Patrick Fenske, Andre Scheder, Tobias Koegel, Konstantin Root, Martin Vossiek, Christian Carlowitz
©SHUTTERSTOCK.COM/INK DROP
Monostatic radar systems as well as radars with separated but nearby-placed transmit and receive antennas are beneficial in terms of cost-efficiency and miniaturization, which is necessary for their integration into compact radar systems or mobile devices [1], [2] as well as for radar networks that demand reciprocal transmission channels between the nodes [3]. However, this lack of separation between the transmit (Tx) and receive (Rx) paths due to miniaturization implies only week isolation between Tx and Rx.
Figure 1 shows a possible monostatic setup, where a nonperfect isolation of the coupler and a mismatch at the antenna input are assumed. Hereby, a frequency-modulated continuous wave (FMCW) signal with a center frequency of ${f}_{\text{c}} = {24.125}{\text{ GHz}}$, a bandwidth of 250 MHz, and a power of ${P}_{\text{Tx}}^{\text{ dBm}} = {20}{\text{ dBm}}$ passes through a branchline coupler and is then radiated by the antenna. Assuming a target at ${R} = {5}{m}$ distance with a radar cross section of ${\sigma}_{\text{rcs}}^{\text{dBsm}} = {-}{10}{\text{ dBsm}}$, the received power at the input of the low-noise amplifier (LNA) can be calculated according to the radar equation [4] \begin{align*}{P}_{\text{Rx}}^{\text{dBm}} = & {P}_{\text{Tx}}^{\text{ dBm}} + {s}_{21}^{\text{dB}} + {2}{G}_{\text{Ant}}^{\text{dBi}} + {\sigma}_{\text{rcs}}^{\text{dBsm}} \\ & + {10}{\log}_{10}\left({\frac{{c}_{0}^{2}}{{\left({{4}{\pi}}\right)}^{3}{R}^{4}{f}_{c}^{2}}}\right) + {s}_{42}^{\text{dB}}{.} \tag{1} \end{align*}
Figure 1. Monostatic radar system example with a one-target scene and an illustration of the leakage effects. LNA: Low-noise amplifier.
Here, ${s}_{21}^{\text{dB}}$ and ${s}_{42}^{\text{dB}}$ account for the branchline coupler transmission losses, ${G}_{\text{Ant}}^{\text{dBi}}$ is the antenna gain, ${c}_{0}$ is the speed of light in a vacuum, and the radar cross section ${\sigma}_{\text{rcs}}^{\text{dBsm}}$ is a measure of the effective target area in square meters, giving the amount of power that is reflected from the target. Using the values of the example in Figure 1, the dominating leakage part results from the reflection of a portion of the Tx signal at the antenna input and can be calculated through \[{P}_{\text{Leak}}^{\text{dBm}} = {P}{}_{\text{Tx}}^{\text{ dBm}} + {s}_{21}^{\text{dB}} + {\Gamma}_{\text{Ant}}^{\text{dB}} + {s}_{42}^{\text{dB}}, \tag{2} \] where $\Gamma{}_{Ant}^{\text{ dB}}$ is the antenna input reflection coefficient. Therefore, the power of the desired signal is ${P}_{\text{Rx}}^{\text{dBm}} = {-}{80}{d}{B}{m}$, while the interferer power amounts to ${P}_{\text{Leak}}^{\text{dBm}} = {-}{2}{d}{B}{m}$, which gives a signal-to-interference ratio (SIR) of ${\text{SIR}}_{\text{dB}} = {-}{78}{\text{ dB}}$. This example emphasizes the significance of the leakage problem in common monostatic radar front ends. Strong leakage signals may lead to different disturbing effects that decrease the target detection performance of the radar system. One general problem of full-duplex radios is the so-called receiver desensitization. This could result from driving the RF receiver components, especially the LNA, into saturation. As consequence, those devices will show nonlinear behavior and generate strong intermodulation (IM) products and reduce the received power level of weak targets, possibly below noise level through compression [5]. In the context of radar signal processing, it could be difficult to distinguish between real target reflections and IM products, or nearby targets could be fully covered by the additional phase noise of the leakage signal [6]. Additionally, [7] shows that strong self-interference signals can raise the receiver noise floor of the sampled signals after the analog-to-digital converter (ADC) by more than 50 dB due to a limited effective dynamic range of the ADC. Therefore, even if the ideal Tx signal is known at the receiver, an all-digital cancellation approach cannot cancel out this additional noise and interference parts. Thus, an analog cancellation mechanism has to be applied [7]. To further push the developments in analog self-interference cancellation, a student design competition (SDC), “Design of a Self-Interference Cancellation Coupler,” was held during the IEEE Microwave Theory and Technology Society (MTT-S) 2022 International Microwave Symposium (IMS) in Denver, CO, USA, sponsored by Technical Coordinating Committees MTT-4 (Microwave Passive Components and Transmission Line Structures Committee) and MTT-24 (Microwave/mm-wave Radar, Sensing and Array Systems Committee). This article discusses the design considerations of the winning self-interference cancellation coupler (SICC) and presents its performance by measuring the figures of merit (FOMs). Since our design is capable of estimating the residual carrier leakage power for adjusting the cancellation signal, we call it Self-Interference Leakage Estimation and Cancellation Element (SILENCE).
