&EPA
EPA/600/R-15/122 | May 2015 | www.epa.gov/ord
U nited States
Environmental Protection
Agency
Next Generation Air
Monitoring (NGAM) VOC
Sensor Evaluation Report
[Office of Research and Development
[National Exposure Research Laboratory
-------
EPA/600/R-15/122 | May 2015 | www.epa.gov/ord
Next Generation Air Monitor (NGAM)
VOC Sensor Evaluation Report
Ron Williams
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC, USA 27711
Amanda Kaufman
ORISE Participant
Oak Ridge Institute for Science and Education
Oak Ridge, TN, USA 37831
Sam Garvey
Alion Science and Technology
Research Triangle Park, NC 27709
-------
Disclaimer
This technical report presents the results of some work performed by Alion Science and
Technology under contract EP-D-10-070 for the Human Exposure and Atmospheric Sciences
Division, U.S. Environmental Protection Agency, Research Triangle Park, NC. It has been
reviewed by the U.S. Environmental Protection Agency and approved for publication. Mention
of trade names or commercial products does not constitute endorsement or recommendation for
use.
-------
Acknowledgments
The NERL's Quality Assurance Manager (Sania Tong-Argao) and associated staff
(Monica Nees) are acknowledged for their excellent contributions to the development of
sophisticated standard operating procedures used in collection of the data. This research was
supported in part by an appointment to the Research Participation Program for the U.S.
Environmental Protection Agency, Office of Research and Development, administered by the
Oak Ridge Institute for Science and Education through an interagency agreement between the
U.S. Department of Energy and EPA (DW 8992298301). Don Whitaker and Karen Oliver (U.S.
EPA) are acknowledged for their contributions involving laboratory analysis of supporting VOC
reference measurements. Sue Kimbrough, Eben Thoma, Bill Squier, and Bill Mitchell (US EPA)
are acknowledged for their provision of the APPCD VOC sensor or their provision of the
ambient monitoring site (Triple Oaks monitoring platform). Chris Hall, Stacey Henkle, and Zora
Drake-Richmond, (Alion Science and Technology) are acknowledged for their contributions in
supporting the U.S. EPA in the execution of data collections and initial data summary.
IV
-------
Abstract
This report summarizes the results of next generation air monitor (NGAM) volatile
organic compound (VOC) evaluations performed using both laboratory as well as field scale
settings. These evaluations focused on challenging lower cost (<$2500) NGAM technologies to
either controlled or ambient conditions. The cost ceiling applied to the technologies selected for
evaluation reflected a value believed to be the limit to what citizen scientists might seek to obtain
for their use. The work conducted here and the summary of findings is not meant to be a
definitive description of all such technologies. It represents a first step in understanding the
capabilities of lower cost VOC technologies and their limitations.
An exhaustive search of commercially-available VOC NGAM products under the $2500
limit yielded a very modest number of devices available for inclusion in the research. Ultimately
a total of five (5) devices were incorporated into the evaluation with one of those being an EPA
developed device which used a commercially-available VOC photoionization detector (PID) as
the sensing element.
The laboratory evaluations involved challenging the devices to a stepwise pattern of VOC
concentrations at levels believed to be environmentally relevant (< 25 ppb) using a chamber.
Reference gas chromatographic (GC) detection was utilized to verify the challenge conditions
being established. The devices were first evaluated for their response to a single VOC (benzene).
If the device revealed some ability to detect benzene at even 25 ppb it was then challenged with
an atmosphere consisting of three VOCs (benzene, 1,3-butadiene, and tetrachloroethylene).
These compounds were selected because of the availability of well qualified test gases and the
fact they represented a variety of VOC moieties (structural variability). The response of the
devices to the various challenge conditions are reported.
NGAM devices were deployed at an outdoor near road test platform for an extended
period where wide variability of VOC conditions were expected to exist. The research plan
involved direct comparison of the NGAM response to GC reference data from collocated
measurements obtained at the test site. Reference data were ultimately not available for the
intended comparisons (instrument malfunction and insufficient resources to conduct a timely
repair). Therefore, field data provided here are limited to non-reference comparisons between
NGAM devices. Such comparisons, providing a non-quantitative assessment of true VOC
response still have the potential of yielding useful information on the relative response
characteristics of the NGAM VOC devices evaluated.
-------
Table of Contents
Abstract v
Tables viii
Figures viii
Acronyms and Abbreviations xi
Executive Summary xii
1.0 Introduction 1
2.0 Materials and Methods 2
2.1 Laboratory Evaluations 3
2.1.1 Laboratory Setup 3
2.1.1.1 Exposure Chamber 4
2.1.1.2 Zero Air Generator 4
2.1.1.3 Gas Chromatograph System 5
2.1.1.4 Dynamic Gas Calibrator 5
2.1.2 Evaluation Methods 6
2.1.2. lUniTec SENS-IT 7
2.1.2.2 AirBase CanarIT 8
2.1.2.3CairPolCairClip 9
2.1.2.4APPCDPID 9
2.1.2.5ToxiRAEPro 10
2.2 VOC Field Evaluations 10
2.2.1 UniTec SENS-IT 11
2.2.2 AirBase CanarIT 12
2.2.3 APPCDPID 13
2.2.4 ToxiRAEPro 13
3.0 VOC Laboratory Evaluation Results and Discussion 15
3.1 Verification of Test Atmospheres 15
3.2 Laboratory Evaluation Results 19
3.2.1 UniTec SENS-IT 19
3.2.2 AirBase CanarIT 25
3.2.3 CairPol CairClip 26
3.2.4 APPCDPID 27
3.2.5 ToxiRAEPro 33
vi
-------
4.0 VOC Field Evaluation Results and Discussion 34
4.1 Field Evaluation Reference Data 34
4.2 Field Evaluation Results 34
4.2.1 UniTec SENS-IT 34
4.2.2 AirBase CanarIT 38
4.2.3 APPCDPID 42
4.2.4 ToxiRAEPro 46
5.0 VOC Sensor Evaluation Summary 50
6.0 Study Limitations 54
6.1 Resource Limitations 54
6.1.1 Intra-sensor Performance Characteristics 54
6.1.2 Test Conditions 54
6.1.3 Sensor Make and Models 54
7.0 Research Operating Procedures and Related Quality Assurance Documents 56
VII
-------
Tables
Table 1-1. VOC Sensors Evaluated under WA 4-03 2
Table 2-1. Sensor Exposure Concentrations Used in Laboratory Experiments 5
Table 3-1. Summary of GC-FID Data 18
Table 3-2. UniTec SENS-IT Summary 25
Table 3-3. APPCD PID Summary 32
Table 5-1. Summary of VOC Sensor Laboratory Performance 50
Table 5-2. Summary of VOC Sensor Field Performance 51
Figures
Figure 2.1-1. Laboratory VOC sensor testing setup 3
Figure 2.1-2. Port with long perforated tube attached 4
Figure 2.1-3. Calibration programs used 7
Figure 2.1-4. Orientation of UniTec SENS-IT in test chamber 8
Figure 2.1-5. Orientation of UniTec SENS-IT unit B, AirBase CanarIT, and CairPol CairClip
inside test chamber 8
Figure 2.1-6. Orientation of APPCD PID and ToxiRAE Pro in test chamber 9
Figure 2.2-1. Triple Oaks sampling site with shelters 11
Figure 2.2-2. UniTec SENS-IT oriented in its sampling shelter with the lid up 12
Figure 2.2-3. AirBase CanarIT installed at the near-road site 12
Figure 2.2-4. APPCD PID oriented on its shelter with the lid up 13
Figure 2.2-5. ToxiRAE Pro oriented in its sampling shelter with the lid up 14
Figure 3.1-1. GC-FID benzene area counts vs. time 16
Figure 3.1-2. GC-FID 1,3-butadiene area counts vs. time 16
Figure 3.1-3. GC-FID tetrachloroethylene area counts vs. time 17
Figure 3.1-4. Final GC-FID benzene area counts vs. set point in all seven single-component
benzene tests 17
Figure 3.1-5. Final GC-FID benzene area counts vs. set point in three-component
mixture tests 18
Figure 3.2.1-1. UniTec SENS-IT voltage, temperature, and humidity during an extended
run of zero air 19
Figure 3.2.1-2. UniTec SENS-IT data and GC-FID benzene data from the single-component
benzene test on April 9,2014 20
Figure 3.2.1-3. Three UniTec SENS-IT tests using the benzene-only tank 20
Figure 3.2.1-4. Four UniTec SENS-IT tests using the three-component mixture 21
viii
-------
Figure 3.2.1-5. UniTec SENS-IT response in both peaks and troughs vs. set point 22
Figure 3.2.1-6. UniTec SENS-IT vs. measured benzene concentration 22
Figure 3.2.1-7. Response factors and offsets for all UniTec SENS-IT tests for both
crests and troughs 23
Figure 3.2.1-8. Response factors and offsets for all UniTec SENS-IT tests for both
crests and troughs, normalized 24
Figure 3.2.1-9. UniTec SENS-IT vs. set point for both benzene-only and three-component
tests 24
Figure 3.2.2-1. AirBase CanarIT signal over time during a three-component mixture test 26
Figure 3.2.3-1. CairPol CairClip overtime during a three-component mixture test 26
Figure 3.2.4-1. APPCD PID response overtime during the benzene test on April 23, 2014 27
Figure 3.2.4-2. APPCD PID response over time during the three-component mixture
test on April 22, 2014 28
Figure 3.2.4-3. APPCD PID response overtime after baseline subtraction 28
Figure 3.2.4-4. APPCD-PID response overtime after baseline subtraction 29
Figure 3.2.4-5. APPCD PID response vs. set point after baseline subtraction 29
Figure 3.2.4-6. Baseline-corrected APPCD PID response vs. benzene concentration
measured by the GC-FID 30
Figure 3.2.4-7. APPCD PID raw data vs. set point for three-component mixture test
on April 22, 2014 30
Figure 3.2.4-8. Raw APPCD PID data vs. set point for all tests 31
Figure 3.2.4-9. All response factors for APPCD PID tests vs. date 31
Figure 3.2.4-10. All offsets for APPCD PID tests vs. date 32
Figure 4.2.1-1. UniTec SENS-IT field data over time 35
Figure 4.2.1-2. UniTec SENS-IT vs. time of day 35
Figure 4.2.1-3. UniTec SENS-IT vs. AirBase CanarIT 36
Figure 4.2.1-4. UniTec SENS-IT vs. APPCD PID 36
Figure 4.2.1-5. UniTec SENS-IT vs. ToxiRAE Pro 37
Figure 4.2.1-6. UniTec SENS-IT vs. temperature 37
Figure 4.2.1-7. UniTec SENS-IT vs. RH 38
Figure 4.2.2-1. AirBase CanarIT field data overtime 39
Figure 4.2.2-2. AirBase CanarIT vs. time of day 39
Figure 4.2.2-3. AirBase CanarIT vs. UniTec SENS-IT 40
Figure 4.2.2-4. AirBase CanarIT vs. APPCD PID 40
Figure 4.2.2-5. AirBase CanarIT vs. ToxiRAE Pro PID 41
Figure 4.2.2-6. AirBase CanarIT vs. temperature 41
ix
-------
Figure 4.2.2-7. AirBase CanarIT vs. RH 42
Figure 4.2.3-1. APPCD PID field data overtime 43
Figure 4.2.3-2. APPCD PID vs. time of day 43
Figure 4.2.3-3 APPCD PID vs. UniTec SENS-IT 44
Figure 4.2.3-4. APPCD PID vs. AirBase CanarIT 44
Figure 4.2.3-5. APPCD PID vs. ToxiRAEPro 45
Figure 4.2.3-6. APPCD PID vs. temperature 45
Figure 4.2.3-7. APPCD PID vs. RH 46
Figure 4.2.4-1. ToxiRAE Pro field data over time 47
Figure 4.2.4-2. ToxiRAE Pro vs. time of day 47
Figure 4.2.4-3. ToxiRAE Pro vs. UniTec SENS-IT 48
Figure 4.2.4-4. ToxiRAE Pro vs. AirBase CanarIT 48
Figure 4.2.4-5. ToxiRAE Pro vs. APPCD PID 49
Figure 5-1. All total VOC sensors vs. time of day 52
-------
Acronyms and Abbreviations
AC alternating current
°C degrees Celsius
FID flame ionization detector
GC gas chromatograph
GSM Global System for Mobile Communication
h hour
in. inch
L liter
|im micrometer
m meter
min minute
mm millimeter
MS mass spectrometer
NERL National Exposure Research Laboratory
NIST National Institute of Standards and Technology
NRMRL National Risk Management Research Laboratory
o.d. outer diameter
ORD Office of Research and Development
PID photoionization detector
ppb parts per billion
ppm parts per million
QAPP quality assurance project plan
r2 coefficient of determination
RH relative humidity, i.e., water vapor content of air expressed as a percentage of vapor
pressure of water at a given temperature and pressure
RSD relative standard deviation
SD secure digital
SIM subscriber identity module
V voltage
VOC volatile organic compound
WA work assignment
XI
-------
Executive Summary
Background
Low cost air quality sensors are indicative of emerging technologies that have a wide
appeal to both professional researchers and citizen scientists. They exist in numerous
configurations (e.g., cell phone, hand-held) and are often available with a wide assortment of
sensor configurations. One area of increasing interest represents NGAM devices that might have
the ability to detect either total VOCs (tVOCs) or specific VOCs (e.g., benzene) in a continuous
fashion and at low cost. While the commercial availability of such devices has increased in the
near recent past, uncertainty about the quality of data such devices might be capable of providing
has been raised.
