The document outlines Clinton J. Smith's final public oral exam for his dissertation on developing laser spectrometers for wireless trace gas sensor networks. The dissertation aims to develop CO2 sensors for deployment in a real-time sensor network to monitor carbon fluxes over a broad area. The outline includes development of a wireless laser spectroscopic sensor node for atmospheric CO2 monitoring, quantifying and improving the accuracy of wireless sensor network nodes, developing a solar-powered distributed wireless CO2 monitoring network, techniques for real-time calibration, and conclusions. Laboratory and field tests show the sensors can accurately measure CO2 concentrations with precision below 1% and linear response over large concentration ranges.
High-accuracy laser spectrometers for wireless trace-gas sensor networks
1. Final Public Oral Exam
10/25/2013
High-accuracy laser spectrometers for
wireless trace-gas sensor networks
Clinton J. Smith
Advisor: Prof. Gerard Wysocki
pulse.princeton.edu
2. Research Goals
• Develop CO2 sensors for deployment as a real-time sensor network
for carbon flux monitoring over a broad geographic area.
• Develop and implement techniques to maintain the long-term
measurement stability of field deployable portable trace-gas sensors
http://www.coas.oregonstate.edu/research/po/satellite.gif
Base
Station
Radio
Range
Sensors
2
3. Dissertation Outline
• Dissertation motivation and introduction to absorption spectroscopy
• Wireless laser spectroscopic sensor node for atmospheric CO2 monitoring
C. J. Smith, S. So, L. Xia, S. Pitz, K. Szlavecz, D. Carlson, A. Terzis, and G. Wysocki, “Wireless Laser
Spectroscopic Sensor Node for Atmospheric CO2 Monitoring – Laboratory and Field Test,” Appl. Phys. B,
vol. 110, no. 2, pp. 241-248 (2013).
• Quantifying and improving wireless sensor network (WSN) node accuracy
http://www.coas.oregonstate.edu/research/po/satellite.gif
• A solar-powered, distributed wireless CO2 monitoring network
C. J. Smith, P. Ramamurthy, M. A. Khan, W. Wang, S. So, M. A. Zondlo, E. Bou-Zeid, and G. Wysocki, “A
Distributed Wireless CO2 Monitoring Network,” Manuscript in Preparation.
• Real-time calibration techniques
C. J. Smith, W. Wang, and G. Wysocki, “Real-time calibration of laser absorption spectrometer using
spectral correlation performed with an in-line gas cell,” Optics Express, vol. 21, no. 19, pp. 2248822503 (2013).
• Electric current tuning of self-oscillation frequency of EC-VCSELs
C. J. Smith, W. D. Li, G. Wysocki, and S. Y. Chou, “Electric Current Tuning the Self-Oscillation Frequency
of EC-VCSELs,” Photonics Technology Letters, IEEE, vol. 25, no. 17, pp. 1707-1710 (2013).
• Conclusions and Future directions
3
4. FPO Outline
• Dissertation motivation and introduction to absorption spectroscopy
• Wireless laser spectroscopic sensor node for atmospheric CO2 monitoring
C. J. Smith, S. So, L. Xia, S. Pitz, K. Szlavecz, D. Carlson, A. Terzis, and G. Wysocki, “Wireless Laser
Spectroscopic Sensor Node for Atmospheric CO2 Monitoring – Laboratory and Field Test,” Appl. Phys. B,
vol. 110, no. 2, pp. 241-248 (2013).
http://www.coas.oregonstate.edu/research/po/satellite.gif
• Quantifying and improving WSN node accuracy
• A solar-powered, distributed wireless CO2 monitoring network
C. J. Smith, P. Ramamurthy, M. A. Khan, W. Wang, S. So, M. A. Zondlo, E. Bou-Zeid, and G. Wysocki, “A
Distributed Wireless CO2 Monitoring Network,” Manuscript in Preparation.
• Real-time calibration techniques
C. J. Smith, W. Wang, and G. Wysocki, “Real-time calibration of laser absorption spectrometer using
spectral correlation performed with an in-line gas cell,” Optics Express, vol. 21, no. 19, pp. 2248822503 (2013).
• Conclusions and Future directions
4
5. Greenhouse Gases and Climate Change
•
Implementation of Kyoto Protocol
(1997) and Successors requires
technology
Carbon credit/Carbon trading
• Quantify and locate all CO2 sources and
sinks requires
Natural concentration variations range
from 370 ppmv to 10,000 ppmv
Global ambient CO2 concentration is
~400 ppmv
Measure with 1:1000 precision
• Need distributed sensor networks
Continuous, long term, wide area
trace-gas sensing
• Multiple greenhouse gases (GHG)
Carbon dioxide
Methane
Nitrous oxide
6. Role of Atmospheric Gases and the GHG Effect
• Simplified representation of energy flow within the atmosphere of the earth
•
~70% of incoming radiation is absorbed and re-emitted by the earth
~30% of radiation is reflected to space
GHG effect is primarily the result of molecules absorbing long-wave light
D. Hartmann, Global Physical Climatology, ch. 2. San Diego, CA: Academic Press, 1994.
6
7. GHG Effect is the Result of Mid-IR Absorption
• Both sun and earth act as black body sources emitting light
• GHGs have a molecular “fingerprint”
•
Primarily absorb long-wave light (beyond 1-2 microns)
Light is absorbed following the Beer-Lambert Law
Governing concept of trace-gas spectroscopy
7
9. Use Lasers to Target Absorption Lines
Vertical Cavity Surface Emitting Laser
- Low Power
Pros
- Moderate cost
Con
- Limited to weaker molecular
transitions in the near-IR
Quantum Cascade Laser
- Probe strongest molecular
Pro
Cons
transitions
- High power consumption
- High cost
9
10. Outline
• Dissertation motivation and introduction to absorption spectroscopy
• Wireless laser spectroscopic sensor node for atmospheric CO2 monitoring
C. J. Smith, S. So, L. Xia, S. Pitz, K. Szlavecz, D. Carlson, A. Terzis, and G. Wysocki, “Wireless Laser
Spectroscopic Sensor Node for Atmospheric CO2 Monitoring – Laboratory and Field Test,” Appl. Phys. B,
vol. 110, no. 2, pp. 241-248 (2013).
http://www.coas.oregonstate.edu/research/po/satellite.gif
• Quantifying and improving WSN node accuracy
• A solar-powered, distributed wireless CO2 monitoring network
C. J. Smith, P. Ramamurthy, M. A. Khan, W. Wang, S. So, M. A. Zondlo, E. Bou-Zeid, and G. Wysocki, “A
Distributed Wireless CO2 Monitoring Network,” Manuscript in Preparation.
