• Photonics Research
  • Vol. 13, Issue 2, 395 (2025)
Jianwei Liu1、2, Ruixuan Wang1、2, Jiyao Yang1, Weichao Ma1、3, Henan Zeng1, Chenyu Liu1、2, Wen Jiang1, Xiangpeng Zhang1, Qinyu Xie1、2, and Wangzhe Li1、2、*
Author Affiliations
  • 1National Key Laboratory of Microwave Imaging, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, China
  • 2School of Electronics, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China
  • 3Guangdong Provincial Key Laboratory of Terahertz Quantum Electromagnetics, GBA Branch of Aerospace Information Research Institute, Chinese Academy of Sciences, Guangzhou 510700, China
  • show less
    DOI: 10.1364/PRJ.533960 Cite this Article Set citation alerts
    Jianwei Liu, Ruixuan Wang, Jiyao Yang, Weichao Ma, Henan Zeng, Chenyu Liu, Wen Jiang, Xiangpeng Zhang, Qinyu Xie, Wangzhe Li, "Photonic-frequency-interleaving-enabled broadband receiver with high reconfigurability and scalability," Photonics Res. 13, 395 (2025) Copy Citation Text show less

    Abstract

    The photonic frequency-interleaving (PFI) technique has shown great potential for broadband signal acquisition, effectively overcoming the challenges of clock jitter and channel mismatch in the conventional time-interleaving paradigm. However, current comb-based PFI schemes have complex system architectures and face challenges in achieving large bandwidth, dense channelization, and flexible reconfigurability simultaneously, which impedes practical applications. In this work, we propose and demonstrate a broadband PFI scheme with high reconfigurability and scalability by exploiting multiple free-running lasers for dense spectral slicing with high crosstalk suppression. A dedicated system model is developed through a comprehensive analysis of the system non-idealities, and a cross-channel signal reconstruction algorithm is developed for distortion-free signal reconstruction, based on precise calibrations of intra- and inter-channel impairments. The system performance is validated through the reception of multi-format broadband signals, both digital and analog, with a detailed evaluation of signal reconstruction quality, achieving inter-channel phase differences of less than 2°. The reconfigurability and scalability of the scheme are demonstrated through a dual-band radar imaging experiment and a three-channel interleaving implementation with a maximum acquisition bandwidth of 4 GHz. To the best of our knowledge, this is the first demonstration of a practical radio-frequency (RF) application enabled by PFI. Our work provides an innovative solution for next-generation software-defined broadband RF receivers.

    1. INTRODUCTION

    Radio-frequency (RF) receivers with large bandwidth and software-defined capability are key enabling blocks for emerging modern RF applications, spanning radar, wireless communication, electronic countermeasures, and next-generation multi-functional integrated RF equipment [14]. To unlock higher detection resolution, faster data transfer rates, and enhanced electromagnetic-environment adaptability in these applications, the receivers are required to have broadband coverage, sufficient accuracy, and high reconfigurability [5,6]. However, today’s pure electronic receivers have been encountering two significant challenges [7,8]: (i) the bandwidth of RF front-end components and the sample-and-hold (S/H) circuits is severely limited; (ii) electronic analog-to-digital converters (ADCs) can work up to only a few gigahertz due to the clock jitter—that is, the deviation of clock edges from their ideal instant—which decreases the signal-to-noise-and-distortion-ratio (SINAD) of the digitized signal quadratically with increasing frequency [7]. Photonic technologies are regarded as promising technologies to overcome these challenges due to the characteristics of photonics, including large bandwidth, high amplitude-phase consistency, immunity to electromagnetic interference, and the ultralow time jitter of the optical sampling pulses [913]. Photonic-assisted broadband RF reception techniques that can simultaneously increase the bandwidth and sampling rate of RF receivers mainly include photonic time-stretch [14], photonic time-interleaving (PTI) [15], photonic channelization [16], and photonic frequency-interleaving (PFI) [1719].

    The photonic time-stretch technique employs optical dispersion to slow down the incoming signal before it enters an electronic ADC; however, it can only capture short pulses with durations in the nanoseconds range due to the limited achievable dispersion by optical dispersive devices [20]. The photonic time-interleaving (PTI) operates within the conventional time-interleaving (TI) paradigm and can effectively reduce the system’s noise floor at high frequency by exploiting the intrinsic ultra-low-jitter characteristic of MLLs [12,21]. However, the PTI technique faces two challenging problems: (i) the severe spurs induced by channel mismatches; (ii) the precise alignment of optical/electrical sampling instants [22,23]. The photonic channelization uses photonic methods to slice a broadband RF signal into multiple spectral segments in the frequency domain for parallel narrow reception [2428]. However, photonic channelization cannot receive broadband RF signals across channels due to uncompensated incoherence between the subband signals, which severely limits its application scope.

