A software defined radar platform for waveformadaptive MIMO radar research

Siddharth Baskar and Emre ErtinDepartment of Electrical and Computer Engineering

The Ohio State UniversityColumbus, Ohio 43210

Abstract—MIMO Radar systems that can adaptively selectindependent waveforms on transmit antennas can provide spatialand temporal diversity gains for improved detection, classificationand tracking capability in challenging propagation environments.In this paper we discuss design, implementation and validationof a Software Defined Radar (SDR) test bed to derive MIMOwaveform adaptive radar research. The OSU micro SDR platformfeatures a custom integrated X-Band RF frontend with twoindependent transmit and a single receive channel multiplexedto four receiver antennas with bandwidth of 250 MHz and adigital back-end with an FPGA and an embedded PC for real-time receive processing and waveform scheduling.

I. INTRODUCTION

Microwave radar systems are crucial components of anystandoff sensor system due to their all-weather capabilitiesand proven performance for tracking, imaging, and situationalawareness. MIMO radar systems which can transmit indepen-dent waveforms on multiple antennas have been suggested forimproving detection, parameter estimation and clutter suppres-sion capabilities. While many traditional multi-antenna radarconcepts such as phased-array, receive beamforming, STAP,polarimetry, and interferometry can be seen as special cases ofMIMO radar, the distinct advantage of a multi-antenna radarsystem with independent transmit waveforms is the increasednumber of degrees of freedom leading to improved resolution,detection and parameter estimation. MIMO system benefits canbe realized in the form of reduced pulse repetition frequency(PRF), larger spot sizes, and/or lower transmit energy.

In the literature, MIMO radars are distinguished basedon the geometry of the receive and transmit centers. Thereare two main categories. MIMO radars with widely separatedtransmit and receive arrays provide statistically independentmeasurements of the illuminated scene and are categorized asa statistical MIMO radars. MIMO radar systems with widelyseparated antennas employs spatially diverse transmitters andreceivers to overcome target fading effects [1], [2] or to esti-mate a target’s location with high resolution [3],[4]. If antennasare relatively close to each other, so that for each scatterer inthe illuminated scene the angle of arrival is approximately thesame for all phase centers, then the system is referred to asa coherent MIMO radar. The main advantage of the coherentMIMO radar is its ability to synthesize a large virtual arraywith fewer antenna elements for improved spatial processing.The OSU micro SDR platforms we developed will enableresearch in both modalities as well as novel hybrid modescombining elements of the two in a multi-static setting.

We focus on development of a low power, short range

versatile radar system that combines a high speed FPGA dig-ital back-end with sideband digital/analog and analog/digitalconverters with a custom built RF Frontend. The operationalprinciple of a software defined radar system is to sample thetransmit/receive waveforms using high speed digital/analogand analog/digital converters and to implement key process-ing stages using programmable digital hardware [7]. Thisallows the SDR to, for example, change modes from detectionto tracking, or adapt its waveform based on environmentalconditions and or information derived from previous radarinterrogations. Increasing number of transmit and receivechannels with the use of antenna switching matrix for MIMOapplications is discussed in [8]. The use of MIMO radar forstudying wideband radar array signal processing in short rangeindoor application has been demonstrated in [9]. Applicationof waveform adaptation for matching transmitted waveformto the target’s impulse response improves target detection andaids in target identification [5],[6].

Our tesbed consists of 14 Micro SDR platforms thatcan support research in active sensing in various settings,including:

1) Micro SDRs can be deployed to perform indepen-dently for non-coherent fusion of backscatter re-turns (also known as statistical MIMO radar) todecrease fluctuations in target returns to selectivefading through spatial diversity. The Micro SDRs canmodify their transmit waveforms and pulse repeti-tion frequencies cooperatively to adapt changes inthe background and target returns as well as scenecomplexity. In addition micro SDRs can be mountedon robotic platforms to optimize the collection geom-etry and derive fusion research with other modalitiessuch as Electro-Optical(EO), Infrared(IR) camerasand acoustic sensors.

