Terahertz radar imaging for standoff personnel screening European Microwave Conference, October 211 Ken Cooper Submillimeter-Wave Advanced Technology (SWAT) Team NASA Jet Propulsion Laboratory California Institute of Technology
The JPL SWAT Team
Active THz Imaging: SNR Potential benefits of active, narrowband THz imaging: - Very high SNR is possible because of high-power sources and low-noise detectors - Video-rate imaging feasible with small number of transceivers clothing: x1-1 two-way loss 1 m diameter aperture 1 mw x 1-8 = 1 pw 7, K noise =.1 pw/µs 1 mw @ 67 GHz body: x1-2 reflection m standoff: x1-4 geometric loss (beam-spreading) x1-1 atmospheric loss (worst case)
Active THz Imaging: SNR Potential benefits of active, narrowband THz imaging: - Very high SNR is possible because of high-power sources and low-noise detectors - Video-rate imaging feasible with small number of transceivers clothing: x1-1 two-way loss 1 m diameter aperture 1 mw x 1-8 = 1 pw SNR = 1 with only 1 µs integration time! 7, K noise =.1 pw/µs 1x1 pixels1 mw in @ 1 ms possible with a single beam body: x1 reflection 67 GHz -2 m standoff: x1-4 geometric loss (beam-spreading) x1-1 atmospheric loss (worst case)
JPL s THz Imaging System, 211 D Tx C 1 cm aperture 2 PA B Rx 67 GHz beam path.5 mw power SHM A/D and signal processing 3.6 GHz LNA PA C E fast-scanning subreflector 1.8-3.4 GHz front/back-end electronics I Q F 18 G 49 MHz A 34.8 GHz 35 GHz ADC 2 MS/s K.B. Cooper, et al., THz Imaging Radar for Standoff Personnel Screening, IEEE THz Sci. & Tech., 211.
Detection Image Gallery small mock bomb belt (2.3 cm thick) replica hand gun (2.5 cm thick) large mock bomb belt (2.5 cm thick) Requirements for detection: solid object, ~1 inch in size (in 3 dimensions) Material of concealed object is irrelevant to detection
Radar Basics Frequency-modulated continuous wave (FMCW) radar: appropriate when available power is limited. frequency K = chirp rate (Hz/s) = target delay f IF IF receive time Range resolution: inversely proportional to chirp bandwidth IF power 2 KR c c r 2 F f IF K THz systems can achieve enormous bandwidths, and hence range resolution. IF frequency
FMCW Waveform Nonlinearity raw point-target range spectrum 5 Power (db) 4 Range resolution from 28.8 GHz bandwidth 3 expected: <1 cm 2 achieved: ~ cm 1 24 26 Target range (m) D Tx Broadening from modulation in amplitude and chirp profile causes modulation in IF signal. But if deterministic, then this can be subtracted out! Rx 2 PA ݐ ݏ A t exp w ݐ dφ ݐ SHM LNA PA ݏ ݐ A ܮ t exp w ݐ dφ ݐ ݏ ݐ w ݐ ή ܣ ሺ ݐ ǡ ሻ ʹߨ ήߜ ሺ ݐ ǡ ሻ
Nonlinearity Compensation 1. Acquire calibration waveform on point target: ݏ ݐ w Ͳ ݐ ή ܣ ሺ ݐ ǡ ሻ ʹߨ ήߜ ሺ ݐ ǡ ሻ 2. Divide subsequent IF signals by (complex) calibration: bandwidth-limited resolution achieved ݏ ݐ ݏ ݏ ݐ ݐ ߜ ݐ ǡ ߜ ݐ ǡ ߜ ݐ ǡ 4 cm ܣ ݐ ǡ ܣ ݐ ǡ ܣ ݐ ǡ 5 Power (db) 4 3 2 1 23.5 24 24 24.5 Target range (m).5 26 26 26.5 24 26 Target range (m)
Achieving Stable, Fast, and Low-Noise Waveforms JPL built and designed chirper 1.6-3.2 GHz up/down chirp in <.1 ms Very low phase noise Digital synthesis stability Careful digital/analog ground isolation -5 Phase noise (dbc/hz) -6 DDS provides good range resolution stability: 5 5 5 4 4 4 3 3 3 2 2 2 1 1 1-7 -8-9 -1-11 -12-13 -1 24.8 24.9.1.2-1 24.8 24.9.1.2-1 24.8 24.9.1.2.1 1 1 1 1 1 Carrier offset (khz)
