1. ElectroScience Lab DIGITAL NOISE RADAR PROTOTYPE 2008 AMTA MEETING Boston, MA November 17-21, 2008 Eric K. Walton The Ohio State University; Electroscience Laboratory Columbus, Ohio walton.1@osu.edu

2. ElectroScience Lab 2 BASIC CONCEPT BASIC NOISE RADAR CONCEPT

3. ElectroScience Lab 3 DIGITAL NOISE RADAR pseudo- noise #1 low pass filter output antennas pseudo-noise #2 clock delay

4. ElectroScience Lab 4 PREVIOUS STUDIES (dual FIFO system) Teoman Ustun MS Thesis, Design and development of stepped frequency Continuous wave and fifo noise radar sensors For tracking moving ground vehicles OSU EE Dept, 2001.

5. ElectroScience Lab 5 DIFFERENT TYPES OF RADAR WAVEFORMS

6. ElectroScience Lab 6 SUPPRESSION OF INTERFERENCE/JAMMING

7. ElectroScience Lab 7 BUILDING PENETRATION UWB NOISE RADAR

8. ElectroScience Lab 8 Conceptual Scenario for Dual Bistatic Building Penetration Radar BUILDING PENETRATION UWB NOISE RADAR

9. ElectroScience Lab 9 Walton - Wall Penetration Example SEQUENCE OF IMAGES SHOWING THE TRACKING OF A HUMAN AS HE WALKS FROM UPPER RIGHT TO LOWER LEFT. THE HUMAN IS INSIDE A CONCRETE BLOCK BUILDING. THE RADAR WAS APPROXIMATELY 50 FEET AWAY ON THE OUTSIDE.

10. ElectroScience Lab 10 Walton - Wall Penetration Example NOISE RADAR

11. ElectroScience Lab 11 TARGET SPECIFIC DIGITAL NOISE RADAR Computer & digital I/O 64K x 9 FIFO Cypress CY7C4282 (100 MHz) 100 MHz D/A 100 MHz D/A 3.4 GHz LO DATA LOAD (TX & RX) TX RX 100 MHz TRANSMIT WAVEFORM 100 MHZ RECEIVE WAVEFORM MODULATED RF RECEIVED RF LOW PASS FILTER AUDIO BAND ANALOG DATA TO COMPUTER Burr Brown DAC900 ANTENNAS 3.4 GHz +/- 50 MHz CLOCK 100 MHZ BP Filt. BP Filt. 64K x 9 FIFO Cypress CY7C4282 (100 MHz) Target Specific Noise Radar RIGHT LEFT SW

12. ElectroScience Lab 12 UWB Noise Radar Example Human Walking Toward the Radar Carrying C-Reflector 3.4 GHz

13. ElectroScience Lab 13 PATENT

14. ElectroScience Lab 14 RECENT STUDIES

15. ElectroScience Lab 15 DOPPLER FROM NOISE

16. ElectroScience Lab 16 SIGNAL TO NOISE CALCULATIONS Here is the model: Let us assume RCS of object:-40 DBSM (1 sq. cm.) Power transmitted: 0.25 watt Center Frequency: 10 GHz Antenna Power Gain: 5 wrt isotropic: (~7 dBi) Radar Range Equation : Where Pr is the received reflected power and L is a set of loss factors lumped together (we use 0.5 here). Eq. 1

17. ElectroScience Lab 17 SIGNAL TO NOISE CALCULATIONS Next, we have signal power received due to thermal noise power at the receiver. Eq. 2 where: k is Boltsman’s constant 1.38*10^-23, T= ambient temperature in deg. Kelvin (taken as 290 deg.) B = bandwidth in Hertz. (We use 300 MHz as the front end frequency bandwidth). This allows us to compute the pre-processing S/N ratio as Pr/Pn. Eq. 3.