    The photonic frequency-interleaving (PFI) technique performs the interleaving in the frequency domain [8,18,19,2933]. It slices the broadband incoming signal into multiple low-frequency narrowband segments using photonic devices, samples the segments simultaneously using a bank of low-speed electronic ADCs, and finally reconstructs the broadband incoming signal digitally. The PFI technique relaxes the requirement for clock jitter and shows less susceptibility to channel mismatch compared to the conventional TI paradigm, making it a superior solution for wideband multi-channel RF reception systems [34,35]. Considering the ever-expanding bandwidth demands and the fact that ADC performance (e.g., effective number of bits and spurious-free dynamic range) improves as its bandwidth decreases, a photonic frequency-interleaving scheme that simultaneously achieves broadband, dense channelization, and high channel crosstalk suppression is critical for RF applications, especially for analog signals. However, current PFI schemes mainly exploit optical frequency combs as multi-wavelength local oscillators (LOs) [18,19,30,31,36,37], and are challenging to meet these performance requirements: the dual-comb scheme enables dense channelization by using the vernier principle of two combs, but the system bandwidth is constrained by the free spectral ranges (FSRs) [19]. In a single-comb scheme, the limited performance of optical filters prevents dense channelization [18,36,37]. In addition, comb-based PFI schemes suffer from high system complexity, as well as limited reconfigurability and scalability, due to the use of narrow linewidth lasers, complex optical comb generation devices, optical filters, optical amplifiers, and challenging system control. Although these issues can be mitigated by using integrated optical comb chips [28,38], the maturity of integrated comb chips still requires further advancement and they still suffer from the issue of low comb line power. Furthermore, previous PFI work focuses on the segment reception of wideband communication signals, with the proposed DSP algorithms primarily addressing inter-channel phase compensation for digital signals, while lacking a comprehensive analysis of system non-idealities and the general capability for distortion-free signal reconstruction. Therefore, the PFI scheme to meet all the key requirements in the next-generation RF receivers, including ultra-wide operational bandwidth, distortion-free signal reconstruction, flexible reconfigurability, and scaling of channels, remains elusive.

    In this paper, we propose and experimentally demonstrate a novel incoherent PFI scheme for distortion-free reception of broadband RF signals, combining free-running laser array-based spectral slicing, real-time system distortion estimation, and a cross-channel signal reconstruction algorithm. Dense broadband RF signal spectral slicing, with high inter-channel crosstalk suppression, is achieved by exploiting multiple independent lasers as local oscillators. To eliminate the signal distortion induced by system non-idealities, a real-time system distortion estimation method based on calibration signals is designed and a dedicated system model is developed through a comprehensive analysis of the time/frequency characteristics and interdependencies of these non-idealities. For the distortion-free signal reconstruction, a cross-channel signal reconstruction algorithm is developed based on precise calibrations of intra- and inter-channel impairments. The system performance is first validated through receiving and reconstructing multi-format broadband signals, both digital and analog, with a detailed evaluation of signal reconstruction quality, achieving inter-channel phase differences of less than 2°. Subsequently, the practicality and reconfigurability of the scheme are validated through a dual-band microwave imaging experiment. Finally, the scalability of the proposed scheme is verified and evaluated based on three-channel interleaving microwave imaging with the total acquisition bandwidth increased to 4 GHz. To the best of our knowledge, our work presents the first demonstration of PFI-enabled practical broadband RF application and shows high integration potential, thereby paving the way for PFI-enabled future high-resolution sensing and high-speed communication systems.

