Speed, Unfiltered
     Erroneous positive
     and negative spikes
                                                Elapsed Time, sec
                                    Figure 5.4 Unfiltered Engine RPM
                                         Bus 11 Engine Speed, Filtered
                                              1000              1500

                                                  Elapsed Time, sec
                                     Figure 5.5 Filtered Engine RPM
Copywright 2002, Sensors, Inc.

-------
       5.4.2  Vehicle Speed Validation
Vehicle speed is a critical parameter from the vehicle interface that is easy to validate with the
GPS. In general, there was good agreement with the GPS on the passenger cars. The
following charts show this comparison for car 2. The slope of the regression line shows that
the speeds agree within 2.5% for this vehicle, although there is some scatter in the
instantaneous data at lower speeds. In addition, the GPS speed was validated manually using
highway mile markers for  10 miles, and found to be within  1% accuracy based on this limited
data.
                                        300        400

                                         Elapsed Time, sec
                       Figure 5.6 GPS vs ECM Vehicle Speed for Car 2
                                       ECM Vehicle Speed, MPH

                   Table 5.3 GPS vs ECM Vehicle Speed Correlation for Car 2

For the busses, the ECM vehicle speed was not always as accurate or reliable.  Figure 5.7
shows the GPS vs ECM comparison for Bus 1.  The regression analysis shows that the ECM
Copywright 2002, Sensors, Inc.	

-------
data is 10% high compared to the GPS. This comparison was not performed for all the buses,
but suggests that GPS data may be more reliable for on-road testing. This may not be
significant for diesel-powered vehicles since the vehicle speed is not used in the  mass
emissions calculations.
                               GPS vs ECM Vehicle Speed Comparison
                                        Bus 1, Trip 1
                                                                  	VEH_SPEED mph
                                                                  	GPS SPEED mph
                   200
                            400
                                    600       800       1000
                                         Elapsed Time, sec
                                                             1200
                                                                      1400
                                                                              1600
                         Figure 5.7 Bus 1 GPS vs ECM Vehicle Speed
Older (model year 1995) buses were equipped with an earlier version ECM that did not
provide vehicle speed.  This was also the case with the  1992 bus with the upgraded 6V92
engine. In these cases, the GPS velocity data was used in place of the ECM data.  This
affected buses 9,12,13, and 11.  This was a relatively easy correction to make, since the GPS
provides velocity data.  The post processor was modified to report the GPS velocity instead of
the ECM vehicle speed for these busses.

There were some erroneous vehicle speed data points (drop-outs) on several of the buses and
two suspect data points for car 9.  To catch bad data and correct it, an algorithm was
implemented that searches for unreasonable accelerations and deletes the suspect velocity data
points. The post-processor then interpolates a new velocity based on the surrounding good
points.
       5.4.3  GPS Dropouts
There were a few instances when the GPS lost communication with the satellite for unknown
reasons. When this occurs, the GPS signal is not necessarily reported as zero, but some fixed,
default position.  It is usually obvious in the data when this occurs. This type of error was not
prevalent, except for bus 11.  The erroneous GPS data for this bus was discovered by the
modelers, who had to remove it manually.

Copywright 2002, Sensors, Inc.	

-------
       5.4.4  Fuel flow QA
As discussed in section 5.2.4, the reported fuel flow on the diesel buses proved to be
unreliable at low levels, (below 1000 RPM). Fuel flows were checked at low power
conditions by comparing CC>2 mass emissions using the fuel-flow method with results based
on calculated engine airflow. Engine airflow can be computed with reasonable accuracy at
low RPM conditions (when the turbo is not engaged) using a simple displacement method
with an assumed volumetric efficiency of 0.85.

Figure 5.8 illustrates this point  for bus  10. At curb idle (600 RPM), we see that the mass
values based on the reported fuel flow method differ significantly from the mass values using
the exhaust flow method. That is because the fuel flow readings become erratic and often
approach zero below 1000 RPM at steady-state, and 1400 RPM on  decels.  When the bus was
set to "high" idle (1000 RPM),  the two methods match very closely, providing validation of
the fuel flow accuracy. At higher power conditions, the turbo-charger is engaged, and engine
airflow cannot be computed with the displacement method.
                             Mass Computation Comparison: AATA Bus 10
                               Fuel flow method vs Exhaust Flow method
              Exhaust flow at idle (< 1100 rpm) based on
              _ RPM and Displacement.

              Exhaust flow above idle back calculated from
              CO2 mass based on Wf
                                    200    250

                                    Elapsed Time
                          Figure 5.8 Fuel Flow Errors at Low RPM

To correct the inaccurate fuel flows at idle, there were two plausible choices. First, the engine
airflow method could have been used to compute mass emissions below 1400 RPM.  We
would still have incorrect fuel flow readings however. The second possible solution was to
use an alternate fuel flow that is computed based on the brake-specific fuel consumption
(BSFC) curve supplied by the engine manufacturer. This also proved to be a reliable
approach, and mass results matched well with the exhaust flow method at low RPMs.

Figure 5.9 shows the BSFC data from the engine manufacturer over the range of engine
speeds. In order to compute fuel flow from this data, the engine torque data is applied. At
Copywright 2002, Sensors, Inc.

