Radar uses reflected radio waves to determine the range, angle, or velocity of objects. These detection systems that were once the exclusive domain of the aerospace and defense industry are now gaining popularity in the consumer industry, most notably automotive radar. Commercial adoption is possible, in part, because of high-volume semiconductor processes such as silicon germanium (SiGe) and CMOS technologies, which are enabling cost-effective systems for mass commercial applications.
This blog presents a 60GHz frequency-modulated continuous-wave (FMCW), frequency-division multiplexing (FDM), multiple-in-multiple-out (MIMO) radar system for commercial radar applications. The unique architecture allows the total number of TX and RX channels to be scaled by the number of chips in the system while still maintaining phase coherence between channels. The approach provides high frame-rate measurement, excellent phase stability, and a large field of view (FoV). The radar architecture and integrated circuits are designed in such a way that the system can be scaled to much larger radar systems.
The intended use for the radar system was short-range, high-resolution detection of nearby moving objects when the radar itself might be moving, in order to capture the flow of people, drones, and other autonomous systems. In addition, the system can support simultaneous localization and mapping (SLAM), object detection (automobiles), and remote multi-target vital sign measurements for medical applications. FDM was determined to be the best choice to address the high-resolution requirements, whereas code division (CDM) was too complicated and the need for accurate tracking of fast-moving objects made time division (TDM) impractical. In addition, FDM allows accurate phase measurements supporting applications such as remote monitoring of heartbeat and breathing rates from the detection of small movements of the chest.
Requirements for the radar system included fast imaging > 200Hz, range resolution < 5cm, multitarget acquisition, moving target capability, and high sensitivity to micromotion, all in a small, lightweight, low-cost footprint.
The specifications for the system were:
The system offers maximum detection range of 20 to 25 meters for stationary human-size objects. By applying background subtraction, this range increases up to 60 meters for moving targets. The system also supports simultaneous detection of multiple moving objects without physical scanning of the antenna. While the radar system offers many potential use models, the multi-person vital sign extraction capability is particularly of interest for future applications. Another key feature of the radar system is its speed. For data analysis, the radar technology can operate at 200 frames per second (FPS) when not supporting visual graphics and 50-100 FPS with visualization.
A traditional pulsed radar detects the range to a target by emitting a short pulse and observing the time of flight of the returned target echo. This requires the radar to have high instantaneous transmit power and often results in a radar with large, expensive physical apparatus.
FMCW radars achieve similar results using much smaller instantaneous transmit power and physical size by emitting a continuous microwave signal that is frequency modulated with a low frequency waveform, such as a sawtooth function of period T, whose duration is much greater than the return time of the echo illustrated in Figure 1.
Unlike pulsed radar systems, FMCW systems transmit and receive simultaneously, eliminating the blind range that occurs when the receiver in a pulsed radar is turned off during pulse transmission. This allows FMCW systems to detect reflected signals from objects very close to the radar, allowing it to measure distances down to a few centimeters. The system achieves excellent range resolution, which is proportional to the reciprocal of the bandwidth, i.e. delta x = c / (2 * delta f) and high signal-to-noise ratio (SNR) with narrow intermediate frequency (IF) bandwidth.
Figure 1. Sawtooth waveforms used in FMCW radar systems
A very simple diagram of the system implemented in Cadence AWR Visual System Simulator (VSS) communications and radar systems design software is illustrated in Figure 2. The signal source is divided between the transmit and receive sides and details of the transmitter power amplifier and receiver LNA chains (not shown) can be further developed. To operate properly, the transmitter and receiver signal paths must be well-isolated. This requirement impacts certain design aspects and limits the acceptable transmit power level. Otherwise, power from the transmit side will leak into the receiver circuit, potentially saturating the LNA and/or down-conversion mixer.
In this simple diagram, the signal is radiated between the transmit and receive antennas through an AWR VSS software radar target model that includes properties such as the radar cross section (RCS), distance, velocity, and ambient conditions. The mixer will down-convert the signal that was reflected from the target, using the swept frequency from the voltage-controlled oscillator (VCO) as the local oscillator. Taking the difference of these two signals creates a beat signal that is directly proportional to the distance to the target. This intermediate frequency (IF) is fed to an analog-to-digital converter for signal processing. This signal processing extracts the target distance using a fast Fourier transform. By using multiple antennas, the Fourier transform can also be used to support digital beam-forming in order to produce a 2D image of the detected object.