2) Each Micro-Radar can serve individually as a 2 TX/4RX co-located MIMO array. Multiple micro SDRplatforms can be combined with same frequencyreference to enable larger co-located transmit arrays.

3) Alternatively the multiple components of the testbedcan be combined to provide multiple illumination andreceive arrays in multi static setting with coherentreferences through the GPS conditioned referencesavailable on each unit.

A system overview is presented in Section II, whileSection III describes RF Frontend Design.We completed testmeasurements of the RF frontend to verify the calculated specs.These are presented in Section IV.

978-1-4799-8232-5/151$31.00@2015IEEE 1590

II. SYSTEM OVERVIEW

The operational principle of a software defined radarsystem is to sample the transmit/receive waveforms usinghigh speed digital/analog and analog/digital converters and toimplement key processing stages using programmable digitalhardware. The block diagram for the proposed software definedradar system is given in Figure 1. A high speed digital wave-form generator is used to construct independent waveformsfor a set of transmit antennas, and produces a synchronizedmulti-channel baseband transmit signal which is mixed andamplified for transmission. In the receive signal chain, thereceived energy is filtered, amplified, and downconverted by anRF module, sampled in the baseband bandwidth synchronouslyacross the multiple channels, and passed to an FPGA-basedreal-time signal processor for multi-channel coherent process-ing. The adaptive operation of the system is controlled by theinformation driven active sensing layer which allocates systemresources to achieve ISR objectives by supplying the user withATR primitives (target detections, target track and ID).

In an idealized model of software defined radar analog-to-digital and digital-to-analog conversion will be accomplishedat the RF frequency band without analog conversion stages.This way, down and up-conversion will be performed inthe digital domain limiting the analog components to high-dynamic range low noise amplifier (LNA) and power transmitamplifiers. Unfortunately, existing ADC performance is farfrom operating with high dynamic range in the radar bands ofinterest above 1 GHz. In addition, real time signal processingtasks of frequency conversion, digital filtering will requiremultiple FPGA/DSPs operating on interleaved data to copewith the large sampling rate of receive and transmit signals.Therefore, RF fronted in a software defined radar systemhave to include up and down conversion stages. There aremany options of implementing conversion stages. The mostcommon architecture is heterodyne receiver structure that usesan conversion stages at multiple intermediate frequencies (IF)to implement image suppression and channel selection. Hetero-dyne receivers can achieve high sensitivity and channel selec-tivity, DC offset is eliminated in the bandpass filters followingeach IF conversion. However, large number of componentsincluding image rejection filters are required for multipleconversion stages increasing the complexity of the design. Atthe other end of the spectrum of the receiver structures areZero-IF receivers that employ single quadrature demodulatorto bring the RF passband signal to complex baseband. Zero-IF receivers are not subject to the image problem commonto receivers with intermediate frequency conversion stages.However, significant DC offsets at the output of quadrature

Fig. 1. Waveform adaptive MIMO SDR

mixers as a result of LO leakage signal mixing with itself.

In our design we employ a hybrid structure relying onthe much higher sampling frequency of the DAC(1 GSps) ascompared to the sampling frequency of the ADC(250 MSps).On transmit, we directly generate a pass band signal arounda lowIF frequency of 250MHz with a bandwidth of 250 MHzusing the high performance DACs in our system. Each DACis used to generate an independent transmit waveform as real-valued pass band signal. For each transmit channel, low-IFpass-band signal is up-converted to X-band using a singlechannel mixer. We use a RF pass-band filter to reject the imageand LO leakage. On receive we use a RF band-pass filter tolimit the wideband noise and a zero-IF receiver with quadraturedownconversion with the same LO that generated the transmitsignal. As a result the received signal is passband signalat the output of down-conversion mixer’s I and Q outputs.Next, we employ standard bandpass sampling in the secondNyquist zone, to alias the digitally generated low-IF signalto the baseband. We note that in our system DAC and ADCuse a single clock -source eliminating the problem of clock-jitter limiting the performance of band-pass sampling systemsin practice. In addition low-IF bandpass sampling systemenables us to use a DC-blocker at the output of quadraturemixer output, eliminating the DC-offset problem common tozero-IF receivers. Figure 2 shows the up and downconversionemployed in our design