-------
each second during the test, engine power is computed based on the real-time torque, and fuel
flow is computed using the BSFC value:

       Wf_BSFC (g/sec) =  Engine Power (BHp)/3600 x BSFC (gramfuel/BHp-hr)

where the BSFC value  is a function of the current RPM. A polynomial curve fit shown on the
chart was used to compute the BSFC values.
                                     Rated BSFC Curve
                           Detroit Diesel Series 50 and 6V92 Engine Models
       0.5


      0.48


      0.46


      0.44


    £  0.42
    Q.
    .C
    £1
    ra   0.4
    o"
    u.
    "*  0.38


      0.36


      0.34


      0.32
y = -9.8801 E-11x + 5.4762E-07X - 9.6020E-04X + 9.0411 E-01
       y = 5.07668E-08x^ - 1.47906E-04X + 4.16085E-01
                R2 = 9.91204E-01
       0.3
         600
                800
                        1000
                                1200
                                       1400     1600

                                           Engine RPM
                                                       1800
                                                               2000
                                                                      2200
                                                                              2400
                      Figure 5.9 BSFC Curves for Detroit Diesel Engines
Figure 5.10 again shows the ECM fuel flow data for bus 10 compared to fuel flow derived
from the BSFC curve. There is good agreement in the two methods at 1000 RPM, but the
BSFC-based flowrate generates mass results that match much better with the exhaust flow
method. Note the good agreement at higher flowrates as well.

The BSFC fuel flow also provides a good cross-check against the ECM reported fuel flowrate.
As you can see, there is good agreement at all conditions above idle.  This comparison was
performed on several other bus tests and was found to be favorable as well.

The solution that was selected to correct the ECM fuel flow was to use the BSFC-based fuel
flow anytime the RPM dropped below 1400. This provided more accurate mass emissions at
these conditions, and also provided more accurate fuel flow results.
Copywright 2002, Sensors, Inc.

-------
                               Fuel Flow Comparison for Bus 10
                           Reported Fuel Flow vs BSFC-based Fuel Flow
     «
                ECM fuel flow and BSFC-based
                fuel flow compare well above 1000
                RPM
         0      50      100      150     200     250     300     350     400     450     500
                                       Elapsed Time

                        Figure 5.10 Corrected Fuel Flows at Low RPM

       5.4.5  Erroneous zero Calibration
Car 12, trip 1 showed an improper zero calibration for HC. The readings were significantly
negative (-48 ppm) for nearly the entire first test, and then were OK on the subsequent test.
The other emissions were normal. This indicates that the analyzer was zero calibrated in the
presence of a significant HC source. A gasoline container in the garage could cause this
problem. The mass emissions are reported as zero for these negative readings, and users
should ignore this data.

       5.4.6  Erroneous NO results
One of the  SEMTECH-G systems had  an intermittent electronics failure on the NDUV NO
analyzer, causing it to occasionally lock up during a test for significant periods. Data was
manually removed in these cases. This only affected cars  12, 15, and 16.

The NDUV NO analyzer was not communicating for Car 13, so NO data was not available for
this vehicle.  The communications were re-established after cycling power to the analyzer. It
is unknown what caused the malfunction.

       5.4.7  FID flame outs
The FID flamed out several times during the bus testing either due to user error or because the
fuel ran out.  The data was manually removed in these cases:
       Bus 1: FID was off for first 6 minutes of trip 1.
       Bus 2: FID was off for entire 1st trip; FID was out during small portion of 2nd trip.
       Bus 4: FID was off for first 7 minutes.
       Bus 9: FID was off for small portion of trip 4 to change fuel bottle.
Copywright 2002, Sensors, Inc.

-------
       5.4.8  Erroneous Throttle Position Data
For the HD diesel bus data, the throttle position values from the ECM were converted
incorrectly.  Reported values need to be multiplied by 10.  This problem has been fixed in
current SEMTECH-D software.
Copywright 2002, Sensors, Inc.

-------
 6  Bus Emissions Summary
Because most of the buses were of the same model with the same engine, the road test data
itself was valuable for comparison.  Some statistics were computed that show good
consistency in results among the buses.

The results of the emissions testing for the 15 series 50 engines is shown in the tables below.
Table 6.1 is a summary of driving conditions, including elapsed test time, average bus speed,
total fuel consumed, total distance traveled, overall fuel economy, and frequency of NTE zone
operation. The frequency of NTE zone operation gives an indication of how the bus was
operated.
Bus#

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
DATE

10/17/01
10/17/01
10/18/01
10/19/01
10/19/01
10/22/01
10/23/01
10/23/01
10/24/01
10/24/01
10/25/01
10/25/01
10/25/01
10/26/01
10/26/01
Elapsed Time
seconds
3140
8461
8431
10347
7951
7295
8023
7888
8091
8069
5644
7858
8061
5283
5688
Average speed
mph
19.93
18.24
16.54
17.31
18.24
16.32
13.03
15.10
12.10
13.82
23.65
15.69
10.50
21.43
17.71
Total fuel
used
gal
3.46
6.99
6.85
8.46
6.88
6.63
7.80
8.24
7.85
6.48
5.81
8.31
6.99
5.12
5.30
Total Distance
Traveled
miles
17.4
42.9
38.7
49.7
40.3
33.1
29.0
33.1
27.2
31.0
37.1
34.2
23.5
31.4
28.0
NTE_ZONE
occurrences
% of total
34.0%
30.6%
25.4%
26.4%
27.0%
25.2%
17.0%
26.4%
20.9%
26.7%
50.8%
25.2%
19.3%
32.3%
29.7%
                          Table 6.1 Overall Trip Summary for each Bus

Table 6.2 shows the average emissions over the entire test in grams/bhp, as well as fuel
economy.  It also shows statistics for the 15 bus sample, including average, standard
deviation, and upper an lower control limits based on two standard deviations. The control
limits give an indication as to whether there were any significant outliers in the bus data.