18. ElectroScience Lab 18 NOTE THAT THE SPECTRAL DISPERSION OF THE SPHERE MAKES THE ABSOLUTE VALUE OF THE PEAK LOWER EVEN WHILE THE TOTAL ENERGY REMAINS HIGH. TIME DOMAIN SIGNATURES SIGNAL TO NOISE CALCULATIONS

19. ElectroScience Lab 19 SIGNAL TO NOISE CALCULATIONS Then the signal processing power gain is based on the final audio filter BW. So:

20. ElectroScience Lab 20  EXTERNAL SOURCE OF WIDE BAND NOISE: Assume an external signal with a flat spectrum but that is incoherent with the transmitted radar signal Power computed as before ( ) For large BW(noise) the total noise power is thus: So the final S/N in this case is:

21. ElectroScience Lab 21 SIGNAL TO NOISE CALCULATIONS We wrote a MATLAB program to evaluate this equation. An example result is shown below for a specific set of conditions: >> nradar G (dB) = 6.9897 Ant. Gain Pt = 0.25 Power trans. R (m) = 30 Range in meters RCS - DBSM = -40Target RCS (DBSM) Pr = 6.999e-016 Pow. Rec. (watts) Pno = 1.2006e-012 Noise Pow. Rec. (watts) SoNraw = 0.00058296 S/N power ratio - raw Gsp = 60000Signal processing gain SoNfin = 34.9777 Final signal to noise power ratio Prdbm = -121.5496 Rec. Power (dBm) Pnodbm = -89.206 Rec. N. Power (dBm) SoNfindb = 15.4379 S/N – post processing - dB If we model it so that we get the post-processing S/N as a function of range, As shown in next figure.

22. ElectroScience Lab 22 SIGNAL TO NOISE CALCULATIONS Post processing S/N for -40 DBSM object (based on parameters from previous slide) Note that at +8 dB S/N, we can see this very small target out to 50 meters.

23. ElectroScience Lab 23 AVAILABLE AS AN OSU REPORT: “Signal to Noise Ratio Calculations and Measurements for the OSU Noise Radar” I. P. Theron, E. K. Walton, S. Gunawan and L. Cai Technical Report 732168-1, The Ohio State Univ. ElectroScience Laboratory, Nov. 1996

24. ElectroScience Lab 24 CONCLUSIONS Note that this is not the best that can be done: The raw data can be further processed using FFT processing. The final bandwidth then is the incremental bandwidth of the FFT. (sometimes called the bin bandwidth) In other words we get the signal processing gain of the FFT where a signal can be extracted from the noise. There are examples in some of my data where the signal can not be seen in the raw data but can be seen in the spectral (Doppler) plots. If we process 128 point FFT, for example, we get another factor of 128 signal processing gain.

25. ElectroScience Lab 25 NOISE RADAR -BACKGROUND Publications: “Ultrawide-Band Noise Radar in the VHF/UHF Band,” (co-authors, I.P. Theron, S. Gunawan and L. Cai), IEEE Transactions Antennas Propagation, Volume 47 Number 6, pp. 1080-1084, Jun. 1999. “Compact Range Radar Cross Section Measurements Using a Noise Radar,” (co-authors, I.P. Theron and S. Gunawan), IEEE Trans. Antennas Propagation, Vol. 46, No. 9, pp. 1285-1288, Sep. 1998.