The custom RF Frontend built at Ohio State features twoindependent Transmit (TX) and a single Receive (RX) channelmultiplexed to four receive antennas. The instantaneous band-width of the system is 250 MHz. The RF Frontend operatesat X-band and includes an onboard Phase Locked Loop (PLL)and Voltage Controlled Oscillator (VCO) to generate Local

Fig. 2. Frequency content of the transmit baseband, transmit passband andreceive baseband signals. Bandpass sampling aliasing is depicted as dashedlines.

978-1-4799-8232-5/151$31.00@2015IEEE 1591

Fig. 3. Two tone model of IIP3

Oscillator (LO) signal from a local GPS conditioned oven-controlled 10 MHz reference. On transmit we use a digitallygenerated low Intermediate Frequency (IF) of 250 MHz withan instantaneous bandwidth of 250 MHz on two independentchannels. On receive this signal is down-converted with a zero-IF baseband I/Q mixer and sampled directly at the secondNyquist zone without a second conversion stage. This hybriddesign eliminates the DC bias problem that often limits the dy-namic range of zero-IF transceiver designs. The 4x1 switchingmatrix on the receiver increases the number of RX antennasfor MIMO applications for observing slowly changing scenes.

A. Critical System Specs

1 dB Compression Point: For any non linear device likean amplifier and mixer, as the input power continues toincrease, at some point the gain begins to decrease. As thispoint the device is said to be at saturation and the outputpower no longer increases linearly with increase in input. 1dB gain compression point or (P1dB) is defined as the inputpower level at which the corresponding gain drops by 1 dB.Mathematically it is expressed as

P1dB =

s

0.145

����↵1

↵3

���� (1)

Ideally RF systems are designed to work 2-3 dB lower thanthe 1 dB compression point.

Third Order Intermodulation: Non Linear devices, espe-cially amplifiers produce harmonics of the amplified inputsignal. Usually these harmonics are of higher frequency rangeand can be easily filtered. However in-addition to harmonics,they also produce mixing effects of two signals. Hence ifthe two input tones are close together in frequency, theirmixing products lies within the band of interest. These sum anddifference frequencies are together known as intermodulationproducts and they are given importance while designing RFsystems.

IIP3 is defined as the extrapolation point where thepower of 2!1-!2 and 2!2-!1 sideband is equal to power offundamental. The cascaded IIP33 of the system is given bythe expression

1

IIP3=

1

IIP31+

G1

IIP32+

G1 ⇤G2

IIP33... (2)

From equation (2) it can be seen that the IIP3 or the systemis most influenced by gain and linearity of the last stage. IM3or Third order intermodulation distortion product is measured

TABLE I. CRITICAL SPECIFICATIONS OF TRANSMITTER ANDRECEIVER

Specification Transmitter Receiver

Gain 23.40 dB 24.20 dB

OIP3 33.99 dBm 29.20 dBm

P1dB(in) -0.3 dBm -7.821 dBm

Noise Figure NA 2.76 dB

Fig. 4. Block Diagram of receiver

Fig. 5. Block Diagram of Transmitter

as the ratio of power of fundamental to the power of third orderintermod product. Higher the IM3 better is the performance ofthe frontend. Figure 3 shows the illustration of IIP3 and IM3of an amplifier for two equal amplitude tones.

