Low-cost OP-FTIR spectrometer for air quality monitoring
Paper # 89
Julia H. Rentz, James R. Engel, David L. Carlson, David J. Mansur, Robert M. Vaillancourt,
George J. Genetti
OPTRA, Inc., 461 Boston Street, Topsfield, MA, USA 01983
Peter R. Griffiths, Husheng Yang
Department of Chemistry, University of Idaho, Moscow, ID, USA 83844
ABSTRACT
We have recently completed a United States Air Force SBIR Phase II contract to develop a low-cost
open-path Fourier transform infrared spectrometer; the intended application is air quality monitoring
inside a hangar where aircraft are being refurbished. We have identified and addressed the following
high-cost components of commercially available OP-FTIR systems: unnecessarily high spectral
resolution, a complex reference metrology, a cooled infrared detector, and lastly, the retroreflector
array. In response to these high-cost attributes, we have demonstrated the utility of a low-resolution
(8 cm-1 ) instrument in measuring medium to large sized organic molecules, offering higher signal to
noise without sacrificing selectivity relative to high resolution (1 cm-1 ) systems. We have
implemented a low-cost encoder based reference metrology in place of a traditional laser
interferometer. We use an uncooled DLATGS detector in place of an LN 2 –cooled HgCdTe detector.
We have successfully manufactured a plastic injection molded retroreflector array at a fraction of the
cost of a typical array composed of individual hollow retroreflectors.
Our Phase II work effort has produced a prototype OP-FTIR spectrometer measuring 10.5 × 8.5 ×
7.25 inches and weighing just under 14 lbs (including a 6-inch telescope). The spectrometer is
powered and controlled through a portable “suitcase” PC which hosts a graphical user interface as
well as two custom-designed PCI boards for the OP-FTIR and one commercially available A to D.
We have also produced a 24- inch plastic retroreflector array weighing just under 16 lbs.
Key Words: OP-FTIR spectrometer, industrial monitoring, air quality monitoring, retroreflector
array
1.0 INTRODUCTION
1.1 Description of the opportunity
In a 2000 SBIR solicitation, the United States Air Force identified the need for a portable, fieldfunctionalized, multi-component organic vapor detector for air quality monitoring in an industrial
setting. The primary application was monitoring of the workspace inside a hangar where aircraft are
being refurbished with the use of paint thinners, strippers, degreasers, etc. The quantification of
these vapors would serve not only for Occupational Safety and Health Administration (OSHA) limit
compliance assessment but also as an overall flux measurement of each vapor. The solicitation
observed that the state of the art vapor sensors fell into three categories: a.) those with separate
dedicated analytical instruments for each vapor of interest, b.) “costly, bulky, and difficult-tomaintain broad spectrum analytical instruments,” or c.) field sampling and off- line laboratory
analysis. The opportunity was identified as the development of sensor type b with the elimination of
the components and characteristics which resulted in the high-cost, large size, and maintenance
difficulties. The OP-FTIR spectrometer was then specified as the desired technology for this
opportunity; the target cost for the system is on the order of $25k.
1.2 Background
The following list identifies a number of key characteristics of commercially available, laboratorygrade OP-FTIR spectrometers which result in their high-cost, bulkiness, and complicated user
requirements.
1.) Complex reference metrology to servo control the scanning mirror and clock the interferogram
Typically these utilize either a helium neon (HeNe) laser or a carefully stabilized laser diode; either
choice typically requires a visible or near- infrared “patch” on the beamsplitter/compensator
assembly, the rest of which is coated for IR operation. This type of reference also measures the
mirror position at the edge of the mirror which will inherently result in abbe error with any tilting.
All of these factors contribute to the overall cost and complexity of the system.
2.) High spectral resolution
While high spectral resolution (∆σ < 1 cm-1 ) may be preferable for some applications, it is not
necessary for the measurement of medium to large organic molecules exhibiting broad spectral
bands in the IR. High spectral resolution is achieved by a longer stroke length of the scanning
mirror which is inherently more difficult to control. High resolution also sacrifices signal to noise
without necessarily the benefit of increased selectivity (relative to the low-resolution system).