    2. PRINCIPLE

    The fundamental concept of the proposed PFI-enabled broadband reception system is illustrated in Fig. 1(a). The system consists of a photonics-assisted channelization front end and a DSP back end. In the front end, the incoming broadband signal is modulated onto a continuous-wave (CW) light from a laser diode (LD) through an electro-optic modulator (EOM). The modulated optical signal is then split into N copies and routed to an array of optical coherent reception channels labeled as CH 1, …, and CH N. In each channel, the corresponding copy is injected into the signal port of an optical coherent receiver (CR), which comprises a 90° optical hybrid and a pair of balanced photodetectors (BPDs). The corresponding LO port of each CR is fed with the selected optical local oscillator (OLO) from a free-running laser. The wavelength of each OLO is tunable and is preset to downconvert the optical signal within a specific frequency range to the baseband, as shown in Fig. 1(b). In each channel, the in-phase and quadrature (IQ) components of the downconverted signal are individually sent to an LPF to implement spectral slicing and then digitized by a low-speed ADC. Importantly, the bandwidth of the LPFs moderately exceeds the frequency spacing of adjacent OLOs, ensuring each pair of neighboring spectral slices shares an overlap region [see the annotated area in Figs. 1(b) and 1(c)]. In the digital back end, the digitized signal slices of all channels are reconstructed to obtain the complete broadband signal. Ideally, the broadband signal reconstruction can be achieved by straightforwardly upsampling all signal slices, digitally upconverting them to their original spectral positions, and then summing them.

    Concept of the proposed photonic-frequency-interleaving (PFI)-enabled broadband receiver. (a) Schematic diagram of the PFI-enabled receiver. LD, laser diode; EOM electro-optic modulator; BPD, balanced photodetector; OLO, optical local oscillator; LPF, low-pass filter; ADC, analog-to-digital converter; DSP, digital signal processing; CH, channel. (b) The relative spectral positions among the optical carrier, the incoming signal, and the OLOs. (c) The digitized baseband spectral slices of the incoming signal in each channel after IQ combination. Overlap n, the overlap region between channel n and channel n+1.

    Figure 1.Concept of the proposed photonic-frequency-interleaving (PFI)-enabled broadband receiver. (a) Schematic diagram of the PFI-enabled receiver. LD, laser diode; EOM electro-optic modulator; BPD, balanced photodetector; OLO, optical local oscillator; LPF, low-pass filter; ADC, analog-to-digital converter; DSP, digital signal processing; CH, channel. (b) The relative spectral positions among the optical carrier, the incoming signal, and the OLOs. (c) The digitized baseband spectral slices of the incoming signal in each channel after IQ combination. Overlap n, the overlap region between channel n and channel n+1.

    However, the non-idealities in the optoelectronic transmission link would lead to intra-channel signal distortion and degrade the inter-channel coherence, thus rendering signal reconstruction quite challenging. In our scheme, the non-idealities can be classified into three categories: (i) time-variant phase deviation, including source phase noise defined by the linewidth of the lasers and link phase noise induced by temperature drift, and mechanical vibration; (ii) time-variant frequency fluctuation induced by the limited frequency control accuracy of lasers and incoherence between the optical carrier and the OLOs; (iii) non-ideal frequency responses of the optoelectronic devices, including non-flat amplitude-frequency responses and nonlinear phase-frequency responses. To eliminate all the signal impairments induced by the non-idealities, different measuring methods and a stepped compensation algorithm are exploited based on their different time-frequency features. For the two types of time-variant impairments, we add a calibration signal to the system to record the time-variant phase deviations and frequency fluctuations between the optical carrier and each OLO and extract the recorded information from the digitized signal for compensation in the DSP unit correspondingly as shown in Fig. 1(a). The non-ideal frequency responses of each channel, however, are observed to be stable over a relatively long duration (30  min) and thus, the corresponding signal impairments can be eliminated using premeasured results, which can be obtained by feeding a signal with a known waveform into the system at regular intervals. After all the impairments are eliminated, the broadband signal can be reconstructed naturally following the ideal reconstruction process mentioned above.

    Mathematically, the principle of the PFI-enabled broadband reception scheme can be illustrated as follows. The incoming broadband signal sRF(t)=s(t)·cos(2πfRFt) is double-sideband suppressed-carrier (DSB-SC) modulated onto an optical carrier generated by a continuous-wave laser. The generated optical signal is then split into N copies and routed to N interleaving reception channels. For the nth channel, the injected copy can be expressed as ES,n(t)=s(t)·Ac,n·{exp[j2π(fc+fRF)t+j2πfflc,c(t)t+jφc(t)]+exp[j2π(fcfRF)t+j2πfflc,c(t)t+jφc(t)]},where Ac,n, fc, fflc,c(t), and φc(t) denote the amplitude, the center frequency, the frequency fluctuation, and the phase noise of the optical carrier, respectively. For simplicity, the paths of each reception channel are considered to be precisely synchronized by utilizing delay-matched optical and electrical cables for all the discussions below. Note that the copies may differ in amplitude due to the imperfection of the optical splitter. However, these differences are invariant over time and thus can be considered as part of the system amplitude-frequency response of each channel.

