mmwave Radar for Automotive and Industrial Applications

Transcription

1 mmwave adar for Automotive and Industrial Appliations Deember 7, 2017 Karthik amasubramanian Distinguished Member of ehnial Staff, exas Instruments 1

2 Introdution adar tehnology has been in existene for several deades Military, Weather, Law enforement, and so on L/M In the past deade, use of radar has exponentially inreased Automotive and Industrial appliations Automotive appliations Front-faing radar (L/M) Adaptive Cruise Control, Autonomous Emergeny Braking Corner radar (S) Blind Spot Detetion, Lane Change Assist, Front/ear Cross raffi Alert Newer appliations Automated parking, 360 degree surround protetion Body/Chassis and In-abin appliations S Industrial appliations Fluid level sensing Solid volume identifiation raffi monitoring and Infrastruture systems obotis, and many others 2

3 77GHz mmwave adar mmwave: F frequenies within 30 GHz to 300 GHz Wavelength is in the order of few millimeters 77GHz mmwave radar bands GHz Alloated for vehiular radar in many ountries Also available for infrastruture systems in ertain regions GHz eently made available for short range radar Legay 24 GHz UWB short range radar to be phased out by GHz: Available for level probing radar mmwave radar sensors an measure adial distane (range) to the objet elative radial veloity to the objet Angle of arrival using multiple X, X Some benefits of radar obust to environmental onditions like dust/fog/smoke Operation in dazzling light, or no ambient light Operation behind plasti enlosure 3

4 FMCW adar Overview Multiple types of radar modulation waveforms used Pulsed radar, CW Doppler radar, UWB, FSK, FMCW, PN-modulated radar FMCW: Frequeny Modulated Continuous Wave FMCW (sometimes alled LFMCW or Linear FMCW) is the most ommonly used sheme in automotive radar today Linear FMCW: X signal has frequeny hanging linearly with time (i.e., hirp) f x (t) f x (t) FMCW Freq vs. ime sawtooth pattern 2 t d f IF (few MHz) hirp t B (in 100 s of MHz or few GHz) Key benefits of FMCW radar Ability to sweep wide F bandwidth (GHz) while keeping IF bandwidth small (MHz) Better range resolution. F sweep bandwidth of 2 GHz an ahieve 7.5m range resolution, while IF bandwidth an still be <15MHz Lower peak power requirement, ompared to pulsed radar 4

5 f (t) FMCW adar System Model (1/3) High-level blok diagram 1. X signal amp waveform gen. he transmitted FMCW waveform (hirp) is x B ( t) os 2f t t 2 B Sweep bandwidth of (t) Chirp duration (s) transmitted hirp (Hz) he instantaneous frequeny of FMCW waveform is 1 d ( t) f ( t) 2 dt f B t f B 5 t

6 FMCW adar System Model (2/3) High-level blok diagram 2. X signal amp waveform gen. eeived signal is a saled and delayed version of transmitted signal x ( ) ( ) os t x t td 2f Path loss attenuation t d ime delay of ( t t B ) refletion (fromobjet) d ( t t d ) 2 f (t) 2 t d ound-trip delay of refletion t d is f (t) 2 t d ange (Distane) of Speed of light the objet f IF (t) 6 t

7 FMCW adar System Model (3/3) High-level blok diagram 3. Beat signal (IF) amp waveform gen. Beat frequeny or IF signal after reeive mixer is as follows y( t) x ( t) x y( t) os 2 ( t) os ( t t ) os ( t) ( t t ) ( t) os ( t t ) ( t) d Can be alulated as 0 2 B t d t d d Filtered out in the reeiver f (t) 2 t d IF signal f (t) f IF (t) Beat frequeny orresponding to the target 7 t

8 FMCW adar How it works (1/2) Stati objets For stati objets, the beat frequeny is simply proportional to the distane (round-trip delay) Beat frequeny is the produt of FMCW frequeny slope (B/ ) and round-trip delay (t d ) For multiple objets, the beat signal is a sum of tones, where eah tone s frequeny is proportional to the distane of the objet he frequenies of these tones gives the distanes to the different objets Detetion of objets and Distane (ange) estimation is done typially by taking FF of reeived IF signal Beat freq = ound-trip delay * Slope f b 2 B 8