Noise Figure (NF): Noise figure is defined as the ratio oftotal noise power output to the portion of noise power outputdue to noise at the input when the input source temperature is290K. The noise figure spec of the receiver directly influencesthe reliable detection range of the radar.

Table I shows some of the critical specifications of theTransmitter and Receiver.

III. RF FRONTEND DESIGN

Figure 4 shows the block diagram of Receiver. The inputsof four RX antennas are fed to a Low Noise Amplifier(LNA) whose output is connected to receive frontend througha four port switch. Since it is likely that the antennas willbe connected to the rest of the RF system by long cables,including the LNA close to the antenna reduces the impactof cable loss on receiver’s overall noise figure. The signalfrom the switch is filtered through a resonant coupled BandPass Filter (BPF) to filter any blocker and image signal. Thefiltered signal is down converted by In-phase and Quadrature(IQ) mixer and the down converted signal is amplified by anIF amplifier. The amplified signal is filtered by a low passfilter and digitized by high speed ADC. In order to minimizethe phase and amplitude imbalance in IQ signal, the I andQ channels are routed symmetrically and dual IF amplifier isused for amplification.

Figure 5 shows the block diagram of Transmitter. The IFsignal is generated by the DAC at 250 MHz and 0 dBmpower. The IF single is filtered by a Low Pass Filter (LPF) toremove digital copies and amplified by a base band amplifier.The amplified signal is up converted to X-Band using adouble-balanced mixer and further amplified by a high Power

978-1-4799-8232-5/151$31.00@2015IEEE 1592

Fig. 6. Block Diagram of PLL and LO Driver

Fig. 7. Momentum 3D view of Resonant Coupled Band Pass filter

Fig. 8. Simulated frequency response of band pass filter

Amplifier (PA). The amplified signal is filtered by a resonantcoupled band pass filter to remove mixer and power amplifierinter-modulation components.

Figure 6 shows the PLL and LO Driver section. The onboard PLL required a reference signal of 50 MHz signal forlocking, however the GPS conditioned reference can generateonly a 10 MHz signal. Hence a 5X frequency multiplier isused and a combination of Low Pass and High Pass filters wereused to remove harmonics from the multiplier. The referencesignal is further amplifier to compensate for the loss duringfrequency multiplication. This resulting 50 MHz signal is usedas reference by the PLL. Output of PLL is divided equally bya 3 dB splitter and further amplified by driver amplifiers andthis amplified signal is used by the mixers for up and downconversion. The PLL used on the SDR works with a standardfour wire Serial Peripheral Interface (SPI). MSP430F1611micro controller is used for configuring the registers on thePLL upon power up.

Figure 7 shows the Momentum 3D view of resonantcoupled band pass filter used on both transmit and receivesection of Ohio State’s RF Frontend. Increasing the number of

Fig. 9. Top side of Ohio State’s RF Frontend

Fig. 10. Spectrum of Transmitter with 250 MHz IF signal at -3 dbm power

sections will provide a sharper cutoff for the filter but at thesame time increase the insertion loss. Hence to optimize bothroll off and insertion loss a four section topology was chosen.Each section of the filter is designed with Microstrip coupledline (MCLIN) and Linecalc software was used to calculate oddand even mode impedance of MCLIN. Microstip to CoplanarWaveguide (CPW) transition was designed and added to bothinput and output ports of the filter so that the filter can beconnected to the components on the board. Figure 8 showsthe frequency response of the filter. From the figure it can beseen that the insertion loss at center frequency is roughly 3 dBand attenuation at 11 GHz (FLO + FIF ) is about 20 dB.