• Real-time calibration techniques
C. J. Smith, W. Wang, and G. Wysocki, “Real-time calibration of laser absorption spectrometer using
spectral correlation performed with an in-line gas cell,” Optics Express, vol. 21, no. 19, pp. 2248822503 (2013).
• Conclusions and Future directions
10
11. Project Goal & Outline
The project goal:
• Develop CO2 sensors for deployment as a real-time sensor network
for carbon flux monitoring over a broad geographic area.
Atmospheric monitoring of CO2 (fluxes, sources, and sinks)
Soil Respiration Monitoring
http://www.coas.oregonstate.edu/research/po/satellite.gif
Outline
• Requirements for a sensor to be used in trace gas sensor networks
• Commercial state-of-the-art
• Overview of sensor design
Overview of control and acquisition electronics
Selection of laser & CO2 absorption line
• Sensor performance tests
• Lab & Field tests
11
12. Requirements for Trace Gas Sensor Network Node
A trace gas sensor node must
provide:
Sensors work autonomously
in the field
Base Station
• Small size/portability
• Low unit/capital cost
• Low maintenance and operating
•
•
•
•
•
•
costs
Robust construction
Low power consumption
High sensitivity (ppb)
High selectivity to trace gas
species
Wireless networking capability
Ease of mass production
Radio
Range
Sensors
12
13. State of the Art Commercial CO2 Sensors
Li-Cor LI-7500A
• Open Path
• Non-dispersive IR
• < 0.1 ppmv precision
• ~1 % accuracy
• 12 W power consumption (30
W at startup)
• Wired interface
• ~30 minute warm-up period
• Unreliable in inclement weather
Vaisala GMP343
• Open Path
• Non-dispersive IR
• (3ppmv + 1 % of reading)
accuracy
• 1-3.5 W power consumption
• Wired interface
• ~30 minute warm-up period
• Seconds-minutes response
time
http://www.licor.com/env/products/gas_analysis/LI-7500A/system_components.html
http://www.vaisala.com/en/products/carbondioxide/Pages/gmp343.aspx
13
14. CO2 Sensor Node Design & Specifications
• Tunable diode laser absorption spectroscopy
•
•
•
•
•
(TDLAS)
Housed within a NEMA enclosure for
environmental protection
3.5 m path Herriott multi-pass cell
2 μm VCSEL & InGaAs photodetector
Custom electronics board (openPHOTONS
platform)
Wireless communication via TelosB Mote
TinyOS & Zigbee
UART communication with the
openPHOTONS board
• Powered by an integrated Li-ion polymer
battery
300 mW power consumption
14
15. Custom Control and Acquisition Board
Direct Digital Synthesizer
TEC driver
MCU
8MHz
Lock-In Amplifier + Front End
Modulated
Current
Driver
So, S., Sani, A. A., Zhong, L., Tittel, F., and Wysocki, G. 2009. “Demo abstract: Laser-based trace-gas chemical sensors for distributed wireless
sensor networks” in Proceedings of the 2009 international Conference on information Processing in Sensor Networks (April 13 - 16, 2009).
Information Processing In Sensor Networks. IEEE Computer Society, Washington, DC, 427-428
15
16. 2 μm VCSEL & CO2 Absorption Spectrum
•
•
•
•
Low power vertical cavity
surface emitting laser
(VCSEL)
Consumes ~5 mW
P=1 atm
power
VCSEL temperature tuning
range of ~5 cm-1
Absorption coefficients in
this range correspond to
~1% absorption over
3.5 m path
Water absorption lines
have limited impact on
CO2 absorption lines
Source: HITRAN 2000 database
Atmospheric
Concentration,
HITRAN/GEISA
I I 0 e ( ) L
16
17. TDLAS CO2 Sensor In-Lab Performance
• VCSEL is wavelength modulated at ~10 kHz
Via current modulation
2nd harmonic peak value will be used for CO2 concentration measurement
Modulation depth of ~0.22 cm-1 is optimized for the 2nd harmonic (m = 2.2)
• Harmonic line profiles are measured by temperature scanning about the
4987 cm-1 absorption line
A lock-in amplifier is used to select each harmonic
• Calibrated 285 ppm CO2 in N2 mixture yields
2nd harmonic SNR of 2530
17
18. TDLAS CO2 Sensor 3rd Harmonic Line Locking
• Control laser temperature so that 3rd
harmonic signal is near zero
This corresponds to the maximum of the
2nd harmonic signal
Measure the CO2 concentration by
continuously monitoring the 2nd harmonic
signal value at the peak
18
19. Sensor Linearity & Startup Behavior
• CO2 measurements are linear over the expected range of
•
•
concentrations to be measured in the field
~1.5% absorption at this line and path length
Electronics are capable of sleep and wake cycles for power
conservation
TDLAS startup and line-locking algorithm is settled within 30 seconds
of startup
Line-locking is started at approximately the 10 second mark
Short startup time is a significant advantage over (conventional) NDIR
which can require up to 30 minutes to “warm up”
19
20. Quantify Precision & Accuracy w/ Allan Deviation
• Use Allan Deviation calculation to quantify sensor sensitivity
• Standard deviation of moving average bins of increasing size
Time domain method of analyzing noise and drift
20
21. In-Lab Long-Term Measurement Stability
• “Constant Temperature”
measurements are compared
to line-locking measurements
Plots of typical results
Values can change by factor of
~2 up/down
• The feedback loop introduces
•
•
some noise at lower averaging
times
Overall stability is improved by
line-locking
Drift still becomes an issue at
longer averaging times
21
22. In-Lab Tests: Compare TDLAS CO2 Sensor
Measurements with Vaisala GMP343
• Calibrated TDLAS measurements are compared with a calibrated
•
Vaisala GMP343 measurements
GMP343 is a non-dispersive IR CO2 sensor
Linear fit yields a slope of 1.00 and intercept of -7.23 ppmv
Curves had to be time shifted due to spatial placement of the sensors
22
23. Soil Respiration Comparison Measurements
• In-lab soil respiration measurements show the TDLAS and Vaisala
have similarly linear response over a large CO2 concentration range.
