December 2004
Environmental Technology
Verification Report
ENVIRONICS USA INC.
M90-D1 -C CHEMICAL WARFARE
AGENT DETECTOR
Prepared by
Battelle
Batteiie
Inn Business of Innovation
Under a contract with
«EPA
U.S. Environmental Protection Agency
-------
December 2004
Environmental Technology Verification
Report
ETV Safe Buildings Monitoring and Detection
Technology Verification Program
Environics USA Inc.
M90-D1-C Chemical Warfare
Agent Detector
by
Kent Hofacre
Tricia Derringer
Dale Folsom
Thomas Kelly
Loraine Sinnott
Zachary Willenberg
Battelle
Columbus, Ohio 43201
-------
Notice
The U.S. Environmental Protection Agency (EPA), through its Office of Research and
Development, has financially supported and collaborated in the extramural program described
here. This document has been peer reviewed by the Agency. Mention of trade names or
commercial products does not constitute endorsement or recommendation by the EPA for use.
11
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the
nation's air, water, and land resources. Under a mandate of national environmental laws, the
Agency strives to formulate and implement actions leading to a compatible balance between
human activities and the ability of natural systems to support and nurture life. To meet this
mandate, the EPA's Office of Research and Development provides data and science support that
can be used to solve environmental problems and to build the scientific knowledge base needed
to manage our ecological resources wisely, to understand how pollutants affect our health, and to
prevent or reduce environmental risks. EPA also addresses responsibilities in homeland security
through the National Homeland Security Research Center, by means of research programs in
Drinking Water Security, Safe Buildings, and Rapid Risk Assessment.
The Environmental Technology Verification (ETV) Program has been established by the EPA to
verify the performance characteristics of innovative environmental technology across all media
and to report this objective information to permitters, buyers, and users of the technology, thus
substantially accelerating the entrance of new environmental technologies into the marketplace.
Verification organizations oversee and report verification activities based on testing and quality
assurance protocols developed with input from major stakeholders and customer groups
associated with the technology area. ETV consists of six verification centers. Information about
each of these centers can be found on the Internet at http://www.epa.gov/etv.
The ETV approach has also been applied to verification of homeland security technologies. The
verification reported herein was conducted by Battelle as part of the Safe Buildings Monitoring
and Detection Technology Verification Program, which is funded by EPA. Information
concerning this specific environmental technology area can be found on the Internet at
http://www.epa.gov/etv/centers/centerl 1 .html.
111
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the
verification test, analyze the data, and prepare this report. We also would like to thank
Donald Stedman of the University of Denver and Bernadette Johnson of the Massachusetts
Institute of Technology for their reviews of this report.
IV
-------
Contents
Page
Notice ii
Foreword iii
Acknowledgments iv
List of Abbreviations viii
1 Background 1
2 Technology Description 2
3 Test Design and Procedures 4
3.1 Introduction 4
3.2 Test Design 5
3.2.1 Chemical Test Compounds 5
3.2.2 Test Matrix 5
3.2.3 Test Locations 6
3.2.4 Test Sequence and Schedule 7
3.2.5 Reference Methods 7
3.2.6 Interferents 9
3.2.7 Materials and Equipment 11
3.3 Test Procedure 14
3.3.1 Response Time 15
3.3.2 Recovery Time 16
3.3.3 Accuracy 16
3.3.4 Repeatability 16
3.3.5 Response Threshold 16
3.3.6 Temperature and Humidity Effects 17
3.3.7 Interference Effects 17
3.3.8 Cold-/Hot-Start Behavior 18
3.3.9 Battery Life 18
3.3.10 Operational Characteristics 18
4 Quality Assurance/Quality Control 20
4.1 Equipment Calibration 20
4.1.1 Reference Methods 20
4.1.2 Instrument Checks 21
-------
4.2 Audits 21
4.2.1 Performance Evaluation Audit 21
4.2.2 Technical Systems Audit 21
4.2.3 Data Quality Audit 22
4.3 QA/QC Reporting 22
5 Statistical Methods 23
5.1 Statistical Analyses 23
5.1.1 Analysis of Response, Response Time, and Recovery Time 24
5.1.2 Analysis of Accuracy 25
5.1.3 Analysis of Repeatability 25
5.1.4 False Negatives and Positives Analysis 26
5.1.5 Analysis of Response to Oscillating Concentrations 26
5.2 Other Analyses 27
6 Test Results 28
6.1 Response Time 29
6.2 Recovery Time 30
6.3 Accuracy 30
6.4 Repeatability 31
6.5 Response Threshold 31
6.6 Temperature and Humidity Effects 32
6.7 Interference Effects 32
6.8 Cold-/Hot-Start Behavior 33
6.9 Battery Life 34
6.10 Operational Characteristics 34
7 Performance Summary 36
8 References 38
Appendix A. Detailed Statistical Analysis Results
VI
-------
Figures
Figure 2-1. Environics USA M90-D1-C CW Agent Detector 2
Figure 3-1. Planned Sequence of TIC Verification Tests 8
Figure 3-2. Logic Diagram for Determining TIC/CW Agent Test Sequence 9
Figure 3-3. Test System Schematic 13
Tables
Table 3-1. Target TIC and CW Agent Challenge Concentrations 5
Table 3-2 Summary of Evaluations Conducted on the M90-D1-C 6
Table 3-3. Test Schedule 9
Table 3-4. Primary Reference Methods 9
Table 3-5. Test Concentrations for Interferents 10
Table 3-6. Temperature and Relative Humidity Conditions 12
Table 4-1. Performance Evaluation Audit Results 21
Table 6-1. GB Test Results 29
Table 6-2. Response Threshold Data 31
Table 6-3. Interference Effects Data 32
vn
-------
List of Abbreviations
AC hydrogen cyanide
CW chemical warfare
DEAE N,N-diethylaminoethanol
EPA U.S. Environmental Protection Agency
ETV Environmental Technology Verification
FID flame ionization detection
FPD flame photometric detection
GB sarin
GC gas chromatography
HD sulfur mustard
HMRC Hazardous Materials Research Center
IDLH immediately dangerous to life and health
IMS ion mobility spectrometer(ry)
L liter
L/min liter per minute
LCD liquid crystal display
|-ig/m3 microgram per cubic meter
I^L microliter
mg/m3 milligrams per cubic meter
mL milliliter
mm millimeter
PE performance evaluation
ppb part per billion
ppm parts per million
ppmC parts per million of carbon
psig pounds per square inch gauge
QA quality assurance
QC quality control
QMP quality management plan
RH relative humidity
THC total hydrocarbon
TIC toxic industrial chemical
TSA technical systems audit
VOC volatile organic compound
Vlll
-------
Chapter 1
Background
The U.S. Environmental Protection Agency (EPA) supports the Environmental Technology
Verification (ETV) Program to facilitate the deployment of innovative environmental tech-
nologies through performance verification and dissemination of information. The goal of the
ETV Program is to further environmental protection by accelerating the acceptance and use of
improved and cost-effective technologies. ETV seeks to achieve this goal by providing high-
quality, peer-reviewed data on technology performance to those involved in the design,
distribution, financing, permitting, purchase, and use of environmental technologies.
ETV works in partnership with recognized testing organizations; with stakeholder groups
consisting of buyers, vendor organizations, and permitters; and with the full participation of
individual technology developers. The program evaluates the performance of innovative tech-
nologies by developing test plans that are responsive to the needs of stakeholders, conducting
field or laboratory tests (as appropriate), collecting and analyzing data, and preparing peer-
reviewed reports. All evaluations are conducted in accordance with rigorous quality assurance
(QA) protocols to ensure that data of known and adequate quality are generated and that the
results are defensible.
Subsequent to the terrorist attacks of September 11, 2001, this ETV approach has been applied
to verify the performance of homeland security technologies. Monitoring and detection
technologies for the protection of public buildings and other public spaces fall within the Safe
Buildings Monitoring and Detection Technology Verification Program, which is funded by EPA
and conducted by Battelle. In this program, Battelle recently evaluated the performance of the
Environics USA Inc. M90-D1-C chemical warfare (CW) agent detector.
-------
Chapter 2
Technology Description
The objective of the ETV Safe Buildings Monitoring and Detection Technology Verification
Program is to verify the performance characteristics of monitoring technologies for chemical
and/or biological contaminants that might be introduced into the building environment. This
verification report provides results for the verification testing of the M90-D1-C CW agent
detector. Following is a description of the M90-D1-C, based on information provided by the
vendor. The information provided below was not subjected to verification in this test.
The M90-D1-C CW agent detector is designed to detect and identify nerve, blister, blood, and
choking agents using Environics' patented open-loop ion mobility spectrometry (IMS)
technology to provide continuous real-time operation without the need for expendable desiccant
cartridges or membranes. The M90-D1-C is fully automatic and provides the operator with
audible and visible alarms upon detecting CW agents. The M90-D1-C display identifies the
agent class (Nerve, Blister, Blood), indicates the relative agent concentration
(Low/Medium/High), and indicates whether the concentration is increasing or decreasing. This
alarm information can be provided to a remote computer/control station through the data
connector on the M90-D1-C. The M90-D1-C can be upgraded to detect new agents by changing
data libraries. It is fully ruggedized to meet appropriate military standards.
The M90-D1-C is a multiapplication
instrument, capable of operating as a point
detector to provide an early warning of
approaching toxic chemical gas or as a
chemical agent monitor to identify and
monitor personnel, vehicles, and equip-
ment for contamination. The M90-D1-C is
generally carried by people, but it can be
installed on vehicles. It also can be used as
a fixed detector, operating without
constant supervision. Both local and
distant alarms are provided, and the
M90-D1-C can be used to automatically
trigger closing down ventilation systems to secure buildings and positions from further agent
contamination.
Figure 2-1. Environics USA M90-D1-C CW
Agent Detector
-------
The M90-D1-C contains two sensor units: an aspiration-type IMS sensor and a semiconductor
sensor. A simulant tube is provided for each sensor to allow performance checks during
operation. The M90-D1-C can operate from 115/240 volts alternating current, from batteries, or
from vehicle power supplies. It weighs 4.7 kilograms (10 pounds, 6 ounces), and it is 28
centimeters (cm) (11.02 inches) long, 10.5 cm (4.12 inches) wide, and 28 cm (11.02 inches)
high. The M90-D1-C is designed to operate in temperatures between -30°C and 55°C (-22°F and
131°F) and at relative humidities up to 99%. The M90-D1-C has a programmed initial startup
delay of less than 10 minutes and not less than a 5-minute delay after power is recycled. It comes
with a carrying case so that the M90-D1-C can be carried over the shoulder or as a front or rear
backpack.