The task of the competition was to design, realize, and demonstrate a SICC for an Industry, Science, Medicine (ISM) frequency band from 24 to 24.25 GHz with three ports with 3.5 mm female coaxial connectors. Port 1 and port 3 are for the Tx and Rx signals, respectively, while port 2 is the antenna port and was connected under two different scenarios:
The goals of the competition were low attenuation between the Tx port and the antenna port, and between the Rx port and the antenna port. High isolation is wanted in the Tx and Rx paths. Passive components, e.g., couplers, dividers, and splitters, were only allowed as self-designed printed circuit board (PCB) components. Active components may only be placed in the cancellation and detector path. The SICC is tested at five equidistant distributed frequencies, and it must find its setting automatically after a maximum of 2 min. All rated criteria of the SICC in detail were as follows:
To fulfill the criteria of the competition, a suitable circuit must be selected first. In principle, two different topologies are suitable to realize a SICC considering the given rules. The first architecture uses an adjustable impedance according to Figure 2; this way, a part of the signal with the desired phase and amplitude is reflected back to the branchline coupler, where it is superimposed with the signals in the receiving path.
Figure 2. A block diagram of the two designs that can be utilized to realize the SICC. DET: detector (RF power detector)
The second architecture, on the other hand, uses phase and amplitude modulation to generate the cancellation signal. In contrast to the first architecture, the signal is not reflected but is passed to the receiving path with an additional directional coupler. This coupler should have a relatively small coupling factor to minimize the losses in the receiving path as much as possible. On the other hand, the amplitude of the cancellation signal must be increased greatly with a low coupling factor as only a small part of the signal is coupled to the receiving path. Therefore, the use of additional amplifiers must be considered.
While the first design seems more feasible because of its reduced hardware complexity, it is based on the principle of an impedance tuner, which is typically realized with varactor diodes. Unfortunately, the performance of discrete varactor diodes in the 24-GHz frequency domain is often not unsatisfactory, while commercial IQ-modulator integrated circuits (ICs) still allow precise modulation of high-frequency signals. For this reason, the second architecture was chosen as more suitable to compete in the contest, and a more detailed description of the design is presented in the following section.
In the last section, a suitable topology was selected, which is shown in Figure 3 as an elaborated block design and then in Figure 4 as a PCB. In this section, the different considerations leading to the overall design are discussed. Therefore, a brief overview of the microwave structures is given, and the possibilities and limitations of the circuit are addressed by investigating the two critical signal paths present in the design.
Figure 3. An overall block diagram of the SICC in a measurement scenario. The detector and cancellation path are depicted in red and blue, respectively. DAC: digital-to-analog converter.
Figure 4. Final RF PCB of the SICC.