3.) Liquid Nitrogen (LN2 ) cooled IR detector
Historically these systems have employed a mercury cadmium telluride (HgCdTe) detector operated
at 77K with either a liquid nitrogen dewer or a closed cycle sterling cooler. Such detectors are
characterized by poor reliability and are a considerable expense. We have found that with the signal
increase allowed by the low spectral resolution, an uncooled detector can be used instead without
sacrificing overall sensitivity.
4.) Hollow mirror retroreflector array
Aside from the high cost of a laboratory- grade OP-FTIR spectrometer itself, the retroreflector array
required to return the open path beam can cost as much as twice the target cost of our low-cost
system. This is because these arrays are composed of individually assembled hollow mirror
retroreflectors, typically two inches in diameter.
5.) Spectral analysis expertise requirement
Complicated analysis is required to extract molecular concentration information from the measured
spectra of the OP-FTIR. Historically this has been done by a spectral analysis expert, which is as
added expense.
With these key areas in mind, the following section summarizes our technical approach to resolving
these short comings.
1.3 Our technical approach
OPTRA has developed a low-cost version of the standard OP-FTIR spectrometer with as compact,
rugged Michelson interferometer and the following cost saving features:
1.) Nanoscale reference system
We have emplo yed our encoder-based Nanoscale position sensor in place of the traditional laser
interferometer. The Nanoscale interpolates the interference fringe pattern between two diffracted
orders off of a small encoder mounted on the shaft of the moving mirror. The sub-nanometer
resolution position word is then used to servo control the mirror and clock the interferogram. This
technique requires no visible patches on the beamsplitter/compensator and significantly reduces abbe
error relative to laser interferometer reference systems which measured off of the edge of the moving
mirror. Our system initializes off of zero optical path difference (OPD) and therefore achieves an
absolute measurement.
2.) Low spectral resolution
Given the broad spectral features of the organic compounds of interest to this application, we have
designed a system with 8 cm-1 spectral resolution; the associated 0.125 cm optical stroke length of
the scanning mirror is accomplished with a rugged flexure assembly driven by a small voice coal
motor with a mirror tilt of less than 10 arc seconds.
3.) Uncooled DLATGS detector
We have replaced the standard HgCdTe detector with a low-cost and reliable uncooled pyroelectric
deuterated L-alanine triglycine sulfate detector. The associated loss in sensitivity by going uncooled
is moderated by the low spectral resolution as well as a more efficient retroreflector array at shorter
open path lengths.
4.) Plastic injection molded retroreflector array
We have successfully fabricated a plastic injection molded retroreflector array (patent pending) at a
fraction of the cost of a standard hollow mirror array which has better efficiency at shorter open path
lengths (up to 100 m).
5.) Automated algorithms to extract molecular concentration information from the measured spectra
Through a cooperative work effort with Dr. Peter Griffiths and Dr. Husheng Yang at the University
of Idaho, we are able to offer an algorithmic engine based on artificial neural networks (ANNs) and
partial least squares (PLS), specifically trained for our instrument and able to automatically identify
and extract concentration information of 105 molecules from the measured spectra.
Other features include a 6- inch telescope and the rugged modulator we originally developed during
our work on the JSLSCAD effort (1). We have also implemented an automatic calibration capability
where the secondary beamsplitter is rotated 90º by a solenoid upon key command; this directs the IR
light exiting the modulator directly onto the detector, inherently creating a zero-path interferogram.
The overall spectrometer module (including telescope) is 10.5 × 8.5 × 7.25 inches in size and weighs
just under 14 lbs. The spectrometer is shown in figure 1.
Figure 1: Photo of Low-Cost OP-FTIR Spectrometer
Figure 1 is a photograph of the low-cost OP-FTIR spectrometer.
The intent of this paper is to present high- level descriptions of the system- level and all subsystem
designs of the low-cost OP-FTIR as well as some instrument characterization data including signal
to noise, spectral range, and spectral resolution. We also present some preliminary results of testing
the University of Idaho algorithms.
2.0 DESIGN
The design of the various subsystems of the OP-FTIR will be described in this section. We will
begin by presenting the system requirements and will follow this with radiometric projections. We
will then show the opto- mechanical, electrical, and software designs.