    Each reception channel is also fed with the selected OLO from an independent free-running laser, written as ELO,n(t)=ALO,nexp[j2πfLO,nt+j2πfflc,LO,n(t)t+jφLO,n(t)],where ALO,n, fLO,n, fflc,LO,n(t), and φLO,n(t) are the amplitude, the center frequency, the frequency fluctuation, and the phase noise of the OLO injected into the nth channel, respectively.

    In the nth channel, an optical signal copy is mixed with the corresponding OLO in an optical CR comprising a 90° optical hybrid (HC) and two pairs of BPDs, where the specific frequency-domain segments of the incoming signal are downconverted to the baseband. Subsequently, the in-phase and the quadrature components of the downconverted signal are sampled and quantized by two low-speed but high-resolution ADCs after being filtered by two LPFs. The passband width of the LPFs determines the bandwidth of the channel, which slightly exceeds the channel spacing Δf. Note that the LPFs also act as anti-aliasing filters for the ADCs to eliminate out-of-band noise and interference.

    The IQ components of the signal slice in the nth channel are digitized by two low-speed ADCs and the received baseband signals can be derived as IFI,n(t)=KI,n×Re{sRF(t)·exp[j2π(f1+(n1)Δf)t]·exp[j2πΔfflc,n(t)t+jφn(t)]}hI,n(t),IFQ,n(t)=KQ,n×Im{sRF(t)·exp[j2π(f1+(n1)Δf)t]·exp[j2πΔfflc,n(t)t+jφn(t)]}hQ,n(t),where KI,n=Ac,n·ALO,n·GI,KQ,n=Ac,n·ALO,n·GQ,Δfflc,n(t)=fflc,c(t)fflc,LO,n(t),fn=fLO,nfc=f1+(n1)Δf,φn(t)=φc(t)φLO,n(t)+φlink,n(t).

    Here, the subscripts I and Q are the abbreviations for the in-phase and quadrature, respectively. The symbol denotes a time-domain convolution, hn(t) is the normalized impulse response of the optoelectrical link in the nth channel, G is the link gain, Δfflc,n(t) is the relative frequency fluctuation between the optical carrier and the OLO of the nth channel, φn(t) is the sum of the phase noise of the nth channel including the relative source phase noise and the link phase noise φlink,n(t) in the nth channel, Δf is the channel spacing of adjacent channels, and fn is the frequency spacing between the OLO of the nth channel and the optical carrier. The link gain G is mainly determined by the optical power, the responsivity of the detectors, and the gain of the trans-impedance amplifiers (TIAs) following the BPDs. In the formula, only the positive first-order modulation sideband is reserved, given that the mixings of the other sidebands with the OLO would exceed the bandwidth of the electrical LPFs. Note that the issue of IQ imbalance in zero-intermediate frequency (IF) mixing is not discussed here, because it can be compensated using a simple Gram-Schmidt orthogonalization method in the DSP unit.

    To simplify the formula expression, the IQ path responses are unified in representation, allowing for further modeling of the system in the frequency domain as follows: [IF1(f)IF2(f)IFn(f)]=[S(f+f1W1(t)(t))S(f+f2W2(t)(t))S(f+fnWn(t)(t))]·[H1(f)(f)H2(f)(f)Hn(f)(f)]+[N1(f)N2(f)Nn(f)],where Wn(t)(t)=Δfflc,n(t)+12πdφn(t)dt,Hn(f)(f)=Kn·+hn(t)exp(j2πft)dt.

    Here, S(f) and IF(f) are the Fourier transforms of the RF signal sRF(t) and the baseband signal IF(t), respectively. N(f) denotes the amplitude noises within each channel, including relative intensity noise (RIN) from the lasers, shot noise in optoelectronic conversion, thermal noise in the electrical devices, and quantization noise in the digitization process. Note that not all the amplitude noises are additive, but this will not affect the analysis of the non-idealities and the corresponding compensation methods. As shown in Eq. (5), the impairments induced by the non-idealities can be decoupled into two parts obviously: the stable system frequency response Hn(f)(f) and the time-variant impairment Wn(t)(t). The stable frequency response Hn(f)(f) can be estimated by feeding the system with a known reference waveform and comparing it with the output signal, which can be performed before the system initialization or at regular intervals. Note that the amplitude term Kn, which is determined by the optical carrier power, the OLO power, and the link gain, is merged into Hn(f)(f) and compensated collectively.