-------
Chapter 3
Test Design and Procedures
3.1 Introduction
When first responders arrive at a potentially contaminated site, they need to immediately and
accurately identify chemicals that may be present. Chemicals and chemical agents that may pose
a threat in a building could include both toxic industrial chemicals (TICs) and CW agents.
The objective of this verification test of the M90-D1-C, a commercially available CW agent
detector, was to evaluate its ability to detect CW agents in indoor air. This verification focused
on the scenario of a detector used by first responders to identify contaminants and guide
emergency response activities after chemical contamination of a building. This verification was
conducted according to a peer-reviewed test/quality assurance (test/QA) plan(1) that was devel-
oped according to the requirements of the Quality Management Plan for the Safe Buildings
Monitoring and Detection Technology Verification Program.(2) The following performance
characteristics of the M90-D1-C were evaluated:
Response time
Recovery time
Accuracy
Repeatability
Response threshold
Temperature and humidity effects
Interference effects
Cold-/hot-start behavior
Battery life
Operational characteristics.
Response time, recovery time, accuracy, and repeatability were evaluated by challenging the
M90-D1-C with known vapor concentrations of one target TIC and two CW agents. M90-D1-C
performance at low target analyte concentrations was evaluated to assess the response threshold.
Similar tests conducted over a range of temperatures and relative humidities (RH) were used to
establish the effects of these factors on detection capabilities. The effects of potential inter-
ferences in an emergency situation were assessed by sampling those interferences both with and
without the target TIC and CW agents present. The M90-D1-C was tested after a cold start (i.e.,
without the usual warm-up period) both from room temperature and from cold storage condi-
tions, and after hot storage, to evaluate the delay time before readings could be obtained. Battery
-------
life was determined as the time until M90-D1-C performance degraded as battery power was
exhausted in continuous operation. Operational factors such as ease of use, data output, and cost
were assessed by observations of the test personnel and through inquiries to the vendor. All
testing was carried out on a single unit of the M90-D1-C.
Testing was limited to detecting chemicals in the vapor phase because that mode of application
is most relevant to use by first responders. Testing was conducted in two phases: detection of
one TIC (conducted in a non-surety laboratory at Battelle) and detection of two CW agents (con-
ducted in a certified surety laboratory at Battelle's Hazardous Materials Research Center
[HMRC]).
3.2 Test Design
3.2.1 Chemical Test Compounds
Hydrogen cyanide (North Atlantic Treaty Organization designation AC) was the only TIC used
in testing, because the vendor indicated prior to the test that this was the only TIC in the
M90-D1-C software library. The CW agents used in testing were sarin (GB) (Lot 7852, 85.1%
purity) and sulfur mustard (HD) (Lot 7864, 95.8% purity).
Table 3-1 summarizes the concentrations of each TIC and CW agent used in this verification
test. For AC, tests were conducted at the immediately-dangerous-to-life-and-health (IDLH)
level. For the CW agents GB and HD, testing was conducted at a single concentration level that
produced less than full-scale readings on the M90-D1-C under normal temperature and humidity
conditions. The concentration used for GB was 0.13 parts per million (ppm) (0.75 milligrams
per cubic meter [mg/m3]), which is approximately four times the IDLH concentration of 0.035
ppm (0.2 mg/m3). No IDLH level has been set for HD, so the concentration used was based on
an alternative toxic effects guideline, as noted in the footnote to Table 3-1.
Table 3-1. Target TIC and CW Agent Challenge Concentrations
Chemical
AC
GB
HD
Challenge Concentrations
50 ppm (50 mg/m3)
0.1 3 ppm (0.75 mg/m3)
0.63 ppm (4.1 mg/m3)
Type of Level
1 x IDLH
4 x IDLH
7 x AEGL-2(a)
(a) AEGL = Acute Exposure Guideline Level; AEGL-2 levels are those expected to produce a serious hindrance to
efforts to escape in the general population. The AEGL-2 value of 0.09 ppm (0.6 mg/m3) for HD is based on a
10-minute exposure.
3.2.2 Test Matrix
Table 3-2 summarizes the evaluations that were conducted in the verification test. As Table 3-2
indicates, except for cold-/hot-start behavior, battery life, and assessment of false positive inter-
-------
ference effects (i.e., the interferent alone), all performance parameters were evaluated during
both the TIC and CW agent testing.
Table 3-2. Summary of Evaluations Conducted on the M90-D1-C
Performance
Parameter
Objective
Comparison Based On
Response Time
Recovery Time
Accuracy
Repeatability
Response
Threshold
Temperature and
RH Effects
Interference
Effects
Cold Start
Hot Start
Battery Operation
Determine rise time of
M90-D1-C response
Determine fall time of
M90-D1-C response
Characterize reliability of M90-D1-C
identification of target chemicals
Characterize consistency of
M90-D1-C readings with constant
analyte concentration
Estimate minimum concentration that
produces M90-D1-C response
Evaluate effect of temperature and
RH on M90-D1-C performance
Evaluate effect of building
contaminants that may
interfere with M90-D1-C performance
Characterize startup performance after
cold storage
Characterize startup performance after
hot storage
Characterize battery life
M90-D1-C readings with step rise in
analyte concentration
M90-D1-C readings with step decrease in
analyte concentration
M90-D1-C identifier display
M90-D1-C readings with constant input
Reference method results
Repeat above evaluations with different
temperature and RH
Sample interferents and target chemicals
together (and interferents alone(a))
Repeat tests with no warm-up(a)
Repeat tests with no warm-up(a)
Observe M90-D1-C duration of operation
on batteries(a)
(a)
Indicates this part of the test performed only during TIC testing.
3.2.3 Test Locations
Two laboratories were used to conduct the verification tests. Testing with the non-chemical
surety materiel—AC and interferents—was conducted in a laboratory at Battelle's Columbus,
Ohio, campus, which has the needed challenge generation, collection, and analysis equipment.
This laboratory has been used previously to conduct IMS instrument and filter tests using AC
under controlled environmental conditions. Testing with CW agents was conducted at the
HMRC at Battelle's West Jefferson, Ohio, campus. Battelle's HMRC is an ISO 9001-certified
facility that provides a broad range of materials testing, system and component evaluation,
research and development, and analytical chemistry services requiring the safe use and storage
of highly toxic substances. Battelle operates the HMRC in compliance with all applicable
federal, state, and local laws and regulations, including Army regulations.
-------
3.2.4 Test Sequence and Schedule
The sequence of tests planned to be performed with the TIC (AC) in this study is outlined in
Figure 3-1. Since analyzer performance was not known a priori, the concentrations used in
testing depended on the results of the first few tests performed. The decision logic used to
determine the actual TIC concentration and the test sequence is shown in Figure 3-2. A similar,
but slightly reduced, set of tests was performed with CW agents. Table 3-3 summarizes the
actual schedule of testing for the TIC and CW agents. As described in Chapter 6, only minimal
testing was conducted with AC because the M90-D1-C did not respond when challenged with
this TIC. A nerve agent simulant was used instead of AC to allow completion of tests such as
battery life and cold/hot start behavior.
3.2.5 Reference Methods
Table 3-4 summarizes the primary reference methods used to determine the challenge
concentrations of the target TIC and CW agents. Listed in the table are the target TIC and CW
agents, the sampling and analysis methods used for each compound, and the applicable concen-
tration range of each method. For AC, low concentration samples were injected directly for
determination by gas chromatography (GC) with flame ionization detection (FID). The CW
agents GB and HD were collected in gas sample bags, and determined by GC with flame
photometric detection (FPD), according to existing HMRC test procedures.
Summaries of these primary methods, and of supplemental methods also used, are as follows.
Hydrogen cyanide (AC)—The reference method for AC was GC/FID, using an Agilent 6890 GC
with a capillary column and FID. This GC was positioned next to the laboratory hood containing
the test system during the TIC testing and sampled automatically from the flow line delivering
the challenge gas to the M90-D1-C.
Sarin (GB) and sulfur mustard (HD)—The analytical method for these CW agents involved
collecting the agents by flowing air from the test apparatus into gas sample bags. The agent
concentrations then were determined using a capillary GC with FPD. Concentrations were
determined based on a linear regression of peak area with the amount of agent.
Total hydrocarbons—A continuous FID was used for the determination of the total hydrocarbon
(THC) content of interferent mixtures provided to the M90-D1-C during testing. The THC
concentrations characteristic of realistic interferent levels in buildings were determined, either
by direct measurement or by interpretation of published data. The interferent delivery systems
were then adjusted to achieve the desired THC indication in parts per million of carbon (ppmC)
for each interferent during testing.
-------
Testl: Vapor challenge with TIC
Alternating clean air with immediately dangerous to life and health (IDLH) level concentration of TIC five times with
M90-D1-C fully warmed up per manufacturer's instructions prior to testing, and room temperature (22 ± 3°C) and 50
± 5% RH.
Test 2: Vapor challenge with TIC at reduced concentration
Test 1 is repeated at a lower concentration giving mid-range on-scale readings (only if off-scale response at IDLH).
The concentration that gives a mid-range on-scale reading is then referred to as the target concentration for all
subsequent tests.
Test 3: Vapor challenge with TIC at increased concentration
Test 1 is repeated at roughly 10 times the IDLH concentration (only if no response at IDLH).
Test 4: Response threshold of TIC
Test 1 is repeated at a concentration below IDLH. If a response is recorded, the concentration is cut in half until no
response is recorded. If no initial response is recorded, the concentration is increased by a factor of 2 until a response
is recorded.
Test 5: Target/low/clean air challenge
Test 1 is repeated by alternating target concentrations, a low concentration (either 0.1 IDLH or response threshold
concentration) and clean air six times and alternating order of low concentration and target concentration.
Test 6: Vapor challenge with TIC at room temperature, low humidity
Test 1 is repeated at room temperature (22 ± 3°C) and less than 20% RH. The test is performed at the concentration
determined via the logic in Figure 3-2.
Test 7: Vapor challenge with TIC at room temperature, high humidity
Test 1 is repeated at room temperature (22 ± 3°C) and 80% RH. The test is performed at the concentration
determined via the logic in Figure 3-2.