For the design shown in Figure 3, a branchline coupler and a directional coupler must be designed. Rogers 3003 was chosen as the substrate as it has important benefits for the previously identified competition criteria of the SDC. First, it has a low dissipation factor (1e-3 at 10 GHz). In addition, the low substrate height of 0.127 mm results in a small stripline width dimension. Further, most of the field is guided inside the substrate, and radiation losses at transmission line discontinuities are reduced. The minimal structure size of 0.1 mm allows accurate geometries and, hence, a small footprint of the RF board. The copper layer has a height of 0.035 mm. The geometry of the couplers can be seen in Figure 4. The arms of the branch directional coupler have a length of ${\lambda} / {4} = {2.15}{m}{m}$, and each of the two opposite branches has a line impedance of ${50}\,{\Omega}$ and ${35.35}\Omega$ or a linewidth of 0.285 mm and 0.495 mm, respectively. The arms of the line coupler have similar lengths of 1.9 mm, and the space between the coupling lines is 0.1 mm, which is the minimal structure size. Figure 5(a) and (b) shows the most relevant measurement and simulation results for both couplers. The measurement setup was through-reflect-line calibrated. Matching is represented by s11 for both structures, and transmission for the directional coupler is represented by s21 and by s31 and s41 for the branchline coupler, respectively. Although the measurement and simulation of s11 show a small deviation, the achieved input reflection is still below ${-}{15}{\text{ dB}}$. The measured values for s21 are close to the simulated values and are ${-}{1.6}{\text{ dB}}$ for the directional coupler, and s31 and s41 are ${-}{3.5}{\text{ dB}}$ and ${-}{4}{\text{ dB}}$ for the branchline coupler. In the simulation, a 15-dB coupling was the aim for the directional coupler, represented by s41. The measurement is in good agreement with 16.2 dB at 24 GHz. The measured isolation of the branchline coupler with ${-}{70}{\text{ dB}}$ at 24.125 GHz is better than the simulated result of ${-}{60}{\text{ dB}}$ and is exactly at the center frequency of the ISM band.
Figure 5. Simulation and measurement results of (a) the directional coupler and (b) the branchline coupler. meas: measured; sim: simulated.
A key part of the SICC system is the cancellation path into which a portion of the Tx signal is fed and used as a local oscillator (LO) signal for a vector modulator IC. The vector modulator adjusts the phase and amplitude of the LO signal, so that the output of the cancellation path interferes destructively with the leakage signal at the Rx port. This modulation can be achieved by utilizing an IQ modulator that can be dc biased at the in-phase and quadrature ports. The MMIQ-1040LSM from Marki Microwave offers this functionality and can be biased over a wide current range from –30 to 30 mA. The IC requires an LO drive level between 10 and 20 dBm. Therefore, a digital stepped attenuator in combination with a fixed-gain amplifier provides a sufficient LO input power over a wide Tx input power range. Figure 6(a) shows the signal power level at the different nodes along the cancellation path for a fixed Tx power of 10 dBm with two different attenuator settings that set the possible LO drive range of the modulator. The origins of the gain and attenuation factors are listed in Table 1. The passive vector modulator can generate phase shifts over the full 360° range and has an insertion loss of about 9 dB with a modulation depth of 41 dB. The effective modulation depth can be increased by also varying the LO drive power in a range of 10 dB, which results in an effective dynamic range of 51 dB, marked by the blue area in Figure 6(a). The output signal of the vector modulator is again amplified by a fixed-gain amplifier before it is superimposed with the leakage signal at the Rx path to ensure that both signals can be adjusted equally in amplitude. The residual signal power is than measured by a power detector, which is described in the next section.
Figure 6. Power level budget of the (a) cancellation path and (b) detector path. The origins of the gain and attenuation factors are listed in Table 1.
Table 1. Variable sources.
In the power detection path it is determined whether the Tx leakage is successfully suppressed by measuring the signal power at the Rx port with a detector. To avoid reducing the power of the receiver signal significantly, only a small portion of the signal is fed into the detector with a 15 dB-directional coupler. Because of the limited dynamic range of the root mean square power detector (Analog Devices LTC5596), from 0 to ${-}{35}{\text{ dBm}}$, as well as the low coupling factor, an adjustable step attenuator, together with an amplifier, was used to increase the effective dynamic range of the detector. In Figure 6(b) the level map of the detector path is illustrated for two cases. The blue line represents the case for the lowest possible detectable power, while the green line represents the maximum power expected at the detector with the adujstable attenuator set to 3 dB for an input power of 10 dBm and a complete reflection of the RF signal at the antenna port with a deactivated cancellation path. It is possible to detect higher powers by attenuating the RF signal more, which for the competition case is unnecessary. If we follow the blue line back from the least detectable power of ${-}{35}{\text{ dBm}}$ and consider the 25-dB amplification of the Analog Devices HMC751, the insertion loss of the step attenuator, and the 15-dB coupling factor by the coupler, we end up with a minimum power detectable at the Rx port of ${-}{42}{\text{ dBm}}$, which results in a possible detectable isolation of 52 dB with an input power of 10 dBm. This is enough to score the highest possible point score for this competition.