Study Objectives
The U.S. Environmental Protection Agency (U.S. EPA) sought to discover what NGAM
VOC devices existed that might be available for citizen scientists and others at a relatively low
cost (<$2500). This cost threshold being the upper ceiling that citizen scientist and others might
consider acceptable for a lower cost (non-reference) continuous monitor. It should be recognized
that the value cited is arbitrary but would seem to capture conversations with citizen scientists
concerning such thresholds.
The study objectives involved both establishment of general findings as well as
analytically-based evaluations. One such objective was the performance of a market survey to
discover representative versions of the VOC sensors available under the aforementioned cost
threshold. All devices considered had to have the capability of detecting and reporting either
tVOCs or specific VOCs in a continuous fashion without any laboratory analyses. Analytical
objectives included the determination of response characteristics of the NGAM devices under
controlled laboratory evaluations. Response characteristics included the response factor, linearity
of the response, and precision of the response across the range of evaluation concentrations. The
relative response of the devices to both a single (benzene) as well as a multi-component test gas
were performed. Lastly, collocated operation at a near road test site was performed to establish
the relative performance of the devices under a widely variable real-world test scenario.
Study Approach
Direct communication with sensor manufactures was pursued to gain an understanding of
the availability, cost, and potential performance features of NGAM VOC sensors. It was
apparent that most of the VOC sensing technology available for purchase fell into only a few
XII
-------
categories based upon an extensive literature review1. The first was the occupational exposure
market. These represented mostly photoionization detector-enabled (PID) devices often costing
in the range of $2-5K per copy. The next category we discovered was high end devices (fast
response gas chromatography or UV-optical) devices with retail costs often exceeding $10K.
The last category we discovered involved devices potentially having retail costs below $2500
and which consumers (citizen scientists) might be able to acquire. This category was extremely
limited in both the number and variety of devices suitable for our purposes. The vast majority of
the true sensing elements in this category consisted of PIDs with one thick film sensor element
(UniTec) identified. Ultimately a total of five NGAM devices were obtained that provided a
general representation of those deemed available at the lower cost range of the market. These
devices consisted of a UniTec SENS-IT (thick film sensor), Cairpol CairClip (NM-VOC-PID),
CanAirlT-PID, Toxi-RAE Pro-PID, and an in-house developed sensor employing the latest
generation miniature PID (APPCD with blue Mocon PID). Standard operating procedures were
developed for each of these devices and telecommunication/data storage/recovery procedures
were established to facilitate data collections in both a chamber and field environment.
Each of the devices were challenged to known concentrations of a challenge VOC gas
containing laboratory established concentrations of benzene and as a follow-up a tri-mix of
multiple VOCs (benzene, 1,3-butadiene, tetrachloroethylene). The follow-up chamber evaluation
was performed only if the device provided satisfactory response to the original benzene
challenge (linearity of response and sensor sensitivity). Data review and reduction was
performed summarizing the sensor's response to the various challenge condition.
A second phase of the proposed research approach was operation of the NGAM devices
for those showing the potential to provide useful (even indicative) indications of changing VOC
concentrations in comparison with collocated reference grade monitors (in-line gas
chromatography) under real world conditions2. In our situation, this was a mobile source test
platform in close proximity to a major interstate highway. The deployment of the NGAM
devices was successful and data collection for approximately 30-60 days for each device was
accomplished. Failure of the collocated research grade reference monitors at this test site
ultimately prevented any direct comparison of the NGAM response with reference data so no
summarization of that comparison could be performed. Intra-comparison of the NGAM devices
and their relative responses were summarized to provide some understanding of how these
sensors might relate to one another.
1 MacDonnell, M., M. Raymond, D. Wyker, M. Finster, Y. Chang, T. Raymond, B. Temple, M. Scofield, D. Vallano, E.
Snyder, AND R. Williams. Mobile Sensors and Applications for Air Pollutants. U.S. Environmental Protection Agency,
Washington, DC, EPA/600/R-14/051 (NTIS PB20I4 105955), 2014.
2 Williams, R., Vasu Kilaru, E. Snyder, A Kaufman, T. Dye, A. Rutter, A. Russell, AND H. Hafner. Air Sensor
Guidebook. U.S. Environmental Protection Agency, Washington, DC, EPA/600/R-I4/I59, 2014.
xiii
-------
Sensor Performance Results Associated with Chamber Challenge
A wide range of potential detection limits would appear to be evident with the NGAM
technologies evaluated. For example, the ToxiRAE-PRO showed no response to challenge
concentration at levels at or below 25 ppb (the maximum chamber concentration established). On
the other hand, the in-house fabricated device employing a Mocon "blue" PID appeared to have
sensitivity capabilities at concentrations below 2.5 ppb. Other devices showed outputs that did
not provide a consistent response to allow determination of detection limits. For example, the
CanAirIT provided a response under known challenge conditions that exhibited a widely
fluctuating series of spikes that had no relationship with the true chamber condition. Likewise
the CairClip NMVOC device provided no data useful for establishing detection limits under the
challenge conditions employed. One of the more interesting devices, the UniTEC SENS-IT,
employing a proprietary thick film sensing element, appeared to be capable of detecting VOC
concentrations between 5 and 25 ppb. Even so, this device had a near sinusoidal electrical output,
especially at concentrations below 10 ppb that made it difficult to truly ascertain detection limits
and precision.
Ease of Use Features Evaluation
The field evaluations (collocation at the near road monitoring platform) provided the best
opportunity to examine ease of use features. The devices deployed had to be provided weather
shielding as they did not come factory-ready for the all-weather environment. In our situation, a
rain helmet was utilized to shield the device from precipitation while ensuring the sensor element
had open access to the ambient environment. It should be stated that the impact of ambient
temperature on the devices deployed at this site was not established. Correlation between
temperature and sensor output was determined but due to the lack of reference data no specific
impact of temperature upon estimating VOC concentrations could be established. It is anticipated
that such effects would exist in some degree due to the nature of the sensor element itself and the
reported range of operation manufacturers specify with these devices.
Of interest to the potential users of these devices are some of the ease of use features
associated with electrical requirements. We needed to establish long time period operational
status so none of the units were operated using any internal battery configuration. The UniTec
SENS-IT required a direct AC power supply and access to a third-party data acquisition system.
Concerning the latter, this represents a significant limitation on the mobility one might have with
deploying this under a variety of field conditions. The CanAirIT device was proven to be easy to
deploy once the initial set-up and operation was established. It too requires AC power for
extended operational times and connection to its proprietary data server. The CairClip NMVOC
proved itself to be easy to operate as many of the CairPol products appear to function. A mini-
USB connection provided the means to establish long-term power to the unit. The unit we tested
had a malfunctioning microfan for its inlet which resulted in a shortened period of operational
status. The in-house produced APPCD PID provided some of the best ease of use features we
encountered. It is a rather large device (6 inches wide) in comparison to many of the other hand-
size sensors we examined. Even so, it was not unwieldy and we were able to easily accommodate
its size in both the chamber and field deployment activities. This sensor does require direct 115V
access and therefore battery or other non-AC power supplies are not applicable. It should be
xiv
-------
clearly stated that the most robust device examined was the ToxiRAE Pro relative to the field
environment. This commercially available product represents one of many similar devices often
sold to occupational safety specialists. It has numerous worthy features such as a high impact
casement, highly adaptable user interface (keypad) to allow operator control of detection
parameters, and built-in calibration programs. This device was tested specifically here to see if
such a device might have any viability in the low concentration (environmentally relevant)
conditions where citizen scientists might wish to attempt VOC measurements.
Conclusions
This early examination of the "best of the best" lower costNGAM VOC devices clearly
indicates that some of these sensors may provide indicative response to environmentally relevant
concentrations of VOCs. As an example, the target VOC benzene often observed in urban
environments at concentrations well below 25 ppb would suggest that only two of the devices
evaluated (APPCD-PID and UniTec SENS-IT) showed detection capabilities in this general area
that would be useful for non-industrial environmental monitoring. Even so, we observed issues
with noisy response outputs that limited the usability of some of the named devices. It should be
clearly stated that there may be circumstances where any of the devices evaluated might provide
some value to the end-users, especially when all of the key features such as ease of use,
simplicity of operation, and ease of data recovery have been considered. In conclusion, lower
cost NGAM VOC technologies would appear to be limited in both their capability and variety of
technologies being employed in this market segment. PID components would still appear to be
the prominent sensing element available and as such, these come with inherent pros and cons
which one must consider in trying to use such a device to estimate VOC concentrations under a
variety of monitoring scenarios.
xv
-------
1.0 Introduction
EPA's Office of Research and Development (ORD) has been continuing a body of
research associated with examining emerging technologies that might prove useful for citizen
scientist as well as professionals interested in use of lower cost next generation air monitor
(NGAM technology. A majority of the aforementioned research has been summarized and has
been made available through a primary data portal, the Air Sensor Citizen Science Toolbox
(http://www.epa.gov/heasd/airsensortoolbox/)3. To this end, NERL has been continually seeking
out novel sensor technologies for the measurement of pollutants of interest. One area of growing
interest is that involving volatile organic compounds (VOCs). ORD has received numerous
inquiries from Regional offices, Program Offices, State Air Offices, industrial concerns, and
community groups about the state of the science for this type of sensing technology. To meet this
stakeholder need, primary research was conducted that attempted to discovery the availability of
lower cost VOC monitors and provide preliminary assessments of its capabilities under known
conditions (chamber evaluations). Operation under ambient conditions was also conducted to
determine some of the ease of use characteristics stakeholders need to understand.