9 FMCW adar How it works (2/2) Moving objets For moving objets, veloity (v) is determined using phase hange aross multiple hirps Phase and frequeny of the reeived beat signal for the n th hirp an be alulated as 2v 2v 0 2 f n 2 ( f n B) Phase hange from hirp-to-hirp that depends only on the veloity (not on range) B t d t New terms in beat frequeny Seond dimensional FF is performed aross hirps to determine the phase hange and thus the veloity he two-dimensional FF proess gives a 2D range-veloity image (FF heatmap) ypially, detetion of objets is done on this image After detetion, the range and relative speed of the objets are easily alulated 9

10 Angle Estimation - Beamforming θ Δ=dsin(θ) A θ d Ae jw Ae j2w Uniform linear array Ae j3w Consider reeived signal for multiple X antennas (say, four) as shown in figure Additional distane (Δ) travelled at suessive antennas depends on the angle of arrival θ his additional distane results in a phase hange (w) aross onseutive antennas his phase hange an be estimated (w est ) using an FF (3 rd dimension FF) One w is estimated, the angle of arrival (θ) an be derived easily

11 ange (m) Longitudinal axis (m) Sample FMCW adar proessing flow (1/2) ypial proessing flow used in FMCW (sawtooth) adar signal proessing ADC output Preproessing (DC removal, Interf mitigation) 1 st dim FF (ange FF) 2 nd dim FF (Doppler FF) Log-Mag & Detetion (CFA-CA / CFA-OS) aw point loud raked objets 2D FF heatmap Angle Estimation (3 rd dim FF) Clustering, raking (Higher layer algorithms) elative adial Veloity Lateral axis (m) 11

12 First-dimension FF (ange FF) Sample FMCW adar proessing flow (2/2) ypial (simple) FMCW hirp onfiguration onsists of a sequene of hirps followed by idle time Frame time (~40ms) Ative transmission time of hirps (10~15 ms) Inter-frame time (upto 25~30ms) 1D FF proessing is done inline. 2D-FF, 3D-FF & Detetion for urrent frame. CFA or other proprietary algorithms often used for detetion. A frame onsists of ative transmission time and idle inter-frame time. time Seond-dimension FF (Doppler FF) adar data ube memory x1,2,3,4

13 Advantages of Fast FMCW modulation Slow FMCW (riangular) waveform used in many legay systems Chirp duration in ms, instead of us Slow (riangular) FMCW Slow FMCW has advantage of low DSP MIPS requirement No two-dimensional FF proessing However, it suffers from ambiguity issues No elegant way of getting range-doppler image Fast (sawtooth) FMCW Fast FMCW (Sawtooth) waveform is preferred in newer systems Fast FMCW has ability to provide range-doppler two dimensional image of objets 13

14 adar system performane parameters Key parameters Max range ange resolution ange auray Max veloity Veloity resolution Veloity auray Field of view Angular resolution Angular auray Cyle time max V v max V max

15 adar devie Max range (1/2) x signal level Based on Friis transmission equation SN = (linear sale) P t G t (CS) (4 2 ) (4 2 ) G r 2 4 f (k)(nf) Noise level P t 12dBm X G t 12 dbi 1/(4 2 ) 80m CS X G r 12 dbi 1/(4 2 ) 1m X signal level: = -121 dbm Noise level: -174 dbm/hz + 16 db + 10*log10(100 Hz) = -138dBm hermal noise floor (k) Noise figure (NF) 10ms ative frame ( f ) (integration time) SN = 17dB

16 Max range (2/2) Max range depends on the below fators P t G t (CS) G r 2 f max = 4 (4) 3 (SN) (k)(nf) X Output power X Antenna gain ypial range 10 dbm 13 dbm 9 dbi 23 dbi (US L) CS of target 0.1m 2 50m 2 (-10 dbsm to 17 dbsm) Depends on Azimuth and Elevation field of view Pedestrian vs. ruk X Antenna gain 9 dbi 23 dbi Depends on Azimuth and Elevation field of view Noise figure Ative frame time Detetion SN 11 db 18 db 2 ms 20 ms 10 db 18 db Implementation dependent Azim FOV (deg) Elev FOV (deg) Antenna gain (db) S M L arget Pedestrian Motorbike Car ruk CS 0.1 ~ 1 sq.m 5 sq.m 10 sq.m 50 sq.m humb rule: 3 db loss = 15% loss of range 12 db loss = 50% loss of range X array beamforming gain an be additionally inluded.