Figure 2. Basic construction of FMCW radar systems in AWR VSS software (amplifier stages [PA, LAN] and individual MIMO channels not shown)
To address the technical requirements (fast imaging, high resolution of multiple targets), the developers chose an FDM MIMO architecture. With MIMO, the number of elements can be reduced significantly. For MIMO radar with Nt TX elements and Nr RX elements, there are Nt × Nr distinct propagation channels from the TX array to the RX array. Therefore, 64 virtual channels can be synthesized with only 8 receive and 8 transmit channels. This greatly reduces the system complexity, size, and cost.
FDM transmits non-overlapping frequencies simultaneously from each transmitter so that different transmitter signals can be separated at the receiver. For this design, the frequency sweep (chirp) was generated outside the transmit/receive channels using a swept 10GHz, phase-locked loop signal generator feeding a 6x frequency multiplier. Direct digital synthesizers (DDS) were used to generate low-frequency IQ modulation signals with frequency offsetting of 1MHz for each individual transmitter channel and external analog-to-digital converters (ADCs) digitized the IF signals from the down-converted receive signals.
Since the FDM MIMO antennas transmit simultaneously, all the RX channels will receive all the TX channels, separated by the constant frequency offset. The de-modulator uses the original chirp frequency as an LO to down-convert the frequency offset signals containing the frequency shift resulting from the delay of the signals reflected off the target(s), and the TX channels are separated at the digital back-end. This approach can handle moving targets due to measuring all the MIMO channels simultaneously, but requires modulators at each transmitter for shifting the transmitter frequency and faster ADCs due to the wider IF signal bandwidth, illustrated in Figure 3.
Figure 3. Conceptual system block diagram with two RX and two TX channels
AWR VSS design software was used to study the main aspects of the MIMO radar at the system level. It provides a block-level representation of the signal sources, low-noise amplifiers (LNAs), mixers, power amplifiers (PAs), frequency multipliers, antennas, and radar targets as illustrated in Figure 4. The software enabled the designers to tune and optimize all the key parameters, and to incorporate real-world operation of the radar system as more circuit-level detail was added.
Figure 4. AWR VSS software system diagram of the FMCW MIMO radar
AWR VSS software simulated the IF output at two of the receiver downconverter channels. From the equation shown in Figure 1, the beat frequency (Fb) is determined to equal 300kHz for a frequency sweep (chirp) of 3GHz bandwidth (BW) from 58.5 to 61.5GHz, pulse duration (t) of 1ms, and target distance (r) of 15 m. Therefore, the de-modulated signal will be the sum of Fb plus the offset frequency per each channel. Simulation of the received down-converted signal in channel one TX 1 (green) is 1MHz + 0.3MHz = 1.3MHz and the received down-converted signal TX 2 (red) is 2MHz + 0.3MHz = 2.3MHz as shown in Figure 5.
Figure 5. Simulation results for the radar
The core of the radar system is the TX and RX RF integrated circuit (RFIC), both of which support four channels occupying a very small area as shown in Figure 6. Additional chips can be added to the system to increase the number of channels. It is advantageous for one RFIC to support multiple channels to reduce the assembly effort and to allow scaling for a system with a very large number of channels. Separate TX and RX chips allow independent TX/RX scaling, lower the TX/RX leakage, and support closer placement to the feed structure to reduce printed circuit board (PCB) losses.
Figure 6. TX chip with four channels (left) and RX chip with four channels (right)
A single external VCO and phase-locked loop (PLL) provide the local oscillator (LO) signal that is distributed to all the RFICs, resulting in excellent phase-noise correlation. A low frequency (10GHz) external signal is used for easier routing on PCB, since in a system with many channels, routing a 60GHz LO signal would be very difficult. This 9.75GHz to 10.25GHz chirp will be multiplied to the desired operating frequency on the die.
Cadence AWR Microwave Office circuit design software was used in combination with the AWR AXIEM 3D planar electromagnetic (EM) simulator to design the TX and RX chips from the transistor level using the IHP SG13S SiGe process design kit (PDK) available for AWR software. The SG13S 130nm SiGe BiCMOS technology for millimeter-wave applications features high-speed HBTs (fT = 240GHz and fmax = 330GHz).
A block-level schematic of the 4-channel Tx RFIC is shown on the left in Figure 7, with the actual TX chip shown on the right. The active balun, 6X (2X and 3X cascaded) harmonic multiplier chain and one of the element’s power amplifiers (PA) are highlighted to indicate their location on the RFIC. Three active power dividers are used to split the signal among four symmetric lines which are each fed into a 2-stage polyphase filter to remove the unwanted products from the harmonic multiplier, followed by the buffer amp, IQ modulator and power amplifier stages.
