DATASET FILES

The file "PMCW2FMCW-interference.zip" contains 8 *.mat files. Each of these contains a single 4-dimensional array called "radarData" which corresponds to the signal samples captured directly after sampling in an AWR1443BOOST radar board. The samples are arranged in the following way, from the leftmost to the rightmost dimension of the array:

-Receiver channels: 1 to 4.
-Sample index (of each sweep): 1 to nº samples/sweep.
-Sweep index (of each burst): 1 to nº sweeps/frame.
-Burst index: 1 to nº frames/measurement run.


MEASUREMENT SETUP

All the measurements were conducted inside an anechoic chamber of inner dimensions (i.e., from the tips of the absorbers) 8.2 m × 2.0 m × 2.2 m. The FMCW radar was located at one end of the chamber, while the PMCW one was placed at the other end, at approximately 7.8 m. A couple of naked metal components constitute relevant static targets. Two mobile targets were considered: one person walking back and forth over the walkable path and a drone flying on the opposite side to the person. The drone is a target with a small radar cross-section, and together with the victim radar was employed in the measurements of our previous work [1], [2]. The PMCW interfering radar is a modified version of the sounder used in [3]. In this dataset, we only present the PMCW-to-FMCW interference, as the transmitted power of the FMCW board was relatively low. Due to this, the FMCW interference was hard to visualize in the RD plots obtained with the PMCW device.

The scenario can be seen in the cover picture. As the floor of the chamber is currently being replaced some of the metallic elements are naked and, thus, will appear as static targets. These are placed at the following distances: 3.9 m (start of the rails in the ground), 5.2 m (iron piece at the left of the image), 6.0 m (turntable), and 6.8 m (support for the interfering radar). These distances have been measured using a manually operated laser telemeter leaning just to the top of the victim radar board. Due to this and the physical extension of these targets, an error of several cm can be expected in these distances. The drone and the person moved across the approximated range of [1.6 m, 5.8 m] with speeds up to about 2.5 m/s and 1.0 m/s approximately. These velocities were carefully chosen in order to avoid damage due to collisions with the chamber elements. Absorbers on the walls and ceiling and covering the half of the floor that is closer to the FMCW radar attenuate multipath propagation.

The configurations of both radars are shown in Table 1 and Table 2. Note that there are two setups for both radars. In setups 1a and 1b, the bandwidth of the victim radar is greater than that of the interfering radar. In setups 2a and 2b, the bandwidth of the FMCW radar is covered by that of the PMCW radar. In all sweeps, the center frequency is close to 78.5 GHz, the same as for the PMCW radar. In setups 2a and 2b, we have decreased the bandwidth of the victim radar instead of increasing that of the interfering radar due to hardware limitations of this last one.
The different sequences for the PMCW radar are either a binary code (PRBS, setups 1a and 2a) or a polyphase sequence (ZC, setups 1b and 2b).

Table 1.  FMCW (victim) radar configuration parameters
|Parameter			|Setup 1     |Setup 2     |
|---				|    ---     |	  ---     |
|Carrier frequency [GHz]	|77.0	     |78.0        |
|Bandwidth [GHz]		|2.81	     |0.94        |
|Sweep slope [MHz/µs]		|52.77	     |52.77       |
|Sweep duration [µs]		|53.33	     |17.78       |
|Idle time [µs]			|5.0	     |5.0         |
|SRI [µs]			|58.33	     |58.32       |
|Sampling frequency [MHz]	|6.0	     |6.0         |
|Samples per sweep		|290	     |82          |
|Sweeps per burst		|128	     |128         |
|Burst repetition period [ms]	|8.0	     |8.0         |

Table 2.  PMCW (interering) radar configuration parameters
|Parameter			|Setup a     |Setup b     |
|---				|    ---     |	  ---     |
|Carrier frequency [GHz]	|78.5	     |78.5        |
|RF Bandwidth [GHz]		|1.0	     |1.0         |
|Chip duration [ns]		|2.0	     |2.0         |
|Sequence			|PRBS	     |ZC          |
|Sequence length		|2047	     |2048        |
|Repetition interval [µs]	|4.094	     |4.096       |