Test 8: Vapor challenge with TIC at high temperature, medium humidity
Test 1 is repeated at high temperature (35 ± 3°C) and 50% RH. The test is performed at the concentration determined
via the logic in Figure 3-2.
Test 9: Vapor challenge with TIC at high temperature, high humidity
Test 1 is repeated at high temperature (35 ± 3°C) and 80% RH. The test is performed at the concentration determined
via the logic in Figure 3-2.
Test 10: Vapor challenge with TIC at low temperature, medium humidity
Test 1 is repeated at low temperature (5 ± 3°C) and 50% RH. The test is performed at the concentration determined
via the logic in Figure 3-2.
Test 11: Interferent false positive tests
Test 1 is repeated alternating interferent only with clean air. The test is repeated for all interferents.
Test 12: Interferent false negative tests
Test 1 is repeated alternating TIC and interferent with clean air. The test is repeated for all interferents.
Test 13: Room temperature, cold start behavior
Repeat Test 1 with the M90-D1-C at room temperature for a minimum of 12 hours and no warm-up.
Test 14: Cold-/cold-start behavior
Repeat Test 1 after the M90-D1-C has been kept refrigerated (5-8°C) overnight for a minimum of 12 hours, with no
warm-up.
Test 15: Hot-/cold-start behavior
Repeat Test 1 after the M90-D1-C has been kept heated (40°C) overnight for a minimum of 12 hours, with no cool-
down or warm-up.
Test 16: Battery test
Repeat Test 1 with the M90-D1-C operating on battery power. The TIC at target concentration is alternated with
clean air once every half hour until the unit stops responding or shuts down due to loss of power.
Figure 3-1. Planned Sequence of TIC Verification Tests
-------
Step 1: Perform Test 1. Depending on the results of this test, go to Step 2a, 2b, or 2c as
appropriate.
Step 2a: If there is no response in Test 1, perform Test 3, then go to Step 4.
Step 2b: If the response in Test 1 is on scale, skip to Step 3 and perform all subsequent
tests at the IDLH concentration.
Step 2c: If the response in Test 1 is off-scale, perform Test 2. Establish the concentration
that gives a mid-range on-scale response and proceed with Step 3, using that established
concentration in all subsequent tests.
Step 3: Perform Test 4 (if not already done), Tests 5 through 10, and Test 12 at the
concentration(s) determined above. For the first TIC, also perform Test 11 and Tests 13
through 16.
Step 4: Repeat Tests 1 through 10 and 12 for all CW agents.
Figure 3-2. Logic Diagram for Determining TIC/CW Agent Test Sequence
Table 3-3. Test Schedule
Chemical
Test Dates (2004)
AC
Nerve simulant
HD
GB
August 6
August 13-24
September 13
September 20 - October 1
Table 3-4. Primary Reference Methods
Analyte
AC
GB
HD
Concentration
Range (ppm)
0.05 to
0.01 to
0.01 to
100
100
100
Air
Air
Air
sample
sample
sample
Sampling Method
injected directly
collected in gas
collected in gas
sampling
sampling
bag
bag
Analysis
GC/FID
GC/FPD
GC/FPD
Method
3.2.6 Interferents
Interferents were selected for testing based upon their prevalence in a building. The interferents
selected were the volatile chemicals in latex paint, air freshener, and ammonia-based floor
cleaner, as well as gasoline engine exhaust hydrocarbons and N,N-diethylaminoethanol (DEAF).
DEAF is a common additive to reduce corrosion in building boiler systems, and is released into
the heating, ventilating, and air conditioning system when boiler steam is used to humidify the
air. These selected interferents were tested for false positives by exposing the M90-D1-C to
-------
selected levels of the interferents in clean air, to see whether the interferents generated a positive
response from the M90-D1-C when no TIC or CW agents were present. Each interferent also
was introduced to the M90-D1-C along with each CW agent, to determine false negatives, i.e.,
whether the interferent prevents the M90-D1-C from indicating that CW agent is present. The
following sections describe the materials and concentrations used for testing.
The interferents are mixtures of chemicals and determining the interferent concentration requires
the quantification of all the chemicals present. However, monitoring each component would be
time and cost prohibitive. For this reason, interferent concentrations were monitored using a
THC analyzer. THC analysis is appropriate because all the interferents consist of a significant
amount of carbon-containing compounds. Because quantification is based on carbon content,
the test concentrations are reported on a per carbon basis in ppmC. The use of the hydrocarbon
analyzer also provided real-time continuous monitoring of the interferent concentration during
testing.
Test concentrations for the interferents were based on direct measurements or published data.
Concentrations found in published data were converted to a per carbon basis as described below.
Table 3-5 is a summary of the interferent test concentrations. The following sections contain a
detailed description of how the test concentrations were determined.
Table 3-5. Test Concentrations for Interferents
Interferent Test Concentration (ppmC)
Latex Paint Fumes 5-10
Floor Cleaner Vapors 10
Air Freshener Vapors 1
Gasoline Exhaust Hydrocarbons 2.5
DEAE 0.02
3.2.6.1 Latex Paint Fumes
The appropriate concentrations of latex paint fumes were established directly by measurements
in and around a freshly painted office. Samples were obtained using a 25-liter (L) Teflon bag
and analyzed for THC content. Each wall in the office was painted, and the room dimensions
were 11 feet by 11 feet with an alcove 4 feet by 10 feet and ceiling 12 feet high. Immediately
after painting, the hydrocarbon concentration was 170 ppmC. After 2.5 hours, the hydrocarbon
concentration in the office fell to 38 ppmC. At this time, the hydrocarbon content was
determined just outside the entrance to the office and in the hallway 80 feet away from the
office. Hydrocarbon content just outside the office was 20 ppmC; in the hallway 80 feet away
from the office, it was 3 ppmC. Based on these measurements, the test concentration was
maintained at 5 to 10 ppmC.
3.2.6.2 Floor Cleaner Vapors
The test concentration for the ammonia-based floor cleaner was inferred from the information
cited in Section 3.2.6.1 on latex paint fumes. Similar to paint, floor cleaner is applied to a
10
-------
surface and allowed to dry. Floor cleaner vapors containing both ammonia and fragrances will
disperse into the hallway. Because of the similarity, a test concentration of 10 ppmC was used
for the floor cleaner.
3.2.6.3 Air Freshener Vapors
Concentration levels of air freshener for interferent testing were based upon values reported at an
indoor air quality conference. Volatile organic compound (VOC) emission for a plug-in air
freshener was reported to be 30 to 80 milligrams per hour, resulting in a concentration of 300 to
500 micrograms per cubic meter (|j,g/m3) for the average room. Assuming the VOC emitted
consists of hydrocarbons similar to limonene, a common fragrance component, the
concentration on a per carbon basis can be calculated. Limonene contains 10 carbons and has a
molecular weight of 136. A concentration of 5.56 |-ig/m3 of limonene is the same as 1 part per
billion (ppb). With a room concentration of 500 |_ig/m3 and limonene as a representative
molecule, the fragrance concentration on a per carbon basis is estimated to be 1 ppmC. This
THC level was maintained for all tests with the air freshener.
3.2.6.4 Gasoline Engine Exhaust
Of the constituents in gasoline engine exhaust fumes, the aromatic components were considered
most likely to interfere with the performance of the M90-D1-C. A recent study reported that
urban areas can have benzene concentrations of over 5 ppb with comparable concentrations of
other aromatics.(3) The test mixture used to simulate exhaust contains 61 compounds ranging in
size from 2 to 10 carbons, with an average concentration of 200 ppb for each component. To
obtain a challenge concentration for the aromatic compounds, the test mixture was diluted 30:1.
Assuming an average size of six carbons, the THC of the mixture was approximated to be
73 ppmC. After dilution, the THC content was 2.5 ppmC, and this target concentration was
maintained for all the experiments.
3.2.6.5 DEAE
DEAE is a common additive to boiler systems to prevent corrosion. When boiler steam is used
to humidify the air in a building, DEAE is released into the building as well. Generally, the
DEAE concentration is kept below 40 ppb, the threshold for odor detection. One study has
shown DEAE concentrations of 1 ppb in a building that uses direct steam injection for
humidification.(4) For testing purposes, the concentration was set at 20 ppbC, which correlates to
3.3 ppb DEAE given that DEAE contains six carbons. This concentration was not detectable by
THC analysis, so the interferent concentration was set by dilution of a concentrated standard.
3.2.7 Materials and Equipment
3.2.7.1 TIC and CWAgents
The commercial gas standard used as the source of AC for testing was a standard of 10,020 ppm
AC in nitrogen (Cylinder B0005506, Scott Specialty Gases). The CW agents GB and HD were
11
-------
obtained as neat materials from the U.S. Army under Bailment Agreement No. DAAD13-03-H-
00-0003.
3.2.7.2 Vapor Delivery Equipment
The compressed gas mixture noted in Section 3.2.7.1 was diluted as the vapor source for AC. A
two-way valve was included in the flow path downstream of the vapor generation source, so that
the dilution and test equipment could be totally isolated from the source. A schematic diagram
of the entire TIC vapor generation, dilution, and delivery system is shown in Figure 3-3. For the
CW agents GB and HD, a diffusion cell containing the pure agent was substituted for the gas
mixture. A temperature-controlled water bath was installed to control the temperature of the
diffusion cell to maintain a stable vapor generation rate.
3.2.7.3 Temperature/Humidity Control
The M90-D1-C was evaluated at the temperature and humidity conditions indicated by an "X"
in Table 3-6. Both the delivered air temperature and the M90-D1-C were maintained within the
specified temperature range. For testing at 35°C, the vapor delivery system was warmed with a
heat-traced line, using an electronic temperature controller. For testing at 5°C, the dilution and
delivery system was enclosed in a cooled chamber to provide approximate temperature control.
For all tests, thermocouples were installed in both the clean air plenum and the challenge plenum
to provide real-time temperature monitoring.
Table 3-6. Temperature and Relative Humidity Conditions
Temperature (°C)
RH (%)
<20
50 ±5
80 ±5
5 ± 3 22 ± 3
X
X X
X
35 ±3
X
X
A commercial Nafion® humidifier (Perma Pure, Inc.) was used to generate controlled high-
humidity air (50 to 100% RH), which was then mixed with dry dilution air and the target vapor
stream to obtain the target RH (< 20% to 80%) in the challenge air.