The CO2 results are important to consider when evaluating the data quality. Typically, the
brake-specific CO2 emissions are consistent for a given engine family, regardless of the
operating conditions.  For the 15 Series 50 engines, the CO2 results were very consistent, with
a standard deviation of only 20.2 g/bhp-hr, or 4.4% of the average. This indicates that both
the bus test population is consistent, and the mass emissions results from SEMTECH are
consistent among the population. Fuel economy varied between 3.37 mpg and 6.39 mpg.
This is not expected to be as consistent as brake-specific CO2, since the route and driving
conditions can greatly affect the result.

Corrected NOx  emissions varied from 3.45 to 6.36 g/bhp -hr for the population, with and
average of 5.25  compared to a certification standard of 5.0 g/bhp -hr.  The standard deviation
was 0.89 g/BHp-hr, or 17%.  While these results fall within 1.5 x the certification standard,
Copywright 2002, Sensors, Inc.

-------
the NTE zone emissions must be evaluated to determine if the busses were in violation.
Average CO and THC emissions were well below the standards at 1.56 and 0.073 g/bhp-hr
respectively.
Bus#

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

average
Std dev
2SUCL
2SLCL
Fuel economy
mpg
5.03
6.13
5.66
5.88
5.85
4.99
3.72
4.02
3.46
4.78
6.39
4.12
3.37
6.15
5.28

4.99
1.03
7.06
2.92
CO2
g/bhp-hr
461.3
460.3
460.0
459.8
460.8
499.1
485.1
447.1
452.0
447.3
415.6
462.2
441.7
472.9
431.0

457.1
20.2
497.6
416.6
CO
g/bhp-hr
2.23
2.03
1.35
2.69
1.40
1.88
2.59
1.30
1.32
1.22
0.78
1.34
1.49
0.83
0.96

1.56
0.60
2.75
0.37
NOx
g/bhp-hr
7.39
7.27
6.81
7.13
7.23
5.64
5.32
4.61
3.58
3.52
7.35
7.02
6.84
5.27
6.59

6.10
1.36
8.82
3.39
KNOx
g/bhp-hr
5.95
5.86
5.53
6.08
6.36
4.94
5.13
4.52
3.45
3.69
6.13
5.74
5.48
4.36
5.51

5.25
0.89
7.03
3.47
THC
g/bhp-hr
0.056
0.077
0.098
0.084
0.097
0.033
0.053
0.077
0.026
0.015
0.057
0.059
0.093
0.118
0.157

0.0733
0.0369
0.1471
0.0000
     Table 6.2 Average Trip Emissions for each Bus and Statistics
Copywright 2002, Sensors, Inc.

-------
 7  Heavy Duty Diesel Correlation Testing
A heavy-duty chasis dynamometer test cell was not available to perform correlation testing on
any of the buses tested in the study.  However, considerable correlation testing has been
performed on SEMTECH-D in engine test cells at various locations.  Extensive correlation
testing of SEMTECH-D was performed at various engine manufacturers, including
Caterpillar, and two other anonymous customers (hereafter referred to as Customer 2 and
Customer 3). While this testing helped to fulfill requirements for the EPA Statement of Work,
the engine manufacturers provided this cooperation mostly for their own evaluation of
SEMTECH. The following is a summary of the correlation testing performed:

   February. 2002. Caterpillar
   •   Six  13-mode steady-state tests on C-10 engine compared to Horiba Mexa 7100.
   •   3 transient FTP tests compared to Horiba Mexa 7100.
   •   Correlation data is public.
   February. 2002. Customer 2
   •   34 steady-state points on undisclosed engine compared to Horiba Mexa 7100.
   •   FTP transient tests compared to Horiba Mexa 7100.
   •   Correlation data is public, but engine information is undisclosed.
   May, 2002. Customer 3
   •   75 steady-state test points compared to laboratory equipment.
   •   Same engine installed in HD truck, then tested on chassis dynamometer and on-road
       from Detroit to Portland,  Oregon.
While much of the test data has been made public, a significant portion has not been released
and is considered proprietary information.  Further, none of the raw data from the test cells
can be released to the public.

Most of the testing has been performed in engine dynamometer cells, and very limited ECM
data was available. Therefore, most of the direct comparisons are for emissions concentrations
only.  Even if the ECM data was available, it was not on the same engine family as tested.
Parameters that contribute to mass emissions (such as engine torque from the ECM) are
unique to each engine family, and cannot be directly validated in a sample study.

Still, there is extensive data to demonstrate the performance of the various SEMTECH-D gas
analyzers, and to quantify the mass emissions ucertainty due to concentration measurements.
When the other measurement uncertainties are determined, the overall system accuracy can
easily be determined through a root-mean-square (RMS) error analysis.

   7.1 Steady-State Correlation Testing
There have been numerous steady-state tests performed on SEMTECH-D at Caterpillar,
Customer 2, and Customer 3. In each case, raw exhaust emissions were compared  against
laboratory instruments during typical 13-mode testing as well as off-cycle points. Figure 7.1
through Figure 7.4 show regression charts for the steady-state correlation testing performed at
Caterpillar. On each chart is 78  data points, representing six 13-mode steady-state  tests
Copywright 2002, Sensors, Inc.	

-------
performed over a two-day period on a C-10 production engine. These results show good
agreement with the Horiba MEXA 7100, and are typical for direct concentration comparisons
performed at other facilities. The repeatability of the SEMTECH data is also demonstrated on
these charts.