2.1 System requirements
The following table lists the requirements against which this system was designed.
Table 1: System specifications
SPECIFICATION
Spectral range
Spectral resolution
Minimum signal to noise (single scan)
Integration time (single scan)
Spectral artifacts
Optical open path range
VALUE
714 to 1428 cm-1
≤ 8 cm-1
≥ 1000
1.25 seconds
≤ 2.5x10-4
1 to 100 m (round trip)
The purpose of the system level requirements is to assure that the resulting design will meet the
application needs. In this case we selected the spectral range in order to include resonance features
of each molecule on the list composed by our customer. The spectral resolution was specified by the
results of the research done by Dr. Griffiths. The single scan signal to noise, when coupled with the
run time of the algorithms (allowing for a certain amount of coadding) assured that we could
measure the target molecules with sensitivity equivalent with OSHA limits. The scan time is
dictated by the detector bandwidth, which is ultimately tied to signal to noise. The artifact
requirement inherently specifies the maximum tolerable sampling errors by the reference metrology;
this is also ultimately tied to signal to noise with coadding in mind. The open path length was
specified by our customer as sufficient for this indoor air quality monitoring application.
2.2 Radiometric projections
The signal to noise of the OP-FTIR is projected by
SNR (σ, TIR Source ) =
N(σ, TIR Source )
(1)
NESR
where N(σ,TIR Source) is the spectral radiance at optical frequency, σ, and IR source temperature, TIR
Source, projected by the Planck function. NESR is the noise equivalent spectral radiance of the OPFTIR given by
NESR =
Ad
(2)
D * ⋅η ⋅ Θ ⋅ ∆σ ⋅ ∆t
The following table summarizes these values and the resulting projected SNR of this system at the
center optical frequency, 1000 cm-1 for an IR source temperature of 1623K.
Table 2: Radiometric Values
VARIABLE
DESCRIPTION
VALUE
UNITS
Ad
D*
η
Θ
∆σ
∆t
NESR
SNR(1000cm-1 ,1623K)
Detector area
Detector detectivity
radiometric efficiency
Etendue
Spectral resolution
Integration time
Noise equivalent spectral radiance
Signal to noise
.071
4.5×108
.009
.022
8
1.25
3.3×10-7
1268
cm2
cmvHz/W
unitless
cm2 ster
cm-1
s
2
W/(cm ster cm-1 )
unitless
Based on these radiometric projections the system design meets the application requirements.
2.3 Opto-mechanical design
Figure 2 shows the optical layout of the OP-FTIR in two sections: the interferometer and the detector
leg. The former shows only one arm of the interferometer, as the other arm is optically identical.
These layouts show that all conventions regarding pupils were observed with our design, insuring
the smallest possible size for the specified optical throughput.
Figure 2a: OP-FTIR optical design – interferometer leg
Telescope
parabolic
primary
Field
stop
Beamsplitter/
compensator
Interferometer
cube
1.27
cm
Secondary
beamsplitter
Moving
mirror
Interferometer
lenses (x2)
16 cm
Limiting
field stop
IR source
field lens
Telescope
hyperbolic
secondary
IR
source
Figure 2b: OP-FTIR optical design – detector leg
Telescope
parabolic
primary
Secondary
beamsplitter
Detector
assembly
16 cm
Telescope
hyperbolic
secondary
Field stop
Detector
field
lens
Detector
0.3 cm
Figure 2 shows the optical layout of the OP-FTIR interferometer leg (2a) and detector leg (2b). The design
obeys the convention of imaging the entrance pupil (the IR source) onto the interferometer mirror, the
interferometer mirror onto the telescope primary mirror, and the telescope primary mirror onto the detector
(i.e. they are all conjugate). This allows the greatest throughput for our 1.27 cm interferometer mirror and
also alleviates any issues that might have been associated with inhomogeneities in the IR source or detector
(i.e. “hot spots”). The layout shows the large field (8.6º full angle) permitted by the low resolution and
associated obliquity limit; the high resulting throughput is a key factor in our sensitivity.