    However, the time-variant impairment term Wn(t)(t), which comprises both the frequency drifts and the phase noise, changes rapidly over time. For a typical DFB laser, the frequency fluctuation is observed to occur in less than 1  μs and the phase noise defined by a 3 dB linewidth is on the order of MHz. Therefore, in the proposed incoherent system, the Wn(t)(t) cannot be simply estimated through a one-time measurement or alleviated by utilizing a similar method reported in Ref. [19]. Therefore, real-time measurements for Wn(t)(t) in every channel are indispensable considering that each channel employs a free-running laser as a local oscillator. To this end, a calibration signal with a frequency range covering all channels is added to the system to record the time-variant impairments for all channels as shown in Fig. 1(a). The addition methods of the calibration signal can exploit common multiplexing techniques such as frequency-division multiplexing and polarization multiplexing. Note that the optical coherent detection devices should be adapted based on the multiplexing method employed. For example, the adoption of the polarization multiplexing method requires the use of polarization diversity coherent receivers (PDCRs).

    Once the time-invariant frequency-dependent impairment Hn(f)(f) and the time-variant term Wn(t)(t) are compensated, the IQ components of the signal slice are combined in each channel and the baseband signal as shown in Fig. 1(c) can be expressed as IFn(t)={sRF(t)·exp[j2π(f1+(n1)Δf)t]}sinc(Bcht),where the sinc(Bcht) is the impulse response of an ideal LPF with a bandwidth of Bch/2 [Bch denotes the channel bandwidth as shown in Fig. 1(b)]. Importantly, to ensure an effective reconstruction of the broadband incoming signal, there must always be an overlap region between adjacent channels and the incoming RF signal must always fall within the receiver’s reception frequency range. To this end, the channel spacing Δf, the channel bandwidth Bch, the minimum frequency of the incoming signal fmin, the maximum frequency of the incoming signal fmax, the frequency fluctuation of the nth OLO fflc,LO,n(t), and the relative frequency fluctuation between the optical carrier and the nth OLO Δfflc,n(t) must satisfy Bch>Δf+max{|fflc,LO,n(t)|n=1,N},f1Bch/2+|Δfflc,1|/2>fmin,fN+Bch/2|Δfflc,N|/2<fmax.

    After IQ combination, the signal slices are filtered using a rectangular filter, digitally upsampled, frequency-shifted back to their original positions, and finally summed to reconstruct the incoming broadband signal, with the data in the overlapping regions processed by averaging. This procedure can be expressed as sRF˜(t)=n=1NUpsample{IFn(t)}·exp(j2πfnt).

    Note that our scheme can be easily scaled to large channel counts and large bandwidth by simply increasing the number of free-running lasers. In addition, all the discussions above assume that all the channels are uniformly distributed in the frequency domain. However, another key feature of the proposed scheme is its high reconfigurability with low hardware cost, allowing for the reception frequency range to be flexibly reconfigured by arbitrarily tuning the frequencies of these OLO lasers. The reception bandwidth of each channel can also be adjusted by employing tunable electrical filters under the premise of adhering to the Nyquist theorem.

    3. EXPERIMENTAL SETUP AND RESULTS

    To demonstrate the viability and practicability of our scheme, a series of experiments including multi-format broadband signals reconstruction, dual-band microwave imaging, and the demonstration of system scalability are performed based on a setup as depicted in Fig. 2(a). The experimental setup comprises a PFI-enabled receiver unit and an electrical transmitter unit, considering that the RF signal generation is not the focus of our research.

    Experimental setup of demonstration of the PFI-enabled broadband receiver. (a) Experimental hardware setup. AWG, arbitrary waveform generator; MSG, microwave signal generator; BPF, band-pass filter; RF chain, the RF filter-amplifier-filter chain; EA, electrical attenuator; TCR, trihedral corner reflectors; EC, electrical coupler; CW laser, continuous-wave laser; DFB lasers, distributed feedback lasers; MZM, Mach–Zehnder modulator; CR, optical coherent receiver; LPF, low-pass filter; ADC, analog-to-digital converter; DSP, digital signal processing; EA, electrical amplifier; IF, intermediate frequency. Inset 1: structure of the RF chain. Inset 2: structure of the optical coherent receiver. (b) The relative spectrum positions setting of the signal reconstruction experiment (Ku-band region) and the dual-band microwave imaging. (c) The relative spectrum positions setting for the experiment of scalability demonstration.