3.2.7.4 InterferentSources
Interference test concentrations were obtained by diluting a concentrated feed with air. For latex
paint, floor cleaner, and air freshener, the concentrated feeds were made by purging the head
space of a large boiling flask containing about 100 milliliter (mL) of the bulk liquid of each
interferent using approximately 0.1 liter per minute (L/min) flow of clean air. THC analysis of
the head space samples found that the concentrated feeds contained 394, 886, and 233 ppmC for
latex paint, floor cleaner, and air freshener, respectively. Gasoline engine exhaust was simulated
using a mixture of 61 organic compounds ranging from 2 to 10 carbon atoms (C2 to C10). This
12
-------
TO
s
I
CD
tt
Challenge
Gas or CW
Agent Source
Temperature
Controlled Chamber
Pressure Sensor
Temperature Sensor
I T/OW I Temperature and Relative
I lmnl Humidity Sensor
[c] One Way Check Valve
-------
mixture was prepared by adding 1 microliter (|^L) of 51 neat liquid components and 250 |_iL of
10 gaseous components into a 15.7-L cylinder and diluting to a final pressure of 1,200 pounds
per square inch gauge (psig) with nitrogen. A concentrated standard of 1 ppm for DEAE was
made by adding zero nitrogen to 6 |_iL of liquid neat DEAE to a final pressure of 1,200 psig. In
all cases these cylinder gases or concentrated vapor streams were diluted to the appropriate level
by addition to the large flows of clean air passing through the test apparatus (Figure 3-3).
3.2.7.5 Performance Evaluation Audit Materials
As part of the quality assurance effort in this verification, a performance evaluation (PE) audit
was performed on reference methods used to confirm the AC concentration provided to the
M90-D1-C. This audit involved conducting analysis on an independent standard, obtained from
a different source than that used for the calibration standard, with the two standards diluted
identically in the test apparatus. The result from the independent standard was then compared
with that from the calibration standard to assess the degree of agreement. The target agreement
in the PE audit was within 20% for AC. For AC, the PE audit standard was 10,000 ppm AC in
nitrogen (Cylinder LL320) obtained from Linde Gas LLC.
A comparable PE audit could not be done for the CW agents because of the lack of independent
standards. In lieu of a PE audit for the CW agents, check samples were prepared at the HMRC
by an analyst other than the staff who conducted routine calibration of the reference method.
These samples were analyzed by the same approach used for analysis of calibration samples
from GB and HD testing, and the results were compared.
3.3 Test Procedure
The test system (Figure 3-3) consisted of a vapor generation system, a Nafion® humidifier, two
challenge plenums, a clean air plenum, an RH sensor, thermocouples, and mass flow meters. The
challenge vapor or gas was generated by the vapor generation system. The challenge vapor was
then mixed with the humid dilution air and flowed into the challenge plenum. Interference
vapors were added to the challenge mixtures as needed for testing.
The RH and target concentration of the challenge vapor were obtained by adjusting the mixing
ratio of the humid air (from the Nafion® humidifier) to the dry dilution air, and the mixing ratio
of the vapor generation stream to the humid dilution air, respectively. To avoid potential
corrosion or malfunction of the RH sensor from exposure to the challenge vapor, the RH meter
was installed upstream of the inlet of the vapor stream. The RH of the challenge vapor stream
was calculated based on the measured RH of the humid dilution air and the mixing ratio of the
vapor generation stream to the humid dilution air.
To establish the baseline reading of the M90-D1-C, a clean air plenum was used. Part of the
humid dilution air was introduced directly into the clean air plenum. When establishing the
M90-D1-C background, the four-way valve connected to the M90-D1-C was switched to the
clean air plenum to collect baseline data.
14
-------
After the baseline measurement, the four-way valve connected to the M90-D1-C was switched
to one of the challenge plenums to allow the M90-D1-C to sample the challenge mixture.
Switching between the challenge and clean air plenums was rapid, and the residence time of gas
in the test system was short to allow determination of the response and recovery times of the
M90-D1-C. The reference methods described in Section 3.2.5 were used to confirm that the
concentration in the challenge plenums was within ± 20% of the target level for AC (or within
35% of the target level for the CW agent). Concentrations outside those tolerance ranges
triggered a repeat of any test procedures conducted since the last analysis.
3.3.1 Response Time
To evaluate M90-D1-C response time, the target test conditions were established at 22 ± 3°C
and 50 ± 5% RH. Initially 10 L/min of clean humidified air were passed through the clean air
plenum. The M90-D1-C sampled the clean air for a minimum of 30 seconds or until a stable
reading was indicated, but not exceeding 10 minutes, to obtain a baseline reading for the
M90-D1-C. The clean air plenum also was sampled with the appropriate reference method.
This sampling took place after the M90-D1-C reading was stabilized.
Concurrent with the baseline measurements the target challenge concentration in the high
challenge plenum was established. The high challenge concentration was generated at the target
test conditions. For AC, the generator operating conditions and the dilution flow were adjusted
as needed to establish a challenge concentration within ± 20% of the IDLH level. For the CW
agents, a delivered concentration within 35% of the target level was acceptable. Reference
samples were collected and analyzed immediately to establish the challenge concentration and
demonstrate stability.
After a stable baseline reading was obtained from the M90-D1-C on clean air, and the challenge
concentration was stable at the target concentration, the four-way valve at the M90-D1-C inlet
was switched to sample from the challenge plenum. The response of the M90-D1-C was then
recorded and the time to produce an alarm was considered the response time. The M90-D1-C
sampled from the challenge plenum for a minimum of 30 seconds, up to a maximum of
10 minutes. The challenge concentration was determined by the reference method as frequently
as possible during the procedure. For AC, a reference sample was taken prior to every challenge
with the M90-D1-C.
After the challenge sampling, the sample inlet four-way valve was switched to again sample
from the clean air plenum. The time required for the M90-D1-C to clear (i.e., the time to return
to its starting baseline or non-alarm reading) was recorded as the recovery time. After a
maximum of 10 minutes, regardless of whether the M90-D1-C returned to baseline, subsequent
cycles of alternating challenge/clean air sampling were carried out, controlled by the four-way
valve. A total of five such challenge/clean air cycles were completed.
The same sampling procedure was carried out at different temperature and RH conditions or
challenge concentration to evaluate temperature and RH effects and response thresholds. For AC
and each CW agent, the initial test was conducted at the levels shown in Table 3-1. If the
instrument gave an over-scale reading when challenged at the initial level at the normal
15
-------
temperature and RH conditions (22°C and 50% RH), a lower challenge concentration was
chosen that provided an on-scale reading. All subsequent tests for that TIC or CW agent used
that lower challenge concentration. If the instrument did not respond to the IDLH or other initial
concentration selected, then the response threshold procedure in Section 3.3.5 was conducted;
but, all subsequent tests planned for that TIC or CW agent were eliminated. Otherwise, testing
proceeded as described.
Following the five challenge/clean air cycles, six cycles were conducted in which the M90-D1-C
sampled sequentially from the high, low, and clean air challenge plenums. The high challenge
plenum provided the respective target concentrations (Table 3-1), and the low challenge plenum
provided a concentration of approximately 0.1 times that level, or the response threshold (see
Section 3.3.5), whichever was greater. Clean air was sampled alternately with sampling from the
challenge plenums, and the order of sampling from the high (H) and low (L) challenge plenums
was also alternated, i.e., the order of sampling was clean air/H/L/clean air/L/H/clean air/H/L/ ..
,. for a total of six such cycles. This procedure simulated use of the M90-D1-C in locations
having different degrees of contamination.
3.3.2 Recovery Time
The time for the M90-D1-C to return to its baseline reading or non-alarm state after removing a
challenge concentration was measured as described in Section 3.3.1.
3.3.3 Accuracy
In all of the response threshold and response time tests, the challenge concentration was
measured using a reference method or monitor. Those measurements confirmed that the target
TIC or CW agent was present at the appropriate challenge concentration. The degree to which
the M90-D1-C correctly identified the challenge TIC or CW agent was evaluated as the measure
of accuracy.
3.3.4 Repeatability
Repeatability was assessed using M90-D1-C responses obtained from the five repeated
challenge/clean air cycles or the high challenge/low challenge cycles. The repeated test results at
the same environmental and concentration conditions were used to quantify the repeatability of
the measurements and the effects of test conditions on repeatability.
3.3.5 Response Threshold
The response threshold of the M90-D1-C was evaluated by repeating the procedure in
Section 3.3.1 at successively lower (or if necessary, higher) concentrations. The response
threshold was determined at the baseline environmental condition of 22 ± 3°C and 50 ± 5% RH,
in the absence of any interfering chemicals. The manufacturer's reported detection limit (± 50%)
was used as the starting concentration. If the manufacturer did not provide a detection limit, a
concentration at least 10 times lower that the IDLH or target concentration was chosen. If there
was no response at the starting test concentration, then the concentration of the challenge was
16
-------
increased by a factor of two. Similarly, if the M90-D1-C responded to the starting concentration,
then the challenge concentration was decreased by a factor of two. The increase or decrease in
concentration was continued accordingly, until the response threshold had been bracketed. The
minimum concentration producing a M90-D1-C response was denoted as the response
threshold.
3.3.6 Temperature and Humidity Effects
The tests described in Section 3.3.1 were repeated at the target concentrations shown in
Table 3-1, over the range of environmental conditions shown in Table 3-6. Five repeat runs were
performed at each set of test conditions (in one case, a sixth run was conducted and recorded at
the operator's discretion, to ensure consistency of the challenges). The data at different
temperature and RH conditions were used to infer whether these conditions affected the
detection (i.e., accuracy, repeatability, response threshold) of the M90-D1-C for the target
chemical. The effect on response time and recovery time also was assessed.
3.3.7 Interference Effects
To evaluate the effects of the interferents described in Section 3.2.6, the test system shown in
Figure 3-3 was modified by adding an interferent vapor generator. The output from this source
was directed as needed to mix with the humidified air flowing to the challenge plenum. The test
chemical generation was independently controlled to generate interferent in the absence or
presence of the test chemical. This allowed interference effects to be evaluated with the inter-
ferent alone and with each interferent and TIC or CW agent together. Testing with the interferent
alone allowed evaluation of false positive responses, and testing with the interferent and
chemical together allowed evaluation of false negative responses caused by the interferents. The
test procedures also allowed observation of interferent effects on the response time and recovery
time of the M90-D1-C. The target concentrations of the planned interferents are shown in Table
3-5. Those concentrations are shown in terms of the equivalent total hydrocarbon concentration
in ppmC. These target concentrations are based on actual indoor measurements by Battelle or on
published data, as described in Section 3.2.6.