The telescope mirrors are gold-coated diamond-turned aluminum (Diversified Optical Products, Inc.,
Salem, NH), and all lenses are spherical and germanium (International Scientific Products, Inc.,
Irvington, NY). The beamsplitters and compensator are zinc selenide (Spectral Systems, Inc.,
Hopewell Junction, NY), and the interferometer beamsplitter coating is designed for low chromatic
phase dispersion. The interferometer mirrors are also diamond-turned aluminum (Aero Research,
Associates, Port Washington, NY). The limiting field stop is in front of the IR source; all others are
slightly oversized.
Figure 3a shows a solid model of the opto-mechanical system without the cover. The majority of the
mounts are black-anodized aluminum to control external reflections. The IR source has a copper
block and heat dissipating fin (not shown) to expel heat from the spectrometer module. Figure 3b
shows the secondary beamsplitter module with rotary solenoid that, upon key commanded 90º
rotation, directs the IR light exiting the modulator directly onto the detector for an automatic zeropath interferogram. This is used for initialization and for the algorithms which ratio an open-path
interferogram to the zero-path interferogram. The gain decreases by a factor of 16 in
initialization/calibration mode (relative to signal mode). We have also implemented (in software) a
means to assess and discard “bad” open-path interferograms in which the open-path beam had been
blocked or otherwise disrupted.
Figure 3a: OP-FTIR opto-mechanical system
Primary
Mirror
Assembly
Backplate
Nanoscale
Secondary
Mirror
Assembly
7.25”
10.5”
8.5”
Baseplate
Baffle
Modulator
Assembly
IR Source
Mount
Figure 3 is a solid model of our compact opto-mechanical system. Most of the
mounts are black anodized aluminum to control extraneous reflections. The IR
source also has a copper block and heat dissipating fin (not shown). Figure 3b
shows our secondary beamsplitter assembly with the rotary solenoid to flip the
beamsplitter between calibration mode and open-path mode.
Figure 3b: Secondary beamsplitter assembly
Secondary
Beamsplitter
Solenoid
DLATGS
Field
Lens
Interface
Plate
Figures 4a through 4c are photos of the assembled and integrated OP-FTIR spectrometer module
with all of the components labeled. These also show the heat sink for the IR source. The cover
holds a small sighting scope (Meade Instruments) to aid in alignment. We have also provided a
commercially available laser ranger (Bushnell) to measure the open-path length between the
spectrometer and retroreflector array.
Figure 4a: OP-FTIR Uncovered
Heat Sink
Figure 4b: OP-FTIR Detector and
Secondary Beamsplitter Assembly
IR Source
Secondary
Beamsplitter
Telescope
Nanoscale
Detector
Interferometer
IR Source
Interferometer
Nanoscale
Lens Mounts
Figure 4c: OP-FTIR Front View
Telescope
Assembly
Baseplate
Figures 4a through 4c are
photos of the integrated OPFTIR spectrometer module with
all of the components labeled.
Figure 4a is a rear view
showing the interferometer and
Nanoscale reference as well as
the IR source with heat sink.
Figure 4b shows a close-up of
the secondary beamsplitter
assembly with detector. Figure
4c shows a front view of the 6inch telescope. The lab bench
in these photos has 1-inch pitch
¼-20 holes to provide a size
reference. The covered system
weighs just under 14 lbs.
Figure 5a and 5b shows the plastic injection- molded retroreflector array. We worked with Fresnel
Technologies, Inc. (Fort Worth, TX) to construct a set of nickel-plated aluminum molds from which
a quantity of 100 mm diameter acrylic cells (each composed of 54, 19 mm corner cubes) were made.
The cells were then coated with protected aluminum, and we then mounted 37 onto a polystyrene
board with silicone to make a large 24- inch array.
Figure 5a: Retroreflector array cell
Figure 5a shows a single 100 mm retroreflector
array and 5b an assembled 24-inch (60 cm)
array. The large array is mounted to a
polystyrene sheet with silicone to provide
flexibility under stress.