This report documents efforts to survey sensor/application technologies for the
measurement of volatile organic compounds (VOCs) through direct contact with inventors and
commercial and research institutions. This project aimed to provide data for identifying which
technologies might prove valuable in conducting air quality measurements under a variety of
conditions. Five VOC sensors were ultimately selected and secured for evaluation under
laboratory and field conditions. This report details the materials and methods used to conduct the
evaluations (Section 2) and results for both the laboratory (Section 3) and the field (Section 4)
components. Section 5 summarizes our evaluation findings.
3 The U.S. EPA's Air Sensor Toolbox for Citizen Scientists. Website available at
www.epa.gov/heasd/airsensortoolbox/. 2015
-------
2.0 Materials and Methods
The five VOC sensors selected for evaluation are listed in Table 1-1. All sensors were
first evaluated in the laboratory under controlled conditions. These same sensors, with the
exception of the CairPol CairClip, were then deployed to the field for assessment of VOC
response under ambient conditions. The CairClip was not evaluated in the field because it
developed mechanical problems during the laboratory evaluation, and resource constraints
prevented acquisition of a new unit.
Table 1-1. VOC Sensors Evaluated under WA 4-03
UniTec SENS-IT
AirBase CanarIT
CairPol CairClip
APPCD PID
ToxiRAE Pro
UniTec SENS-IT
AirBase CanarIT
APPCD PID
ToxiRAE Pro
The evaluation focused on the following performance characteristics of each sensor:
Linearity of response
Precision at each concentration range point
Lowest established concentration in which a response was detected
Concentration resolution
Response time (if it could be easily established)
Suggested range of operation to achieve best practical operation conditions
The effects of temperature, relative humidity (RH), pollutant atmospheres, and interfering
species on sensor performance were also considered in the evaluation.
Unless otherwise noted in the text, the values used for statistical analysis were the last
5-min average values at each input concentration level, or set point, which is the target or desired
value as opposed to the measured concentration. This ensured maximum stabilization time and
generated one value per set point per test.
The coefficient of determination (r2) is the square of the sample correlation coefficient
and was used as a measure of linearity. Microsoft Excel was used to calculate r2 by plotting all
measured values against their set points, which displayed the linear regression. The same plot
was used to determine response factors and offsets.
Relative standard deviation (RSD) was used to define precision at a given set point. For
each sensor, measured values at each set point for all tests were examined together. For example,
all measured values for the 25 ppb target concentration for the UniTec SENS-IT comprised one
data set. For each data set, the standard deviation was divided by the mean to calculate the RSD.
-------
2.1 Laboratory Evaluations
The five sensors were independently challenged in triplicate with four concentrations
each of benzene only or a three-component VOC mixture containing benzene, 1,3-butadiene, and
tetrachloroethylene in a temperature and RH controlled exposure chamber. The sensors were
evaluated for responsiveness, linearity, and precision. Data ultimately in hand revealed that our
ability to establish response time (in a very definitive degree) was not a practical statistical
operation due to the pattern of response or the conditions of the evaluation parameters (chamber
characteristics). Concerning this latter statistical output, the graphical data provided to the reader
might still provide some information of use in practical response times but we make no attempt
to formally calculate these values as specified above. This section describes the laboratory setup
and the methods used for testing the sensors in the laboratory.
2.1.1 Laboratory Setup
The laboratory setup shown in Figure 2.1-1 was used for evaluating the performance of the
sensors. The setup consisted of an exposure chamber, a model 111 zero air generator (Thermo
Scientific, Waltham, MA), and a 146C dynamic gas calibrator (Thermo Scientific). The
temperature was maintained during test runs at the ambient conditions of the climate-controlled
laboratory (-20 to 25 °C). An all-glass impinger was used to humidify the zero air. The test
atmospheric conditions within the exposure chamber were verified using a temperature/humidity
probe based on the Honeywell Hffl-4602-C. An Agilent Technologies (Santa Clara, CA) 7890A
gas chromatograph with flame ionization detector (GC-FID) using an Entech Instruments (Simi
Valley, CA) 7200 pre-concentrator was used to monitor the gas concentrations within the chamber.
Figure 2.1-1. Laboratory VOC sensor testing setup: exposure chamber (top), Thermo Scientific
146C gas calibrator (middle), and Thermo Scientific model 111 zero air generator (bottom). The
impinger is shown to the left of the setup
-------
2.1.1.1 Exposure Chamber
The exposure chamber was a 52-L stainless steel test chamber with two ports that were
used as the inlet and outlet ports for the experiments. Through each port, a long perforated steel
tube extended deep inside the chamber, as shown in Figure 2.1-2. The perforated tube assisted
with even diffusion of the challenge gas in the chamber.
Figure 2.1-2. Port with long perforated tube attached
Four ports were added to the lid of the chamber to allow 1/8-, 1/4-, and 3/8-in. (two ports)
o.d. Teflon lines to be connected to the chamber. Any cables needed for operating the sensors
were run through the two 3/8-in. sampling ports. The number and size of cables required varied
significantly depending on which sensors were being tested. Adapters and fittings were added to
the ports to reduce the diameter to a size that would more closely fit the cabling. In addition,
parafilm was used inside compression fittings and wrapped around cables and the opening where
the cables emerged from the fittings to achieve a seal. A chamber test was conducted that
confirmed the parafilm did not outgas detection limit of concentration of VOCs (< Ippb) and
thus could be used for sealing ports without affecting measurements.
The chamber was pressurized and a leak test was performed. The chamber was cleaned
prior to exposures. The inside of the chamber was first wiped with acetone, then with ethanol,
and finally three times with deionized water. The empty test chamber atmosphere was measured
via GC-FID, which showed no measurable VOCs at the 1 ppb threshold as being present in the
chamber.
2.1.1.2 Zero Air Generator
Zero air was introduced into the exposure chamber by passing in-house pressurized air
through a Thermo Scientific model 111 zero air generator. The zero air was then passed through
an all-glass impinger filled with deionized water to humidify the air. A bypass valve attached to
the impinger allowed the operator to maintain the humidity in the test chamber at approximately
50%. Exposure chamber conditions were monitored and verified via a custom-designed
-------
temperature/humidity probe based on a Honeywell (Morristown, NJ) 4602C sensor. Temperature
and humidity data were recorded using a Personal Daq/55 data acquisition system (Measurement
Computing Corp., Cleveland, OH). The zero air generator temperature was set at either 350 or
425 °C as part of the determining best operational status. Both temperature set points provided
equivalent baseline characteristics and the vast majority of all chamber evaluations were
performed using the 350 °C scrubber temperature.
2.1.1.3 Gas Chromatograph System
Gas was introduced into the CG-FID system via a 1/8-in. o.d. Teflon line run through the
smallest of the test chamber's ports. An Agilent Technologies 7890A GC-FID system with a
60 m x 0.32 mm x 1 jim Rxi-lMS capillary column and a 1.5 mL/min carrier gas flow rate was
used to monitor the concentration of each test gas mixture introduced into the test exposure
chamber. The initial oven temperature was set to 35 °C and was held for 5 min. The oven
temperature was ramped up to 220 °C at a rate of 10 °C/min. The temperature was set to hold for
5 min to allow good resolution of the peaks of interest.
2.1.1.4 Dynamic Gas Calibrator
A Thermo Scientific 146C dynamic gas calibrator was used to dilute gas standards. The
two standards used were benzene at 0.99 ppm as a single component (Linde, NIST traceable by
weight) and a three-component mixture of benzene (1.03 ppm), 1,3-butadiene (0.96 ppm), and
tetrachloroethylene (1.03 ppm) (Air Liquide, traceable to Scott reference standard). The three-
component mixture was used to determine if the sensors were specific to benzene. Sensors that
are specific to benzene should respond the same to both the benzene challenge and the three-
component mixture. Sensors that are equally sensitive to all three components of the three-
component mixture might be expected to respond in some fashion to approximately three times
as much to the mixture as to benzene only.
Four concentrations were chosen for testing that covered two orders of magnitude while
being evenly spaced on a logarithmic scale. This allowed for lower concentrations to be spaced
closer together and thus more precise visualization of limits of detection. Test concentrations
were < 25 ppb to reduce the possibility of carryover effect when trace-level analysis resumed for
other chamber projects. The VOC test concentrations for each analyte are listed in Table 2-1.
The use of the dynamic gas calibrator limited the automation of the test procedure.
Limitations on the range of the mass flow controllers prevented use of a single total flow rate.
The flow rates used to obtain each concentration are shown in Table 2-1.
Table 2-1. Sensor Exposure Concentrations Used in Laboratory Experiments
Benzene Three-Component Mixture (ppb) Coort p=tQ
from Tank
, .. Benzene 1,3-Butadiene Tetrachloroethylene Total VOCs (L/min)
24.75
5.25
1.14
0.25
25.75
5.46
1.18
0.26
24.00
5.09
1.10
0.24
25.75
5.46
1.18
0.26
75.50
16.01
3.47
0.76
2
5
5
8
-------
2.1.2 Evaluation Methods
Each sensor was exposed to the VOCs in the chamber over 15 hours while the
concentration was stepped through each set point and the sensor response to the VOCs was
recorded. The concentration of each gas was prepared using the Thermo 146C, which allowed
automation of a single cycle of 10 steps. Each step could be held for a maximum of 99 min. The
sequence to produce the test concentrations began with a period of zero air for at least 2 hours
before the program started. The zero air was introduced at a flow rate of 5 L/min. The flow into
the dynamic chamber equaled the flow out of the chamber, which prevented any buildup of
pressure.
During the first sensor test (UniTec SENS-IT), the Thermo 146C program started at the
highest set point and then stepped down to each subsequently lower set point until reaching zero
air and then repeated the sequence. Each of the 10 steps was 90 min, which was only long
enough to allow two GC-FID runs per step. This was insufficient to determine if equilibrium had
been reached in the test chamber. In addition, data from previous internal research not reported
here showed that the test chamber required 6.9 air exchanges, or a total of 359 L, to equilibrate to
99.9%.
Based on data from the initial test, changes were made to the original calibrator program.
The highest concentration (25 ppb) in the calibrator program could be achieved only by reducing
the total flow rate to 2 L/min. Thus, approximately 180 min was required to equilibrate to 99.9%
of 25 ppb. The calibrator program was modified to allow for 180 min at each concentration set
point. EPA staff who oversee the laboratory where the work was conducted suggested reversing
the order of concentrations used. As such, subsequent tests reversed this order, starting at low
concentrations and stepping up while repeating each concentration once. Thus, each
concentration was held for two back-to-back steps of 90 min for a total of 3 h. The two calibrator
programs used are shown in Figure 2.1-3.
At times, multiple sensors were tested in the chamber simultaneously. Analysis of zero
air showed no sign of outgassing from any of the sensors that might affect data acquired by the
other sensors.
-------
Calibrator Program
30
20
"g
'o
Q- 15
10
0
-180 -90 0 90 180 270 360 450 540 630 720 810 900 990
Time (min)
Test 1 All Other Tests
Figure 2.1-3. Calibration programs used
2.1.2.1 UniTec SENS-IT
The UniTec SENS-IT is a small VOC sensor (approximately 5 by 5 by 8 cm) that was
claimed to have benzene specificity. Two UniTec SENS-IT units, designated A and B, were tested.
