Networked Radar System: Waveforms, Signal Processing and. Retrievals for Volume Targets. Proposal for Dissertation.

Transcription

1 Proposal for Dissertation Networked Radar System: Waeforms, Signal Processing and Retrieals for Volume Targets Nitin Bharadwaj Colorado State Uniersity Department of Electrical and Computer Engineering Committee : Dr. V. Chandrasekar (Adisor) Dr. Anura Jayasumana Dr. Branisla Notaros Dr. Paul Mielke April 15, 2009

2 Networked Radar System Radar Nodes SOCC System Operation & Control Center Internet MCC Meteorological Command & Control Signal Processor User Rules Real-time Data Internet Internet Obtained oer the Internet ia LDM Timeseries Storage Data Archie Internet Data Storage End-users 2

3 RESEARCH QUESTION: The goal of this research is 1. To deelop waeform systems which will enable operation of the radar nodes in the networked radar system 2. To proide networked retrieal algorithms to enhance the radar obserations 3. To proide both qualitatie and quantitatie inferences about the waeforms and retrieal algorithms.

4 OBJECTIVES OF THE RESEARCH: Waeforms and signal processing Pulsing scheme & processing Study the performance of spectral processing in estimating the Doppler spectral moments and polarimetric ariables Ealuate spectral filtering methodology for polarimetric radars Design the waeform for a single node to minimize the impact of range- elocity ambiguity and ground clutter Ealuate the performance of the waeform with the CASA s first generation radar network Waeforms for solid-state transmitters Design and ealuate wideband pulse compression waeform for meteorological radar Design and ealuate the performance of frequency diersity pulse compression waeform 4

5 OUTLINE Spectral processing for polarimetric radars Spectral ground clutter filtering Waeform design for X-band radars Waeforms for solid state radars 5

6 OUTLINE Spectral processing for polarimetric radars Spectral ground clutter filtering Waeform design for X-band radars Waeforms for solid state radars 6

7 Φ Φ Φ Φ Φ Estimates of Spectral Moment and Polarimetric Variables, h and Rˆ Time domain Spectral domain : Normalized DFT matrix : Receied signal in horizontal w : Normalized processing window function ertical polarization channel = diag 1 N k = 2 ( n ) = ( k + n ) ( ) power elocity spectral N : 0 h h k P ˆ h = h, P ˆ = N N : differential differenti λ π { ( ) } 2 s s h m = = power λ ˆ ˆ k = = R 1 elocity : ˆ 4 width al arctan ˆ T s :σ = λ ˆ ln T s reflectiity : phase polar correlation : R 2 π 2 R ρˆ ˆ Z dr :ψˆ = h ( 0 ) = = 10 log h H h 10 ( ) 0 ( 1 ) { } H dp arctan h P ˆ P ˆ h m m spectral : h differential differenti co- co- polar : : { w } Spectral Spectral P ˆ h = s h, P ˆ = s N N N 1 = 0 = N 1 k = 0 width al : k s k σˆ s k phase 2 2 reflectiity : correlation coefficients in H - channel coefficients in V - channel N 1 k k = 0 = N 1 :ψˆ = : ( ˆ ) k = 0 s k s k 2 { s } H dp arctan h s ρˆ h ˆ Z dr ( 0 ) = = 10 s s h log H h s s 10 P ˆ P ˆ h 7

8 Standard Deiations Vs No. Pulses λ = 11 cm, σ = 4 m/s, SNR=20 db, PRT=1.0 ms, ρh(0) = (a) (b) SD(ˆp) (db) SD(ˆ) (m/s) pulse pair spectral Hamming spectral Chebyshe 80 db (c) 3.5 (d) Number of samples (N) Number of samples (N) 8 SD(Ẑdr) (db) SD( ˆψdp) (deg)

9 Standard Deiations Vs Spectral width λ = 11 cm, N = 64, SNR = 20 db, PRT= 1.0 ms, ρh(0) = (a) (b) SD(ˆp) (db) SD(ˆ) (m/s) (c) (d) Doppler Spectral Width (m/s) Doppler Spectral Width (m/s) SD(Ẑdr) (db) SD( ˆψdp) (deg) pulse pair spectral Hamming spectral Chebyshe 80dB

10 OUTLINE Spectral processing for polarimetric radars Spectral ground clutter filtering Waeform design for X-band radars Waeforms for solid state radars 10

11 Ground Clutter Filtering Ground clutter is the radar return from non-meteorological targets that bias radar parameters. Clutter filtering is performed by applying a notch filter centered at zero Doppler elocity Ground clutter filtering techniques IIR notch filter centered at zero Doppler elocity Adanced spectral filter Adanced time domain filters 11

