Results of the first in-filed test of the proposed photonics-assisted radar network
To confirm the enhanced performance of the proposed photonics-assisted MIMO radar, with respect to traditional distributed multistatic radars, some preliminary tests have been run in a real open-door scenario, with a down-scaled 2 TX x 2 RX system working in X band.
The demonstrator of the photonic distributed radar is composed by the photonic core, a co-located acquisition system and two remoted SPs, for three SPs in total, each divided in one TX and one RX. The architecture of the demonstrator is depicted in Fig. 39. In this preliminary system, the optical master clock contained in the photonic core is implemented by a solid-state MLL with a repetition rate of 400MHz. The master clock is split in two arms, one of which is connected to the RXs in the SPs through a span of optical single-mode fiber (SMF), represented by the coils sketched in Fig. 39, whereas the other is used for the E/O conversion of the radar waveform in transmission. The photonic core receives the radar waveform at IF and operates its conversion to the optical domain.
The optical output of the E/O conversion stage is split equally over two arms, which are fed into the TXs of the SPs. The E/O conversion is motivated by the inherent advantages in photonic-assisted RF signal generation, which entails a high stability and a high coherence between the signals transmitted by different SPs. Moreover, among the benefits of the Radio-over-Fiber (RoF) signal distribution between the photonic core and the SPs are the small propagation losses and absence of electromagnetic interference, along with preservation of coherence especially in case of signals with large BW. Once the E/O converted radar waveform is received by a TX, this operates its O/E conversion implying the up-conversion of the IF signal to the RF carrier, in this case at 9.7GHz.
The employed antennas at the TXs and RXs are ultra-wideband (2 – 18GHz) horn antennas with a ~50° half power beam width (HPBM) aperture and 12dB gain. It is worth underlining the presence of an ODL before the TX of the second SP, to implement the TDM of the two SPs. The ODL is realized by means of another 1 km-long spool of optical fibre, corresponding to a delay ΔT = 5 µs, which introduces orthogonality between the waveforms transmitted by SP1 and SP2.
The target echoes received by the SPs are amplified, pass-band filtered and E/O converted before the transmission back to the photonic core on a SMF span. Once in the photonic core, the received signal is transferred back to the electrical domain by O/E conversion which, at this stage, entails the optical sampling of the signal and its down-conversion to IF. After this operation, the signal is low-pass filtered and fed into the acquisition system, where it is digitized by an ADC with a sampling rate of 800 MBs and, eventually, processed thanks to the digital signal processing (DSP).
The coherent photonic radar network has been deployed, as depicted in Fig. X1, with the RX and TX antennas aligned over a 21 m-long baseline. These are oriented upwards, in order to mitigate clutter and multipath returns due to surrounding structures, buildings and ground. The antennas are tilted to ensure the simultaneous illumination of the target. The experiments have been carried out considering two possible target configurations: (i) single target, (ii) two closely spaced targets. In the case (i), two possible targets have been considered: a small, cooperative target consisting of a cylinder, with 17 cm radius and 50 cm height, made of a tight-mesh metal net and hanging from a mini-drone, hovering above the baseline; an extended, non-cooperative target, represented by airplanes taking off in the proximity of the test field. The sketch of the experimental setup, and the picture of the targets are reported in Fig. X2.
A first system validation has been conducted to test the system imaging capabilities. The multistatic ISAR processing, based on a Doppler Parameter Estimation Algorithm, has been adopted for testing the imaging capability of the photonic radar network on moving targets. Experiments on extended targets have been conducted employing a single TX and both RXs with B = 20 MHz, with a range resolution of 15 m, with an observation interval of 0.1 s which limited the Doppler resolution to 10 Hz. The chosen targets were airplanes taking off in the proximity of the building. In Fig. X3 are shown the images of the target before (a, c) and after (b, d) the focusing in the Range-Cross range domain, for the RX1 (a, b) and RX2 (c, d). The selected target is moving at radial velocity of about 130 km/h, at a distance from the baseline of around 770 m. Although it is difficult to identify the shape of the airplane, the visual inspection demonstrates the capability of the algorithm to achieve well focused images, despite the short acquisition time, thanks to the stability of the optically generated and distributed signals.
Tab. 1 reports typical quality parameters to evaluate imaging performances for both RXs. As apparent from the results, all the parameters improve after the application of the ISAR imaging algorithm, thus confirming the capability of the system to guarantee good performances. Even though the analysis were carried out on a premature version of the radar system which is not tuned to ISAR imaging, the achieved results are rather promising.
The second test in the single target scenario, employing the above-described cooperative small target, has been conducted employing a radar signal with a bandwidth B = 100 MHz. Combining the detection maps for all the channels, it is possible to evaluate the range/cross-range cumulative detection map, as depicted in Fig. 43, in which the ellipsoidal crowns represents the iso-range regions of positive detections at each channel. The region of intersect, i.e. the area in which all iso-range ellipses overlap, extends for about 3x3 m around the central point of coordinates (-12 m, 16.5 m). During the acquisition the target position/height has been constantly measured using a laser range finder. After the data acquisition, GPS positions (red circles) extracted from the drone log-file have also been considered. They show that the area of intersect is compatible with the actual target position.
In the two-target scenario, two identical cylinders (as the one employed in the single-target experiments), held by two mini-drones have been employed as cooperative targets. The experimental setup and the picture of the two drones hovering above the system baseline are reported in Fig. X4. The two targets were always kept at a distance of around 3 m from each other, at 18 m above the baseline. The axes of the cylinders were perpendicular to the baseline, to enhance the radar cross-section seen from the antennas.
The acquired echoes have been processed with non-coherent and coherent MIMO radar algorithms. As we can observe in Fig. 17 a), the two targets are too close to be correctly separated in cross-range with a non-coherent MIMO processing. Indeed, being the in-range distance from the baseline of the two targets 18 m, and given the antenna HPBW ~50°, the expected monostatic cross-range resolution is 15.7 m. Fig. 17 b), instead, depicts the output of the coherent MIMO processing. Here, the two targets are correctly separated in cross-range, around 3 m apart from each other, demonstrating a cross-range resolution improvement by a factor of > 5. As expected, the 1.1 km of SMF between the PC and one of the SP does not negatively affect the coherence and the resolution of the radar network.