    Figure 2.Experimental setup of demonstration of the PFI-enabled broadband receiver. (a) Experimental hardware setup. AWG, arbitrary waveform generator; MSG, microwave signal generator; BPF, band-pass filter; RF chain, the RF filter-amplifier-filter chain; EA, electrical attenuator; TCR, trihedral corner reflectors; EC, electrical coupler; CW laser, continuous-wave laser; DFB lasers, distributed feedback lasers; MZM, Mach–Zehnder modulator; CR, optical coherent receiver; LPF, low-pass filter; ADC, analog-to-digital converter; DSP, digital signal processing; EA, electrical amplifier; IF, intermediate frequency. Inset 1: structure of the RF chain. Inset 2: structure of the optical coherent receiver. (b) The relative spectrum positions setting of the signal reconstruction experiment (Ku-band region) and the dual-band microwave imaging. (c) The relative spectrum positions setting for the experiment of scalability demonstration.

    In the transmitter unit, broadband RF signals are first generated by an arbitrary waveform generator (AWG, Keysight M8190A) with a sampling rate of 12 GSa/s and then sent to different RF links to obtain the required RF signals with sufficient power and spectral purity in different bands. Specifically, to generate the C-band signal, the signal generated by the AWG is directly fed into the RF chain built by the canonical filter-amplifier-filter architecture. However, to obtain the Ku-band signal, the output of the AWG is first upconverted by employing an electrical mixer (Marki MM1-1140H) before being sent to the RF chain. The output signals of the RF chains are then either fed into the antenna for the microwave imaging experiment or directly attenuated by the electrical attenuator (EA) to implement closed-loop experiments or system frequency response premeasurements.

    In the receiver unit, the optical carrier with a power of 13 dBm and a linewidth of 1 kHz, is generated by a 1550 nm CW laser (TeraXion PureSpectrum-TNL) and then modulated by an incoming signal and a calibration signal through a null-biased Mach–Zehnder modulator (MZM, AFR AM20). Note that the calibration signal consists of two or three coherent tones generated by the different channels of a microwave signal generator (MSG, Sinolink SLFS2220T) and is frequency-division multiplexed with the incoming signal using an electrical coupler. The optical signal is then equally split into three copies and is sent to the signal ports of three optical CRs (Optilab DPCR-23-R). The CRs are also fed with the OLOs from three free-running DFB lasers with a linewidth of 1 MHz. In addition, the max relative frequency fluctuation between the optical carrier and the selected OLO lasers is 50 MHz under stable state. Note that the output power of each LO laser is attenuated to about 4 dBm to comply with the input optical power constraints of the CRs. In each coherent receiver, the +1st-order sideband of the optical signal is mixed with the corresponding OLO. The outputs of the CRs are then filtered by LPFs with a bandwidth of 1.8 GHz and digitized by three 8-bit ADCs with a sampling rate of 4 GSa/s. Note that the LPFs are designed with steep roll-offs to ensure a 60  dB out-of-band suppression ratio at the 2 GHz frequency to meet the Nyquist sampling theorem. Note also that the clocks of the AWG and the ADCs are locked to the same reference 10 MHz clock inside MSG to synchronize the whole system.

    To achieve optimal imaging, we compensate for the transmitter’s non-ideal frequency response by incorporating it into the receiver’s response using a unified closed-loop calibration. Alternatively, this can be achieved through pre-compensation of the transmitted waveform at the transmitter, but this requires a high-speed oscilloscope, and the pre-compensation may affect the evaluation of the algorithm’s performance.

    In addition, a digital algorithm for image interference cancellation must be implemented, given that the CRs used are not IQ mixers. Under this condition, the output of the nth channel can be expressed as x˜n(t)=Re[s(t)hrx,n(t)×exp(j2πfnt)]htx,n(t)=[s(t)hrx,n(t)]×exp(j2πfnt)htx,n(t)+[s(t)hrx,n(t)]*×exp(j2πfnt)htx,n(t).