Figure 5b: 60-cm retroreflector array
2.4 Nanoscale Reference
Figure 7 shows a solid model of the Nanoscale reference assembly mounted to the top of the rear
cube of the interferometer. This nano-scale position sensor was originally developed by OPTRA
under a commercial contract for monitoring the position of the read/write arm in servo track writing
for computer hard drives. The Nanoscale emits a NIR laser beam which is diffracted off of the
encoder mounted to the shaft of the scanning mirror. The Nanoscale interferes the +/- 1 orders and
interpolates the fringe pattern measured by a linear silicon array to discern the encoder position to
very high accuracy. We have added a relay lens assembly to extend the workable standoff from the
standard 3 mm. Figure 7 also shows the motor assembly in detail.
Figure 7: Nanoscale Reference Assembly
Nanoscale
Relay lens
assembly
Interferometer
cube
Interface
plate
Beamsplitter /
compensator
Motor
assembly
Motor
shaft
Back iron
Flexures
Base plate
Encoder
and mount
Moving
mirror
The Nanoscale linear array has every third element wired in parallel. The fringe imaging is such that
each spans exactly three detector elements; the result is three signals, R, S, and T, each separated by
120º in phase. While our manufacturing procedure balances the three signals with respect to gain
and offset in order to minimize interpolation error (IE) resulting from unbalanced signals, our
systems analysis suggested that the process of installing the Nanoscale on the OP-FTIR would most
likely imbalance them enough to create unacceptable IE for our sampling requirements (section 2.1).
We therefore made it part of this development effort to add an auto-balancing capability to the
Nanoscale interpolation electronics and demonstrated the ability to automatically drive the IE down
to an acceptable level for our requirements.
2.5 Electrical design
The OP-FTIR electrical design effort produced two custom PCI boards which provide all of the
functionality of the system. The OP-FTIR processor board hosts the Nanoscale interpolation
electronics as well as the IE correction circuits. This board also handles the IR source drive voltage
regulated from a resident DC to DC converter. The OP-FTIR servo board hosts the servo electronics
for the moving mirror and also provides the clock and trigger to the A to D for sampling the
interferogram. We used a National Instruments 16-bit A to D (PCI-6034E Low Cost Multifunctional
I/O Board). Figure 8 details the flow of the OP-FTIR electronics.
Figure 8: OP-FTIR Electronics
Spectrometer Module
Scanning
Mirror
Interferometer
Preamp
Actuator
IR
Source
High Res Position Word
Processor
Mirror
Control
Clock
Trigger
Digital Interferogram
IE
Correction
IR Source
Driver
DC to DC
Converter
Analog Interferogram
Nanoscale
Software
FFT (apodization, zero-filling, phase correction, etc)
Algorithms
Data storage
Molecular Concentration display
Portable PC
Figure 8 is a block diagram
of the OP-FTIR electronics
which are primarily located
on two custom PCI boards
which will fit in any standard
PC with appropriate slots.
The Nanoscale electronics as
well as a voltage regulator
for the IR source are resident
to the Processor PCB; the
mirror control electronics are
found on the Servo PCB. We
employ an off-the-shelf
National Instruments 16-bit
A to D to digitize the
interferogram using an
external clock and trigger
provided through the
Nanoscale. All data
collection is handled through
LabVIEW , and most of the
processing (transforms and
algorithms) is handled by
MatLAB.
2.6 Software
OP-FTIR software is responsible for receiving the digital interferogram and interfacing with the
algorithms (a MatLAB executable). The standard operation of the system is to coadd interferograms
while the algorithms are running on the previous coadded interferogram. The algorithm sequence of
events is as follows:
1. Receive N digital interferograms where N is the user specified number to coadd
2. Check for “bad” interferograms where the beam was blocked, discard these, and coadd
the “good” interferograms.
3. Execute FFT with apodization, zero-filling, and phase correction
4. Execute ANN on spectra to determine which molecules are present
5. Pass “molecule present” list to PLS
6. Execute PLS and quantify present molecules
7. Pass results back to OP-FTIR software for comparison with OSHA limits and display
Figure 9 shows our GUI with the functions labeled.