The sensors were placed in conditions with a very large CO2
concentration gradient.
138 ppmv/min for TDLAS and 125 ppmv/min for Vaisala measured
• Field measurements showed much less CO2 respiration
A black plastic chamber was used to suppress photosynthesis
7 ppmv/min for TDLAS and 7.22 ppmv/min for Vaisala measured
23
24. Outline
• Dissertation motivation and introduction to absorption spectroscopy
• Wireless laser spectroscopic sensor node for atmospheric CO2 monitoring
C. J. Smith, S. So, L. Xia, S. Pitz, K. Szlavecz, D. Carlson, A. Terzis, and G. Wysocki, “Wireless Laser
Spectroscopic Sensor Node for Atmospheric CO2 Monitoring – Laboratory and Field Test,” Appl. Phys. B,
vol. 110, no. 2, pp. 241-248 (2013).
http://www.coas.oregonstate.edu/research/po/satellite.gif
• Quantifying and improving WSN node accuracy
• A solar-powered, distributed wireless CO2 monitoring network
C. J. Smith, P. Ramamurthy, M. A. Khan, W. Wang, S. So, M. A. Zondlo, E. Bou-Zeid, and G. Wysocki, “A
Distributed Wireless CO2 Monitoring Network,” Manuscript in Preparation.
• Real-time calibration techniques
C. J. Smith, W. Wang, and G. Wysocki, “Real-time calibration of laser absorption spectrometer using
spectral correlation performed with an in-line gas cell,” Optics Express, vol. 21, no. 19, pp. 2248822503 (2013).
• Conclusions and Future directions
24
25. Project Goal & Outline
The project goal:
• Quantify and improve field-deployed WSN node accuracy to < 1%
• Develop a solar powered autonomous wireless sensor network with each
node capable of measuring local CO2 concentration changes in a footprint
area of 1 m to 100 m radius.
http://www.coas.oregonstate.edu/research/po/satellite.gif
Outline
• Existing technology cannot accurately monitor diverse CO2 sources and
sinks
• Network deployment plans
• Field deployment and measurements
• Improving sensor node accuracy and ruggedness
• Sensor network deployment around E-Quad
25
26. Chamber measurement of CO2 exchange
• Chamber measurements are used for measuring concentration at the
smallest spatial scales of areas < 1 m3.
• Due to the size of the chamber measurement area, they result in
geographically sparse CO2 data points.
Flow-through chamber designs can have errors of as much as ±15%
In accumulation chamber designs, concentration gradients are degraded over
time as CO2 accumulates in the chamber
LI-COR Flux Chamber
Accumulation chamber & TDLAS node
26
27. Eddy Covariance measurement of CO2 exchange
• Eddy Covariance can measure the CO2 exchange of entire ecosystem
Commonly used for spatial scales on the order of 100 m to several kilometers
Uses micrometeorological theory to interpret the covariance between vertical
wind velocity and a scalar CO2 concentration measurement
Sample at as much as 20 Hz, which enables great temporal resolution in
monitoring for low time-duration events
• Limitations with the Eddy Covariance method
Most accurate during steady environmental conditions
Measurement areas with uneven terrain, diverse vegetation, or buildings cause
errors to be introduced into the measurement
27
28. Goal: Wireless Sensor Network Around E-Quad
E-QUAD at Princeton University
350 m range directional
antennas are used
Test Sight: Crop field in Princeton, NJ
• Sensor node network addresses disadvantages of Chamber and Eddy CoVariance Techniques
Complementary measurement technique
• Three locations selected to monitor coupled local environments:
1. Adjacent to the local road: car traffic
2. In the inner courtyard: local vegetation
3. On the roof of the building
28
29. First: Single Node Long-term Field Measurements
• TDLAS sensor measurements are compared with the LI-COR LI-7500A NDIR CO2
•
analyzer (40 sec. averaging)
TDLAS is placed 0.5 m above the ground
LI-COR is placed 0.8 m above the ground
Both short term and long term trends are observable
CO2 changes as small as 3 ppmv are observed by both instruments
29
30. Multi-Day Field Measurement Comparison
• Over two days unattended operation
•
•
LI-COR consumes 12 W; AC powered DC supply used
TDLAS consumes ~1 W (with fan); powered off of single Li-Ion battery
TDLAS volume is approximately ¼ that of LI-COR
Nearly all environmental conditions experienced
High & low humidity; rain & sun; high & low temperature
BUT TDLAS suffers from background “fringes”
Approximately 4% uncertainty
30
31. What are Fringes??
• Fringes are the (parasitic) filter effect of an etalon in a (coherent) beam’s path
•
Etalon can be formed from any two surfaces which reflect light
Environmental temperature changes affect etalon length
Te
1
1 F sin 2 (2 l )
31
33. Laser Injection Responsible for Most Drift
• Temperature controlled
•
•
Temperature Controlled Vessel
environment
Laser and detector temperature
is locally perturbed
Isolate sensor components most
sensitive to temperature
Detector shows no
dependence
Laser shows sensitive
dependence
33
34. Zero Gas Scans Yield Fringe Information
FFT
~Laser-MPC
separation
3.8%
standard
deviation
(15ppmv)
~Half the MPC
length
• Temperature scan VCSEL wavelength in a zero gas environment
•
•
Any signals must originate from sensor itself
FFT of zero gas scans reveal two major sources of fringing
Multi-Pass Cell (MPC) scattering
Laser to MPC entrance hole scattering
Higher frequency fringes can be ignored
Over-modulated or quickly averaged out over time
34
35. Design Solutions to Improve Accuracy
• Re-designed VCSEL-MPC mount
Robust: resistant to torque & temperature change
5 degrees of freedom for precise placement of
injection position
• MPC mirror channel widened to cone
Eliminated sidewall scatter of laser beam
35
36. Single Node Improved Field Performance
Nor’easter
• Multi-day field deployment at Broadmead Eddy-Covariance Station
• More robust opto-mechanical design and improved optical alignment
•
•
Fringing from earlier campaign is not observed here
TDLAS and Li-Cor show similar noise (~1%)
TDLAS shows higher concentration because it is near to the ground
Several diurnal cycles observed
36
37. E-Quad Field Campaign Layout & Locations
Node 1
Node 3
Princeton University Engineering-Quad (E-Quad) building
• Node 1 was deployed in the E-Quad courtyard
LICOR
•
Node 2
•
~0.5 m above the ground
Node 2 was deployed to B-wing rooftop
~23.5 m above the ground.