26. ElectroScience Lab 26 NOISE RADAR BACKGROUND Papers: “Development and Applications of a 16 Channel UHF/L-Band Noise Radar,” (co-author, S. Gunawan), Twentieth Annual Meeting and Symposium of the Antenna Measurement Techniques Association, pp. 210-213, Montreal, Canada, Oct. 26-30, 1998. “Signatures of Surrogate Mines using Noise Radar,” (co-author, L. Cai), AeroSense PIERS Meeting, Orlando, Florida, Apr. 13-17, 1998. “Future concepts for Ground Penetrating Noise Radar,” (invited paper) PIERS Workshop on Advances in Radar Methods, Joint Research Centre of the European Commission (Space Applications Institute), Baveno, Italy, Jul. 20-22, 1998. “Noise Radar A-P Mine Detection and Identification,” Demining MURI review, Night Vision Laboratory, Delphi, Maryland, Aug. 10- 13, 1998. “Comparative Analysis of UWB Underground Data Collected Using Step-Frequency, Short pulses and Noise,” (co-author, S. Gunawan), Ultra-Wideband, Short Pulse Electromagnets 3, Proceedings of the Third International Conference on Ultra-Wideband, Short Pulse Electromagnetics, May 27-31, 1996, Albuquerque, New Mexico. (refereed; digest released as book for, Jan. 1997) “Moving Vehicle Range Profiles Measured Using a Noise Radar, “ (co-authors, I.P. Theron, S. Gunawan and L. Cai), 1997 IEEE AP-S Symposium and URSI Meeting, Montreal, Canada, Jul. 13-18, 1997. “Use of Fixed Range Noise Radar for Moving Vehicle Identification,” ARL 1997 Sensors and Electron Devices Symposium, College Park, MD, Jan. 14-15, 1997. “UWB Noise Radar Using a Variable Delay Line,” (co-authors, I. Theron and S. Gunawan), Nineteenth Annual Meeting and Symposium of the Antenna Measurement Techniques Association, Boston, Massachusetts, Nov. 17-21, 1997. “Comparative Analysis of UWB Underground Data Collected using Step-Frequency, Short Pulse and Noise Waveforms,” (co-author, S. Gunawan), AMEREM ‘96 International Conference on “The World of Electromagnetics” Albuquerque, New Mexico, May 27-31, 1996. “ISAR Imaging Using UWB Noise Radar,” (co-authors, V. Fillimon and S. Gunawan), Antenna Measurement Techniques Association Symposium, Seattle, Washington, Sep. 30-Oct. 3, 1996. “Comparison of Impulse and Noise-Based UWB Ground Penetrating Radars,” (co-author, F. Paynter), URSI Radio Science Meeting (Joint with AP-S), Seattle, WA, Jun. 19-24, 1994. “High Resolution Imaging of Radar Targets using Narrow Band Data,” (co-author, A. Moghaddar), Joint URSI Meeting and International IEEE/AP-S Symposium, London, Ontario, Jul. 1991. “Use of Stepped Delay Line Noise Radar for ISAR Imaging in the OSU Compact RCS Measurement Range,” (co-authors S. Gunawan), Joint Tech. Report 732168-2 and 727723-12, The Ohio State University ElectroScience Laboratory, Sep. 1997. I know this is too small to read; see me afterward and I can e-mail a copy of this talk

27. ElectroScience Lab 27 External source of narrow band signal: Using similar logic, we can show that the average power at each frequency due to a narrow band source is based on the average power over the BW of the radar (IE: N s A 2 ). Thus we obtain the same total noise power as for the wide band external noise signal with a flat spectrum. (We simply must compute the total external signal power in the BW of the radar.)

28. ElectroScience Lab 28 The signal power must be averaged in the t-domain and the noise power must be computed relative to the signal peak (IE: Lets assume a threshold target detection algorithm). If we plan on using a threshold level to determine the detection of a target, then we must consider the peak signal response to the peak noise response. (not the average noise power). We can thus compute the peak response of a random noise with a known power based on a useful number of statistical (sigma) widths. We usually find that the result yields a requirement of a factor of as much as 10 in the ratio of signal to noise to reliably “detect” a target. Of course, this depends on the desired ratio of false alarms to missed detections. (As does nearly all radar target detection processes.)

29. ElectroScience Lab 2008 NOISE RADAR PROTOTYPE Eric Walton Videl Smith (undergrad) as of 7/10/2008

30. ElectroScience Lab 30 FIFO Serializer LPF A/D ~66MHz Out Clock In clock enable x 9 bit data in x 18 bits parallel x 20  Serial up up/ down 3.5 GHz 2008 NOISE RADAR PROTOTYPE 66 MHz clock becomes 1,320 MHz clock

31. ElectroScience Lab 31 l The Noise Radar transmits a signal that resembles random noise spread over a very wide bandwidth. l The information sent is pc controlled, and interpreted by most radio receivers as random noise. l The data is sent through the FIFO, then is serialized and output. The sent signal is compared with the original by a mixer and goes through a low-pass filter before going back to the pc for additional processing. 2008 NOISE RADAR PROTOTYPE