Table 7.1 summarizes the majority of steady-state correlation testing performed at the various
facilities. The regression slopes, intercepts, and correlation coefficients are very near perfect,
and very consistent for the three data sets. All of this data has been independently verified by
the manufacturers who hosted the studies. Customer 2 and Customer 3 elected to keep the
engine family undisclosed, but are willing to release the regression statistics.

The standard error of each regression line was determined for each constituent in each study.
Multiplying this value by 2 gives an approximation of the 95% CI measurement uncertainty
for the full range of steady state concentrations. The standard errors were typically 17 ppm for
CO, 7 ppm for NOX, and 3  ppm for THC. The NOX standard error was higher (12 ppm) for the
Customer 2  study, but still well within acceptable limits.
                              SEMTECH-D CO2 Correlation to Horiba Mexa 7100
                                 Six 13-Mode Tests, Caterpillar C10 Engine
        o
        o
        m
        t£
        o
           9.00
           7.00
           5.00
           3.00
           0.00
                                               y = 0.998x + 0.078
                                                  R2 = 0.999
                    1.00     2.00     3.00     4.00     5.00     6.00     7.00     8.00     9.00     10.00

                                            SEMTECH-D CO2, %


                       Figure 7.1 SEMTECH-D Steady-State CO2 Correlation
Copywright 2002, Sensors, Inc.

-------
                                   SEMTECH-D CO Correlation to Horiba Mexa 7100
                                      Six 13-Mode Tests, Caterpillar C10 Engine
         o
         o
             400
             200
                                                           y=1.110x-12.248
                                                               R2 = 0.960
                       100      200      300      400      500      6(

                                                  SEMTECH-D CO, ppm
                                                                         700
                                                                                 800
                                                                                                 1000
         O
         I
             400
             300
             100
                           Figure 7.2 SEMTECH-D Steady-State CO Correlation

                                   SEMTECH-D NOx Correlation to Horiba Mexa 7100
                                      Six 13-Mode Tests, Caterpillar C10 Engine
                                                        y = 0.992x- 0.813
                                                             ^ = 0.999
                      NOx measurement is the sum of NO
                      and NO2, measured seperately in
                      SEMTECH.
^
                0       100      200      300      400      500      600      700      800      900      1000

                                                  SEMTECH-D NOx, ppm


                          Figure 7.3 SEMTECH-D Steady-State  NOX Correlation
Copywright 2002, Sensors, Inc.

-------
         o
         E
         o
         I
            400
            200
                                  SEMTECH-D THC Correlation to Horiba Mexa 7100

                                     Six 13-Mode Tests, Caterpillar C10 Engine
                                                                  y = 0.981x +2.683
                                                                     FT = 0.999
                      100
                              200
                                      300
                                              400
                                                      500
                                                                      700
                                                                              800
                                                                                              1000
                                                SEMTECH-D THC, ppmC





                         Figure 7.4 SEMTECH-D Steady-State THC Correlation
Copywright 2002, Sensors, Inc.

-------

Caterpillar
C-10 Engine
C02
CO
NOX
FIDTHC
Customer 2
Engine
CO2
CO
NOX
FIDTHC
Customer 3
Engine
C02
CO
NOX
FIDTHC
n

78
78
78
78

34
34
34
34

74
74
74
74
Slope

.998
1.110
.992
.981

.982
1.00
1.022
1.055

.963
1.02
.999
.73
Intercept

.078%
-12 ppm
-.81 ppm
2.68 ppm

-.076%
-20.9 ppm
-10.2 ppm
-1.8 ppm

.074%
-6.4 ppm
-5.4 ppm
5.9 ppm
Data Range

1.1 -9.4%
70-451 ppm
160-900 ppm
26-515 ppm

4.1 -12.7%
10-760 ppm
175 -1470 ppm
7-49 ppm

1 .4 - 9%
23-163 ppm
185 -1175 ppm
7-31 ppm
Correlation
Coefficient, r2

.999
.960
.999
.999

.999
.980
.999
.943

.999
.934
.999
.934
Std
Error

.067%
16.7 ppm
6.6 ppm
3.3 ppm

.083%
16.7 ppm
12.8 ppm
2.8 ppm

.054%
8.4 ppm
7.0 ppm
1 .4 ppm
                Table 7.1 SEMTECH-D Steady-State Correlation Results Summary

   7.2 Transient FTP Comparisons
     In addition to the steady-state testing, eleven transient FTP cycles were also performed
     in order to assess SEMTECH-D accuracy and repeatability under these conditions.
     Figure 7.5 through Figure 7.8 show SEMTECH-D transient raw emissions compared to
     a Horiba MEXA 7100 during one of the FTP tests. These charts are typical of all of the
     transient tests performed. While the transient response time of SEMTECH meets the
     requirements of the RFP, it is evident that the Horiba achieves higher peaks in some
     cases. However, the error of the integrated mass due to response time differences is
     negligible, as demonstrated below.

     In order to quantify the error associated with the SEMTECH-D transient concentrations,
     mass emissions results were calculated using engine airflow and engine torque as
     reported by the test cell instrumentation. Exhaust flow was computed by determining the
     chemical air/fuel ratio of the exhaust, and adding the fuel mass to the airflow. Also, dry
     concentrations were converted to wet using appropriate equations. The integrated brake-
     specific emissions results were then computed and compared to the Horiba MEXA 7100
     results. Figure 7.9 through Figure 7.12 show the transient mass comparison for one of
     the FTP tests. Since the MEXA 7100 uses the same data compute mass emissions, the
     differences are due  solely to the concentration measurements. Note that the overall
     integrated mass results were within 2.5% of the Horiba MEXA 7100 for all of the
Copywright 2002, Sensors, Inc.