Node 3 was deployed at the northwest outside corner of
E-Quad near the intersection of Olden St. and a service
road leading to a parking lot
~1 m above the ground and ~1.5 m from the service road
37
38. Wireless network specifications
• TinyOS ActiveMessage used for transmission of data
•
Single-hop only
Transmission rates as fast as 250 kbps
6 Hz transmission of sensor data packets (30 bytes each, ~1 kbps)
MultiHopRouter, Tymo (Dynamic MANET On-demand implementation)
available for multihop
Built on ActiveMessage protocol
Node bandwidth is reduced due to aggregate bandwidth limit and increased
overhead
TDMA with data update
every 15 seconds
Node 3
Base Station
Node 2
Node 1
38
39. Solar Irradiance Calculations
• Calculations based on historical
Princeton, NJ solar irradiance data
Found a 100 Ah battery with 35 W
panels is needed for areas of shade
(1/3 direct sunlight per day)
Corresponds to 3 sq. ft. of solar panels
• Solar panel power is rated based on 1
•
•
kW/m2 irradiance
Enabled Nodes 1 and 3 to be solar
powered indefinitely
For comparison, Eddy Covariance
stations typically consume a minimum
of 12 W power
Would require a minimum of 350 W
solar panels
Corresponds to > 30 sq. ft. of solar
panels
39
40. Field Campaign Measurements Over a Week
• 30 minute averages shown
• 5 minute rolling average σ is 1.6 – 3.7 ppmv (depending on the node)
• TDLAS measurements compared against commercial LI-COR Non-Dispersive
•
•
Infrared (NDIR) CO2 sensor
Node 2 on rooftop
Large changes such as diurnal cycles are common to all three nodes
Node 3 is largely decoupled from Nodes 1 & 2
Street corners have increased turbulence
40
41. LI-COR and Node 2 Correlation
Perfect correlation
• All sensor nodes calibrated a priori with known CO2 concentration.
• Scatter plot of the LI-COR and TDLAS sensor Node 2, computed for Jan. 11
• The measurements are in good agreement
• A robust regression (with downweighting of outliers) between the two
measurements produces a slope of 0.9966 and an offset of 8.1 ppmv
Approximately the calibration accuracy of the two instruments.
41
42. Vignette of Jan. 11 Network Measurements
• The network captures some of the localized
•
•
•
•
effects induced by the landscape geometry
The low wind speed (< 1m/s) and ustar (<.2
m/s) indicate low turbulence
Less mixing during this period.
These conditions lead to a gradual build up of
CO2 (from approximately 11.4 to 11.6)
At the courtyard, aided by low ventilation, the
buildup of CO2 is higher/more gradual
compared to other nodes.
Sources and sinks vary from within the EQuad courtyard to out on the street
The sharp dip a little past 11.5 UTC is only
visible at the Courtyard and Rooftop node.
The Street Corner node does not pick up this
dip.
12 AM
7 AM
2:15 PM
42
43. Outline
• Thesis motivation and introduction to absorption spectroscopy
• Wireless laser spectroscopic sensor node for atmospheric CO2 monitoring
C. J. Smith, S. So, L. Xia, S. Pitz, K. Szlavecz, D. Carlson, A. Terzis, and G. Wysocki, “Wireless Laser
Spectroscopic Sensor Node for Atmospheric CO2 Monitoring – Laboratory and Field Test,” Appl. Phys. B,
vol. 110, no. 2, pp. 241-248 (2013).
• Quantifying and improving WSN node accuracy
http://www.coas.oregonstate.edu/research/po/satellite.gif
• A solar-powered, distributed wireless CO2 monitoring network
C. J. Smith, P. Ramamurthy, M. A. Khan, W. Wang, S. So, M. A. Zondlo, E. Bou-Zeid, and G. Wysocki, “A
Distributed Wireless CO2 Monitoring Network,” Manuscript in Preparation.
• Real-time calibration techniques
C. J. Smith, W. Wang, and G. Wysocki, “Real-time calibration of laser absorption spectrometer using
spectral correlation performed with an in-line gas cell,” Optics Express, vol. 21, no. 19, pp. 2248822503 (2013).
