Hardware Platforms for RF Imaging

SDR platforms provide flexible, programmable RF front-ends for radar and imaging experiments:

• USRP (Ettus/NI): 70 MHz -- 6 GHz, up to 200 MHz bandwidth, MIMO support (up to $8\times 8$). Cost: USD 3k--USD 30k. Use case: sub-6 GHz OFDM radar, channel sounding, WiFi sensing.

• Xilinx RFSoC: Integrated ADC/DAC with FPGA, up to 5 GHz bandwidth at mmWave (with external mixers). Cost: USD 5k--USD 15k. Use case: real-time radar processing, wideband imaging.

• ADALM-PLUTO (Analog Devices): 325 MHz -- 3.8 GHz, 20 MHz bandwidth. Cost: USD 200. Use case: education, proof-of-concept.

Commercial mmWave radar modules provide integrated radar front-ends at 60--79 GHz:

• TI AWR1642/IWR6843: 77 GHz, 4 GHz bandwidth, $N_t = 3$, $N_r = 4$ MIMO. Cost: USD 50--USD 200 (evaluation boards). Use case: automotive imaging, gesture recognition, vital signs.

• Infineon BGT60TR13C: 60 GHz, 7 GHz bandwidth, $N_t = 1$, $N_r = 3$. Cost: USD 30. Use case: short-range imaging, presence detection.

• Vayyar: 62--69 GHz, custom MIMO array (up to $N_t = 48$, $N_r = 48$). Cost: custom. Use case: through-wall imaging, medical imaging.

These modules include on-chip signal processing (FFT, CFAR) and can output raw ADC data for custom algorithm development.

Channel sounders are measurement systems designed for characterising wireless propagation, directly applicable to imaging:

• Keysight/R&S VNA-based: Frequency-domain measurements with extreme dynamic range ($> 100$ dB). Bandwidth up to 110 GHz. Use case: high-precision indoor imaging, material characterisation.

• Sliding correlator sounders: Time-domain measurements with wide bandwidth using pseudo-noise sequences. Use case: outdoor channel measurement, delay spread characterisation.

• Phased array sounders: Combine channel sounding with electronic beam steering for angle-resolved measurements. Use case: mmWave/sub-THz angular spectrum imaging.

The antenna array geometry determines the k-space coverage and hence the imaging resolution and sidelobe structure:

• Uniform Linear Array (ULA): $N_r$ elements at spacing $d$ along one axis. Provides 1D angular resolution $\Delta\theta \approx \lambda / (N_r d)$. Simple but limited to azimuth-only imaging.

• Uniform Planar Array (UPA): $N_r$ elements on a rectangular grid. Provides 2D angular resolution (azimuth + elevation). Standard for mmWave automotive radar.

• Circular array: elements equally spaced on a circle. Provides $360^\circ$ coverage with approximately uniform angular resolution. Useful for rotational SAR.

• Random/sparse array: elements at non-uniform positions. Can achieve wider aperture with fewer elements but suffers from higher sidelobes. Requires CS-based reconstruction.

Calibration removes systematic errors before imaging:

• Phase calibration: measure the phase of each channel with a known reference target (corner reflector). For MIMO: calibrate each Tx-Rx pair independently using $c_m = e^{j(\phi_m^{\mathrm{exp}} - \angle h_m)}$.

• Timing synchronisation: measure a known target at known range to determine the system delay $\tau_0$: $\hat{\tau}_0 = \hat{\tau}_{\mathrm{meas}} - 2R_{\mathrm{true}}/c$.

• Amplitude calibration: normalise received power against a reference target with known RCS: $G_{\mathrm{cal}} = P_{\mathrm{rx}} / (P_{\mathrm{tx}} \sigma_{\mathrm{ref}} / (4\pi R^2)^2)$.

• Mutual coupling correction: measure the coupling matrix $\mathbf{C}$ in an anechoic chamber and apply $\mathbf{y}_{\mathrm{cal}} = \mathbf{C}^{-1}\mathbf{y}_{\mathrm{raw}}$.

• Self-calibration via redundant baselines: for arrays with repeated element spacings, the gain/phase of each element can be estimated from the data itself without an external reference.