-------
     constituents. The NOX mass result was an exact match, indicating that the response time
     differences average out over the cycle.

     Note that in the case of CO and CC>2, the integrated mass values using the SEMTECH-D
     concentrations were within 1% and 2% of the Horiba system respectively. This was
     consistent with other tests as well, as discussed below. This data demonstrates that the
     .83 Hz data rate for these analyzers is sufficient.
        s
        o
                                 SEMTECH-D CO2 Correlation to Horiba Mexa 7100
                                   FTP Transient Test 2, Caterpillar C10 Engine
                       200
                                  400
                                                        800
                                                                   1000
                                                                              1200
                                              Time, sec

                        Figure 7.5 SEMTECH-D Transient CO2 Comparison
Copywright 2002, Sensors, Inc.

-------
                                      SEMTECH-D CO Correlation to Horiba Mexa 7100
                                        FTP Transient Test 2, Caterpillar C10 Engine
            1600
            1400
            1200
         a.  1000
         o
         o
             800
             600
             400
             200
            1200
            1000
             800
             600
                                                  Time, sec

                                    Figure 7.6 Transient CO Comparison
                                      SEMTECH-D NOx Correlation to Horiba Mexa 7100
                                        FTP Transient Test 2, Caterpillar C10 Engine
                                   Figure 7.7 Transient NOX Comparison
Copywright 2002, Sensors, Inc.

-------
                                       SEMTECH-D THC Correlation to Horiba Mexa 7100
                                         FTP Transient Test 2, Caterpillar C10 Engine
             400
             350
             300
             250
             200
             150
             100
                                    Figure 7.8 Transient THC Comparison
            90
                                      SEMTECH-D CO2 Mass Correlation to Horiba Mexa 7100
                                           FTP Transient Test 2, Caterpillar C10 Engine
                                             Mass computed from test cell airflow
                SEMTECH-D Mass = 687 g/BHP-hr
                Horiba Mass = 696 g/BHP-hr
                Ratio = .
                                 Figure 7.9 Transient CO2 Mass Comparison
Copywright 2002, Sensors, Inc.

-------
             0.40000
         o
         o
             0.20000
             0.15000
                                        SEMTECH-D CO Correlation to Horiba Mexa 7100
                                         FTP Transient Test 2, Caterpillar C10 Engine
                                            Mass computed from test cell airflow
                                                         SEMTECH-D Mass = 1.40 g/BHP-hr
                                                         Horiba Mass = 1.37 g/BHP-hr
                                                         Ratio = 1.02
                         Figure 7.10 SEMTECH-D Transient CO Mass Comparison
             1200
             1000
             800
                                       SEMTECH-D NOx Correlation to Horiba Mexa 7100
                                         FTP Transient Test 2, Caterpillar C10 Engine
                                            Mass computed from test cell airflow
                                                          SEMTECH-D Mass = 3.86 g/BHP-hr
                                                          Horiba Mass = 3.86 g/BHP-hr
                                                          Ratio = 1.00
                                Figure 7.11  Transient NOX Mass Comparison
Copywright 2002, Sensors, Inc.

-------
                                SEMTECH-D THC Correlation to Horiba Mexa 7100
                                  FTP Transient Test 2, Caterpillar C10 Engine
                                    Mass computed from test cell airflow
           500
           450
           400
           350
           300
           250
SEMTECH-D Mass = 0.450 g/BHP-hr
Horiba Mass = 0.461 g/BHP-hr
Ratio = 0.976
           200
           150
           100
                                   400         600

                                           Time, sec
                     Figure 7.12 SEMTECH-D Transient THC Mass Comparison
     The other FTP tests demonstrate similar results. Figure 7.13 shows a control chart of the
     percent mass emissions error for each of the constituents for each of the eleven FTP
     tests. Three of the tests were performed over a two-day period at Caterpillar on a C-10
     engine. The other eight were all performed on a single Customer 2 engine over a 2-day
     period. All of the results are well within the required 10% error limit. Again, the errors
     shown are primarily attributed to the concentration measurement. These results
     demonstrate quantitatively the accuracy and repeatability of the concentration
     measurements over a transient cycle.

     This data also substantiates over the entire set of transient tests, that the data rate for the
     CO and CO2 measurements are sufficient at .83  Hz.
Copywright 2002, Sensors, Inc.

-------
                     SEMTECH-D Emissions Concentration Accuracy and Repeatability
                            Eleven FTP Tests at Caterpillar and International
           30.0
           10.0
          -30.0 J
                Mass emissions computed using SEMTECH-D
                concentrations and labarotory airflow and engine
                torque. Results compared to test cell values.
 HC

-NOx

-CO

-C02
                                 Demonstrates that mass emissions uncertainty due to
                                 concentration measurement are within 10% target.
                                                                         10
                                                                                     12
                                                Test#
             Figure 7.13 SEMTECH-D Transient Emissions Accuracy and Repeatability
     Table 7.2 summarizes the mass emissions uncertainties due to the concentration
     measurements. The standard deviation of the mass error was computed for each
     constituent; the 95% CI uncertainty was determined by doubling the standard deviation.
     As the table shows, the 95% CI uncertainty is less than 10% of the average measured
     value for each constituent. However, the uncertainty was 8.9% for HC and 6.4% for CO,
     which does not leave much room for other uncertainties in the system. However, the
     uncertainties are quite small relative to the current emissions standards.