17 ange resolution and ange auray ange resolution Ability to separate two losely spaed objets in range ange resolution is a funtion of F bandwidth used ange auray Auray of range measurement of one objet Depends on SN ypially range auray is a small fration of range resolution Sweep BW (MHz) ange resolution (m) B f B f B f b 2 therefore,, 1 But 2 2 b B f 2 f 1 3dB ) H( f f b B f f ) ( 2 SN B 2 3.6

18 Max veloity Max unambiguous veloity in Fast FMCW modulation depends on hirp repetition period Higher veloity needs faster ramps v max 4 = wavelength = otal hirp duration (inl. inter-hirp time) otal Chirp duration (us) Max unamb veloity (+/-kmph) For a given max range and range resolution, higher max veloity needs higher IF bandwidth Max range ange resolution IF bandwidth Max unamb. veloity Advaned tehniques are often used to inrease the max veloity Ambiguity resolution tehniques an be used to resolve aliased veloity into true veloity

19 Veloity resolution and Veloity auray Veloity resolution Ability to separate two objets in veloity Depends on the ative duration of the frame v 2N = wavelength N = number of hirps in the frame = otal hirp duration (inl. inter-hirp time) Ative Frame duration (ms) Veloity resolution (+/-kmph) Veloity auray Auray of veloity measurement of one objet Depends on SN ypially a fration of veloity resolution v 3.6N SN

20 Benefits of 77GHz mmwave Wide F bandwidth (4 GHz) provides good range resolution and range auray 20X better than legay 24GHz narrowband sensors (whih use ~200MHz bandwidth) High F frequeny (small wavelength) provides good veloity resolution and auray 3X better than 24GHz sensors Better resolution performane with 77GHz Improved performane More foused beam with 77GHz Smaller form-fator for the sensor 20

21 Angular resolution Angular resolution Ability to separate objets in angle (for same range and veloity) adar sensors have poorer angular resolution typially (ompared to LIDA for example) However, in many real life situations, objets get resolved in range or veloity, due to good resolution in those dimensions Angular resolution (in radians) for K-length array is given by: = (in radians) = 2 K λ Kdos(θ) Note the dependeny of the resolution on θ. esolution is best at θ=0 esolution is often quoted assuming d=λ/2 and θ=0 Array length Ang. esolution (deg)

22 Use of Multiple X MIMO radar Multiple X along with multiple X (MIMO radar) to inrease angular resolution eg. 2 X, 4 X an give 8 virtual hannels x1 x2 x1 x4 8 virtual hannels 2 /2 /2 Multiple X time-interleaved Frame time (~40ms) Ative transmission time of hirps (x1 and x2 hirps interleaved) x1 x2 x1 x2 x2 Inter-frame time Multiple X BPM-multiplexed Frame time (~40ms) time Ative transmission time of hirps (x1 and x2 BPM-oded, simultaneous transmission) x1+x2 x1-x2 x1+x2 Inter-frame time Multiple X an also be used for X beamsteering (simultaneous transmission with linear phase shifter based steering of beam) time

23 Casaded multi-hip radar X X X X Measurement results with 2 orner refletors at ~4deg separation 8 hannels (2 X, 4 X) 12 hannels (3 X, 4 X) Single hip Single hip Slave devie (LO syn ed to master) Master devie 24 hannels (3 X, 8 X) 40 hannels (5 X, 8 X) 2-hip asade 2-hip asade wo orner refletors 2-hip asade radar 2-hip asade enables better separation of the orner refletors 23

24 I mmwave radar devies I offers a family of 77GHz radar devies for automotive and industrial appliations Highly integrated devies based on FCMOS High auray, Small form-fator, Sensing simplified For more information, visit: I.om/mmWave

25 eferenes M. Skolnik, Introdution to adar Systems, MGraw-Hill, 1981 Donald E. Barrik, FM/CW adar Signals and Digital Proessing, NOAA ehnial eport EL 283-WPL 26, July 1973 A. G. Stove, Linear FMCW radar tehniques, IEE Proeedings F, adar and Signal Proessing, vol. 139, pp , Otober 1992 M. Shneider, Automotive adar Status and rends, in German Mirowave Conferene (GeMiC), pp , Ulm, Germany, April 2005 I whitepapers he fundamentals of millimeter wave Available at Highly integrated 77GHz FMCW adar front-end: Key features for emerging ADAS appliations Available at I s smart sensors enable automated driving Available at AW1642: adar-on-a-hip for short range radar appliations Available at Fluid-level sensing using 77 GHz millimeter wave Available at obust traffi and intersetion monitoring using millimeter wave radar Available at 25