The SENS-IT was connected directly to the Personal Daq/55 that was used for recording
temperature and humidity, and the voltage (V) output of the SENS-IT was recorded at a rate of
one data point per minute. All data used in this evaluation of the SENS-IT was collected with unit
A in the chamber by itself, as shown in Figure 2.1-4. A single qualitative test of unit B in the
chamber alongside the AirBase CanarIT and the CairPol CairClip (Figure 2.1-5) was run, which
confirmed that both UniTec SENS-IT units performed similarly. Three benzene-only tests and
three tests of the three-component mixture were successfully completed with the SENS-IT.
-------
Figure 2.1-4. Orientation of UniTec SENS-
IT in test chamber
Figure 2.1-5. Orientation of UniTec SENS-IT
unit B (left), AirBase CanarIT (center), and
CairPol CairClip (right) inside test chamber
2.1.2.2 AirBase CanarIT
The AirBase CanarIT (approximately 25 by 25 by 10 cm) is a multi-sensor unit capable
of measuring total VOCs (ppb) once every 20 seconds in addition to measuring several other
analytes not covered in this report. This sensor transmits all data to a Web server where it can be
accessed online using a Global System for Mobile Communication (GSM) subscriber identity
module (SUV!) card and data plan. The CanarIT is no longer commercially available and was
reestablished under the Perkin Elmer ELM branding. Since an actual Perkin Elmer ELM product
has not been evaluated, no linkages should be made between the performance of the CanairIT
and the ELM should be attempted as they may contain significantly different sensing
components and estimating algorithms.
The AirBase CanarIT was placed in the chamber alongside the CairClip and UniTec
SENS-IT unit B to conserve time and resources (Figure 2.1-5). Early attempts to use the CanarIT
involved attaching an external antenna to the unit via a cable run through one of the 3/8-in.
sampling ports in the test chamber in order to establish communication with its Web service. It
was thought that the steel test chamber might be causing a Faraday cage effect, i.e., the steel
chamber was acting as a shield that blocked all radio or electromagnetic signals. However, even
after the antenna was placed outside the chamber, a reliable connection could not be established.
Many variations of antenna length and positioning were tried, but these efforts also failed to
achieve a reliable connection to the GSM network from inside the EPA building. The
manufacturer was contacted and disclosed that all data awaiting transmission to the server were
stored on a secure digital (SD) card on the underside of the main circuit board inside the
CanarIT. With this knowledge, testing resumed and all previously acquired data were accessible.
As all other sensors had undergone initial testing, further testing of the CanarIT was
conducted with no other sensors sharing the chamber. The CanarIT was tested three times with
the three-component mixture. Testing was terminated when it was determined that there was
little if any response to test concentrations under any challenge condition.
8
-------
2.1.2.3 CairPol CairClip
The CairPol CairClip is an extremely compact sensor (approximately 6.4 cm long, 3.2cm
diameter) for measuring total VOC concentration (ppb). While this sensor can operate on battery
power for approximately 24 hours, for this study it was operated continuously plugged into a
powered mini-USB cable so that it would remain powered throughout testing. The CairClip
generates data once per minute. Initial data showed no observable trends that would be expected
to correlate with the test conditions. In addition, every other data point provided zero response
values. Mechanical problems were suspected, but resource constraints prevented a thorough
investigation of the problems or acquisition of a replacement unit. It was noted that the microfan
used to bring fresh air over the sensing element sometimes did not self-activate (turn on when
power was applied) and this might have been one component of the malfunctioning we observed.
Therefore, testing of the CairClip was stopped to preserve study resources. This is further
elaborated on in the results section.
2.1.2.4 APPCD PID
The APPCD PID sensor (approximately 11.5 cm tall and 11.5 cm in diameter), developed
internally by EPA staff (ORD's National Risk Management Research Laboratory), is a
photoionization detector (PID)-based sensor that measures total VOC concentration (V). The
device they developed employed a Mocon (blue -pID-TECH eVx 10.6 eV). This particular
Mocon PID sensor is no longer available as the manufacturer has developed new products, but
at the time of its purchase it had reported detection capabilities down to 0.5 ppb (isobutylene).
Raw 1-second data were recorded and then processed into 5-min averages because the large
volume of 1-second data was difficult to manage. The APPCD PID was tested five times with
the benzene-only tank and three times with the three-component mixture. The APPCD PID
sensor was tested alongside the ToxiRAE Pro, which is also a PID-based sensor (Figure 2.1-6).
The sensor head was placed in the test chamber upside down to allow sufficient ventilation to its
inlets, and its data-logging module remained outside the test chamber with the cable connecting
them run through one of the 3/8-in. sampling ports.
Figure 2.1-6. Orientation of APPCD PID (left) and ToxiRAE Pro (right) in test chamber
-------
2.1.2.5 ToxiRAE Pro
The ToxiRAE Pro is a PID-based sensor (approximately 15 by 10 by 10 cm when on its
charging cradle) that measures total VOC concentrations. It can operate on battery power or be
plugged into a wall outlet via a proprietary charging cradle. For these tests, the unit was installed
in its charging cradle connected to power. The ToxiRAE Pro can be programmed to measure
concentrations of a specified compound automatically and has a real-time reading of VOC
concentrations (ppm). The ToxiRAE Pro was designed for industrial hygiene use and is one of
the more market-established sensors evaluated in this study. The sensor was designed with
occupational safety issues in mind. Thus, its focus is on alerting the user of VOC levels (ppm)
above preset thresholds rather than accurately measuring at trace levels. The ToxiRAE Pro was
tested five times with the benzene-only tank and three times with the three-component mixture.
It should be mentioned that the manufacturer does have devices more sensitive than the item
evaluated but at substantially greater cost.
2.2 VOC Field Evaluations
The Triple Oaks near-road sampling site, located on Interstate 40 (1-40) in Raleigh, NC,
and maintained by EPA's National Risk Management Research Laboratory (NRMRL), was
selected for all ambient VOC testing. Three of the custom-made "bowl on pole" shelters were
attached via zip tie to the railing of the Triple Oaks sampling trailer roof as shown in Figure
2.2-1. The design and utility of these sensor weather shields have been discussed in depth
elsewhere4. A "bowl on pole" shelter consists of a pole with mounting flanges, a grating that acts
as a stable floor for the samplers while allowing airflow to pass through, and a bowl-shaped rain
shield connected to the grating with a hinge. In order from left to right, these contained the
APPCD PID, the ToxiRAE Pro PID, and the UniTec SENS-IT. The AirBase CanarIT, was
attached to a laboratory stand with wire and was C-clamped into place at the same height as the
other sensors. A sampler shelter was used to house a laptop computer for data recovery and most
of the electrical connections. Any connections that could not be made inside the shelter were
instead encased in a zip-lock bag, which was closed with a zip tie to further protect against
water.
4 Williams, R., Kaufman, A., Hanley, T., Rice, J., Garvey, S. Evaluation of Field-deployed Low Cost PM Sensors.
EPA/600/R-I4/464. December 4, 2014. Research Triangle Park, NC.
10
-------
Figure 2.2-1. Triple Oaks sampling site with shelters
Because the gas chromatographic systems anticipated to provide collocated reference
grade data comparisons at the Triple Oaks site during the field deployment time period was
ultimately determined to be non-operational, the VOC sensor data were compared with one
another and with NRMRL-supplied meteorological data. Meteorological data was available for
just the first 13 days of the study due to other instrumentation problems the operator of the site
encountered.
2.2.1 UniTec SENS-IT
The UniTec SENS-IT was oriented with its inlet protruding through the sample grating as
shown in Figure 2.2-2. The SENS-IT experienced frequent failures of its data logger, the
Personal Daq/55, throughout the study. When these failures occurred, the Personal Daq/55 would
suddenly stop recording data until it was manually reset. The Personal Daq/55 was replaced with
a DATAQ Instruments (Akron, OH) DI-145, but data collected with the DI-145 featured a
strange sinusoidal curve that appeared to overwhelm the VOC response (see section 3.2.1).
Voltage would rise to 1 V over the course of about a minute and then fall to -1 V over the course
of another minute and repeat. The Personal Daq/55 unit was used for the remainder of the effort
to record data at the near road site.
11
-------
Figure 2.2-2. UniTec SENS-IT oriented in its sampling shelter with the lid up
2.2.2 AirBase CanarIT
The AirBase CanarIT was deployed on the opposite side of the trailer from the other
sensors, but at the same height as shown in Figure 2.2-3. It was held in place on a laboratory
stand with wire, and the stand was C-clamped to the railing.
The CanarIT maintained excellent uptime throughout the field study. The only time it went
down was during a power outage on June 20, 2014. When the circuit breakers were turned back on
a few days later, the sensor resumed operation until it was removed at the conclusion of the study.
Figure 2.2-3. AirBase CanarIT installed at the near-road site
12
-------
2.2.3 APPCDPID
The APPCD PID was placed on the grating with its ventilation holes oriented downward
as shown in Figure 2.2-4. The 9-pin connector was too large to fit through the holes in the
grating, so the cable was cut, fed through the holes, and spliced back together.
The APPCD PID sensor maintained excellent uptime throughout the field study. The only
time it went down was during a power outage on June 20, 2014. When the circuit breakers were
turned back on a few days later, the sensor resumed operation until it was removed at the
conclusion of the study.
Figure 2.2-4. APPCD PID oriented on its shelter with the lid up
2.2.4 ToxiRAEPro
The ToxiRAE Pro was deployed upside down so that its inlet would be protruding
through the grating (Figure 2.2-5), thus maximizing exposure to ambient air. The ToxiRAE
developed a fan error prior to deployment in the field and had to be sent to the manufacturer for
repairs. Upon its return, it performed without error. Because it started sampling late in the study,
it was left at the field site two weeks later than the other sensors. No meteorological data were
available during the period that the ToxiRAE was sampling.
13
-------
Figure 2.2-5. ToxiRAE Pro oriented in its sampling shelter with the lid up
14
-------
3.0 VOC Laboratory Evaluation Results and Discussion
3.1 Verification of Test Atmospheres
Data from the exposure chamber were collected using a GC-FID during 17 days of
testing. The dynamic calibrator supplied four VOC concentration levels beginning with the
highest concentration (25 ppb). Each input concentration level is referred to as the "set point,"
which is the target or desired value as opposed to the measured concentration. In this case, the
dynamic calibrator is that system. Each concentration was held for 180 minutes. For least-
squares regression only, the last data point for each set point for each test was used. This was
done to minimize conditioning effects on the data used for least-squares regression. The first test
run with this calibrator program yielded an area count for benzene 18% lower than the average at
25 ppb. However, all other concentration levels were within 5% of average. The low area count
was likely the result of conditioning effects in the lines to the GC system. The other anomalous
GC result was on the first day of testing using the three-component mixture. The GC signal of all
three components began to level off as normal for the 25 ppb set point but then spiked
approximately 30%. The cause of this spike is currently unknown.
With the exception of two tests, GC area counts (concentration) for benzene and
tetrachloroethylene were highly consistent (RSD < 3% at 25 ppb) and linear with respect to set
point (r2 > 0.99). Measurements for 1,3-butadiene were highly variable (RSD = 15% at the
25 ppb set point, r2 = 0.975). Based on these data, the atmospheres produced in the test chamber
were deemed consistent and precise throughout the study.