    Here, for simplicity, the time-variant impairment term is omitted. As it can be seen, due to the non-IQ downconversion, the output of the CR in the nth channel comprises both the desired signal component within the frequency range [fn, fn+BLPF] and the image interference within the frequency range [fnBLPF, fn]. Here, BLPF is the bandwidth of the LPFs used. However, note that the desired component in the (n1)th channel becomes the image interference of the nth channel. Besides, our experimental setup is a cooperative signal reception system, where the generated signal is consistently band limited, given that proper RF filtering ensures sufficient out-of-band suppression. Therefore, there would be no image interference within the signal slice in the first channel by simply setting f1fmin, where fmin denotes the minimum frequency of the generated RF signal. Given these conditions, the image interference in the nth channel can be eliminated by subtracting the corresponding component included in the (n1)th channel, starting from n=2.

    A. Multi-format Broadband Signal Reconstruction

    To evaluate the performance of the proposed PFI-enabled reception scheme, the experiments of multi-format broadband signal reception and digital reconstruction are carried out by sending the signal generated in the transmitter unit to the closed-loop link. Four typical waveforms in radar and communication applications are tested, including LFM, NLFM [39], QPSK, and 16QAM signals. The LFM signal and NLFM signal are individually defined as sLFM(t)=exp(2πfct+πBTt2),sNLFM(t)=exp[2πfct+πBTt2πBTn=1NKnπncos(2πntT)],where fc is the carrier frequency, B is the bandwidth, T is the pulse width (T/2tT/2), Kn denotes the nth coefficient of the Fourier series, and N is the total number of the coefficients. The NLFM signal is designed to have low sidelobe characteristics with an S-shaped frequency function of time, including peak-to-sidelobe level ratio (PSLR) and integrated sidelobe level ratio (ISLR). Note that the NLFM signal cannot be processed through the most used deramp reception scheme in MWP radar systems.

    The experiment is conducted using a two-channel frequency-interleaving reception scheme, and the relative spectrum positions are illustrated in the Ku-band region of Fig. 2(b). The center frequency differences between the optical carrier and the two OLOs are set to 13.6 GHz and 16.8 GHz, while the frequencies of the calibration signals are 13.8 GHz and 16.6 GHz, respectively to facilitate subsequent signal separation. Consequently, the reception frequency ranges of the two channels are set as 13.95 GHz to 15.25 GHz and 15.15 GHz to 16.45 GHz, respectively, corresponding to a channel overlap frequency range of 100 MHz. The overlap frequency range can be reduced by using more stable LO lasers, and for longer-term frequency drift (e.g., 1  day), we believe regular feedback calibration can effectively address this issue.

    The broadband signal is initially generated by the AWG and mixed with a 17.8 GHz sine signal from the MSG. The center frequency of both the generated LFM signal and NLFM signal is 15.2 GHz, with a bandwidth of 2.4 GHz and a pulse width of 100 μs. Moreover, the modulation rate of the generated QPSK and 16QAM signal is 1.5 GBaud (roll-off factor, 0.5) with a carrier frequency of 15.2 GHz and a pulse width of 40 μs. The RF signals are amplified by the RF chain, adjusted to a power level of 0 dBm, and then attenuated to 60  dBm, which is the basic level of actual echo signal.

    For the precise compensation of the incoming signal, the amplitude-frequency and phase-frequency responses of the system are first measured by setting the AWG to produce an ideal chirp signal pulse. Note that the frequency-dependent non-idealities of both the transmitter unit and the receiver unit are included in the measured results. The frequency-dependent responses of each channel are obtained by comparing the signal slices with the ideal chirp waveform after the time-variant phase deviations are eliminated. The measured normalized amplitude-frequency and phase-frequency responses of the two channels in our system are shown in Fig. 3. To mitigate the measurement deviations induced by the system noise, 20 pulses are exploited to perform the measurement at one time and the averaged results are adopted as the valid system frequency-dependent responses. After the initial measurement, additional measurements are conducted at the following 5 min and 30 min moments based on the same method. Both the amplitude-frequency curves and phase curves obtained at different moments present a high degree of consistency, which validates our stability assumption of the frequency-dependent responses in Section 2.

    System frequency-response measurements. (a) The amplitude-frequency response of the first channel. (b) The phase-frequency response of the first channel. (c) The amplitude-frequency response of the second channel. (d) The phase-frequency response of the second channel. The gray “background” is the superposition of 20 results curves measured by feeding 20 pulses with a known waveform into the system at the T0 moment. The red curve is the averaged result of the gray curves and is adopted as the reliable system response. The blue and green curves are the averaged response recorded at the moments of the T0+5 min and T0+30 min.