12 Spectral Clutter Filtering Obtain spectral coefficients and power spectral density of receied signal Obtain adaptie noise floor by sorting spectral coefficients by power Design notch filter in spectral domain Estimate clutter model based on Gaussian model fit to zero Doppler region Estimate notch width based on clutter model and noise Notch the clutter signal with a spectral clipper Interpolate the notch filtered region by iteratiely fitting a Gaussian model to the weather signal Replace the clutter region with model and subtract noise power 12

13 Spectral Clutter Filtering Error in estimated spectral moments as a function mean Doppler elocity Error in estimated polarimetric ariables as a function mean Doppler elocity 13

14 Spectral Clutter Filtering: Implementation with CSU-CHILL Data collected with PRT =1 ms; N=64 on Dec 20, 2006 at 23:58:19 UTC Seere clutter from Rocky mountains Reflectiity after spectral clutter filtering Velocities biased due to clutter Velocities after clutter filtering 14

15 Spectral Clutter Filtering: Implementation with CSU-CHILL Data collected with PRT =1 ms; N=64 on Dec 20, 2006 at 23:58:19 UTC Clutter contaminated differential reflectiity Filtered differential reflectiity Biased differential phase due to clutter Filtered differential phase Clutter contaminated co-polar correlation coefficient Filtered co-polar correlation coefficient 15

16 OUTLINE Spectral processing for polarimetric radars Spectral ground clutter filtering Waeform design for X-band radars Waeforms for solid state radars 16

17 S X r a Range-Velocity Ambiguity a r a In a pulsed Doppler weather radar scattering is due to precipitation particles that spatially extend oer a large area r a a = cλ 8 ; Fundamental limitation of pulsed Doppler radar transmitting uniformly spaced pulses a r a If is increased, decreases correspondingly ( Range-elocity ambiguity) The trade off between maximum range and maximum elocity is more stringent for X- band radar 17

18 Dual-PRF Waeform A dual-prf waeform with is recommended for operational use Phase coding of the transmitted pulse is used to suppress oerlaid echoes Staggered PRF unfolding to measure high elocities A waeform look-up table proides the necessary tool to adaptiely select the waeform PRT 2 PRT 1 Horizontal Polarization Dwell Time Vertical Polarization Random Phase Processing with Dual PRF 18

19 Results from IP1 Radar Network Coers an area of 7000 square km The deployment of this four-node network represents a unit-cell of a larger deployment 19

20 Design Considerations : Hardware Requirements (First Generation) The first generation CASA radar systems are magnetron based systems with limited agility on duty cycle and supported waeforms. The transmitter can delier a maximum peak power of 25 kw at a duty cycle of 0.1 %. The transmitter can be tuned below its maximum peak power allowing one to increase the duty cycle, which is used to accommodate higher PRF. Random phase coding is the only scheme that can be implemented because a magnetron based system has a random startup phase. 20

21 Design Considerations : Hardware Requirements (First Generation) Sensitiity better than 10 dbz at 30 km and with an accuracy on the order of a 1dB The Nyquist elocity should be no less than 25 m/s with an accuracy of 1m/s Increase in PRF beyond 1.5 khz has to compensated by reducing the peak power. Increasing PRF will lead to range oerlay. Minimize impact of range oerlay on the first 30 km 21

22 Design Considerations : Operational Requirements (First Generation) Minimum clutter suppression capability of 30 db The bias and standard deiation of elocity after clutter filtering will be no greater than 2 m/s for an SNR of 20 db Maximum allowable bias in reflectiity after clutter filtering 1 db 22

23 Ground Clutter Filtering Impact of Phase Noise 23

24 Ground Clutter Filtering 24

25 Oerlaid Echo Suppression 25

26 Velocity measurements on a radial-by-radial basis with dual- PRF Velocity Unfolding Unfolded elocity with dual-prf 26

27 OUTLINE Spectral processing for polarimetric radars Spectral ground clutter filtering Waeform design for X-band radars Waeforms for solid state radars 27

28 Sensitiity of Weather Radar The receied signal at the antenna reference port is used to measure reflectiity The minimum detectable reflectiity is a function of pulse width for a fixed transmitter and antenna The sensitiity can be mapped to the transmit pulse length Long pulses are transmitted to achiee sensitiity 28

29 Pulse compression Waeform Pulse Compression a mature technology for hard-target radar Improed range resolution Reduced peak-power requirement Weather radar use has been limited Range sidelobe impact for olume targets Not operationally proen Offers seeral benefits Increased sensitiity Improed range resolution Improed accuracy of estimates through range aeraging Enables low-power solid-state transmitter technology Dynamic range beyond RF hardware limitations 29