Imaging Algorithm: Surfaces of Last Scattering High range resolution at cm-scale is critical for being able to digitally peel away layers of clothing and reveal potential threats. 24.2 24.3 relative 24.4-5 range 5 cm
Phase Noise Limitations target: 3 metal ball 5 Power (db) 4 measured SNR: 34-44 db (phase noise floor) 3 2 1 potential SNR: 67 db (thermal noise floor) 23.5 24 24 24.5.5 26 26 26.5 thermal noise floor at -13 db on this scale IF frequency (MHz) Idealized phase noise model: Phase noise is multiplied in frequency multipliers For simple heterodyne measurement, transmit and LO phase noise is uncorrelated delay to target Tx 18 Expected phase-noise floor: -97 dbc/hz + 3 db 2log(18) db + 1log(1kHz) db = -29 db Rx 2 ଶ(d 2 + 2 ) 18 We get -44 dbc above; why so good? Can we do even better?
Phase Noise Cancellation via Homodyne delay to target: Tx Rx 18 2p ݐ D Tx 2 PA B Rx SHM» ܨܫ a 2p t sin 2p ݐ 18 2p ݐ 3.6 GHz PA Phase noise in IF will vanish if the electrical path delays are balanced One source of imbalance is the time-of-flight of 17 ns = 6 MHz -1 At 1 khz offset, expect cancellation of (2 1kHz 17 ns)2 = -19 db Expect -29 dbc -19 db = -48 dbc phase noise floor; close agreement w/ msmt C E This makes the heterodyne effectively homodyne! 1.8-3.4 GHz LNA C I Q F 18 G 49 MHz A 34.8 GHz 35 GHz ADC 2 MS/s J.L. Doane, Broadband superheterodyne tracking circuits in millimeter-wave measurements, 198.
Phase Noise without Homodyne Cancellation 44 db Power (db) 4 3 29 db 5 Increased close-in phase noise without homodyne cancellation 2 D Tx PA Beyond ~1MHz, time of flight delay is too long for noise cancellation to be effective. B SHM 3.6 GHz PA C E -1 1 IF signal offset (MHz) 1.8-3.4 GHz LNA -2 Rx 1 A 2 3.6 GHz Confirmation: 2 C -29 dbc noise floor restored without homodyne circuit in place. G I Complicated by chirp source common to Tx and LO chains, and by dispersion in submm electronics Can tunable cancellation be improved? Q 34.8 GHz 35 GHz 49 MHz ADC 2 MS/s
Goal: Near-Video Rate THz Radar Imaging Higher frame rates wanted: rapid crowd scanning, subjects in motion, wider field of view Better THz components will not help: bottleneck from mechanical scanning Path to video rates: scanning multiple beams simultaneously N beams = N x speed-up But it s not practical to duplicate the front-end module 8 times! Concept: 3D stacks of precision micromachined silicon waveguide: extremely compact integration input waveguide wafer #1 wafer #2 wafer #3 wafer #4 ~5 inches output waveguide
Conclusions 67 GHz imaging radar is effective at detecting concealed threats Good penetration allows for covert operation and seeing through even thick clothing Chirp stability and calibration is critical for achieving high range resolution and hence high contrast imagery Careful circuit architecture can improve noise floor and dynamic range Transceiver array development is underway to reach near-video frame rates This work was carried out by the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration, and was supported by the United States Naval Explosive Ordnance Disposal Technology Division. Copyright 211 California Institute of Technology