Figures 3.1-1 through 3.1-3 show the GC-FID area counts recorded for all tests for each
analyte plotted against time overlaid with the set-point concentration plotted against time.
Figures 3.1-4 and 3.1-5 show the GC-FID area counts plotted against set-point concentration for
all tests of the benzene-only and three-component mixture tests, respectively. Table 3-1 tabulates
the coefficient of determination, response factor, and offset for each analyte as well as the
relative standard deviation (RSD) at each set point. Response factors are equal to the slope of the
least-squares regression line plotted for the data in area counts per ppb. Offsets are the
y-intercept of the same scatterplot in area counts. RSDs were found by taking the standard
deviations of the area counts found for each set point across all tests and dividing them by the
average of the same set. High RSDs are generally found only at the zero air and 25 ppb set
points. At zero air, this represents the variability one would expect from an instrument
attempting to measure below its detection limit. At the 25 ppb set point, this is likely due to the
long equilibration time combined with the poor time resolution of the GC-FID. If some of the
25 ppb GC-FID data points were taken before equilibrium had been established, this might
explain the increased variability.
15
-------
Benzene Analysis
180
360 540
Time (min)
720
900
35
Figure 3.1-1. GC-FID benzene area counts vs. time. All 17 tests are shown together. The red stair
step pattern is the set-point concentration in ppb.
450
1,3-Butadiene Analysis
180
360 540
Time (min)
720
900
Figure 3.1-2. GC-FID 1,3-butadiene area counts vs. time. All 10 three-component mixture tests are
shown together. The red stair step pattern is the set-point concentration in ppb
16
-------
400
~ 350
Tetrachloroethylene Analysis
180
360 540
Time (min)
720
900
Figure 3.1-3. GC-FID tetrachloroethylene area counts vs. time. All 10 three-component mixture
tests are shown together. The red stair step pattern is the set-point concentration in ppb
Single-Component Benzene Tests
/ uu
finn
~V)
t± cnn
^
o
o
CD 4UU
-5-
m ^nn
(/5
c
o
°- 9nn
CD
o:
inn
n t
,"
,
.*""
."
,
»''
y = 25.418x+C
r2 n QQR
....,.-
,"
.7557
2.
.,--"
-
..1
..-"'
10 15
Set point (ppb)
20
25
Figure 3.1-4. Final GC-FID benzene area counts vs. set point in all seven single-component
benzene tests
17
-------
Three-Component Mixture Tests
IUUU
Qnn
^-s Qf)f)
-i *
^ vnn
o
cc 600
&
m ^nn
in Ar\r\
o
Q_ ^nn
Q. ,300
O>
a: 200
100
n
_..«'
»..
'
....-'"
y = 26.453X
r2 = 0.9
f.
..""
..'
y =
-0.3448
842
...'
.--" y=14.
r2
3.9618X-0.68
r2 = 0.9849
1
479x + 0.907
= 0.9751
33
10
20
25
30
15
Set point (ppb)
Benzene »1, 3 Butadiene Tetrachloroethylene
Figure 3.1-5. Final GC-FID benzene area counts vs. set point in three-component mixture tests
Table 3-1. Summary of GC-FID Data
Single-
Corn pone
0.9962
Three-
tmponent
T
1,3-Butadi
0.9842
0.9751
achloroethylene
0.9849
Response factor
(area counts / ppb)
0.0393
0.0378
0.0691
0.1004
Offset
(area counts)
0.7557
-0.3448
0.907
-0.6833
RSD at 25 ppb
5.8%
11.6%
14.6%
11.4%
RSD at 5.3 ppb
0.5%
8.8%
8.1%
3.0%
RSD at 1.15 ppb
1.7%
3.7%
10.4%
3.1%
RSD at 0.25 ppb
3.0%
4.9%
12.7%
16.6%
RSD at zero air
27.2%
30.4%
200.7%
93.9%
18
-------
3.2 Laboratory Evaluation Results
3.2.1 UniTec SENS-IT
Concentration data (V) for the UniTec SENS-IT were collected using the Personal
Daq/55 and recorded in 1-min averages. The tests performed on April 12 and April 14, 2014,
were both cut short when the Personal Daq/55 system malfunctioned. When these malfunctions
occurred, the Personal Daq/55 would simply stop recording data. The cause remains unknown.
The reference GC-FID failed on April 13 and could not be used to verify test conditions on that
day.
The primary observation of the SENS-IT performance in the laboratory challenge was a
wave-like pattern in the data that obscured the response to the challenge concentration. This
cyclic pattern had a period of approximately 80 min, which can be most easily seen over a long
period of sampling zero air, as shown in Figure 3.2.1-1.
UniTec SENS-IT Zero Air Test
0.6
0.5
0.4
0.3
^^
70
60
50
n>
73
-i
Q)
i-h
C
40 ^
30
u
£ 0.2
'c
0.1
0
T
20 ^
10 '
Hr^roCTiL/i^Hr^roCTiL/i^Hr^roCTiL/i^Hr^roCTiL/i^Hr^roCTiL/i^H
L/i*-HiD(Nooroa>^-oio<-ir^(Noo^-a>L/ioiO!Nr^rooo^-o
rHrH(N(Nroro^-LnLniDiDr^r^ooooaioOrHrHiN(Nro^-
Time (min)
^^ UniTec Sens-IT ^^Temperature ^^ Humidity
Figure 3.2.1-1. UniTec SENS-IT voltage, temperature, and humidity during an
extended run of zero air
When temperature and RH are graphed alongside the voltage output of the SENS-IT, the
cyclic pattern of the SENS-IT output does not correlate with either temperature or RH. Because
the amplitude of the cycle is significant compared to the response to the challenge VOC, it must
be accounted for. However, we believe the effect was likely not so great as to obscure the
response to challenge completely. Therefore, an effort was made to mathematically compensate
for this cycle, but it could not be determined conclusively if the cycle was additive to the signal
or subtractive from it. Therefore, all data points were measured for a given set point at both the
peaks and the troughs of the cycle. Figure 3.2.1-2 plots the UniTec SENS-IT and the GC-FID
benzene response against time. GC-FID area counts were converted to ppb using the response
19
-------
factor and offset from Table 3-1. Figures 3.2.1-3 and 3.2.1-4 show the SENS-IT response from
initiation of the testing program for benzene and three-component mixture tests, respectively.
0.65
0.35
-180
UniTec SENS-IT and GC-FID on April 9
180 360 540
Time (min)
UniTec Sens-It
720
GC-FID Benzene
CD
03
to
£
3
Q.
O
o
3
O
CD
o
3
a
u-
-5
900
Figure 3.2.1-2. UniTec SENS-IT data and GC-FID benzene data from the single-component benzene
test on April 9, 2014
0.7
0.
CO
z.
LU
V)
O
0.
0.35
0.3
-180
UniTec SENS-IT Benzene Tests
180 360 540
Time (min)
8-Apr ^^9-Apr 11 -Apr -
720
900
-Setpoint
Figure 3.2.1-3. Three UniTec SENS-IT tests using benzene challenge
20
-------
UniTec SENS-IT Three-Component Tests
V)
LU
V)
o
0)
-8
-180
180 360 540 720
Time (min)
12-Apr
13-Apr
14-Apr
15-Apr
900
Setpoint
Figure 3.2.1-4. Four UniTec SENS-IT tests using the three-component mixture
In these graphs, the SENS-IT appears to be losing sensitivity as evidenced by its
diminishing response to challenge conditions. Figure 3.2.1-3 shows that the response to the
challenge concentration was decreasing with each subsequent day of testing. This apparent
declining sensitivity might artificially inflate the measured precision of the SENS-IT. The
linearity and precision of any one day of testing might be much better than that of the aggregate
data.
This effect was investigated by plotting each test individually to display trends over time.
An example from April 15, 2014, is shown in Figure 3.2.1-5. Figure 3.2.1-6 shows the UniTec
SENS-IT response against benzene concentration as measured by the GC-FID. Area counts were
converted to concentration using the benzene response factor and offset measured, as described
in Section 3.1. The response factor and offset of Figure 3.2.1-6 is nearly identical to that of the
trough response factor and offset in Figure 3.2.1-5 as expected.
21
-------
UniTec SENS-IT vs. Set Point on April 15, 2014
0
0
> 0
5 °
y = 0.0169x +0.4413
r2 = 0.9754
y = 0.0179x +0.3772
r2 = 0.9909
20
25
5 10 15
Set point (ppb)
^Troughs BCrests
Figure 3.2.1-5. UniTec SENS-IT response in both peaks and troughs vs. set point
UniTec SENS-IT vs. Measured Concentration on April 15, 2014
».uuc-u i
_ 8.00E-01
>
oT 7.00E-01
(/)
c
g. 6.00E-01
(/)
£ 5.00E-01
CO 4.00E-01 j
LU
CO 3.00E-01
0
0)
H 2.00E-01
c
D 1.00E-01
O.OOE+00
0.
r
B,«»"
'"y = 0.017J
r - C
,.-
3x+ 0.3974
.9798
DO 5.00 10.00 15.00 20.00 25.00 30.
Measured benzene concentration (ppb)
Figure 3.2.1-6. UniTec SENS-IT vs. measured benzene concentration. Benzene concentration was
calculated using the response factors and offsets in Table 3-1
22
-------
The response factors and offsets for all data were plotted on one graph by date (Figure
3.2.1-7). Note that the tests on the April 8, 9, and 11 used only a benzene challenge. All other
test days used the three-component mix.
Response Factors and Offsets for Each SENS-IT Test
o>
"o
H
LJJ
O
o>
0.01
Qi'
0.005 o
w
8-Apr 9-Apr 10-Apr 11-Apr 12-Apr 13-Apr 14-Apr 15-Apr
Test date
-^Trough Offset ^Crest Offset ^Trough RF ^Crest RF
Figure 3.2.1-7. Response factors and offsets for all UniTec SENS-IT tests
for both crests and troughs
T3
T3
CT
The tests performed with the three-component mixture featured a response factor
between two and three times the response factor of the benzene-only tests. It is difficult to
quantify this ratio precisely because the test on April 13 is possibly an outlier. However, the tests
on April 14 and 15 feature response factors more than two times that of the benzene-only tests.
This suggests that the UniTec SENS-IT might not be specific to benzene, as suggested. It is
likely sensitive to at least one of the other two components and possibly both. If the SENS-IT
was sensitive to all three components equally, then one could measure the benzene signal by
dividing the total signal by the ratio of the total concentration of the three-component mixture to
that of the benzene tank, as shown in Figure 3.2.1-8.
Figure 3.2.1-8 shows that the offsets fall steadily over the course of the benzene tests, but
rise again during the three-component tests. The response factors behave just the opposite. It
should be noted that the test on April 13 did not have any corresponding GC-FID data, and
therefore test conditions could not be verified. While a net downward trend for both response
factor and offset can be established, the correlation with time is not as strong as was initially
suspected. Therefore, all of the troughs were averaged with their corresponding crests and
graphed together against set point in Figure 3.2.1-9.