32. ElectroScience Lab 32 Components l Analog to digital converter –Measurement Computing USB – 1408fs –Interfaces with pc software for control and loading of FIFO memory –Two 8-bit digital I/O ports l FIFO memory –Texas Instruments SNV-273 –166 MHz operation (we use 66 MHz to be compatible with the serializer) –Able to write 32768 9-bit words to memory –Capable of simultaneously reading/ writing data (x9-bit in/ x18-bit out) l Serializer/ Deserializer –National Semiconductor DS92LV18 –18-bit bus parallel in  1 bit serial out –15-66 MHz clock multiplied by 20

33. ElectroScience Lab 33 Components l Mixer –Mini-Circuits ZP-5H Mixer –20 – 1500 MHz –Used to match the delayed signal to a reference l Low Pass Filter l Up/ down convertor

34. ElectroScience Lab 34 ADC Power Supply FIFO, Ser/Des and other components 2008 NOISE RADAR PROTOTYPE

35. ElectroScience Lab 35 l The power supply delivers 3.3v DC to the FIFO. TO FIFO 2008 NOISE RADAR PROTOTYPE

36. ElectroScience Lab 36 l The pcb contains the FIFO, oscillator, and Ser/ Des. FIFO Oscillator 2008 NOISE RADAR PROTOTYPE

37. ElectroScience Lab 37 l We are using the ADC’s digital output ports for sending the information to the FIFO. To FIFO From PC Testing Full Flag 2008 NOISE RADAR PROTOTYPE

38. ElectroScience Lab 38 Programming l The software used to interface with the ADC and control the FIFO is written in C++ and allows for user controlled data output. l The random waveform generation program first creates an array with ~297k (32768 x 9) random 1’s and 0’s. The program then randomly selects nine values from the array and sends the data to the FIFO. 2008 NOISE RADAR PROTOTYPE

39. ElectroScience Lab 39 Start Reset/ Initialize settings Enable writing or quit? Load Data or quit? Load Data Start output or quit? Output Data End quit

40. ElectroScience Lab 40 Initial Component testing l We used three different output waveforms to test ADC and FIFO operation. –The first program outputs a square wave pattern and completely fills the memory. Completely loading the FIFO allows us to verify the operation of the memory status flags. –The next program only fills about 33% of the FIFO memory and outputs another square wave. This shows the FIFO doesn’t have to be completely full for operation. –The third program was the random wave pattern. The sequence begins with 2KB of zeros, used as a marker, followed by the random waveform. l The FIFO’s is configured for zero-latency retransmit and we are reading output pin Q4. 2008 NOISE RADAR PROTOTYPE

41. ElectroScience Lab 41 l The first square wave program readout using oscilloscope: We loaded 6000 9-bit words (3000 all zero, then 3000 all ones) into memory. Our output clock runs at 66MHz, and outputs data 18- bits (2 words) at a time. This should result in ~22.7 micro- seconds each of ones and zeros. Experimentally, we measured 22 micro- seconds of ones and zeros. 2008 NOISE RADAR PROTOTYPE

42. ElectroScience Lab 42 l The second square wave program used in testing the FIFO output: We loaded 6000 9-bit words (1000 all zero, then 5000 all ones) into memory. Our output clock runs at 66MHz, and outputs data 18- bits (2 words) at a time. This should result in ~37.87 micro- seconds of ones. Experimentally, we measured 37.6 micro- seconds of ones. 2008 NOISE RADAR PROTOTYPE

43. ElectroScience Lab 43 l The random waveform output: The same method was used in the random waveform generation program. 2000 zeros were loaded, followed by random data. Experimentally, we measured 29.2 micro- seconds of zeros. 2008 NOISE RADAR PROTOTYPE

44. ElectroScience Lab 44 l Spectral analysis (random waveform) results indicate a strong output signal is observed from 10KHz to ~3GHz. 2008 NOISE RADAR PROTOTYPE