• Conclusions and Future directions
43
44. Project Goal & Outline
The project goal:
• Develop and implement a technique for real-time calibration of
portable trace-gas sensors
Using spectral correlation with a revolving in-line gas cell
Minimizes sources of drift in real time
http://www.coas.oregonstate.edu/research/po/satellite.gif
Outline
• Key challenges to long-term sensor measurement stability
• Conventional calibration methods
• Overview of the rotational reference cell implementation
• Experimental results
44
45. Measurement Noise & Drift Reduce Sensitivity
Measurement drift can be induced
by many factors:
• Fabry-Perot Fringing
Opto-mechanical instability
Allan Deviation
(sensitive to ambient
temperature change)
Scattering
• Electronics instability
• Optical power fluctuation
Averaging Time (sec)
Recurring Calibration Required
45
46. Conventional Calibration Methods
• Send beam to separate optical
branch
Use separate reference cell
Subject to different parasitic
I0
IDet,1
Ambient
Detector 1
fringes
• Multiple detectors
Reference Cell
With different noise & drift
IDet,2
Detector 2
• Single gas cell
• Single detector
• Cycle between reference and
•
sample gases
Maintenance challenge
Lack of portability and
autonomy
Ambient
Ref. Gas
I0
Inlet
Outlet
IDet
Detector
46
47. Permanent In-Line Reference Cell
Separate reference cell signal and ambient signal
using gas parameters and WMS
• Permanently insert a low-pressure reference cell in the beam path
•
•
•
Contains the same gas as sampled
Reference beam experiences the same fringes as the ambient/sampling beam
No complex gas handling required
Only single detector is needed
47
48. Key WMS Parameters
Laser frequency change
Amplitude = β
Key characteristics of Wavelength
Modulation Spectroscopy (WMS)
Absorption Line Shape
νHWHM
Laser
Detector
• Modulate laser
•
wavelength at a
high frequency (f) and with
modulation depth/amplitude (β)
Avoids 1/f noise
Demodulate and filter at
multiples of f
Low-noise, “derivative-like”
1f
2f
spectral envelopes
• WMS Signal is proportional to
laser power and depends on
3 key parameters:
Line-width, νHWHM (cm-1)
Modulation depth, β (cm-1)
Harmonic (e.g., 1f, 2f, 3f, …)
m
3f
HWHM
48
49. Simultaneous Detection of Sample and Reference
IDet Detector
2f ambient
6f reference
2f
I0
IDet Detector
• Simultaneous 2f & 6f demodulation
• Selectively suppress ambient sample or low pressure reference signal
• Real-time, in-line calibration is possible
6f
IDet Detector
I0
IDet Detector
49
50. Optimum Reference Signal Isolation
• Experimentally varied the modulation index for pressures from 50 to 1000 Torr
• Line-center value used for comparison
• Good agreement between experiment and simulation
• Ambient signal into reference cell signal cross-talk of ~4%
Ambient signal is at or below the observed noise floor
50
52. Suppress Drifts by Division of Background Signal
Raw Spectra Scans
Reference Path:
Raw Scan with Spectral Fit
I ref ( ) Tref I 0e
( b ( ) Lb ref ( ) La )
Corrected Reference Scan
Background Corrected
Reference Signal
Tref , c ( )
Background Path:
I zero ( ) Tzero I 0 e b ( ) Lb
I ref ( )
I zero ( )
Tref
TZero
e
ref ( ) La
• Baseline and fringes are suppressed through division
Same process applies to the sample signal
Spectral fitting removes baseline but not fringes
52
53. Long-term Suppression of Drift
Time Series of Uncorrected and Corrected Signals
Allan Deviation of Uncorrected and Corrected Signals
1 ppmv
1sec, 1σ = 1.75 ppmv
• Measure away from absorption line
•
•
Assess instrument stability
1 sec. 1σ sensitivity is 3.510-4 (1.75 ppmv)
Sensitivity of 610-5 (0.3 ppmv) after 100 sec. averaging & sustained
past 3000 sec.
53
54. Incomplete Correction of Absorption Peak Signal
Time Series of Signals at the Peak and Away
Allan Deviation of Signals at the Peak and Away
1 ppmv
• Measured reference gas stability at and away from absorption peak
• Long-term drift remains for on-line measurements
• Full spectral fit of corrected reference also shows drift
•
Improves 1 sec sensitivity ~2 to 210-4 (1 ppmv)
Uses full spectral information
Background signal 4-5 reference signal
Difficult to suppress but can calibrate sample against reference
54
55. Baseline Drift & Error Correction
Background-Corrected Scans at
Different Experiment Times
Scatter Plots Tsample,c(Tref,c)
Before & After Baseline Correction
• Baseline drift introduces error into single spectral point measurements
•
Differences in reference & sample baseline also introduce error
Use Sample-Reference regression + fundamental principles to correct
At 100% transmission the fit of Tsample,c(Tref,c) should intersect (1,1) coordinate
Scale spectra to meet this condition
55
56. Calibration Through Spectral Correlation
Scatter Plots Tsample,c(Tref,c)
Before & After Baseline Correction
TS (t n ) mTR (t n ) y0
Baseline
Corrected
Transmissions
TR (tn ) Tref , c (tn )
Bsim
Bmeas
TS (t n ) Tsample,c (t n )
Bsim,meas are simulated & measured baselines
β is a modeled transmission correction factor
• Use slope (m) of TS(TR) to calibrate sample concentration.
Slope is proportional to the ratio of analyte concentrations.
[CO2 ]sample m [CO2 ]ref
• Point-by-point spectral correlation of time domain data.
Uses all spectral data (like a spectral fit).
Frequency calibration not needed.
56
57. Single point, spectral fit, & spectral correlation
Time Series Showing Effects of Calibration
Allan Deviation Showing Effects of Calibration
Drift
reduction
• Single point & spectral correlation calibration suppress drift
• Full spectral fit shows ~6 % offset
•
Consistent with sample and reference baseline differences
1.7 increase in 1s sensitivity (2.35 ppmv to 1.33 ppmv)
Drift remains
Spectral correlation calibration has accuracy & precision of fit.
Removes baseline error without frequency calibration.
57
58. Conclusion
Solar-powered Distributed Wireless Trace-gas Sensor Network
• Built a sensor network to monitor atmospheric trace-gases.
Captured events on different time and spatial scales.
• Network sensor nodes operated autonomously .
Placed where CO2 exchange is difficult to quantify.
• Sensor nodes have similar sensitivity as commercial sensors.
Consume 1/10th the power and more compact.
Greater sustained operation than commercial sensors.
Spectral Correlation with Revolving Reference Cell
• Developed a novel in-line drift suppression & calibration technique.
Uses spectral correlation with a revolving in-line gas cell.
Provides real-time calibration.
• Divide the sample (reference) signals by the background spectrum.
Three sub-cells share the same optical interfaces.
Parasitic interference fringes are minimized.
• Spectral correlation calibration technique.
Maintains concentration retrieval accuracy.
Improves measurement precision.
Uses entire spectrum without costly wavelength calibration.
58
59. Future Work
Solar-powered Distributed Wireless Trace-gas Sensor Network
• Incorporate the sensor network as an environmental model input .
• Change communications interface to forward over 3G or equivalent network.
Ubiquitous and robust network infrastructure.
• Redesign MPC to accommodate a larger sized beam to further reduce clipping.
• Develop an electronics board with improved capabilities
On-board temperature and pressure sensors.
Laser drive current ramping
Spectral modeling and measurement of multiple species
Spectral Correlation with Revolving Reference Cell
• Minimize background signal magnitude with solid optical waveguides
Calibration in a controlled atmosphere
Operation at the Brewster’s angle to further suppress fringes
Field deployment to act as local calibration source
59
60. List of Publications (1)
Journal
M. Nikodem, D. Weidmann, C. Smith, and G. Wysocki, “Signal-to-noise ratio in chirped laser dispersion spectroscopy,” Opt. Express, vol. 20, no. 1, pp.