Interferent testing involved only one interferent at a time. Testing was done by alternately
sampling clean air and the interferent mixture, for a total of up to five times each, in a procedure
analogous to that described in Section 3.3.1. However, if no interferent effect was observed
after three such test cycles, the test was truncated. Testing with interferents alone involved
alternately sampling from the clean air plenum and then from the challenge plenum, to which
only the interferent in clean air was delivered. The same process was used for testing with
interferents and TIC or CW agents together, with the two compounds diluted together in
humidified air delivered to the challenge plenum. The same TIC and CW agent concentrations
used in the initial testing under Section 3.3.1 were used in this test, i.e., the levels shown in
Table 3-1.
A response from the M90-D1-C with the interferent alone was recorded as a false positive; and
the absence of a response, or a reduced response, to the TIC or CW agent in the presence of the
interferent was recorded as a false negative.
17
-------
The replicate test runs conducted with the interferent plus TIC or agent also allowed the
response time and recovery time of the M90-D1-C to be assessed with interferents present.
Differences in response and recovery times, relative to those in previous tests with only the TIC
or agent present, were attributed to the effect of the interferent vapor.
3.3.8 Cold-/Hot-Start Behavior
The cold-/hot-start tests were conducted in a manner similar to the response time test in
Section 3.3.1. Prior to these tests, however, the M90-D1-C was not allowed to warm up per the
manufacturer's recommendation. Only one cold-/hot-start test was performed per day.
The cold-start test was conducted twice, once with the M90-D1-C at room temperature and,
subsequently, at reduced temperature, prior to start-up. In the former test, the M90-D1-C was
stored with the power off at 22 ± 3°C for at least 12 hours prior to testing. The cold-start effect
was assessed by observing the time from powering up the M90-D1-C to its first readiness to
provide readings. This was considered the start-up delay time.
For the reduced temperature cold start, the M90-D1-C was placed in a refrigerated enclosure (5
to 8°C) with the power off for at least 12 hours overnight. At the start of the next test day, the
cold-start test was repeated to record the start-up delay time.
For the hot-start test, the M90-D1-C was placed in a heated enclosure at 40 ± 3°C for at least
12 hours overnight. At the start of the next test day, the hot-start test was conducted in the same
fashion as the cold-start tests to determine start-up delay time.
In initial runs per Section 3.3.1 procedures, the M90-D1-C was found not to respond to AC.
Therefore, the response time, recovery time, repeatability, and accuracy could not be determined
with AC after a cold/hot start. Instead, qualitative evaluations of cold/hot start behavior were
conducted, using a nerve agent simulant to obtain a response from the M90-D1-C.
3.3.9 Battery Life
Battery life was evaluated by assessing the duration of continuous M90-D1-C operation on
battery power. Fully charged batteries were installed, and the M90-D1-C was turned on and
allowed to fully warm up. For this test, a nerve agent simulant was used to produce an alarm on
the M90-D1-C. The M90-D1-C then sampled clean air for 30 minutes, and the simulant was
sampled again. This procedure was repeated with the M90-D1-C operating continuously until it
no longer responded to the simulant challenge. The total time of operation was recorded as the
measure of battery life. Any warnings of impending battery failure provided by the M90-D1-C
are noted in this test.
18
-------
3.3.10 Operational Characteristics
Key operational characteristics of the M90-D1-C were evaluated by means of the observations
of test operators and by inquiry to the M90-D1-C vendor. Ease of use was assessed by operator
observations, with particular attention to the conditions of use by emergency first responders.
Signal or data output capabilities were assessed by observations of the personnel who operated
the M90-D1-C during testing. The type of data that was output was noted on the data sheets
(e.g., audio and/or visual alarm, bar graph, low/med/high indication, and/or quantitative
measure of concentration). In addition, the clarity and readability of the output were noted,
especially in low light conditions or when holding the M90-D1-C while walking, as in use by a
first responder. The availability of multiple forms of data output or display also was assessed
(e.g., the availability of both a visual display and an analog voltage output for recording
purposes).
The vendor was asked for the purchase and operational costs of the M90-D1-C as tested.
Estimates for key maintenance items also were requested from the vendor.
19
-------
Chapter 4
Quality Assurance/Quality Control
QA/quality control (QC) procedures were performed in accordance with the program quality
management plan (QMP)(2) and the test/QA plan for this verification test.(1)
4.1 Equipment Calibration
4.1.1 Reference Methods
The reference methods used for determining AC and the CW agents are summarized in
Section 3.2.5. The analytical equipment needed for these methods was calibrated, maintained,
and operated according to the quality requirements of the reference methods and Battelle's
normal documentation. Procedures for blank sampling during testing and for calibration of
reference methods are described below.
For AC testing, blank reference samples were run before each challenge concentration. The
sequence of reference sampling thus included establishing the concentration prior to testing the
M90-D1-C, running a blank on clean air, switching to challenge gas and taking a reference
sample immediately prior to challenging the M90-D1-C with the challenge gas, and again
running a blank when the M90-D1-C was once more sampling clean air. In testing with GB and
HD, blank gas sample bags were run at the start of each test day.
Calibration procedures for the reference and other analyses were as follows:
The GC reference method for AC was calibrated by preparing gas mixtures in 1-L gas sampling
bags. For AC, calibration standards were prepared by diluting 0.5 to 4 mL of commercial
concentrated AC gas standards (e.g., 10,000 ppm AC in nitrogen) in 800 mL of clean air in a
bag. Three samples from each bag were injected by syringe into the GC, and the peak area was
recorded. Several such calibration standards ranging from 12.5 to 50 ppm AC were prepared and
analyzed over a three-day period. The regression of peak area versus AC standard concentration
had the form Peak Area = 0.7192 x (AC, ppm), with an r2 value of 0.9961.
Calibration standards for the CW agents were prepared by diluting stock agent to micrograms
per milliliter concentrations and then injecting a l-|iL volume of each standard into the
GC/FPD. Calibration was based on a regression of peak area versus amount of agent injected.
20
-------
The THC analyzer used to document the interferent levels provided in testing was calibrated by
filling a 25-L Tedlar bag with 33 ppm of propane in air from a commercial gas standard. Since
propane is a three-carbon molecule, this standard constitutes a THC concentration of 99 ppmC.
This standard was used for calibrating the THC analyzer throughout the verification. Clean air
from the room was used for zeroing.
4.1.2 Instrument Checks
The M90-D1-C was operated and maintained according to the vendor's instructions throughout
the verification test. Maintenance was performed according to predefined M90-D1-C
diagnostics. Daily operational check procedures for the M90-D1-C were performed with two
vendor-supplied simulant tubes. Proper response of the M90-D1-C to the simulants was required
before testing could proceed.
4.2 Audits
4.2.1 Performance Evaluation Audit
As described in Section 3.2.7.5, a PE audit was conducted to assess the quality of reference
measurements made in the verification test. For AC, the PE audit was performed once prior to
the verification test by diluting and analyzing a standard that was independent of the standards
used during the testing. The acceptable tolerance for this PE audit was ±20%. Table 4-1 shows
that the result of the PE audit was within the target tolerance. For the CW agents, check
standards of GB and HD were prepared by individuals other than the staff conducting the
reference analyses. The reference data obtained for these standards were compared. For GB,
standards were prepared at concentrations of 1.0, 0.75, 0.50, 0.25, 0.1, and 0.05 |ig/mL. All
results were within 5% for the separate standards made by the two individuals. For HD,
standards were prepared at concentrations of 5, 2.5, 1.0, and 0.5 |ig/mL. All results were within
9% for the separate standards made by the two individuals.
Table 4-1. Performance Evaluation Audit Results
Date of Agreement
TIC Sample Audit Concentration Result (%)
AC Standard 7/12/04 10,020 ppm 43.2 ppm 9.8
(Cylinder B0005506)
PE Audit Std 10,000 ppm 47.8 ppm
(Cylinder LL320 )
4.2.2 Technical Systems Audit
The Battelle Quality Manager also conducted a technical systems audit (TSA) to ensure that the
verification test was performed in accordance with the test/QA plan(1) and the ETV QMP.(2) As
part of the audit, the Battelle Quality Manager reviewed the reference sampling and analysis
21
-------
methods used, compared actual test procedures with those specified in the test/QA plan,(1) and
reviewed data acquisition and handling procedures. Observations and findings from this audit
were documented and submitted to the Battelle Verification Test Coordinator for response. The
records concerning the TSA are permanently stored with the Battelle Quality Manager.
4.2.3 Data Quality Audit
At least 10% of the data acquired during the verification test was audited. Battelle's Quality
Manager traced the data from the initial acquisition, through reduction and statistical analysis,
to final reporting, to ensure the integrity of the reported results. All calculations performed on
the data undergoing the audit were checked.
4.3 QA/QC Reporting
Each assessment and audit was documented in accordance with the test/QA plan.(1) Once the
assessment report was prepared, the Battelle Verification Test Coordinator ensured that a
response was provided for each adverse finding or potential problem and implemented any
necessary follow-up corrective action. The Battelle Quality Manager ensured that follow-up
corrective action was taken.
22
-------
Chapter 5
Statistical Methods
To extract the most information about the M90-D1-C performance from the test procedures, a
statistical analysis of the test results was performed whenever appropriate. Such an analysis used
all available data to explore the impact of test parameters on the M90-D1-C performance.
Section 5.1 summarizes the statistical approaches and the parameters tested. The performance
parameters of response threshold and battery life were assessed with simple comparisons that did
not require statistical analysis. Section 5.2 describes the analyses used for these performance
parameters.
5.1 Statistical Analyses
For a given chemical and test condition, several successive readings of the M90-D1-C response
to the chemical were recorded. The chemical exposures alternated with clean air samples. Test
conditions included a range of temperatures, relative humidities, and starting environments.
Performance was also assessed in the presence of interferents alone and in the presence of
interferents and GB. These data were the basis for the statistical analysis of M90-D1-C
performance.
The statistical analyses focused on the following performance parameters:
Response time
Recovery time
Accuracy
Repeatability
False positives/false negatives
and considered the following explanatory variables:
Identity of the target TIC or CW agent
Temperature
Humidity
Identity and presence/absence of interferent
Fluctuation in chemical concentration.