The control logic is integrated onto a separate PCB which comprises an STM32 microcontroller, the Texas Instruments DAC8775, and power conditioning for the amplifiers on the RF board. The digital-to-analog converter (DAC) has four channels that can be individually configured to act as either voltage or current DACs, each in different voltage or current ranges. Two channels of the DACs are configured in the $\pm{24}{-}{mA}$ current range to generate the I and Q signals for the modulator. The other two channels are configured in the $\pm{5}{\kern0.0em-}{V}$ range to generate negative bias voltages for two amplifiers. The additional power conditioning circuit consists of linear regulators to supply the drain contacts of the amplifiers as well as the step attenuators. The control PCB is connected to the RF PCB via a flat ribbon cable, over which all of the necessary control signals, power rails, and the analog voltage from the power detector are transmitted. The microcontroller provides a serial USB connection to a control PC. A program on the PC can then be used to send commands to the microcontroller, which interprets them and then, for example, sets different I and Q currents or changes the attenuation of the step attenuators. The control PCB can be seen in Figure 7. Because this competition allowed plenty of time to set the right I and Q currents for maximum isolation, an incremental grid search algorithm was used. The algorithm in the first run does a coarse grid search from the minimum to maximum settable currents and records the isolation reached for each set current value. Then, it selects the point for the least power detected and does another grid search with a finer grid around this point. After the third run, the I and Q currents that caused the detected power to reach its minimum are selected. In Figure 8 the results of the three runs can be seen. The algorithm takes less than 2 min to run and therefore satisfies the requirement for this competition.
Figure 7. The control board for the SICC. IO: input–output.
Figure 8. Successively sampled ADC values of the detector output voltage over a (a) coarse, (b) medium, and (c) fine current IQ grid during an optimization run.
As previously described, the FOMs of the SICC are low insertion loss in the Tx and Rx paths, sufficient input matching at the Tx and antenna ports, and a high isolation between the Tx and Rx paths. To account for different possible antenna matching characteristics, different measurement scenarios are considered. To get a full three-port characterization of the SICC, the Tx, antenna, and Rx ports are directly connected to the VNA. This setup can be considered as an almost perfect matching at the antenna port. The insertion losses of the Tx and Rx paths as well as the antenna port matching were solely measured in this scenario. This is also an indicator for the performance of the canceller in the presence of weak reflection signals. To validate the performance for realistic antenna reflection coefficients, the antenna port was not connected to the VNA, but instead a coaxial attenuator was mounted and left open. The Tx signal was hence attenuated by the attenuator, almost totally reflected at the open termination, and again attenuated. So, an antenna matching of about twice the attenuation was emulated. Both described setups are depicted in Figure 3. For each scenario, a two-step measurement procedure was conducted on a four-port VNA. In the first step, the VNA was set to CW mode at different frequencies of the respective ISM band, where the SICC element can perform a cancellation optimization algorithm to find the optimum phase and amplitude for the cancelling signal. After an optimal operation point was found, the VNA was set to the frequency sweep mode to get the required S-parameters in a frequency range from 21 to 27 GHz. Figure 9 shows the measurement results for a CW optimization at 24 GHz and the three-port, ideally matched scenario as well as for a 6-dB open termination at the antenna port, which can be considered as an antenna with approximately 12-dB input matching. In both scenarios, an isolation of more than 40 dB was achieved, while the insertion loss of the Tx and Rx paths comprised 4.5 and 6.5 dB, respectively. The Tx as well as the antenna ports showed good input matching of about 20 dB each. It should also be noted that the SICC performs better for more poorly matched antennas. This could be because of the vector modulator nonidealities. Carrier leakage has a higher impact when the cancellation signal is weak, i.e., the modulator input currents are very low (close to zero). Similar measurements were conducted at the 2022 IMS, and the results are shown in Table 2. The terms ${s}_{31,\text{d}\text{i}\text{r}}$ and ${s}_{31,{\text{emu}}}$ stand for the direct connection to the VNA and the antenna emulator, respectively.