644-653 (2012).
C. J. Smith, S. So, L. Xia, S. Pitz, K. Szlavecz, D. Carlson, A. Terzis, and G. Wysocki, “Wireless Laser Spectroscopic Sensor Node for Atmospheric CO2
Monitoring – Laboratory and Field Test,” Appl. Phys. B, vol. 110, no. 2, pp. 241-248 (2013).
C. J. Smith, W. D. Li, G. Wysocki, and S. Y. Chou, “Electric Current Tuning the Self-Oscillation Frequency of EC-VCSELs,” Photonics Technology
Letters, IEEE, vol. 25, no. 17, pp. 1707-1710 (2013).
C. J. Smith, W. Wang, and G. Wysocki, “Real-time calibration of laser absorption spectrometer using spectral correlation performed with an in-line gas
cell,” Optics Express, vol. 21, no. 19, pp. 22488-22503 (2013).
C. J. Smith, P. Ramamurthy, M. A. Khan, W. Wang, S. So, M. A. Zondlo, E. Bou-Zeid, and G. Wysocki, “A Distributed Wireless CO2 Monitoring Network,”
Manuscript in Preparation.
Proceedings
Stephen So, Evan Jeng, Clinton Smith, David Krueger, and Gerard Wysocki, “Next generation infrared sensor instrumentation: remote sensing and sensor
networks using the openPHOTONS repository (Proceedings Paper),” Infrared Remote Sensing and Instrumentation XVIII, Marija Strojnik; Gonzalo
Paez, Editors, vol. 7808, no. 780818, 2010.
Clinton J. Smith, Stephen So, Amir Khan, Mark A. Zondlo, and Gerard Wysocki, “Low-power wireless trace gas sensing network (Proceedings Paper),”
Sensing Technologies for Global Health, Military Medicine, Disaster Response, and Environmental Monitoring; and Biometric Technology for Human
Identification VIII, Sárka O. Southern; Kevin N. Montgomery; Carl W. Taylor; Bernhard H. Weigl; B. V. K. Vijaya Kumar; Salil Prabhakar; Arun A.
Ross, Editors, vol. 8029, no. 80291G, 2011.
M. Nikodem, C. Smith, D. Weidmann, and G. Wysocki, “Remote mid-infrared sensing using chirped laser dispersion spectroscopy (Proceedings Paper),”
Advanced Environmental, Chemical, and Biological Sensing Technologies VIII, Tuan Vo-Dinh; Robert A. Lieberman; Gnter Gauglitz, Editors, vol 8024, no.
80240F, 2011.
60
61. List of Publications (2)
Conference
C. J. Smith, W. Li, S. Bai, and S. Y. Chou, “High Frequency Polarization Switching VCSEL Clock Using Subwavelength Quarter-Wave Plate,” in Conference on
Lasers and Electro-Optics/International Quantum Electronics Conference, OSA Technical Digest (CD) (Optical Society of America, 2009), paper CMP6.
C. J. Smith, S. So, and G. Wysocki, "Low-Power Portable Laser Spectroscopic Sensor for Atmospheric CO2Monitoring," in Conference on Laser ElectroOptics: Applications, OSA Technical Digest (CD) (Optical Society of America, 2010), paper JThB4.
C. J. Smith, S. So, A. Khan, M. A. Zondlo, and G. Wysocki, "Temperature Impact on the Long-Term Stability of a Portable Laser Spectroscopic CO2 Sensor,"
in Quantum Electronics and Laser Science Conference, OSA Technical Digest (CD) (Optical Society of America, 2011), paper JWA117.
C. J. Smith, W. Li, G. Wysocki, and S. Y. Chou, “Drive-Current Tuning of Self-Oscillation Frequency of External Cavity VCSEL,” in CLEO:2011 – Laser
Applications to Photonic Applications, OSA Technical Digest (CD) (Optical Society of America, 2011), paper CTuP3.
M. Nikodem, C. J. Smith, D.Weidmann, and G. Wysocki, “Chirped Laser Dispersion Spectroscopy with baseline-free 2nd harmonic detection,” in
CLEO:2011 - Laser Applications to Photonic Applications, OSA Technical Digest (CD) (Optical Society of America, 2011), paper CThT4.
C.J. Smith, S. So, and G. Wysocki, “Wireless sensor network for environmental CO2 monitoring,” in Field Laser Applications in Industry and Research
(FLAIR), 2011.
D. Richter, S. Meyer, S.M. Spuler, C.J. Smith, S. So, and G. Wysocki, “Ultra compact VCSEL based CO2 spectrometer for flux measurements,” in Field Laser
Applications in Industry and Research (FLAIR), 2011.
M. Nikodem, C. Smith, D. Weidmann, and G. Wysocki, “Open-path remote sensing of nitrous oxide using chirped laser dispersion spectroscopy,” in Field
Laser Applications in Industry and Research (FLAIR), 2011.
W. Wang, C. Smith, S. So, E. Bou-Zeid, and G. Wysocki, "Wireless Sensor Networks for Monitoring of Atmospheric Chemicals," in Renewable Energy and
the Environment, OSA Technical Digest (CD) (Optical Society of America, 2011), paper JWE19.
C. J. Smith, M. A. Khan, M. A. Zondlo, and G. Wysocki, "In-Line Reference Cell for Real-Time Calibration of Laser Absorption Spectrometers," in CLEO:
Science and Innovations, OSA Technical Digest (online) (Optical Society of America, 2012), paper CW3B.1.
C. Smith, W. Wang, and G. Wysocki, “A Rotational Sample/Reference Cell for High-accuracy Real-time Spectroscopic Trace-gas Sensing,” in CLEO:
2013, OSA Technical Digest (online) (Optical Society of America, 2013), paper CW1L.3.