Flips BS to Flips BS to
cal mode signal mode
Initializes
off of zero
OPD
Figure 9: OP-FTIR GUI
# of igrams
to coadd
Measured
range
Logs
concentration
over time
Allows user to
save igrams
Enables
algorithms
# discarded
“bad” igr ams
List of compounds
present and
quantities
Comparison
with OSHA
limit
3.0 TESTING AND RESULTS
3.1 Tests description
The OP-FTIR instrument was characterized for SNR (and at the same time spectral range), spectral
resolution, and spectral artifacts. The SNR and spectral range measurement was straight forward.
Spectral resolution and artifacts were measured using a 10.6 µm CO2 laser with an integrating sphere
(which insured that the full field of the interferometer was filled). Spectral artifacts refer to any
periodic noise source which shows up in the spectra or spectral satellites which appear
symmetrically about a spectral feature (such as the laser line) and are due to periodic sampling
errors, amplitude or phase modulation, or fold noise from higher frequencies. This test allows for us
to assess for all of these simultaneously.
In addition to instrument characterization, we detail the preliminary testing of the algorithms using a
relatively controlled release of methanol. The experimental set-up for this test was a 3 m open-path
with a 1.5 m section of PVC gutter underneath, serving as the methanol receptacle; the methanol was
let to evaporate into the open path beam.
3.2 Tests results
Figure 10a shows the measured SNR for three coadded scans. The measured single scan SNR is
then about 690. The target SNR is then met in closer to two scans than one. The spectral range is
close to 800 to > 1400 cm-1 .
Figure 10a: SNR and vs optical frequency
1400
1200
1000
800
600
400
200
0
600
800
1000
1200
1400
1600
1800
2000
optical frequency (cm-1)
Figure 10b shows the instrument response to the 10.6 µm CO2 laser. The spectral resolution or fullwith at half- height of the instrument function is 7.6 cm-1 .
Figure 10b: CO2 laser spectrum
1.1E-04
9.0E-05
AU
7.0E-05
5.0E-05
3.0E-05
1.0E-05
-1.0E-05
900
910
920
930 940 950 960 970
optical frequency (cm-1)
980
990 1000
Figure 10c shows the instrument function over the spectral range on a log plot and clearly shows that
there are no spectral artifacts.
9.0E-05
1.E-04
7.0E-05
1.E-05
AU
1.E-03
5.0E-05
1.E-06
3.0E-05
1.E-07
1.0E-05
1.E-08
-1.0E-05
700
800
900 1000 1100 1200
optical frequency (cm-1)
1300
AU
Figure 10c: CO2 laser spectrum (log plot)
1.1E-04
1.E-09
1400
Figure 11 is a jpg of the GUI display during the methanol measurement. Using 60 coadded
interferograms, the ANN/PLS combination was able to successfully discriminate methanol from a
list of 104 other compounds. Given the experimental set up and the associated uncertainties, the
reported methanol concentration seems reasonable. It did not exceed the OSHA limit.
Figure 11: GUI from Methanol Measurement
4.0 CONCLUSION
In conclusion, OPTRA has successfully met the design objectives of this SBIR solicitation to
produce a significantly lower cost OP-FTIR spectrometer than is currently on the market. We have
demonstrated some novel design economies including an encoder based reference system, uncooled
detector, low spectral resolution with discrimination/ concentration algorithms specifically tailored
for low resolution, and finally a plastic injection molded retroreflector array. Further development
efforts will be in validation and additional field hardening for outdoor use.
ACKNOWLEDGMENTS
This research was conducted under an SBIR Phase II contract funded by the U.S. Air Force. We
would like to thank our Technical Monitor, Dr. Daniel A. Stone, Environmental Management
Directorate OO-ALC/EM, Hill AFB, UT.
REFERENCES
1. OPTRA is presently under contract to Intellitec, a recent acquisition of General Dynamics, to
manufacture the sensor portion of the Joint Services Lightweight Standoff Chemical Agent Detector
(JSLSCAD). We have designed and successfully fabricated 15 Engineering Design Test units and
50 Product Quality Test units. Production quantities are estimated at upwards of 1400 over the next
eight years.
2. World Wide Web < http://www.epa.gov/ttnemc01/ftir/refnam.html#v> EPA - TTN EMC Spectral Database - Fourier Transform Infrared (FTIR) Reference Spectra