Figure 9. Measurement results of the SICC system optimized at ${f}_{\text{Tx}} = {24}{\text{ GHz}}$ for a matched termination at the antenna port $({s}_{ij,match})$ and for an open 6-dB attenuator ${(}{s}_{31,6dB,open}{)}$.
Table 2. SILENCE measurement results at IMS 2022.
This article presented a SICC for the ISM frequency band from 24 to 24.25 GHz with a carrier cancellation much better than 30 dB. In the beginning, a short introduction and description of the competition criteria for the IMS SDC 2022 was given, and two possible circuit designs were discussed. The 2022 IMS SDC Winner icon is shown in Figure 10. The chosen RF system includes self-designed microwave structures, an active cancellation path, and a power detection path. The superposition of the cancellation signal and reflected Tx signal from the antenna is detected, and the phase and amplitude of the cancellation signal are varied until a power minimum is reached. The measured results at the student competition showed a carrier cancellation better than 30 dB for all frequencies between 24 and 24.25 GHz and a matching of at least 23 dB. With a footprint of only 16 cm2, a very small and powerful device was designed. To enhance the performance of the circuit further, additional improvements can be made to the circuit. Very low power levels occur during the minimum search, but the current design of the detection path allows a minimum detection power of only ${-}{42}{\text{ dBm}}$. Using an Rx stage with an LNA, IQ mixer, and ADC, a higher dynamic range is achievable, lower power levels can be detected, and additional phase information can be used for faster optimization. As a result, lower power minima can be found, and a better self-cancellation is possible. For the competition, the designed hardware was tested at five discrete frequencies. Nevertheless, an application in an FMCW radar also seems possible. For this purpose, the necessary phase shift for the IQ modulator must be determined once for multiple points of the frequency ramp of the radar. Synchronization between the FMCW ramp and determined phase shift would then allow a carrier cancellation for continuous frequency sweeps.
Figure 10. The 2022 IMS SDC Winner icon for “Design of a Self-Interference Cancellation Coupler.”
This work was supported by the Deutsche Forschungsgemeinschaft (the German Research Foundation) GRK 2680, Project ID 437847244.
[1] J. Lu, Z. Shao, C. Li, C. Gu, and J. Mao, “A portable 5.8 GHZ dual circularly polarized interferometric radar sensor for short-range motion sensing,” IEEE Trans. Antennas Propag., vol. 70, no. 7, pp. 5849–5859, Jul. 2022, doi: 10.1109/TAP.2022.3142303.
[2] S. Praveen et al., “Miniature radar for mobile devices,” in Proc. IEEE High Perform. Extreme Comput. Conf. (HPEC), Piscataway, NJ, USA, 2013, pp. 1–8, doi: 10.1109/HPEC.2013.6670337.
[3] M. Gottinger, F. Kirsch, P. Gulden, and M. Vossiek, “Coherent full-duplex double-sided two-way ranging and velocity measurement between separate incoherent radio units,” IEEE Trans. Microw. Theory Techn., vol. 67, no. 5, pp. 2045–2061, May 2019, doi: 10.1109/TMTT.2019.2902553.
[4] M. I. Skolnik, Ed. Radar Handbook, 2nd ed. New York, NY, USA: McGraw-Hill, 1990.
[5] S. Sadjina, C. Motz, T. Paireder, M. Huemer, and H. Pretl, “A survey of self-interference in LTE-advanced and 5G new radio wireless transceivers,” IEEE Trans. Microw. Theory Techn., vol. 68, no. 3, pp. 1118–1131, Mar. 2020, doi: 10.1109/TMTT.2019.2951166.
[6] R. Feger, C. Wagner, and A. Stelzer, “An IQ-modulator based heterodyne 77-GHz FMCW radar,” in Proc. IEEE German Microw. Conf., 2011, pp. 1–4.
[7] A. Sabharwal, P. Schniter, D. Guo, D. W. Bliss, S. Rangarajan, and R. Wichman, “In-band full-duplex wireless: Challenges and opportunities,” IEEE J. Sel. Areas Commun., vol. 32, no. 9, pp. 1637–1652, Sep. 2014, doi: 10.1109/JSAC.2014.2330193.
Digital Object Identifier 10.1109/MMM.2023.3256380