30 Nonlinear FM Pulse Compression The complex enelope of the transmitted pulse is The FM characteristic, f(t), is decomposed into linear and nonlinear components, as Characteristics of Nonlinear FM, implemented using a smooth FM law = t T d f j t g t u 2 2 / ) ( )exp ( ) ( τ τ π 30 into linear and nonlinear components, as B is the total bandwidth of the chirp, 0<k T <1 and 0<k B <1 are parameters that control the nonlinearity > = ) ( ) ( ) ( ) ( T T T B k T t t k T t k k T B t f ϕ

31 Sidelobe Suppression Low range sidelobe leels are essential for weather radar applications Strong reflectiity gradients (40 dbz/km) can result in contamination of adjacent range gates Integrated Sidelobe Leel (ISL) are important for olume targets ISL is controlled by Adjusting the nonlinearity parameters, k T and k B Applying amplitude weighting functions on transmit Using suitable compression filters Compression filters Window functions Minimum ISL filter 31

32 Sidelobe Suppression: Min ISL Filter Let G be the transmit conolution matrix based on the discrete transmit complex enelope g and h be the FIR filter Output of the filter is when transmit pulse is used as input is Let G m be the modified transmit conolution matrix obtained by deleting the columns that corresponds to mainlobe The ISL is gien by The ISL is minimized with a constraint on the peak of the output as 32

33 Sidelobe Suppression: Min ISL Filter The filter is obtained by soling the Lagrangian The min ISL filter is gien by The filter is obtained by normalizing it to hae zero DC gain 33

34 Compression Filter Performance Sidelobe leel as a function of Doppler elocity for a Chebyshe and minimum ISL compression filter Frequency response of the transmit pulse and compression filter along with the ambiguity function 34

35 Compression Filter Performance Impact of Phase Noise Sidelobe leel as a function of Doppler elocity and system phase noise obtained using a minimum ISL compression filter 35

36 Impact of Sidelobe Leel on Reflectiity Error in reflectiity with a 50 db rise with 20 db/km gradient using a 40 microsecond pulse compression waeform with phase noise = 0.0 deg Error in reflectiity with a 50 db rise with 20 db/km gradient using a 40 microsecond pulse compression waeform with phase noise = 0.5 deg. Phase noise of 0.5 deg results in ~ -45 db ISL 36

37 Impact of Sidelobe Leel on Reflectiity Error in reflectiity with a 25 db rise with 20 db/km gradient using a 40 microsecond pulse compression waeform with phase noise = 0.5 deg. Phase noise of 0.5 deg results in ~ -45 db ISL Error in reflectiity with a 50 db rise with 20 db/km gradient using a 40 microsecond pulse compression waeform with phase noise = 0.5 deg. Phase noise of 0.5 deg results in ~ -45 db ISL 37

38 Frequency Diersity Pulse Compression 38

39 Frequency Diersity Pulse Compression Simulations (phase noise 0.25 deg) based on CSU-CHILL data 3.45 km 9.45 km 39

40 Summary(1/3) The primary motiation was the deelopment of waeforms and retrieal algorithms for X-band system The performance of spectral processing in terms of standard deiations were presented The standard deiation of the estimated spectral moments and polarimetric ariables increase due to the application of processing widow It is recommended to use time-domain methods where applicable Spectral clutter filtering for polarimetric radars was presented Clutter suppression is improed by using spectral processing The errors in estimated polarimetric ariables is within acceptable leel in the presence of spectral clutter filtering Implementation with CSU-CHILL radar shows good performance in suppressing high clutter leels 40

41 Summary(2/3) A dual-prf waeform is suggested for use on a X-band radar system The dual-prf waeform is used to measure high elocities Spectral clutter filtering is performed on each PRF block and satisfactory results are obtained if more than 40 pulses are used Phase coding the transmit pulses along with spectral processing is suggested to suppress oerlaid echoes A waeform table is proided to change the waeform based on scan speed The dual-prf waeform was implemented with CASA s IP1 radars The statistics of clutter suppression were presented from data collected during 2007 and A clutter suppression of 30 db was achieed with the IP1 radars and the limiting factor is the phase noise of the system Statistics of oerlaid echoes suppressed with random phase coding were presented. About 22 db SNR of oerlaid echoes were suppressed The dual-prf waeform proided unambiguous elocities in excess of 25 m/s 41

42 Summary(3/3) A frequency diersity pulse compression waeform was presented to mitigate the low sensitiity and blind range problems associated with solid-state radars A non-linear pulse compression waeform was presented Minimum ISL Compression filter proided low sidelobe leel The phase noise performance of pulse compression waeform was presented for a waeform Frequency diersity is recommended to mitigate blind range A simulation based on CSU-CHILL obseration indicates good performance of the frequency diersity pulse compression waeform 42