23
-------
Response Factors and Offsets for Each UniTec SENS-IT
Test
0)
£
o
(0
z
LU
(0
O
0.015 2
0.012
0.009
0.006
0.003
(D
O
C/)
m
(A
o
O
CO
(D
5T
O
-Apr 9-Apr 10-Apr 11-Apr 12-Apr 13-Apr 14-Apr 15-Apr
Test date
-Trough Offset
Crest Offset
Trough RF
- Crest RF
Figure 3.2.1-8. Response factors and offsets for all UniTec SENS-IT tests for both crests and
troughs. Three-component mixture data have been normalized against the ratio of the total VOC
concentration in the three-component tank to the concentration of benzene challenge
1.2
0.8
CO
0.2
UniTec SENS-IT vs. Concentration
y = 0.0213x +0.3942
R2 = 0.9328
y = 0.0081x +0.4051
R2 = 0.8973
5 10 15 20
Set point (ppb)
^Benzene Average 3-Component Average
25
Figure 3.2.1-9. UniTec SENS-IT vs. set point for both benzene-only and three-component tests
24
-------
As can be seen from Table 3-2, the response factor for the three-component mixture was
2.6 times that of the benzene-only tests, although this might be inflated by the unusually strong
response to challenge during the three-component mixture test on April 13. In addition, the RSD
held fairly steady at all concentration set points except at 25 ppb for the three-component
mixture. This suggests there were no limit of detection issues with this experiment. The high
RSD at the 25 ppb concentration in the three-component mixture was likely due to the unusually
high test results on April 13, as discussed earlier.
Table 3-2. UniTec SENS-IT Summary
0.8973
0.9328
Response factor (V/ppb)
0.0081
0.0213
Offset (V)
0.4051
0.3942
RSD at 25 ppb
3.5%
15.3%
RSD at 5.3 ppb
8.0%
4.4%
RSD at 1.15 ppb
5.3%
2.8%
RSD at 0.25 ppb
6.8%
6.6%
RSD at zero air
7.7%
5.1%
3.2.2 Air Base CanarIT
CanarIT data are transmitted via a GSM cellular signal to a Web server. However, due to
poor signal strength, we had to download data from the internal SD card, with guidance from the
manufacturer. The CanarIT was tested three times with the three-component mixture, all with
similar results.
As seen in Figure 3.2.2-1, the CanarIT displays a similar 80-min cyclical response as
seen with the UniTec SENS-IT, but to a much greater degree relative to the response to
challenge. As such, no correlations can be made between the CanarIT and the challenge
conditions at this time. It is interesting to note that this cyclic signal is extremely weak during the
challenge at a concentration of 25 ppb. It is possible that the cyclic signal is related to flow rate,
which is at the minimum of 2 L/min at the 25 ppb concentration.
25
-------
Q.
325
0)
M
§20
Q.
15
re
re 10
O
-------
3.2.4 APPCDPID
The APPCD PID sensor is a photoionization detector (PID)-based sensor that measures
total VOC concentration in volts (V). The raw 1-second data were processed into 5-min averages
to ease data processing and interpretation.
The test data for the APPCD PID, such as that from April 23, 2014, shown in Figure
3.2.4-1, show several interesting phenomena. The unit sampled zero air at both the beginning and
the end of the exposure. The response (V) of the zero air samples collected at the end of the test
had drifted far lower than what was found at the beginning. With the exception of the 25 ppb
concentration level, all responses began to decay after reaching equilibrium. The most likely
reason the sensor responded to the 25 ppb set point in this manner is that at the lower flow rate
the chamber was just achieving equilibrium at the end of the 180-min sampling period..
An attempt was made to compensate for the signal decay exhibited by the APPCD PID
sensor. This was done by taking slopes of 30-min segments of data over time to approximate the
slope of the baseline at that time. These slopes were then compiled with the beginning and end
zero air set points. The end result was an approximation of a smooth curve in which the slope at
any given point was approximately the slope of the decay of the signal. This smooth curve was
also developed to intersect the data during the zero air set points, thus forming a baseline. These
baselines were overlaid on data from April 22, 2014, and the APPCD PID data were then
subtracted from these baselines, as shown in Figure 3.2.4-2.
Raw APPCD PID Data vs. Time
-180 0 180 360 540 720 900
Time (min)
A APPCD-PID ^Setpoint
Figure 3.2.4-1. APPCD PID response overtime during the benzene test on April 23, 2014
27
-------
Raw APPCD PID Data and Baseline vs. Time
-180
720
900
180 360 540
Time (min)
A APPCD-PID ^Baseline
Figure 3.2.4-2. APPCD PID response overtime during the three-component mixture test on
April 22, 2014
Figures 3.2.4-3 and 3.2.4-4 show the results of this baseline correction. The baseline was
subtracted from the APPCD PID raw data and graphed over time. Figure 3.2.4-3 compares the
corrected APPCD PID response with set point. The results are nearly parallel with the set point.
Figure 3.2.4-4 compares the corrected APPCD PID response with GC-FID measurements. The
GC-FID response was converted from area counts to concentration using the benzene response
factor and offset in Table 3-1.
Corrected APPCD PID Data vs. Time
0.18
-180
180 360
Time (min)
A APPCD-PID
540
720
900
Setpoint
Figure 3.2.4-3. APPCD PID response overtime after baseline subtraction
28
-------
a
Q.
a
o
a.
Q.
Corrected APPCD PID Data and Measured
Concentration vs. Time
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
-180
180
360 540 720
Time (min)
*APPCD-PID GC-FID
r 30
- 25
- 20
- 15
900 1080 1260
CD
D)
CO
£
I
f 8
13
o
o
o
CD
r+
3
p*
5'
3
Figure 3.2.4-4. APPCD-PID response overtime after baseline subtraction
The corrected APPCD PID data are shown below versus concentration set point (Figure
3.2.4-5) and GC-FID response (Figure 3.2.4-6). The GC-FID response in Figure 3.2.4-6 was
converted from area counts to concentration using the benzene response factor and offset in
Table 3-1. These two figures have nearly identical slopes and intercepts.
APPCD PID vs. Set Point (Baseline Corrected)
APPCD PID response (V)
n 1R
n 14
n 19
n 1 .
OflR
Onfi
On9
^
^^
^^
^s^"^ y = 0 0062x + 0 001 7
^^ r2 = 0.9996
^^
^^
n Jr^
20
25
0 5 10 15
Set point (ppb)
Figure 3.2.4-5. APPCD PID response vs. set point after baseline subtraction
29
-------
APPCD PID vs. Measured Benzene
Concentration (Baseline Corrected)
PPCD PID response (V)
U. 1 O
n 1R
014 -
n 1
OfM .
*^w
^s^
^^
^^
^^
.^ y = 0.0062X + 0.0024
^^ r O.OOCO
^*
fe
5 10 15 20 25
Measured benzene concentration (ppb)
30
Figure 3.2.4-6. Baseline-corrected APPCD PID response vs. benzene concentration measured by
the GC-FID. The response factor and offset listed in Table 3-1 were used to calculate
concentration from area counts
While this process yields a very tight best-fit line, the level of effort required might not
be suitable for field operators. Therefore, uncompensated data were examined. Uncompensated
data consisted of the last 5-min average per set point with no further modifications, as shown in
Figure 3.2.4-7 for the test on April 22, 2014. A very strong coefficient of determination (r2 =
0.9385) is seen, and all variability appears to be restricted to the extreme low end of the
concentrations used.
a
Q.
a
o
a.
a.
0.2
0.1
0
APPCD PID vs. Set Point (Uncompensated)
y = 0.0058x +0.3834
r2 = 0.9385
10 15
Set point (ppb)
20
25
Figure 3.2.4-7. APPCD PID raw data vs. set point for three-component mixture
test on April 22, 2014
30
-------
Figure 3.2.4-8 shows data for all tests graphed together. The ratio of the slopes for the
three-component tests and the benzene-only tests is 2.73:1. This is slightly less than the ratio of
total VOC concentration in the three-component test to the benzene concentration in the
benzene-only tests (3.05:1). This suggests that the APPCD PID is slightly less sensitive to
1,3-butadiene and/or tetrachloroethylene than it is to benzene.
Figures 3.2.4-9 and 3.2.4-10 show the response factors and offsets, respectively, of all
tests by date so that any trends over time can be determined. It can be seen from these graphs that
the offsets are falling steadily for all eight days of testing for reasons that are as of yet unknown.
The slopes appear to hold steady throughout testing, however.
0.6
c- 0.5
APPCD PID vs. Set Point (Uncompensated)
y = 0.006x + 0.3454
r2 = 0.7912
E 0.2
a
o 0.1
a.
< n
_ y = 0.0022X + 0.3379
r2 = 0.7799
10
15
20
25
Set point (ppb)
*3-Component BBenzene
Figure 3.2.4-8. Raw APPCD PID data vs. set point for all tests
All APPCD PID Response Factors
APPCD PID response factor (V/ppb
Onn?
0 005
Onnj.
Onn9 -
n nm
n -
>
21-Aor 23-Aor 25-Aor 27-Aor 29-Aor 1-M
Date
Figure 3.2.4-9. All response factors for APPCD PID tests vs. date. The higher response factors on
April 22, 28, and 29 correspond to three-component mixture tests
31
-------
All APPCD PID Offsets
0.37
> 0.36
0)
w
o
Q.
(/)
£
a
Q.
a
o
Q.
Q.
<
0.35
O.G
0.33
O.G
O.G
y = -0.0057x + 237.35
r2 = 0.8746
21-Apr 23-Apr 25-Apr 27-Apr 29-Apr 1-May
Date
Figure 3.2.4-10. All offsets for APPCD PID tests vs. date, demonstrating a
downward trend over time
Table 3-3 was populated from the following sources: Response factors, offsets, and r2
derived from Figure 3.2.4-8. RSDs were found by taking the standard deviation of all last 5-min
averages at each set point across all tests and dividing it by the average of that same set.
Table 3-3. APPCD PID Summary
0.7799
0.7912
Response factor
(V/Ppb)
0.0022
0.0060
Offset (V)
0.3379
0.3454
RSD at 25 ppb
2.6%
5.7%
RSD at 5.3 ppb
3.2%
7.5%
RSD at 1.15 ppb
3.2%
9.6%
RSD at 0.25 ppb
3.2%
11.2%
RSD at zero air
3.3%
9.6%
32
-------
The decay of APPCD PID offsets shown in Figure 3.2.4-10 and the comparison between
baseline-corrected and uncompensated data (Figures 3.2.4-5 and 3.2.4-7) suggest that
imprecision in the APPCD PID is a result of signal decay over time. Post study discussions with
the APPCD research team indicates that the sensor element used in the device had revealed drift
characteristics in their separate evaluations, thus confirming our data conclusions. The RSD data
in Table 3-3 support this because the three-component data has much greater variability than the
single-component data taken over five consecutive days. The gap between data points for the
three-component tests allowed for a significant amount of decay to the offsets, which in turn
resulted in greatly increased variability in the data set. The nature of this decay requires further
study. It is possible that it is the result of a prolonged equilibration period of the APPCD PID
while adjusting to an environment that has lower total VOCs than ambient conditions. The
steady, approximately 3% RSD at all single-component concentration set points suggests that the
APPCD PID had no problems with limit of detection during testing.
3.2.5 ToxiRAEPro
The ToxiRAE Pro PID sensor has a lower limit of detection at 0.1 ppm total VOCs. This
is not necessarily a technical limitation of the sensor itself, but rather a software-defined limit on
what the ToxiRAE Pro PID can display and record. The sensor was exposed to eight challenges
at the various concentrations and yielded a 0 response to each. Because test concentrations were
limited to < 25 ppb to reduce the possibility of carryover effect when trace-level analysis
resumed for other chamber projects, it was expected that the ToxiRAE Pro would not respond to
total VOC concentration even at our highest set point of 25 ppb.