IV. MEASUREMENTS

Figure 9 shows top side of RF Frontend Printed CircuitBoard (PCB) designed at Ohio State. All the RF and IF compo-nents are placed on the top side of the board. To minimize thesignal cross coupling and to reduce the effect of power supplyinduced signal distortion, no two RF/IF component share thesame bias network. The RF Frontend is fabricated on a 4layer RO4350B substrate and all the RF traces were designedand simulated on Advanced Design System (ADS) software.The board was designed using Allegro PCB editor software.All measurements were performed in lab environment withIF signal generated with Agilent Analog Signal generator andmeasured with Agilent PXA Spectrum Analyzer. The receiveris characterized with transmit signal looped back to receiver

978-1-4799-8232-5/151$31.00@2015IEEE 1593

Fig. 11. Transmitter Two Tone test result (tones centered at 250 MHz with25 MHz frequency separation)

with 40 dB attenuation. Figure 10 shows the transmit spectrumwith 250 MHz signal at IF. The desired signal appears at(FLO � FIF ). Along with the desired signal, LO signal isalso present due to the finite LO-RF isolation of the mixer.However on the receiver side, during down conversion thisLO leakage signal will be down converted to Direct Current(DC) and will be blocked by capacitors. In addition to thisFLO + FIF and FLO � 2FIF (third order inter-modulation)can also be seen but their power levels are significantly lowerthan the fundamental signal.

Figure 11 shows two tone test result of transmitter. Thetwo tones are -3 dBm in power and are centered at 250MHz and with 25 MHz separation between them. As seen inany upcoversion transmitter, due to the non linearity of mixerand power amplifiers, the two tones interact with each otherand produce second and third order inter-modulation products.The important component for signal analysis is third orderintermodulation products. From the figure the third order intermodulation products are at 2F2�F1 and 2F1�F2. Measured

IM3 value is 16.44 dB which is consistent with the theoreticalvalue calculated from specification of the components.

V. ACKNOWLEDGMENT

This research was partially supported by Army ResearchOffice grants W911NF-11-1-0391 and W911NF-13-1-0280.

REFERENCES

[1] E.Fishler, A. Haimovich, R. Blum, L. Cimini, D. Chizhik, and R. Valen-zuela Spatial diversity in radars-models and detection performanceIEEE Transactions on Acoustics, Speech, and Signal Processing, vol.54, no. 3, pp. 823–838, March 2006.

[2] E. Fishler, A. Haimovich, R. Blum, L. Cimini, D. Chizhik, andR. Valenzuela Performance of MIMO radar systems: advantages ofangular diversity Conference Record of the Thirty-Eighth AsilomarConference on Signals, Systems and Computers vol. 1, pp. 305–309,Nov. 2004

[3] N. H. Lehmann, A. M. Haimovich, R. S. Blum, and L. Cimin Highresolution capabilities of MIMO radar Fortieth Asilomar Conferenceon Signals, Systems and Computers, pp. 25–30, Oct. Nov. 2006

[4] E. Fishler, A. Haimovich, R. Blum, D. Chizhik, L. Cimini, and R.Valen-zuela MIMO radar: an idea whose time has come Proceedingsof the IEEE Radar Conference. pp. 71–78, Apr. 2004.

[5] M. R. Bell Information theory and radar waveform design IEEETransactions on Information Theory, vol. 39, no. 5, pp. 1578–1597,Sep 1993.

[6] J. R. Guerci and S. U. Pillai Theory and application of optimumtransmit-receive radar IEEE International Radar Conference, pp. 705–710, 2000.

[7] C.W. Rossler, E.Ertin, and R.L. Moses A software defined radarsystem for joint communication and sensing IEEE Radar Conference(RADAR), pp.1050–1055, 23-27 May 2011.

[8] Mark T. Frankford, Ninoslav Majurec, and Joel T. Johnson Software-Defined Radar for MIMO and Adaptive Waveform Application IEEERadar Conference, pp.724–728, 10-14 May 2010 .

[9] Jason Yu, Matthew Reynolds, Jeffrey Krolik An Indoor S-band RadarReceive Array Testbed IEEE Radar Conference (RADAR), pp.712–717, 10-14 May 2010.

978-1-4799-8232-5/151$31.00@2015IEEE 1594