Sample Size
Standard Deviation of Error, g/Bhp-hr
95% CI Uncertainty, g/Bhp-hr
95% CI Uncertainty % of average measured values
95% CI Uncertainty % of Current Emissions Standards
C02
11
11.14
22.28
3.55

CO
11
0.041
0.0826
6.42
0.5
NOX
11
0.087
0.1743
5.54
4.4
THC
11
0.008
0.0169
8.86
1.3
            Table 7.2 Transient Mass Uncertainty Due to Concentration Measurements

   7.3 ECM Validation Data
     It is not possible to validate all ECM data for all engine families with a small sample.
     However, we did collect some data during the correlation testing that is useful in both
     heavy-duty (SAE-J1708) ECM protocols and light-duty OBD II (SAE-J1850) protocols.
Copywright 2002, Sensors, Inc.

-------
     Figure 7.14 shows an example of the heavy-duty (SAE-J1708) ECM torque correlation
     to the engine dynamometer on the Caterpillar C-10 engine. The ECM torque was
     determined by applying the ECM Engine Load parameter, engine RPM, and the
     appropriate engine lug curve supplied by the manufacturer. At loads above idle, the
     ECM data appears reasonably accurate. However, the ECM torque is inaccurate at lower
     loads, and cannot be used under these conditions. That does not pose a problem for NTE
     zone testing. These results are typical of other correlation results on this engine, and
     consistent with on-road data observed in the bus study with Customer 3 engines.
                                ECM Torque Correlation to Test Cell Dynamometer
                              FTP Transient Test 2, Caterpillar C10 Engine, J1708 Protocol
          14001
          1000
       Figure 7.14 ECM Torque Correlation to Engine Dynamometer on Transient Test Cycle

   7.4 Overall System Accuracy
The SEMTECH-D correlation results presented above quantify the emissions measurement
uncertainty, with a 95% confidence interval, attributable to concentration measurements in
both steady-state tests and transient emissions tests. The SEMTECH-D measurement
uncertainty due to concentration is expressed in terms of concentration and brake-specific
mass emissions as describe above. However, this does not create the complete picture for the
overall system accuracy. To do this, we need to know the effect of the other primary
measurements that contribute to mass emissions computations. These include:

•  Either the fuel-flow data from the  ECM, or a direct exhaust flow
   measurement

•  ECM torque data.
Copywright 2002, Sensors, Inc.

-------
The 95% CI uncertainty of these measurements over the FTP cycle must be quantified in
order perform an error stack-up analysis. The error stack-up due to the three primary
measurements is computed using the root-mean-square (RMS) technique, as shown below.
         System _ error = ^ (concentration _ error)2 + (flow _ error]2 + (ECMtorque_ error]2

For example, if the error in integrated mass over the FTP cycle attributable to concentration
measurement uncertainty is 5%, and the error in integrated mass associated with flow
measurement uncertainty is 5%, and the error in integrated mass associated with the ECM
torque is 7%, then the overall system error would be 9.9%.
Copywright 2002, Sensors, Inc.

-------
 8  Gasoline Vehicle Correlation Testing
This section presents the result of the SEMTECH-G correlation testing compared to the FTP
bag and modal data from EPA.  The purpose of the correlation testing was to establish validity
of the SEMTECH-G emissions measurements, as well as mass emissions computations using
ECM data.

As discussed in section 4.3, GM and Chrysler vehicles had not previously been testing in this
manner by either EPA or Sensors, Inc., so the precise definition of some of the vehicle
interface parameters were unknown prior to the testing.  Since the FTP correlation tests were
the first test for these vehicles, this data was first used to validate the emissions computations
using new ECM parameters. In this sense, the data was not collected and processed in a
"blind" fashion as often prescribed for the most rigorous correlation analysis. In the three
cases where Type 2 Load Factors were identified, the FTP data was required to actually
"calibrate" the SEMTECH-G emissions calculation algorithms, by empirically determining
the reference engine speed used for the reference airflow.  It is up to the reader to determine
the validity of the correlation under these circumstances. Nevertheless, we are confident that
the mass emissions calculations for SEMTECH-G are correct based on the FTP data, and that
the on-road data is valid.

  8.1 FTP Composite Correlation Data
The composite FTP emissions results are shown in table Table 8.1. For the SEMECH-G data,
two files had to be manually combined into one since the vehicle was shut off between bags 2
and 3.  The SEMTECH automatically starts a new data file at key-on events.
Test
#

1
2
3
6
7
8
9
11
12
13
14
16
18
Vehicel Info
Disp
3.1
3.0
3.0
3.1
1.9
2.0
2.0
3.0
2.0
3.0
3.0
2.2
3.0
Model
Lumina
Ls
Taurus
GL
Sable
LS
Malibu
LS
Saturn
Escort
Escort
Taurus
SE
Escort
Sable
Taurus
Wagon
Cavalier
Tarurus
Year
1998
1997
1996
1999
1999
1999
2000
1998
1997
1998
1998
1998
1996
Semtech-G
CO2
398
419
412
386
299
273
292
401
349
408
457
353
387
CO
2.396
5.120
2.889
2.432
1.283
0.936
0.680
2.35
1.71
1.562
0.91
10.51
7
3.651
NO
0.243
0.797
0.434
0.369
0.736
0.181
0.061
0.260
0.342
0.107
0.20
0.483
0.626
HC
0.138
0.154
0.179
0.152
0.074
0.063
0.057
0.121
0.072
0.086
0.15
0.116
0.172
HC
FID