Figure 7. A schematic-level design example of a PA and a 6X harmonic multiplier chain within the TX chip
The detailed schematic of the PA was developed using components from the foundry’s PDK for AWR Microwave Office, shown in Figure 8 along with the transformer, differential transmission line, and inductor broken out. These passive structures are electrically large (proportionate to the wavelength) and therefore required EM analysis and optimization using the AWR AXIEM planar EM solver. The EM components are embedded as subcircuits in the schematic for co-simulation with AWR Microwave Office software. By including EM analysis of these structures in combination with the PDK models, the measured vs. simulation results for the chip-level amplifier show excellent agreement, shown in Figure 9.
Figure 8. PA schematic with the transformer, differential transmission line, and inductor
Figure 9. PA simulation and measurement results for gain and return loss
Simulations of the active balun, harmonic multiplier chain, and active power divider on the output (Figure 10) were performed with a 10GHz input signal and 2.4V of bias voltage. The simulation results shown in Figure 11 provided extremely useful insight into the multiplier operation, allowing the designer as well as the end user to understand the power levels of the spurious signals that are being generated. From a radar design perspective, it is very beneficial to have this information in order to take design steps and fine-tune the circuit to suppress these three signals.
Figure 10. Detailed schematic of the 6X multiplier
Figure 11. 6X multiplier measurement results (left) and simulation results (right)
The RF section of the receiver, in Figure 12, shows the total of eight RX channels supported by two RX RFICs mounted as flip chips on a PCB. The PCB front and back (high-speed signal processing back-end) mounted on a stand for lab testing is shown in Figure 13. Phases and amplitudes of the receivers are calibrated from a single-point target measurement. Phase and amplitude correction factors are determined such that the point target measurement gives an image of a point target at the correct angle.
Figure 12. Eight RX channels with two RX chips flip-chip mounted on a PCB
Figure 13. System front (left) and back (right) showing the chip mounted on the PCB
The “people flow” (Figure 14) measurement results of the 2D MIMO radar verified that it can detect multiple people at the same time with 100-200 FPS. The range resolution was 3-5cm and the angular resolution was 3.5 degrees. During the image formation, the Hamming windowing function is applied in the range direction and a −25dB sidelobe level Taylor window was applied in the azimuth direction. The Taylor windowing function slightly degrades the angular resolution but improves the sidelobe level, allowing the image to be formed with a higher dynamic range. Different targets are well separated in the generated image.
Figure 14. “People flow” setup for radar measurement testing
Accurate phase measurement is useful in measuring very small movements of a target, for example, in determining the heartbeat and breathing rate of a human from the displacement of the chest. Frequency multiplexing has an advantage over time multiplexing for the measuring phase since all the channels are measured at the same time. This is demonstrated in the results of the radar-based multi-person heart rate variability (HRV) extraction shown in Figure 15. Vital signs such as heart rate, HRV, and breathing can be observed from the radar signal.
Figure 15. Results of multi-person HR, HRV, and breathing extraction
The design of a novel 60GHz MIMO RMCW radar system for commercial applications that uses FDM has been described. The system provides high frame-rate measurement, excellent phase stability, and a large FoV, and the unique architecture and integrated circuits were designed in such a way that the system can be scaled to much larger radar systems while still maintaining phase coherence between channels. Both 2D and 3D imaging systems have been demonstrated, and, to the best of the authors’ knowledge, this is the first 3D imaging mmWave frequency-division MIMO system of its kind.
 J. Hasch, E. Topak, R. Schnabel, T. Zwick, R. Weigel, and C. Waldschmidt, “Millimeter-wave technology for automotive radar sensors in the 77 GHz frequency band,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 3, pp. 845–860, Mar. 2012.
 C. Pfeffer, R. Feger, C. Wagner, and A. Stelzer, “FMCW MIMO radar system for frequency-division multiple TX-beamforming,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 12, pp. 4262–4274, Dec. 2013.
By: Dr. Tero Kiuru and Henrik ForsténVTT Technical Research Centre of Finland Ltd.
For questions, general feedback, or suggestions for future blog topics, write to email@example.com.
To receive announcements about new software releases and blogs in this series, click Subscribe Now and enter your email address in the Subscriptions box.