23
-------
As described in Chapter 6, statistical analysis could only be conducted using the data from
testing with GB.
5./J Analysis of Response, Response Time, and Recovery Time
The effects of temperature and humidity on the actual response were investigated using the
Jonckheere-Terpstra test.(5) This non-parametric method tests the hypothesis of no association
between response and temperature or humidity versus the hypothesis that the response increases
or decreases as temperature or humidity increases. The test accommodates the categorical nature
of the dependent (response) and independent (temperature or humidity) variable and is appro-
priate when both the dependent and the independent variables have a natural ordering (low,
medium, and high in this case). Furthermore, the Jonckheere-Terpstra test is appropriate when
the sample size is small or the data are sparse.
Unlike temperature and humidity, start state has no natural ordering. The Kruskall-Wallis test(5)
was used to determine whether start state has an effect on machine response. This test is
equivalent to an analysis of variance (ANOVA) performed on the ranked data. Unlike the
Jonckheere-Terpstra test, the Kruskall-Wallis test simply tests for differences in response among
the start state alternatives (i.e., the alternative hypothesis is not ordered). Like the Jonckheere-
Terpstra test, it accommodates small sample sizes and sparse data.
For the analysis of response time, a standard ANOVA was used. This allowed testing for the
effect of temperature, humidity, and start state on the response time. To investigate the effect of
temperature, for example, the following model was fit:
Yy = [i + ttj + ey (1)
Here Yy denotes the log of the 1th response time for a given TIC under temperature j. The term [i
is a constant common to all observations, the term Oj denotes the effect of temperature j, and the
term e^ accounts for variation not explained by the model components [i and Oj.
The log response time was modeled because time-to-event measurements are typically skewed to
the right. The log transformation is a standard technique used to achieve normality of error(6)
effects when the data are skewed in such a manner. This model provided the average log
response time under a given set of conditions. This average was transformed back into the
original scale (as opposed to log scale) by exponentiating it. Thus, the modeled geometric mean
of the response times was reported under the given set of conditions. The significance of effects
of interest was tested by evaluating the corresponding coefficients in the model. Thus, to test
whether temperature had an effect on log response time, a standard F test was used to test
whether «j is equal to zero for all j. For more information on the ANOVA approach, see Kirk.(6)
The analysis of recovery time was similar to that of response time unless there were recovery
times that were "censored." When the M90-D1-C did not recover within the maximum allotted
time of 600 seconds, that particular recovery time was considered censored. In a censored
model, instead of assuming that the log recovery times, Yy, have a joint normal density function,
the likelihood for the vector of recovery times, Y, is assumed to be:
24
-------
where C is the collection of censored observations and C* is the collection of uncensored
observations. Here g is a normal density function and S is the "survival" function:
where O is the standard normal distribution function and a is the standard deviation for the
recovery times. The parameter \i represents the common constant; the parameter Oj represents
the effect of treatment j. Once again, effects were investigated by testing the parameters of the
model. Because the model addressed the log recovery times, the geometric mean of the recovery
times was reported.(7)
5.7.2 Analysis of Accuracy
The M90-D1-C response was defined as "accurate" under a given set of conditions if the M90-
Dl-C:
1. Alarmed in the presence of a TIC or CW agent challenge
2. Correctly identified the TIC or CW agent.
The M90-D1-C accuracy was modeled under a given set of conditions via a binomial logit
model.(8) The significance of an effect can be determined by investigating the corresponding
coefficient(s) in the model. For example, to investigate accuracy under different temperatures,
the following model applies:
log(pi/(l-pi)) = n + ai (3)
where p; is the proportion of accurate responses under temperature i. Here a4 again denotes the
effect of temperature i and |_i is the common mean. By testing the significance of the a/s using a
likelihood ratio test, the effect of each factor was tested.
5.1.3 Analysis of Repeatability
For testing the repeatability of response and recovery times for the M90-D1-C, a test of equal
variances was used. Where there is a difference between the variability in response or recovery
times for the different levels of temperature or humidity, there is evidence that temperature or
humidity has an effect on the repeatability of the response or recovery time. The specific test
used to test for equal variances was the Brown-Forsythe test.(6) This test is essentially an
ANOVA run on the absolute deviation from the treatment (level of temperature or humidity)
medians.
For testing repeatability of the M90-D1-C response, an approach was used that took into
account the categorical nature of the response data. For all responses observed under a given set
of conditions, the mode (the most common response) was computed. The number of observed
25
-------
responses that equaled that mode was then determined. Thus, the proportion of responses
equaling the most common response was the measure for the M90-D1-C response repeatability.
This proportion was modeled using a binomial logit model.
5.1.4 False Negatives and Positives Analysis
To test whether interferents caused false negatives in the M90-D1-C response, Dunn's non-
parametric multiple comparisons procedure was used.(9) To employ this procedure, the responses
for all interferent tests are ranked (ties receive average ranks). The test statistic, which is
asymptotically normal, is then:
R -Rr
1 C (4)
where R4 is the average rank for interferent i, Rc is the average rank for no interferent, n4 is the
number of tests for interferent i, nc is the number of tests for no interferent, and N = n; + nc. The
smaller this test statistic is, the greater the evidence that the given interferent is creating a false
negative response.
To investigate the proportion of false positives, a Clopper-Pearson approach(10) was used. To
estimate the rate of false positives, the sample proportion was used (i.e., the number of false
positives divided by the number of trials). Along with this point estimate, a measure of its
uncertainty was calculated in the form of a 95% confidence interval. Simply because the process
did not register a false positive for a particular interferent does not guarantee that it would never
register a false positive for that interferent. This methodology makes an effort to quantify such a
possibility by determining bounds for the false positive rate estimate based on its value and the
number of trials. By assuming that the response obtained was representative of M90-D1-C
performance, the individual tests may be modeled as a binomial distribution, and standard
methods of confidence interval estimation may be employed. The Clopper-Pearson "exact"
interval is commonly used in such instances. Its endpoints are directly calculated from the
binomial distribution without approximation.
5.7.5 Analysis of Response to Alternating Concentrations
This analysis addressed the M90-D1-C response to varying concentrations of a target TIC or
CW agent. As described in Section 3.3.1, the test procedure involved sequentially sampling
clean air followed by high (H) and low (L) concentrations in varying order (i.e., clean
air/H/L/clean air/L/H/clean air/H/L. ..). The data analysis involved two separate analyses. The
first analysis assessed whether response to a high (low) concentration preceded by a low (high)
concentration is different from response to a high (low) concentration preceded by clean air. The
assessment was accomplished using a Cochran-Mantel-Hansel statistic.(8) Empty cells for this
contingency table analysis were filled with counts of 0.01 to allow for convergence of the test
statistic. In the second analysis, the difference between the response for the two challenge levels
was investigated. More specifically, when challenged by a high concentration after being
26
-------
challenged by a lower concentration, the machine response should increase. Similarly, when
challenged by a low concentration after being challenged by a higher concentration, the machine
response should decrease. The proportion of tests exhibiting this behavior for each target
chemical was recorded. Clopper-Pearson bounds were placed on the probability that the machine
response would increase or decrease as appropriate. Results of this analysis are presented in
Section 6.3.
5.2 Other Analyses
The data used to evaluate the response threshold were the replicate M90-D1-C readings obtained
at each succeeding TIC or CW agent concentration. These data were tabulated, along with the
corresponding reference method data that established the challenge concentration. The response
threshold was determined by inspection as the lowest reference method concentration that
produced a positive M90-D1-C response in all replicate runs. In this evaluation, any positive
M90-D1-C response was taken as detection of the target TIC or CW agent, i.e., M90-D1-C
response of L (low) was sufficient in terms of the response threshold evaluation.
Battery life was assessed. Battery life is reported as the total time from start-up to battery
exhaustion when the M90-D1-C unit was warmed up and operated continuously solely on
battery power at room temperature and 50% RH. This time was measured from initial start-up to
the point when the M90-D1-C no longer responded to a challenge mixture of a nerve agent
simulant in air. Any warnings of impending battery failure provided by the detector are noted in
this test.
27
-------
Chapter 6
Test Results
As discussed in Chapter 5, statistical approaches were used to test for the effects of different
conditions on the M90-D1-C performance. The following sections summarize the results from
this verification. A more detailed presentation of the modeled statistical results is included in
Appendix A of this report.
One M90-D1-C unit was used in all testing. Environics chose to provide a single engineering
unit for this test because Battelle's surety license decontamination requirements prevent
returning the entire instrument after exposure to CW agents. As stated in Chapter 3, the M90-
Dl-C was tested with AC, HD, and GB. In initial testing with AC, it was found that the M90-
Dl-C did not respond, though tested at concentrations up to 346 ppm AC (nearly seven times
the IDLH level). This result was discussed with the vendor, who indicated (contrary to previous
information) that the tested M90-D1-C was not programmed to alarm for AC. It should be noted
that there may be M90-Dl-Cs in use that are programmed to respond to AC. However, this can
only be established definitively for a given unit by contacting the vendor or testing directly with
an AC challenge. No further testing was conducted with AC, but a nerve agent simulant was
used to carry out cold/hot start tests and the battery life test.
The M90-D1-C unit tested also was found not to respond to HD, although it was programmed to
do so and did respond to the vendor-supplied simulants. This was an unexpected result, since a
test report(11) written for the Domestic Preparedness Program by the Soldier and Biological
Chemical Command lists results in which the M90-D1-C responded and alarmed to HD at
concentrations above and below the AEGL-2 level. After discussion with the vendor of the M90-
Dl-C, at the vendor's direction the M90-D1-C was put through a "decontamination" program
provided as part of the M90-D1-C software. The M90-D1-C was connected to a laptop
computer, and the decontamination program was initiated. The M90-D1-C was then subjected to
the decontamination program for one hour. At the conclusion of that hour, the M90-D1-C was
tested again with 5.5 mg/m3 of HD. Again, the M90-D1-C unit did not respond. After this
observation, no further HD testing was performed with the M90-D1-C.
The M90-D1-C was provided to Battelle with Environics' M90-UIP (User Interface Program),
which allows logging of raw detector data during operation via a data cable and computer. After
running the decontamination program, data were collected during the HD challenge and
subsequently e-mailed to Environics for analysis. Environics stated that its review of this raw
data showed that the spectral signature produced by the M90-D1-C did not match the
programmed profile for HD.Environics suggested that this indicated that the HD used in testing
28
-------
may have been contaminated It should be noted that the HD used in this testing was determined
to have a purity of 95.8%.