0.355

0.136

0.060
0.202
0.125
0.137
0.28
0.237
0.220
FTP Bag results
CO2
377
392
381
381
298
290
280
386
355
392
460
351
384
CO
2.275
4.750
2.882
2.270
1.283
0.946
0.699
2.224
1.685
1.460
0.955
9.890
3.434
NO
0.239
0.795
0.444
0.362
0.701
0.189
0.066
0.289
0.347
0.107
0.207
0.493
0.606
HC
0.184
0.244
0.362
0.190
0.137
0.100
0.061
0.209
0.133
0.163
0.260
0.276
0.231
FTP Modal Results
CO2
392
397
384
392
289
279
276
389
356
404
494.6
365
405
CO
2.344
4.535
2.813
2.250
1.219
0.974
0.635
2.224
1.651
1.428
0.993
10.27
2
3.542
NO
0.179
0.719
0.486
0.379
0.733
0.192
0.067
0.297
0.360
0.112
0.216
0.522
0.614
HC
0.096
0.247
0.335
0.183
0.148
0.088
0.064
0.222
0.133
0.175
0.297
0.325
0.261
                  Table 8.1 SEMTECH-G FTP Composite Correlation Results
Copywright 2002, Sensors, Inc.

-------
More detailed results are provided in the appendix, including distance traveled, fuel
consumption, and fuel economy comparisons.

As mentioned in section 4.4, additional HC measurements were made through the SEMTECH
system by attaching an external heated FID through an analog data connection.  The external
FID is identical to the one integrated in the SEMTECH-D system used in this study. It is also
available fully integrated into SEMTECH-G systems. It was desirable to evaluate the
performance improvement from using a heated FID instead of the NDIR HC measurement.

   8.2 Omissions
There were several tests omitted from the correlation results due to suspect data.

Vehicles 4 and 5 were omitted because the FTP modal CO2 results were more than 10%
greater than the bag data.  EPA indicated that a 5% tolerance was commonly used, however
even more tests would have to be omitted in that case.  There was no apparent reason for these
discrepancies, although EPA had noted that the particular test cell had not been used for
several years.

 One notable observation from the FTP modal data is that the exhaust flow, computed using
the CC>2 tracer method, is unstable and often drops to zero at idle  conditions. It is unknown if
this indicates a potential problem with the bag data also. An example is shown in figure
Figure 8.1. While this may have caused some of the bag vs modal disagreement on the tests
in questions, it should be noted that this behavior was observed on nearly all of the tests.
                          Exhaust Volume on Car 15 ('96 Chevy Cavalier)
                                    FTP Bags 1 and 2
  5
  in
     40.0
     35.0
     30.0
25.0
     20.0
     15.0
     10.0
                         200
                                  300       400       500

                                       Elapsed Time, sec
                                                            600
                                                                     700
                   Figure 8.1 FTP Modal Exhaust Volume Instability at Idle
Copywright 2002, Sensors, Inc.

-------
Vehicles 10 and 15 were omitted due to a very large number of data dropouts observed on the
FTP modal data. There was no known reason for the data dropouts, but they were far worse
on these tests than any other. There was no cause identified for these occurrences, and it was
agreed that these tests should be treated as suspect. Figure 8.2 shows the FTP data dropouts
for the CC>2 concentrations on car 10.  Data for car 15 was similar.
                         CO2 Concentration on Car 10 ('99 Geo Prizm)
                                   FTP Bags 1 and 2
    16.0 !
    14.0
    12.0
    10.0
                        400
                                600       800       1000

                                     Elapsed Time, sec


                              Figure 8.2 FTP Data Dropouts
                                                           1200
                                                                   1400
                                                                            1600
There was one additional omission: Car 17 was omitted because it failed a SEMTECH-G
post-test zero audit by a significant margin. It appears that the analyzer was accidentally zero
calibrated while sampling span gas or exhaust prior to the test. This could have been user
error, or a mechanical failure of a solenoid valve.

A post test zero check is standard procedure for on-road testing, and this would always
indicate a failure.

   8.3 US06 Correlation Data
The US06 emissions results were puzzling. Correlation to bag data deteriorated significantly
for both the SEMTECH-G and FTP Modal data.  While bag vs modal CO2 was generally in
agreement, other emissions showed significant discrepancies.  For example, on all but one of
the valid tests, CO emissions differed between 15% and 40% between bag and modal results.
SEMTECH-G results were similar.
Copywright 2002, Sensors, Inc.

-------
US06 test results are shown in table Table 8.2.
Test
#

1
2
6
8
9
11
12
13
14
16
18
Vehicel Info
Disp
3.1
3.0
3.1
2.0
2.0
3.0
2.0
3.0
3.0
2.2
3.0
Model
Lumina
LS
Taurus
GL
Malibu
LS
Escort
Escort
Taurus
SE
Escort
Sable
Taurus
Wagon
Cavalier
Tarurus
Year
1998
1997
1999
1999
2000
1998
1997
1998
1998
1998
1996
Semtech-G
CO2
402
370
386
279
294
386
317
369
425
347
364
CO
10.95
15.09
4.56
4.38
12.85
11.97
32.38
6.44
4.76
17.62
16.75
NO
0.217
1.079
0.400
0.715
0.142
0.348
0.966
0.180
0.230
0.880
0.639
HC
0.063
0.084
0.075
0.032
0.045
0.099
0.082
0.018
0.025
0.134
0.097
HC
FID