45. ElectroScience Lab 45 Time Shifted Waveforms l The next stage of experimentation involved testing the random waveforms and adding time delays. The mixer and low-pass filter were also added, allowing us to test the components as an actual radar. l The focus in this section involved testing two outputs, being able to add a software time-delay to one channel, and add a hardware delay to the other. –The hardware delay was first created using various lengths of wire, corresponding to varying signal delays. –The next step of the hardware delaying process is to use two feed horns to send the signal instead of the wire. l Being able to control the delay is important because if we can’t match the signals, then our outputted data just appears to be random noise. Knowing the time-shift allows us to decode the information. 2008 NOISE RADAR PROTOTYPE

46. ElectroScience Lab 46 Time Shifted Waveform l Setup with additional wire l Setup with wire, mixer and LPF Wire Mixer LPF 2008 NOISE RADAR PROTOTYPE

47. ElectroScience Lab 47 Time Shifted Waveform l The Mixer… To LPF From Hardware Delay From Software Delay 2008 NOISE RADAR PROTOTYPE

48. ElectroScience Lab 48 l The LPF… From mixer To Oscilloscope 2008 NOISE RADAR PROTOTYPE

49. ElectroScience Lab 49 Time Shifted Waveform l Here is an example of the random waveform with no hardware delay, and a 2-bit software delay corresponding to ~30ns: ~34ns 2008 NOISE RADAR PROTOTYPE

50. ElectroScience Lab 50 Time Shifted Waveform l With the added wire (~20 ft = 30 ns) the two random waveforms line up: 2008 NOISE RADAR PROTOTYPE

51. ElectroScience Lab 51 Swept Time Shifted Waveform l Next we created a program that does a 1-bit to 7-bit delay ‘sweep.’ –The software first sends a 1-bit delay for 60 micro-seconds. Next, a 2-bit delay for 60 micro-seconds, then a 3-bit, and so on. –The two outputs were input to the mixer and LPF. l This allows for us to see a spike when the time delays match up, corresponding to a stronger signal. l By varying the length of wire, a change in output was observed. 2008 NOISE RADAR PROTOTYPE

52. ElectroScience Lab 52 Time Shifted Waveform l Results with 33 ft. of coax: 2008 NOISE RADAR PROTOTYPE Note the location of the spike.

53. ElectroScience Lab 53 Time Shifted Waveform l Results with 18 ft. of coax: 2008 NOISE RADAR PROTOTYPE

54. ElectroScience Lab 54 Time Shifted Waveform l Results with 3 ft. of coax: 2008 NOISE RADAR PROTOTYPE

55. ElectroScience Lab 55 FIFO Serializer LPFA/D ~66MHz Out Clock In clock enable x 9 bit data in x 18 bits parallel x 20  Serial up down 3.5 GHz 2008 NOISE RADAR PROTOTYPE 66 MHz clock becomes 1,320 MHz clock mixers Antenna coupling targets t Recycle time

56. ElectroScience Lab 56 2008 NOISE RADAR PROTOTYPE MODULE 1 OF DUAL WAVEFORM NOISE RADAR FIFO 66 MHZ CLOCK SERALIZER (X 20) BALUN OUTPUT @ 66 MHZ CLOCK WITH X20 SERIALIZER; OUTPUT BITRATE IS 1.32 GBPS

57. ElectroScience Lab 57 2008 NOISE RADAR PROTOTYPE TESTING OF DUAL WAVEFORM NOISE RADAR MODULE 1 (with clock 1) MODULE 2 (with clock 2) A/D & DIO INTERFACE OSCILLOSCOPE MURPHY’S LAW;  IT DID NOT WORK IN TIME FOR THIS MEETING

58. ElectroScience Lab 58 Time Shifted Waveform l Results: –The ‘sweeping delay’ program showed the response vs. waveform delay –NEXT: u Do this with an antenna and a free space target u Add a power amplifier u Add an up/down converter u Test modifying the waveform to create a target identification radar 2008 NOISE RADAR PROTOTYPE

59. ElectroScience Lab 59 QUESTIONS? / DISCUSSION?