The provided files combine these setups in the following way:

"b3g-clean": Setup 1. NOTE: due to a human error during the measurement process, for this specific file, the parameters varied a little bit from the Table 1:
	-Sweep slope: 56.25 MHz/µs instead of 52.77 MHz/µs.
	-Bandwidth: 3.00 GHz instead of 2.81 GHz, resulting in a resolution of 5.33 cm instead of 5.00 cm.
	-Samples per sweep: 306 instead of 290.
	-The ADC starting time was earlier than in the rest of the measurements and therefore some artifacts appear at the first samples of each ramp. By using windows before the FFT, the impact of this is greatly attenuated.
	-The resulting difference in spatial resolution can be neglected when comparing the results, as it is only a difference of 3.3 mm. The measurements could not be repeated as the floor of the chamber was replaced after this campaign took place.
"b3g-prbs": Setup 1a.
"b3g-zc" Setup 1b.
"b1g-clean": Setup 2.
"b1g-prbs": Setup 2a.
"b1g-zc": Setup 2b.

For the "b1g-prbs-qc" and "b3g-zc-qc", the parameters of the victim radar have been slightly modified so that the SRI of the FMCW radar is an integer multiple of the repetition period of the corresponding interfering radar. This has been done with the aim of causing a certain coherence of the interference in the slow-time dimension of the victim radar, as some works have studied [4], [5]. The setups for these files are mostly those of the tables with a couple of changes:

"b3g-zc-qc": Setup 1b but with chirp duration 56.44 µs and 312 samples per sweep.
"b1g-prbs-qc": Setup 2a but with idle time 43.63 µs.

All of the radars are configured in MISO operation. The main lobe of the interfering radar is aimed towards the victim radar. Neither of them move during the whole measurement process.


DATA PROCESSING

After loading the variable "radarData", typical FFT processing can be conducted. An FFT across the second dimension (i.e the fast time) produces the range spectrum. A second FFT, applied across the third dimension (i.e. the slow time), produces the range-Doppler (RD) profiles. The fourth dimension depicts the evolution of the captures against time. The first dimension (i.e. the Rx channel) can be used to extract angular information. In the attached script, "loadMeasurementResults.m", a very simple example of how to work with the data can be seen. It reads the file and plots the raw data of a single burst and Rx channel, as well as its associated range and RD profiles.


REFERENCES

[1] L. A. López-Valcárcel, M. García Sánchez, F. Fioranelli, and O. A. Krasnov, “An MTI-like Approach for Interference Mitigation in FMCW Radar Systems,” IEEE Trans. Aerosp. Electron. Syst., pp. 1–16, 2024, doi: 10.1109/TAES.2023.3345263.
[2] L. A. López-Valcárcel, M. García Sánchez, F. Fioranelli, and O. A. Krasnov, “Raw ADC data from FMCW radar at 77 GHz with interference.” IEEE DataPort, Jun. 30, 2023. doi: 10.21227/E47T-P857.
[3] L. A. López-Valcárcel and M. García Sánchez, “A Wideband Radio Channel Sounder for Non-Stationary Channels: Design, Implementation and Testing,” Electronics, vol. 10, no. 15, p. 1838, Jul. 2021, doi: 10.3390/electronics10151838.
[4] U. Kumbul, F. Uysal, C. S. Vaucher, and A. Yarovoy, “Automotive radar interference study for different radar waveform types,” IET Radar Sonar & Navi, vol. 16, no. 3, pp. 564–577, Mar. 2022, doi: 10.1049/rsn2.12203.
[5] A. Bourdoux, K. Parashar, and M. Bauduin, “Phenomenology of mutual interference of FMCW and PMCW automotive radars,” in 2017 IEEE Radar Conference (RadarConf), Seattle, WA, USA: IEEE, May 2017, pp. 1709–1714. doi: 10.1109/RADAR.2017.7944482.