0.094
0.191
0.225
0.063
0.081
0.197
0.260
FTP Bag results
CO2
385
371
389
299
296
396
340
382
443
371
389
CO
13.91
19.24
6.85
6.33
13.30
16.05
34.78
7.00
5.01
21.26
24.35
NO
0.210
1.024
0.393
0.769
0.162
0.390
0.894
0.211
0.242
0.974
0.673
HC
0.150
0.224
0.116
0.045
0.104
0.209
0.268
0.076
0.087
0.206
0.280
FTP Modal Results
CO2
383
368
384
293
292
391
338
373
441.8
367
390
CO
8.57
14.78
4.29
4.73
9.69
12.43
31.33
4.94
3.11
18.68
18.29
NO
0.168
1.073
0.403
0.933
0.207
0.417
0.993
0.223
0.162
1.024
0.706
HC
0.108
0.145
0.082
0.037
0.082
0.150
0.296
0.062
0.060
0.225
0.203
                             Table 8.2 US06 Correlation Data
Two additional tests were omitted from the US06 results.  Car 7 was omitted simply because
the US06 cycle was not performed on this vehicle due to time constraints. Car 3 was omitted
because EPA results showed an unrealistic 27 mpg fuel economy for the 3.0 liter Ford Taurus.
This was a 22% higher fuel economy than measured on the FTP. All other tests showed a
noticeable decrease in fuel economy with the US06 cycle.
Copywright 2002, Sensors, Inc.

-------
    Appendix
    Statement of Work documentation
    Documentation for test procedures
    Test file database
    Correlation test charts-gasoline
    Correlation test charts - diesel
    Diesel engine spec sheets
Copywright 2002, Sensors, Inc.

-------
The various gas analyzers have different types of calibrations and span
adjustments. The following describes the general approach for each.

Heated FID:  The FID is fully linear, so that all concentrations are
interpolated from a single point calibration and zero offset.  This it
typical of most FIDs.   We used a single range of 0 - 1000 ppmC during the
bus study, but the analyzer has other ranges available.  Each range has its
own zero offset, but all ranges scale the concentration from a single point
span on one of the ranges.  There is an option to account for any non-
linearity by manually adjusting the calibration at various concentrations,
but we have never found enough error to bother with this.

NDIR C02 and CO:  These are both single range devices.  The C02 range is 0
- 16% and the CO is 0  - 8%.  There is a "Factory" multi-point calibration
curve that is hard-coded into each analyzer when it is first made.  This
factory calibration curve is typically never changed.  This is a 4th order
polynomial, I believe, but I will have to verify.  To account for short
term drift and linearity changes, the user performs periodic "span
adjustments", or "user calibrations".  A single point user-calibration will
adjust the entire factory calibration curve to match the gas bottle at a
particular concentration  (typically near the maximum of the range) while
holding the zero point fixed.  A two-point user calibration uses a high
concentration gas to adjust the high concentration range, and a lower
concentration gas to adjust the mid-range if the linearity does not meet
specification. This increases or decreases the midpoints of the calibration
curve, while holding both the zero and upper span points fixed.

NDUV NO and N02.  These are also single range devices.  The NO range is 0 -
3000 ppm, and the N02  is 0 - 500 ppm.  The calibration is exactly the same
as for the NDIR analyzers.  There is detailed, multi-point factory
calibration hard-coded into the analyzer.  Again, this is a 4th or maybe
5th order polynomial  (higher orders are not significantly more accurate).
Single or two-point user calibrations adjust the factory calibration curve
as described above.  The NDUV analyzers are typically much more linear than
the NDIR analyzers, and almost never need more than a single point span
adjustment.

02:  This is a galvanic cell that produces a voltage that is linearly
proportional to the 02 concentration.  No zero offset is typically
required.  A single-point calibration using ambient air or bottled air is
sufficient.


Audit/Calibration procedures used during the bus study:

At a minimum, the SEMTECH-D was audited at the beginning of each day of
testing during the bus study using a BAR-97 Low quad-blend carried in
Scotty III portable bottles with the following concentrations:
C02:  6%
CO:  0.5%  (5000 ppm)
NO:  300 ppm
HC:  200 ppm propane

A data file was recorded during the audits.  If an audit failed, then a
single-point user calibration was performed on the analyzer that failed.  A
3% audit limit is typically used for C02 and CO, and 2% for NO and THC.
These evaluations were performed visually during the study. N02 required a
full gas cylinder, and was therefore only audited on several occasions when
the analyzer was briefly returned to Sensors.
Copywright 2002, Sensors, Inc.	

-------
We did not keep records on the frequency of calibrations that were required
during the study, but I can say that the C02, CO, and NO required single
point adjustments on only a few occasions.   The FID required a single
point calibration more often, nearly every day.  That is not unusual for
FIDs, however, especially when routinely cycling power as we did.

When a single-point user calibration was required for C02, CO, or NO, a
BAR97 High quad blend was used (also carried in a Scotty III portable
bottle)  with the following concentrations:

C02:  12%
CO:  8%
NO:  3000 ppm
HC:  3000 ppm propane

When a single-point calibration was required for the FID, the BAR97 Low
bottle was used, since it better matched the operating range we were
interested in.
For the gasoline car tests, the SEMTECH-G analyzers were audited in the
same manner after they were installed in the vehicles and allowed to warm
up for approximately 1 hour.  The same portable bottles were used for the
audits and calibrations.

Note:  Since the bus study, we have switched to different audit and
calibration blends that better match the range we observe in diesel
exhaust.  Our current bottles are:

SEMTECH-D Audit:
C02:  6%
CO:  200 ppm
NO:  300 ppm
HC:  50 ppm propane

SEMTECH-D Calibration:
C02:  12%
CO:  1000 ppm
NO:  1500 ppm
HC:  200 ppm propane
Copywright 2002, Sensors, Inc.

-------