The M90-D1-C unit did respond as expected to the presence of GB, and the following sections
provide the results from testing the M90-D1-C with GB. The GB used in this testing was
determined to have a purity of 85.1%.
6.1 Response Time
Results of the response time analysis for GB are summarized here and detailed in Appendix A,
Section A. 1. Table 6-1 summarizes data used for the analysis of response time and other
performance parameters.
Table 6-1. GB Test Results
Environmental
CW Agent(a) Conditions
GB Control (22°C/50% RH)
22°C/<20% RH
22°C/80% RH
35°C/50% RH
35°C/80% RH
5°C/50% RH
M90-D1-C
Response
Low (4) - Medium (1)
Low (3) - Medium (2)
Medium
Low
High
Alarms
(Indicated
Chemical)
5/5 (Nerve)
5/5 (Nerve)
5/5 (Nerve)
5/5 NR(C)
4/5 (Nerve)
1/5 NR
6/6 (Nerve)
Response
Time Range
(s)
8-10
8-10
6-7
13-25
3-17
Recovery
Time Range
(s)
15-199
12-110
600(b)
12-28
600(b)
-------
6.2 Recovery Time
Results of the recovery time analysis for GB are summarized here and detailed in Appendix A,
Section A.2. Recovery time results are also illustrated in Table 6-1.
In both the temperature and RH tests of recovery times for GB, recovery times in excess of
600 seconds were observed. The recovery times for all runs of GB at low temperature and
medium humidity exceeded 600 seconds. The recovery times for all runs of GB at room
temperature and high humidity also exceeded 600 seconds.
At medium RH, the recovery time at low temperature was in excess of 600 seconds, whereas the
modeled recovery time at room temperature was about 57 seconds (range of 15 to 199 seconds).
However, the magnitude of the latter mean was greatly influenced by one outlier. For 4 of the 5
runs of GB at medium (50%) RH and room temperature, the M90-D1-C response was "low,"
and the recovery time took 32 seconds or less. For the other run of GB at those conditions, the
M90-D1-C response was "medium," and the recovery time was 199 seconds. These results
indicate longer recovery times for GB at the lower temperature. At the medium RH/high
temperature level, the M90-D1-C did not alarm for GB.
At room temperature (22°C), the modeled recovery time at both low and medium humidity was
approximately 30 seconds (excluding the outlier noted above). However, at high humidity the
recovery time for GB exceeded 600 seconds in all runs. These results indicate longer recovery
times for GB at higher humidity.
6.3 Accuracy
Results of the accuracy analysis for GB are summarized here and described in Appendix A,
Section A.3. Results of tests that involved alternating different challenge concentrations, as
opposed to alternating clean air and a single challenge concentration, are summarized below and
detailed in Appendix A, Section A. 8. Accuracy results are also illustrated in Table 6-1. The
M90-D1-C was considered to be accurate if it alarmed in the presence of the agent and correctly
identified the agent class.
At medium humidity (50% RH), the M90-D1-C performed with 100% accuracy to GB on all
runs at low temperature and room temperature. However, at the medium humidity and high
temperature, 0% accuracy was achieved, since the M90-D1-C did not respond to the presence of
GB.
There was no evidence that relative humidity at room temperature had an effect on the M90-D1-
C accuracy. Over all of the humidity conditions at room temperature for the GB testing, the
M90-D1-C performed with 100% accuracy. At the high temperature and high humidity
condition, an accuracy of 80% (4/5) was found.
30
-------
For the alternating concentration tests (described in Section 3.3.1), the high concentration
challenge for GB was 4 x IDLH and the low concentration challenge was 0.5 x IDLH. The
purpose of the alternating concentration test is to assess whether instrument response to a given
concentration is affected by initial exposure to an alternate concentration of GB, as compared to
an initial exposure to clean air. There was no evidence that the M90-D1-C response to GB at a
given concentration was affected by a preceding alternate concentration. The response to GB at
both the high and the low concentrations tended to be "Low" whether or not the GB challenge
was preceded by the alternate GB concentration or by clean air (which produced no response).
6.4 Repeatability
Results of the repeatability analysis are summarized below and detailed in Appendix A,
Section A.4.
Repeatability addressed the consistency of the "Low," "Medium," and "High" readings of the
M90-D1-C for GB. Even though temperature has an effect on the level of the M90-D1-C
response to GB (Section 6.6), there was little evidence of a dependence of variation of response
on temperature. At the low temperature setting, the M90-D1-C consistently alarmed as a "High"
response for GB. At the room temperature setting, the M90-D1-C alarmed as a "Low" response
for GB for four out of the five runs. At the high temperature setting, the M90-D1-C did not
respond at all to the presence of GB at medium humidity. The same is true for the repeatability
of response for the different RH conditions. For the low and medium humidity runs, the
response tended to be "Low" in the presence of GB. For the high humidity response, the
response was consistently "Medium."
There was no evidence that variation in either temperature or humidity had an effect on the
repeatability of the response time for GB. Also, there was insufficient data to assess whether
temperature had an effect on the repeatability of recovery time for GB. At the low temperature/
50% RH setting, the recovery time exceeded the 600 second threshold; and at the high
temperature/50% RH setting, there was no response to the GB and thus no recovery time. An
assessment of the effect of humidity on repeatability of recovery time at room temperature
showed no evidence of a humidity effect for the medium and low humidity settings, but for the
room temperature/high humidity condition, the recovery time exceeded the 600-second
threshold. However, tests at high temperature/high humidity (35°C/80% RH) showed recovery
times of 12 to 28 seconds.
6.5 Response Threshold
Response threshold for GB was determined by challenging the M90-D1-C unit with successively
lower concentrations until it no longer responded. Table 6-2 provides the results for the response
threshold test. The responses listed in the table give the results for three successive challenge/
clean air cycles. For GB, the response threshold was between 0.05 and 0.1 mg/m3 (0.008 and
31
-------
0.017 ppm). This response level is below the IDLH concentration for GB of 0.2 mg/m3 (0.035
ppm).
Table 6-2. Response Threshold Data
Agent (Concentration)
GB (0.75 mg/m3) (0.13 ppm)
GB (0.1 mg/m3) (0.017 ppm)
GB (0.05 mg/m3) (0.008 ppm)
M90-D1-C Response
Low - Medium (Nerve)
Low (Nerve)
No Response
6.6 Temperature and Humidity Effects
The results of investigating temperature and humidity effects on the M90-D1-C response are
summarized here and are detailed in Appendix A, Section A.5. Table 6-1 also illustrates
temperature and humidity effect data.
Temperature had an effect on the M90-D1-C response to GB in the presence of medium (50%)
RH. The M90-D1-C unit responded with a "High" response to all six runs at low temperature.
As the temperature increased, the level of the response decreased. At the high temperature, the
M90-D1-C unit did not respond to the presence of GB.
Humidity over the range of 20 to 80% RH did not appear to affect the M90-D1-C response to
GB at room temperature.
6.7 Interference Effects
The results of investigating interference effects on M90-D1-C response are summarized here and
are detailed in Appendix A, Sections A.6 and A.7. Table 6-3 summarizes data used for the
analysis of interference effects in tests with both GB and an interferent present.
Table 6-3. Interference Effects Data
CW Agent(a)
GB
Interferent
Control
Latex Paint Fumes
Ammonia Floor Cleaner
DEAF
Gasoline Engine Exhaust
Air Freshener
M90-D1-C
Response
Low (4) - Medium (1)
Low
Low
Low (3) - Medium (2)
Low
Low
Alarms
(Indicated
Chemical)
5/5 (Nerve)
5/5 (Nerve)
5/5 (Nerve)
5/5 (Nerve)
5/5 (Nerve)
5/5 (Nerve)
Response
Time
Range (s)
8-10
11-19
11-13
8-12
9-11
9-13
Recovery
Time
Range (s)
15-199
17-249
19-187
14-369
16-124
11-288
1 Results shown are with GB at the target concentration level in Table 3-1.
32
-------
A false positive response would occur if the M90-D1-C responded and provided an alarm in the
presence of an interferent, but in the absence of GB. A false positive was defined as any alarm
under those conditions. For the five interferents tested, false positive alarms occurred
consistently in the presence of ammonia floor cleaner vapors and latex paint fumes. In only one
of the five DEAE challenges, the M90-D1-C responded with an alarm. The M90-D1-C did not
respond to the presence of engine exhaust hydrocarbons or air freshener vapors.
False negative responses would occur if the presence of an interferent masked the presence of a
TIC or CW agent and the M90-D1-C provided a lower response or did not respond to the TIC or
CW agent. The M90-D1-C responded to all GB challenges when interferents were present; thus,
false negative responses were not observed. Changes in response, response time, and recovery
time due to interferences are discussed in the following paragraphs. The interferents did not
affect the response or identification accuracy of the M90-D1-C to the presence of GB. The
response tended to be a "Low" alarm in the presence of any interferent, and the correct
indication of "Nerve" was always obtained (Table 6-3).
Some of the interferents did have a small affect on the response time of the M90-D1-C to the
presence of GB. The modeled response time for GB went from a control of 9.4 seconds to
12 seconds in the presence of the ammonia floor cleaner and to 13.4 seconds in the presence of
the latex paint fumes. These small increases in response time for GB are of no practical conse-
quence. The modeled response times in the presence of DEAE, gasoline engine exhaust hydro-
carbons, and air freshener vapors were within 2.5 seconds of the control result.
Overall, the interferents also did not affect the recovery time of the M90-D1-C after sampling
GB. There was, however, a great deal of variability in the recovery time data. In general, during
the GB and interferent testing, the recovery time for the first run was greater than 200 seconds.
After the first run, the recovery time significantly decreased. By the third of the five runs in all
cases except with DEAE, the recovery time was less than 60 seconds. The trend showed a
decrease in recovery time upon each successive run, with the shortest recovery time being the
fifth run for four of the five interferents
6.8 Cold-THot-Start Behavior
Qualitative analysis of the effects of insufficient warm-up time, under start-up conditions
ranging from cold (5 to 8°C) to hot (40°C), is summarized here. These tests were conducted with
a nerve agent simulant, and general observations are provided about start-up delay time and
response.