33
-------
4.0 VOC Field Evaluation Results and Discussion
Four of the five sensors under evaluationUniTec SENS-IT, AirBase CanarIT, APPCD
PID, and ToxiRAE Prowere placed at the Triple Oaks sampling site for approximately two
months during the summer of 2014. The CairPol CairClip, which was tested initially in the
laboratory, experienced mechanical problems and was unable to be deployed in the field, and
resource constraints prevented acquiring a replacement unit.
4.1 Field Evaluation Reference Data
The GC-FID operated by NRMRL at the Triple Oaks site, which was to be used as a
reference instrument, failed to produce viable data. Research staff were unaware that the GC
reference data provided by NRMRL was not viable until late in the data-processing stage. As
such, all data had already been organized to match the 30-min averaged data acquired by the GC,
even though no GC data are presented in this report. As such, all sensors are compared with each
other, meteorological data (provided by NRMRL), and time of day.
Correlations with time of day were examined to determine if the sensors showed daily
patterns of behavior that might correspond with local conditions. All data for a given hour across
all days of the field study were averaged in order to identify any daily cyclic trends.
4.2 Field Evaluation Results
4.2.1 UniTec SENS-IT
Figure 4.2.1-1 shows the UniTec SENS-IT data over the course of the study. The x-axis
major unit labels are all at midnight of the date shown. The data show variation both over the
course of the day and across days. This suggests that it is measuring actual pollutants.
In Figure 4.2.1-2, the SENS-IT data were split into blocks based on the hour the data were
acquired. The SENS-IT response appears to dip to its minimum value at 04:00 and peak at 10:00. It
dips slightly before holding an intermediate value from 13:00 to 24:00. The cause of this pattern is
currently unknown and cannot be determined without reference GC data and further investigation.
Figures 4.2.1-3 through 4.2.1-5 compare the APPCD PID with the other sensors in the
study. The only correlation is a very strong one (r2= 0.57) with the ToxiRAE Pro.
Figures 4.2.1-6 and 4.2.1-7 compare the SENS-IT to temperature and RH, respectively.
Only a slight correlation is observed with temperature (r2 = 0.1) and none with RH.
34
-------
UniTec SENS-IT Over Time
Date
Figure 4.2.1-1. UniTec SENS-IT field data overtime
0.2500
0.2000
c
o
Q.
£ 0.1500
V)
jjj 0.1000
V)
o
0.0500
0.0000
UniTec SENS-IT vs. Time of Day
»*
8 12
Time of day (hour)
16
20
Figure 4.2.1-2. UniTec SENS-IT vs. time of day
-------
UniTec SENS-IT vs. AirBase CanarIT
y = 0.0003x +0.1628
r2 = 0.0342
100 200 300 400
AirBase CanarIT response (ppb)
Figure 4.2.1-3. UniTec SENS-IT vs. AirBase CanarIT
500
UniTec SENS-IT vs. APPCD PID
y = 0.0195x +0.1723
r2 = 0.0054
234
APPCD PID response (V)
Figure 4.2.1-4. UniTec SENS-IT vs. APPCD PID
36
-------
UniTec SENS-IT vs. ToxiRAE Pro
^"
0)
V)
f
0
Q.
(/)
-------
UniTec SENS-IT vs. RH
response
0.7
> 0.6
V)
LU
V)
O
t
c
D
0.5
0.0
0.2
0.1
y = -0.001 2x + 0.469
r2 = 0.0789
20 40 60
Relative humidity (%)
Figure 4.2.1-7. UniTec SENS-IT vs. RH
80
100
4.2.2 Air Base CanarIT
Figure 4.2.2-1 shows the AirBase CanarIT data over the course of the study. The data are
very erratic with no observable pattern.
In Figure 4.2.2-2, data for the CanarIT have been split into blocks based on the hour the
data were acquired. All data for a given hour were averaged in order to identify any daily cyclic
trends. The concentrations begin rising at 06:00, peak at approximately 16:00, and return to
baseline by 20:00. The cause of this pattern is currently unknown and cannot be determined
without reference GC data and further investigation.
Figures 4.2.2-3 through 4.2.2-5 compare the AirBase CanarIT with the other sensors in
the study. Only the ToxiRAE Pro shows any sort of correlation with the CanarIT. The correlation
is low (r2 = 0.15), however.
Figures 4.2.2-6 and 4.2.2-7 compare the AirBase CanarIT with temperature and RH as
supplied by NRMRL, respectively, for which correlations (r2 = 0.37 for temperature and r2 = 0.23
for humidity) can be observed. It should be noted, however, that the reference temperature and
humidity data were limited. Temperature might be related to the CanarIT response coincidentally
since there is more traffic during the day. Still, a positive correlation with temperature might
explain why the CanarIT reports a greater response for the evening commute than it does for the
morning commute.
38
-------
AirBase CanarIT Over Time
~ 450
£1
Q.
c
o
Q.
re
c
re
O
0)
w
re
m
400
Date
Figure 4.2.2-1. AirBase CanarIT field data overtime
AirBase CanarIT vs. Time of Day
.Q 1 OU
Q.
0)
= 140
0
Q.
0) '^U
oi
i_ mn
c 80
CD
° 60
^ <
< on
n .
* *
* * * *
4 8 12 16 20
Time of day (hour)
Figure 4.2.2-2. AirBase CanarIT vs. time of day
39
-------
450
AirBase CanarIT vs. UniTec SENS-IT
0.1
0.2 0.3 0.4 0.5
UniTec Sens-It response (V)
0.6
0.7
Figure 4.2.2-3. AirBase CanarIT vs. UniTec SENS-IT
AirBase CanarIT vs. APPCD PID
450
400
£1
Q.
? 350
c
o
Q.
re
re
O
0)
w
re
m
-50
-100
1234
APPCD PID response (V)
Figure 4.2.2-4. AirBase CanarIT vs. APPCD PID
40
-------
AirBase CanarIT vs. ToxiRAE Pro
£1
Q.
O
Q.
re
c
re
O
0)
w
re
m
0.02 0.04 0.06 0.08
ToxiRAE Pro response (ppm)
0.1
0.12
Figure 4.2.2-5. AirBase CanarIT vs. ToxiRAE Pro PID
500
£1
Q.
Q.
-100
AirBase CanarIT vs. Temperature
y = 10.057x-160.74
r2 = 0.3743
10
30
15 20 25
Temperature (°C)
Figure 4.2.2-6. AirBase CanarIT vs. temperature
35
40
41
-------
AirBase CanarIT vs. RH
y = -2.308x + 223.17
r2 = 0.2342
20
40 60
Relative humidity (%)
80
100
Figure 4.2.2-7. AirBase CanarIT vs. RH
4.2.3 APPCDPID
Figure 4.2.3-1 shows the APPCD PID data over the course of the study. The data show a
daily cycle of rising and falling. Approximately one-third of all days sampled showed a large
spike in the data above 1.5V.
In Figure 4.2.3-2, data for the APPCD PID have been split into blocks based on the hour
the data were acquired. All data for a given hour were averaged in order to find any daily cyclic
trends. Thus, the data point for 12:00 represents an average of all field data taken between 12:00
and 12:59. For the APPCD PID sensor, which measures data once per second, this is
approximately 100,000 measurements in each block. The APPCD PID response peaks at
approximately 07:00. The cause of this pattern is currently unknown and cannot be determined
without reference GC data and further investigation.
Figures 4.2.3-3 through 4.2.3-5 compare the APPCD PID with the other sensors in the
study. None of the other sensors show any correlation with the APPCD PID.
Figures 4.2.3-6 and 4.2.3-7 compare the APPCD PID with temperature and RH,
respectively. Only a slight correlation is observed with RH (r2 = 0.12).
42
-------
APPCD PID Over Time
~ 5
o
Q.
(/)
£ 3
a
Q.
Q 2
O
Q.
Q.
< 1
^
fe^ %
%4ARMMARfe PP%%^ ^PWPHflfc
en
o5
en
en
fo
Date
en
fo
en
Figure 4.2.3-1. APPCD PID field data overtime
APPCD PID vs. Time of Day
o5
APPCD PID Response (V)
1 .0 -
1 fi -
1 4 -
19-
-\ .
00 d
n R -
04 -
OO .
n .
«
8 12
Time of day (hour)
16
20
Figure 4.2.3-2. APPCD PID vs. time of day
43
-------
> 5
0)
(/)
i 4
Q.
APPCD PID vs. UniTec SENS-IT
y = 0.276x + 0.5872
r2 = 0.0054
0.1
0.2 0.3 0.4 0.5
UniTec Sens-It response (V)
0.6
Figure 4.2.3-3 APPCD PID vs. UniTec SENS-IT
APPCD PID vs. AirBase CanarIT
y = -0.002x +0.853
r2 = 0.0509
100
200 300
AirBase CanarIT response (ppb)
400
Figure 4.2.3-4. APPCD PID vs. AirBase CanarIT
0.7
500
44
-------
APPCD PID vs. ToxiRAE Pro
0)
(/)
§
Q.
(/)
0)
y = -1.3909x +0.5931
r2 = 0.026
0.02
0.04 0.06 0.08
ToxiRAE Pro response (ppm)
Figure 4.2.3-5. APPCD PID vs. ToxiRAE Pro
0.12
APPCD PID vs. Temperature
y =-0.0136x +0.8995
r2 = 0.0174
15 20 25
Temperature (°C)
30
Figure 4.2.3-6. APPCD PID vs. temperature
35
40
45
-------
APPCD PID vs. RH
~ 5
y = 0.0102x-0.0682
r2 = 0.1177
20
40 60
Relative humidity (%)
Figure 4.2.3-7. APPCD PID vs. RH
80
100
4.2.4 ToxiRAEPro
Figure 4.2.4-1 shows the ToxiRAE Pro PID sensor data over the course of the study.
Most of the data are less than 0.1 ppm, the reported limit of detection. This limit of detection is
not the limit on the hardware, but rather a software limit for what the ToxiRAE is able to display
and record. Data below this level are reported as 0.0. For this study, the data were compiled into
30-min averages. The values reported as less than 0.1 ppm are a result of averaging some data
points at 0.0 and some at 0.1. For example, a value of 0.05 ppm might represent that half of the
data taken in a 30-min average was approximately 0.1 ppm and the other half was less and
therefore recorded as 0.
In Figure 4.2.4-2, data for the ToxiRAE Pro PID have been split into blocks based on the
hour the data were acquired. All data for a given hour were averaged in order to identify any
daily cyclic trends. Thus, the data point for 12:00 represents an average of all field data taken
between 12:00 and 12:59. The ToxiRAE Pro data appear bell shaped with a peak at 12:00. The
cause of this pattern is currently unknown and cannot be determined without reference GC data
and further investigation.
Figures 4.2.4-3 through 4.2.4-5 compare the ToxiRAE Pro with the other sensors in the
study. A slight correlation (r2 = 0.15) is observed with the CanarlT. A much stronger correlation
is seen between the ToxiRAE Pro and the SENS-IT (r2 = 0.57), which suggests the two sensors
have similar relative sensitivities to the VOCs present at the Triple Oaks sampling site.
46
-------
ToxiRAE Pro Over Time
0.12
I 0.1
0.08
0)
«
0
Q.