61
62. Acknowledgements
Prof. Gerard Wysocki
Prof. Craig Arnold
Prof. Naveen Verma
Prof. Claire Gmachl
Prof. Elie Bou-Zeid
Prof. Mark Zondlo
Prof. M. Amir Khan
Prof. Stephen Chou
Dr. Stephen So
Dr. Tracy Tsai
Yin Wang
Wen (Eve) Wang
Dr. Michal Nikodem
Dr. Brian Brumfield
Dr. Andreas Hangauer
Dr. Prathap Ramamurthy
Genny Plant
Eric Zhang
Alex Hallden-Abberton
ELE Staff
Friends & Family
Haining Helen Yu
This work was sponsored in part by:
The National Science Foundation’s MIRTHE Engineering Research Center
An NSF MRI award #0723190 for the openPHOTONS systems
An innovation award from The Keller Center for Innovation in Engineering Education
National Science Foundation Grant No. 0903661 “Nanotechnology for Clean Energy IGERT”
62
65. Interband Emission vs. Intersubband Emission
M. Razeghi, S. Slivken, A. Evans, J. Nguyen, Y. Bai, J. S. Yu, S. R. Darvish, and K. Mi, “High power semiconductor lasers make mid-infrared wavelengths accessible,” SPIE Newsroom, 2006.
65
66. Vignette of Street Corner
• When the Node 3 data (near the street-corner) is examined with only 15
•
seconds of averaging, the influence of passing cars can be detected
Direction of the tail-pipe and the size and model of the car correlate with the
degree of the increase in CO2 concentration
Traditional internal combustion engine based cars with a tail-pipe facing the
direction of the sensor cause much higher concentration spikes than hybrid
vehicles (for which there is no measurable concentration change).
• Larger vehicles have a much greater impact on the local CO2 concentration.
66
67. Wireless Communications Interface
•
•
•
Commercial Xbow TelosB wireless interface card
IEEE 802.15.4/ZigBee compliant communications
Running TinyOS
Communicates with acquisition & control board via UART
Communicates with the base station PC via USB
Labview used for control and data logging
http://moodle.utc.fr/file.php/498/SupportWeb/co/Module_RCSF_35.html
67
69. Lock-in Detection
V psd VsigVL sin( r t sig ) sin( Lt ref )
1/ 2VsigVL cos([r L ]t sig ref )
1/ 2VsigVL cos([r L ]t sig ref )
If L r then..
V psd 1 / 2VsigVL cos( sig ref )
Vpsd is proportional to signal
amplitude
69
Editor's Notes
The goal of our project is to develop CO2 sensors that can be deployed into a network in the field for real-time monitoring of carbon flux over a large geographic area. The outline of this talk is as follows: I’ll talk about the requirements for a sensor to be used in such a network, I’ll overview our sensor design by talking about the optics and electronics, I”ll review sensor performance tests of SNR and Allan variance, and finally I’ll review different lab and field tests we performed in collaboration with our colleagues at JHU.
A sensor must ideally have all of these features to be used in a trace gas sensor network.
A sensor must ideally have all of these features to be used in a trace gas sensor network.
A sensor must ideally have all of these features to be used in a trace gas sensor network.
A sensor must ideally have all of these features to be used in a trace gas sensor network.
The goal of our project is to develop CO2 sensors that can be deployed into a network in the field for real-time monitoring of carbon flux over a large geographic area. The outline of this talk is as follows: I’ll talk about the requirements for a sensor to be used in such a network, I’ll overview our sensor design by talking about the optics and electronics, I”ll review sensor performance tests of SNR and Allan variance, and finally I’ll review different lab and field tests we performed in collaboration with our colleagues at JHU.
A sensor must ideally have all of these features to be used in a trace gas sensor network.
A sensor must ideally have all of these features to be used in a trace gas sensor network.
We are using TDLAS spectroscopy for CO2 detection. The optical path is 3.5 m within a Herriott multi-pass cell. The sensor uses a 2um VCSEL as the laser source and an off the shelf InGaAs photodetector. The multi-pass cell components are mostly stock commercial parts with custom Al adapters for housing the laser and detector. All of these parts are easily bought/made and assembled. The VCSEL is in a TO5 can package and has its own TEC for temp control. The detectors is in a TO18 package and is mounted on a custom PCB with an integrated pre-amp. Both the laser and the detector interface with a custom electronics board that does all the laser control and data acquisition – based on the openPhotons platform. The custom electronics board mates with a commercial wireless card. Another wireless card is plugged into a computer and the two communicate through a wireless link. All of the electronics (including real-time wireless transmission) consume approximately 300 mW of power. The system is run off of a 10Ah Li-ion polymer battery that will last for approx. 100 hours in this manner. For fieldwork, the Herriott cell and electronics were mounted inside a water-tight NEMA enclosure. The total size of the system is about that of a shoebox. In this case, a pump is used to pump outside air (that is passed through a desiccant) into the chamber and then back out. This allows for sampling of environmental CO2 while keeping the electronics protected from humidity/water/dew. With the pump running, the battery will power the system for about 10 hours.
Our custom control and acquisition board developed by Dr. Stephen So provides all the functionality for controlling the VCSEL and processing the data from the detector. The board is designed to communicate with a Telos mote via the standard UART protocol. Updated control board with all of the functionality integrated.TEC driver provides 0.001C stability, precision modulation frequency to match to photoacoustic or faraday rotation magnetic coils
The VCSEL is wavelength modulated at 10 kHz at a modulation depth that corresponds to optimum SNR when modulating over the absorption line. (amplitude/HWHM = 2.2). The 1st, 2nd, & 3rd harmonic line profiles are measured by temperature scanning about the 4987 cm-1 absorption line and a lock-in amplifier was used to select each harmonic. Harmonic SNR measurement of a calibrated 285 ppm CO2 in N2 mixture produced 1F SNR of 3247, 2F SNR of 2530, & 3F SNR of 1052.
The VCSEL is wavelength modulated at 10 kHz at a modulation depth that corresponds to optimum SNR when modulating over the absorption line. (amplitude/HWHM = 2.2). The 1st, 2nd, & 3rd harmonic line profiles are measured by temperature scanning about the 4987 cm-1 absorption line and a lock-in amplifier was used to select each harmonic. Harmonic SNR measurement of a calibrated 285 ppm CO2 in N2 mixture produced 1F SNR of 3247, 2F SNR of 2530, & 3F SNR of 1052.