In the room temperature cold-start test, the delay time, or time from powering the M90-D1-C on
until it reached a ready state, was 8 minutes and 17 seconds, as programmed. The M90-D1-C
responded to the simulant as a "Low" nerve alarm. Response times were generally less than 20
seconds, and recovery times ranged from 4 minutes 24 seconds to 33 seconds, showing a
downward trend similar to that in tests conducted with GB (Section 6.7).
33
-------
In the cold temperature cold-start test, the M90-D1-C was powered on and produced a "Failure"
alarm after 8 minutes and 17 seconds. The M90-D1-C maintained the "Failure" alarm for
8 minutes and never reached a ready state. The M90-D1-C was then powered off and left off for
2 minutes. The M90-D1-C was turned on again, and the delay time was 6 minutes and
20 seconds. The M90-D1-C responded to the simulant as a "Low" nerve alarm, with response
times ranging from about 30 to 80 seconds. Recovery times were short, ranging from 6 to
26 seconds.
For the hot temperature cold-start test, the delay time was 8 minutes and 20 seconds. The
M90-D1-C responded to the simulant as a "Low" nerve alarm, with variable response and
recovery times.
6.9 Battery Life
The M90-D1-C can be powered by several types of batteries (see Section 6.10). For this test, the
M90-D1-C was powered with rechargeable nickel metal hydride (NiMH) batteries. The battery
life test was conducted by placing a fully charged battery pack provided by the vendor in the
M90-D1-C. The M90-D1-C was then powered on and allowed to warm up fully according to the
manufacturer's directions. The battery life tests were conducted with a nerve agent simulant.
The M90-D1-C responded to the simulant as a "Low" nerve alarm. The "Low Battery" light
came on 1 hour and 47 minutes after start-up and quickly switched to a "Failure" alarm. At this
time, the M90-D1-C did not respond when challenged with the simulant. After completion of
verification testing, the vendor stated that the battery life documented in this report did not
correspond with Environics' data and suggested that the reason for this difference was that the
rechargeable battery provided by Environics for testing was not new.
6.10 Operational Characteristics
General performance observations noted during verification testing:
• Instrument Operation—The M90-D1-C has two caps that must be removed for the
M90-D1-C to operate properly, the air inlet cap and the air outlet cap. After these caps
are removed, the M90-D1-C can be powered on by switching the power/test switch from
the "Off to "On" position. This switch has two other options, SCCell and EVICell. These
options are to be used when testing the M90-D1-C with a simulant to ensure proper
operation. The M90-D1-C also has a separate switch to control the volume of the audio
alarm.
• Instrument Indicators—The M90-D1-C has several lighted indicators to show the status
of the detector. These indicators include Nerve, Blister, Blood, High, Med, Low, Low
Batt, Failure, and Power Mode. When the M90-D1-C alarms to a challenge, it will
indicate both the type of alarm (Nerve, Blister, or Blood) and the level of the alarm
(High, Medium, or Low) by lighting up the lights that correspond to these alarms and by
34
-------
producing an audible alarm. When the M90-D1-C detects a failure within its system, the
"Failure" light and a different audible alarm occur.
Warm-Up—The M90-D1-C generally took 8 minutes or more to reach a ready state after
being turned on, whether starting from cold (5 to 8°C) room temperature or hot (40°C)
storage conditions.
Batteries—The M90-D1-C can operate on several types of batteries. There are two types
of rechargeable batteries (NiMH and nickel cadmium) and two types of one-time-use
batteries (lithium and magnesium).
Errors—The M90-D1-C produced one "Failure" alarm during testing. In that case, after
soaking at a cold temperature overnight, the M90-D1-C produced a "Failure" alarm
during its warm-up period. The M90-D1-C was turned off, then restarted, and reached a
ready state within about 6 minutes.
Vendor Support—Before the verification testing, a vendor representative trained Battelle
employees to operate the M90-D1-C. Testing proceeded according to the vendor's
recommendations on how to operate the M90-D1-C for testing. The vendor also
responded promptly when information was needed during the verification testing.
Cost—The list price of the M90-D1-C, as used in this verification test, is approximately
$17,500.
35
-------
Chapter 7
Performance Summary
This chapter summarizes the overall performance results found in testing of one unit of the
M90-D1-C with one TIC and two CW agents. This summary focuses on aspects of the
performance that are most important in field use of the M90-D1-C by first responders.
Consistent with that use, test procedures were conducted with challenge levels of the target
chemicals that were at or near IDLH concentrations.
The M90-D1-C was tested with AC, HD, and GB. However, contrary to prior indications from
the vendor, the M90-D1-C was not programmed to respond to AC. Also, the unit did not
respond to HD challenges, although it was programmed to do so and did respond to the vendor-
supplied simulants.
For GB, the M90-D1-C response time was minimally affected by temperature or humidity, with
response times usually 10 seconds or less. However, in six of ten runs at the high temperature
level (35°C), the M90-D1-C did not alarm for GB. The recovery times were about 30 seconds in
most cases, but exceeded 600 seconds for all runs at low temperature (5°C) and medium
humidity (50% RH) and for all runs at room temperature (22°C) and high humidity (80% RH).
The M90-D1-C identified GB accurately in most temperature and humidity conditions tested,
and in all tests with interferents also present. The overall accuracy of identification in all tests
was 91% (60/66) for GB. However, at the high temperature (35°C), the M90-D1-C did not
respond to the presence of GB in six of 10 test runs (0/5 accurate responses at medium humidity
and 4/5 accurate responses at high humidity). The M90-D1-C response at a given GB concen-
tration was unaffected by a preceding higher or lower concentration.
Except for the absence of GB response in some tests at 35°C, there was no evidence that
variation in either temperature or humidity had an effect on the repeatability of the response or
response time for GB. Data were insufficient to assess whether temperature had an effect on the
repeatability of recovery time for GB.
For GB, the M90-D1-C response threshold was between 0.05 and 0.1 mg/m3 (0.008 and
0.017 ppm), which is below the IDLH concentration for GB of 0.2 mg/m3 (0.035 ppm).
Temperature had an effect on the M90-D1-C response to GB. As the temperature increased with
a 50% RH, the level of the response decreased. At the high temperature (35°C), the M90-D1-C
unit did not respond to the presence of GB. Humidity did not affect the M90-D1-C response to
GB.
36
-------
Ammonia floor cleaner vapors and latex paint fumes consistently produced false positive alarms
when sampled by the M90-D1-C. However, none of the interferents had an effect on the
response to GB when the agent and interferent were sampled together. Interferents did not
significantly affect the response time or recovery time of the M90-D1-C in sampling GB. A
decrease in recovery time was observed upon each successive run, with the shortest recovery
time occurring in the last test run for four of the five interferents.
In the room temperature cold-start test using a nerve agent simulant, the delay time was
8 minutes and 17 seconds, as programmed. In the cold temperature cold-start test, the M90-D1-
C produced a "Failure" alarm after 8 minutes and 17 seconds and never reached a ready state.
After being powered off for 2 minutes, the delay time was 6 minutes and 20 seconds. For the hot
temperature cold-start test, the delay time was 8 minutes and 20 seconds. In all three tests, the
M90-D1-C responded to the simulant as a "Low" nerve alarm.
The battery life test was conducted by powering on a fully charged NiMH battery pack and
allowing the M90-D1-C to warm up fully, then operate continuously until battery power was
depleted. The battery life test was conducted with a nerve agent simulant. The M90-D1-C
responded to the simulant as a "Low" nerve alarm. At 1 hour and 47 minutes after start-up the
"Low Battery" light came on, followed immediately by a "Failure" alarm. At this time, the M90-
Dl-C did not respond when challenged with the simulant.
The M90-D1-C has two caps that must be removed for it to operate properly. The power/test
switch has two options other than On/Off, which are to be used when testing the M90-D1-C
with a simulant to ensure proper operation. The M90-D1-C also has a separate switch to control
the volume of the audio alarm. The M90-D1-C has several lighted indicators (Nerve, Blister,
Blood, High, Med, Low, Low Batt, Failure, and Power Mode) to show the status of the detector
and took 8 minutes or more to reach a ready state after being turned on. It can operate on two
types of rechargeable batteries (NiMH and nickel cadmium) and two types of one-time-use
batteries (lithium and magnesium). The M90-D1-C produced only one "Failure" alarm during
testing, i.e., that during the cold temperature cold start noted above.
37
-------
Chapter 8
References
1. Test/QA Plan for Verification of Portable Ion Mobility Spectrometers for Detection of
Chemicals and Chemical Agents in Buildings, Version 2, Battelle, Columbus, Ohio,
June 2004.
2. Quality Management Plan (QMP)for the ETV Safe Buildings Monitoring and Detection
Technology Verification Program, Version 1, Battelle, Columbus, Ohio, June, 2004.
3. Spicer, C. W., Gordon, S. M., Holdren, M. W., Kelly, T. J., and Mukund, R., Hazardous Air
Pollutant Handbook: Measurements, Properties, and Fate in Ambient Air, ISBN 1-56670-
571-1, CRC Press, Boca Raton, Florida, 2002.
4. Turpin, J. "Direct Steam Injection Humidification: Is It Safe for Building Occupants?"
Engineered Systems, www.esmagazine.com/CDA/ArticleInformation/features/BNP
Features_Item/0,2503,23246,OO.html.
5. Sprent, P. and Smeeton, M. C. Applied Non-Parametric Statistical Methods, Chapman and
Hall, New York, 2001.
6. Kirk, R. E. Experimental Design: Procedures for Behavioral Sciences, Third Edition,
Brooks/Cole Publishing Co., 1995.
7. Klein, J. P. and Moeschberger, M. L., Survival Analysis: Techniques for Censored and
Truncated Data, Springer, New York, 1997.
8. Agresti, A. Categorical Data Analysis, John Wiley and Sons, New York, 1990.
9. Dunn, O. J., "Multiple comparisons using rank sums." Technometrics 6, 241-252, 1964.
10. Clopper, C. J. and Pearson, E. S.,"The use of confidence or fiducial limits illustrated in the
case of the binomial," Biometrika 26, 404-13, 1934.
11. Testing ofM90-Dl-C Chemical Warfare Agent Detector Against Chemical Warfare Agents
Summary Report, Soldier and Biological Chemical Command, Aberdeen Proving Ground,
MD, http://hld.sbccom.army.mil/ip/md90_dl_detector_download.htm.
38
------- |