£ 0.06
o
m 0.04
0.02
IT
Tt
n
.*
CD
en
Date
Figure 4.2.4-1. ToxiRAE Pro field data overtime
ToxiRAE Pro vs. Time of Day
?
-------
ToxiRAE Pro vs. UniTec SENS-IT
U. 1 O
- * 016
Q. n 14
a> n 19
w u- '^
o n 1
Q. U. I
M
a) n OR
*- n HR
Q_ U.UO
UJ oo4
Oi
3 n no
° 0
n 09
01 O -
E n 1
Q. U. I
&
>r^ t
* v^^i 4 *
100 200 300 400
AirBase CanarIT response (ppb)
Figure 4.2.4-4. ToxiRAE Pro vs. AirBase CanarIT
500
48
-------
ToxiRAE Pro vs. APPCD PID
0.15
-0.1
y =-0.0193x +0.0249
r2 = 0.0268
1234
APPCD PID response (V)
Figure 4.2.4-5. ToxiRAE Pro vs. APPCD PID
49
-------
5.0 VOC Sensor Evaluation Summary
The performance of the VOC sensors tested in the laboratory is summarized in Table 5-1.
The field evaluations are summarized in Table 5-2. Additionally, Figure 5-1 shows all total VOC
sensors deployed in the field against time of day alongside one another. The terms used in the
tables are as follows:
Benzene r2: coefficient of determination; linearity of the sensor response for benzene
only.
Three-component r2: coefficient of determination; linearity of the sensor response for
total VOCs in the three-component mix.
RH: if a direct relationship exists between relative humidity and the sensor's signal,
the r2 of that relationship is displayed.
Temperature: if a direct relationship exists between temperature and the sensor's
signal, the r2 of that relationship is displayed.
Time resolution: the length of time between data points.
Uptime: qualitative assessment by the operator about the frequency of data loss.
Ease of installation: qualitative assessment by the operator about the level of effort
required to bring the sensor to operational status in the field.
Ease of operation: qualitative assessment by the operator about the level of effort
required to operate the sensor, take data, and process the data.
Mobility: qualitative assessment by the operator about the level of infrastructure
required to operate the sensor in the field using the current operating procedure listed
in Table 5-2. (Other procedures might have different requirements.)
Table 5-1. Summary of VOC Sensor Laboratory Performance
Sensor
UniTec SENS-IT (V)
AirBase CanarIT (ppb)
CairPol CairClip (ppb)
APPCDPID(V)
ToxiRAE Pro (ppm)
»-,-.,,. o~n,~n~ Three- Three-
enzener2 Component Response Component Component:
r r Response Benzene Ratio
0.8973
NA
NA
0.7799
NA
0.9328
NA
NA
0.7912
NA
0.0081
NA
NA
0.0022
NA
0.0213
NA
NA
0.0060
NA
2.63
NA
NA
2.73
NA
50
-------
Table 5-2. Summary of VOC Sensor Field Performance
Sensor
UniTec
SENS-IT 0.1036 0.0789
(V)
AirBase
CanarIT 0.3743 0.2342
(PPb)
UniTec AirBase
SENS-IT CanarIT
(V) (ppb)
APPCD
PID (V)
ToxiRAE
Pro PID
(ppm)
NA 0.0342 0.0054 0.5665
0.0342 NA 0.0509 0.1509
APPCD
PID (V)
0.0174 0.1177 0.0054 0.0509 NA 0.0268
ToxiRAE
Pro PID NA NA
(ppm)
0.5665 0.1509 0.0268 NA
Time (s) Uptime
60
20
20
Poor
Fair
Ease of
install
Poor
Excellent Poor
Good
Ease of
Very
Good
Excellent Good Excellent
Mobility
Poor
Very
good
Good Good
Fair Excellent
-------
VOC Sensors vs. Time of Day
4 8 12 16 20
Time of Day
-CanarIT -A-ToxiRAE -B-APPCD-PID -*-Sens-IT (X 10)
Figure 5-1. All total VOC sensors vs. time of day. Note that the UniTec SENS-IT data have all been
multiplied by 10 to scale it to be more visible next to the other sensors
UniTec SENS-IT: The UniTec SENS-IT must be wired directly into power and a third-
party data acquisition system, greatly reducing its mobility. This also necessarily increases the
difficulty of initial setup. Data were recorded in tab-delimited 1-min averages for very simple
processing. In the laboratory evaluations, there was evidence of both response to challenge and
an 80-min cyclic signal. In the field evaluations, the SENS-IT was significantly correlated with
the ToxiRAE Pro. This suggests the two sensors have similar sensitivity to the same mixture of
VOCs.
AirBase CanarIT: Once the AirBase CanarIT has been set up, all it requires is power
and the occasional reboot when it loses connection to the server. Even in the event connection is
lost, the sensor continues recording and saving data for transmission once connection has been
reestablished. In this round of testing the only interruption in field data collection was when a
storm knocked out power to the sampling site. Upon restoration of power, the CanarIT
automatically resumed taking data without incident. The requirement that it be furnished with a
GSM SIM card and data plan adds a recurring expense to operations.
In laboratory testing the CanarIT data included the cyclic signal displayed by the SENS-
IT but to a greater degree. This cyclic signal completely overwhelmed the response to
challenges. In the field, the CanarIT showed some significant correlations with temperature and
humidity and a minor correlation with the ToxiRAE Pro. The daily cyclic patterns to the field
data do imply that the CanarIT was responding to real phenomena present at Triple Oaks. The
CanarIT correlated most closely with the ToxiRAE Pro than with the other sensors.
CairPol CairClip: The CairPol CairClip NMVOC is light, portable, can run on battery
power for 24 hours, and requires only a mini-USB connection for prolonged operation. However,
52
-------
the unit tested in this study had apparent mechanical problems and no viable data could be
gleaned from it.
APPCD PID: The installation of the APPCD PID sensor is hampered by the extremely
large diameter of its cable and connector. Once installed, the device is robust and reliable. Data
are transferred easily from the micro SD card in a usable format, but its 1-s time resolution leads
to data files quickly becoming large and cumbersome. Because the data logger is not in any way
weather shielded, care must be taken to protect it from water intrusion. This sensor also requires
110 V AC power. In this round of testing, the only interruption in field data collection was when
a storm knocked out power to the sampling site. Upon restoration of power, the APPCD PID
automatically resumed taking data without incident.
Laboratory testing demonstrated falling sensitivity over time as evidenced by decaying
offsets on each subsequent day of testing. It is unknown if this is due to issues with the sensor
itself. After compensating for the decreasing sensitivity, the linearity of the APPCD PID was
extremely high at r2= 0.9996. Further research might be able to characterize, compensate for, or
prevent this falling sensitivity, which would make the APPCD PID extremely precise. In the
field, the only correlation with APPCD PID data was a minor (r2 = 0.1177) correlation with RH.
This suggests that the APPCD PID is sensitive to a different mixture of VOCs than the other
sensors. The APPCD PID signal regularly peaked at approximately 07:00 each day. It is not
currently known what this this was indicating relative to its environment.
ToxiRAE Pro: Because the ToxiRAE was designed for industrial hygiene uses, its limit
of detection and resolution are relatively high. It proved to be easy to install and simple to
operate and can run for approximately 14 hours on a fully charged battery. It can also run
continuously when plugged in to a power supply or outlet. Data are not well organized for
processing or review of extended collection periods. The challenge conditions for laboratory
trials were all below the limit of detection. In field testing, the ToxiRAE Pro correlated slightly
with the AirBase CanarIT (r2= 0.1509). It had the strongest correlation in the entire study with
the UniTec SENS-IT, suggesting that the ToxiRAE and the SENS-IT are both sensitive to the
same mixture of VOCs.
The robustness of the sensor was less than expected based on previous testing. One
difficulty with its operation was its tendency to slip out of the charging cradle, and frequent
reseating of the sensor in the cradle caused the connection to break during the laboratory
evaluation. The cradle had to be replaced. Later when being deployed for field trials, a fan error
developed and the unit had to be sent for repair.
53
-------
6.0 Study Limitations
6.1 Resource Limitations
It must be recognized and clearly stated that a variety of resource limitations influenced
both the depth and overall dimension of the work performed and ultimately its conclousion. The
findings of this report were limited to resources available to obtain NGAM VOC sensor
technology and perform the necessary evaluation, especially that involving laboratory-based
chamber efforts which incur greater overall research expense. To the greatest extent possible,
attempts were made to leverage this effort with other ongoing ORD activities. This including the
evaluation of the APPCD PID and use of an active VOC analysis system that was not designed
specifically for this effort. Therefore, study staff had inherent limitations on both the availability
of equipment and some of the operating parameters that had to be observed.
6.1.1 Intra-sensor Performance Characteristics
With the exception of the UniTec device, no replicate analyses to determine intra-sensor
precision was possible. Resource limitations prevented direct purchase of replicates as well as
the fact that such efforts were beyond the ability of the resources in conducting the more costly
laboratory evaluations. Therefore it should be recognized that these data represent a snap shot of
potential performance characteristics of the devices evaluated and are not to be considered
definitive.
6.1.2 Test Conditions
The test conditions were limited in both their scope and depth. As examples, we limited
the range of testing to cover only the 0 to 25 ppb range. This was due to the fact we believe such
a range to be generally reflective of urban concentrations of relevance for target pollutants like
benzene. Of course, there could be industrial and even other hot spot situations were VOC
concentrations in the ppm range might be evident and where some of the devices evaluated here
might have shown more potential. We also limited laboratory testing to a narrow range of both
temperature and relative humidity (~ room temperature) because resources were not available to
fully examine the wide range of environmental conditions that might be possible. It was planned
that the collocated field tests would provide some benefit in investigating RH and temperature
impact upon sensor performance but as stated elsewhere, ultimately reference data were not
available as originally planned.
6.1.3 Sensor Make and Models
The findings reported here are limited to a very narrow window of NGAM VOC devices.
Only a total of five (5) devices were obtained as these appeared to cover a range of some of the
most commercially available types of products associated with the low cost tier (<$2500). While
literature review indicated that some of the more exotic devices such as those involving
advanced nanotechnologies might be more prevalent, market review did not support that being
54
-------
the case. PID technology still appears to be the primary sensor element of choice for this
component of the market and with some of the same inherent benefits/limitations that PID
technologies can provide. The benefits include the PID ability to respond to a wide variety of
VOCs but simultaneously offering a lack of specificity.
55
-------
7.0 Research Operating Procedures and Related Quality Assurance
Documents
1. Alion Science and Technology (Alion). 2013. Quality Assurance Project Plan: PM and VOC
Sensor Evaluation, QAPP-RM-13-01(1), November 14, 2013. Research Triangle Park, NC.
2. U.S. Environmental Protection Agency (EPA). July 2013. Quality Assurance Project Plan:
Raleigh Multi-Pollutant Near-Road Site: Measuring the Impact of Local Traffic on Air
Quality. Research Triangle Park, NC.
3. EMAB-153.0. 2013. Research Operating Procedure for the UniTec SENS-IT Air Sensor.
4. EMAB-154.0. 2013. Research Operating Procedure for the APPCD PID Sensor.
5. EMAB-155.0. 2013. Research Operating Procedure for the ToxiRAE Pro PID Sensor.
6. EMAB-156.0. 2013. Research Operating Procedure for the CairPol CairClip NM-VOC Air
Sensor.
7. EMAB-161.0. 2013. Research Operating Procedure for the AirBase CanarIT PM and VOC
Sensor.
56
------- |