The TDLAS CO2 sensor was placed in an environmental chamber with a commercial sensor which uses NDIR (nondispersive infrared sensor). The temperature was set to 0C and the CO2 concentration was ramped up and down by ~100 ppm. The TDLAS sensor performed exactly as the TDLAS (in terms of precision, not accuracy). Additionally, another lab test of measuring soil respiration over time was performed by the TDLAS sensor. In this case the sensor showed good responsivity to a larger change of CO2 concentration and reported a CO2 concentration increase slope consistent with soil respiration measurements. NEED REFERENCE FOR THIS…
Next, the TDLAS CO2 sensor was used in an experiment to detect CA isopod respiration. 10 isopods were placed into a test tube which was a part of a closed path system containing the CO2 sensor. During the course of this experiment CO2 out-gassing was observed. This is likely due to CO2 being trapped in the desiccant. Nevertheless the CO2 out-gassing rate was constant over several measurements, so it was used as a basline. When the isopods were placed in the test tube several times for different measurements of CO2 concentration increase. When this data was compared to the baseline, an increase in CO2 concentration was observed. The isopods were detectable after 2 minutes of CO2 concentration increase. This corresponds with a detected rate of increase of approx. .021 ppm/sec.
We are using TDLAS spectroscopy for CO2 detection. The optical path is 3.5 m within a Herriott multi-pass cell. The sensor uses a 2um VCSEL as the laser source and an off the shelf InGaAs photodetector. The multi-pass cell components are mostly stock commercial parts with custom Al adapters for housing the laser and detector. All of these parts are easily bought/made and assembled. The VCSEL is in a TO5 can package and has its own TEC for temp control. The detectors is in a TO18 package and is mounted on a custom PCB with an integrated pre-amp. Both the laser and the detector interface with a custom electronics board that does all the laser control and data acquisition – based on the openPhotons platform. The custom electronics board mates with a commercial wireless card. Another wireless card is plugged into a computer and the two communicate through a wireless link. All of the electronics (including real-time wireless transmission) consume approximately 300 mW of power. The system is run off of a 10Ah Li-ion polymer battery that will last for approx. 100 hours in this manner. For fieldwork, the Herriott cell and electronics were mounted inside a water-tight NEMA enclosure. The total size of the system is about that of a shoebox. In this case, a pump is used to pump outside air (that is passed through a desiccant) into the chamber and then back out. This allows for sampling of environmental CO2 while keeping the electronics protected from humidity/water/dew. With the pump running, the battery will power the system for about 10 hours.
Redundant phrases2 (electronics and optics) points which include the others
Get rid of multi-pass cells
Too much textSample – ambientZero (non-absorbing)3D exploded view of rotational cell
Combien this slide with previous using optimized dataStart with corrected and uncorrected then flip to red, green, and blue
Maybe split into two slides
Maybe split into two slides
We are using TDLAS spectroscopy for CO2 detection. The optical path is 3.5 m within a Herriott multi-pass cell. The sensor uses a 2um VCSEL as the laser source and an off the shelf InGaAs photodetector. The multi-pass cell components are mostly stock commercial parts with custom Al adapters for housing the laser and detector. All of these parts are easily bought/made and assembled. The VCSEL is in a TO5 can package and has its own TEC for temp control. The detectors is in a TO18 package and is mounted on a custom PCB with an integrated pre-amp. Both the laser and the detector interface with a custom electronics board that does all the laser control and data acquisition – based on the openPhotons platform. The custom electronics board mates with a commercial wireless card. Another wireless card is plugged into a computer and the two communicate through a wireless link. All of the electronics (including real-time wireless transmission) consume approximately 300 mW of power. The system is run off of a 10Ah Li-ion polymer battery that will last for approx. 100 hours in this manner. For fieldwork, the Herriott cell and electronics were mounted inside a water-tight NEMA enclosure. The total size of the system is about that of a shoebox. In this case, a pump is used to pump outside air (that is passed through a desiccant) into the chamber and then back out. This allows for sampling of environmental CO2 while keeping the electronics protected from humidity/water/dew. With the pump running, the battery will power the system for about 10 hours.
We are using TDLAS spectroscopy for CO2 detection. The optical path is 3.5 m within a Herriott multi-pass cell. The sensor uses a 2um VCSEL as the laser source and an off the shelf InGaAs photodetector. The multi-pass cell components are mostly stock commercial parts with custom Al adapters for housing the laser and detector. All of these parts are easily bought/made and assembled. The VCSEL is in a TO5 can package and has its own TEC for temp control. The detectors is in a TO18 package and is mounted on a custom PCB with an integrated pre-amp. Both the laser and the detector interface with a custom electronics board that does all the laser control and data acquisition – based on the openPhotons platform. The custom electronics board mates with a commercial wireless card. Another wireless card is plugged into a computer and the two communicate through a wireless link. All of the electronics (including real-time wireless transmission) consume approximately 300 mW of power. The system is run off of a 10Ah Li-ion polymer battery that will last for approx. 100 hours in this manner. For fieldwork, the Herriott cell and electronics were mounted inside a water-tight NEMA enclosure. The total size of the system is about that of a shoebox. In this case, a pump is used to pump outside air (that is passed through a desiccant) into the chamber and then back out. This allows for sampling of environmental CO2 while keeping the electronics protected from humidity/water/dew. With the pump running, the battery will power the system for about 10 hours.
We are using TDLAS spectroscopy for CO2 detection. The optical path is 3.5 m within a Herriott multi-pass cell. The sensor uses a 2um VCSEL as the laser source and an off the shelf InGaAs photodetector. The multi-pass cell components are mostly stock commercial parts with custom Al adapters for housing the laser and detector. All of these parts are easily bought/made and assembled. The VCSEL is in a TO5 can package and has its own TEC for temp control. The detectors is in a TO18 package and is mounted on a custom PCB with an integrated pre-amp. Both the laser and the detector interface with a custom electronics board that does all the laser control and data acquisition – based on the openPhotons platform. The custom electronics board mates with a commercial wireless card. Another wireless card is plugged into a computer and the two communicate through a wireless link. All of the electronics (including real-time wireless transmission) consume approximately 300 mW of power. The system is run off of a 10Ah Li-ion polymer battery that will last for approx. 100 hours in this manner. For fieldwork, the Herriott cell and electronics were mounted inside a water-tight NEMA enclosure. The total size of the system is about that of a shoebox. In this case, a pump is used to pump outside air (that is passed through a desiccant) into the chamber and then back out. This allows for sampling of environmental CO2 while keeping the electronics protected from humidity/water/dew. With the pump running, the battery will power the system for about 10 hours.