    Figure 3.System frequency-response measurements. (a) The amplitude-frequency response of the first channel. (b) The phase-frequency response of the first channel. (c) The amplitude-frequency response of the second channel. (d) The phase-frequency response of the second channel. The gray “background” is the superposition of 20 results curves measured by feeding 20 pulses with a known waveform into the system at the T0 moment. The red curve is the averaged result of the gray curves and is adopted as the reliable system response. The blue and green curves are the averaged response recorded at the moments of the T0+5  min and T0+30  min.

    After the signal slices are compensated by utilizing the premeasured frequency responses and the digitized calibration signal, the signal reconstructions of multi-format signals can be achieved through the operations of upsampling, frequency shifting, and summing, with the data in the overlap regions processed by averaging. To evaluate the performance of our PFI-enabled reception scheme, the qualities of the reconstructed signals are analyzed and compared to the original signals as shown in Fig. 4. As illustrated in Figs. 4(a-i)–4(d-i), all the reconstructed signals—whether constant-envelope modulated signals or non-constant-envelope modulated signals—precisely align with the original RF signals, which indicates that our scheme is not constrained by the formats of the incoming signals. Note that the non-idealities of the whole closed-loop link are taken into consideration and thus the original signal here is equivalent to the corresponding ideal signals. The normalized mean squared errors (NMSEs) of the reconstructed LFM signal, NLFM signal, QPSK signal, and 16QAM signal are 1.52×102 (equivalent SNR, 18.18 dB), 1.58×102 (equivalent SNR, 18.01 dB), 1.08×102 (equivalent SNR, 19.68 dB), and 1.61×102 (equivalent SNR, 17.92 dB), respectively. Note that the equivalent SNR calculated here comprises the influence of the system noise and the signal distortion. Moreover, we believe that the SNR performance of our system can be further improved by increasing the BPDs’ saturating power, replacing the OLOs with lower-noise ones, and improving modulation efficiency. The normalized power spectra of the reconstructed signals are shown in Figs. 4(b-ii)–4(d-ii), where the overlap regions indicate the high amplitude consistency between the two channels. Please note that for better identifying the frequency range of each channel and the overlap region, the figure retains the data in the overlap regions of both channels, which are plotted with different colors. In addition, for the LFM and NLFM signals, we employ a common time-frequency two-dimensional filtering method in the overlap regions to extract the phase consistency between the two channels as shown in Figs. 4(a-v) and 4(b-v). The calculated inter-channel root-mean-square phase difference (RmsPD) in the overlap region for LFM and NLFM signals is 1.61° and 1.76°, which indicates a high degree of phase consistency between the two channels, thereby validating the effectiveness and accuracy of our DSP algorithm.

    The experiment results of multi-format broadband signals reconstruction, which are widely used in radar and communication applications. RmsPD, root-mean-square phase difference; EVM, error vector magnitude. (a) The results and analyses of the reconstructed LFM signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the time-frequency plot of the reconstructed signal; (iv) the pulse-compression curve of the reconstructed signal; (v) the inter-channel phase difference in the overlap region. (b) The results and analyses of the reconstructed NLFM signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the time-frequency plot of the reconstructed signal; (iv) the pulse-compression curve of the reconstructed signal; (v) the inter-channel phase difference in the overlap region. (c) The results and analyses of the reconstructed QPSK signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the constellation diagram of the reconstructed signal. (d) The results and analyses of the reconstructed 16QAM signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the constellation diagram of the reconstructed signal.

    Figure 4.The experiment results of multi-format broadband signals reconstruction, which are widely used in radar and communication applications. RmsPD, root-mean-square phase difference; EVM, error vector magnitude. (a) The results and analyses of the reconstructed LFM signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the time-frequency plot of the reconstructed signal; (iv) the pulse-compression curve of the reconstructed signal; (v) the inter-channel phase difference in the overlap region. (b) The results and analyses of the reconstructed NLFM signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the time-frequency plot of the reconstructed signal; (iv) the pulse-compression curve of the reconstructed signal; (v) the inter-channel phase difference in the overlap region. (c) The results and analyses of the reconstructed QPSK signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the constellation diagram of the reconstructed signal. (d) The results and analyses of the reconstructed 16QAM signals. (i) The temporal waveform of the reconstructed signal compared to the original signal; (ii) the power spectra of the compensated signal for two channels; (iii) the constellation diagram of the reconstructed signal.

    The time-frequency diagrams of radar signals and the constellation diagrams of communication signals provide a more intuitive representation of signal quality [Figs. 4(a-iii)–4(d-iii)]. The darker regions, existing in the frequency range of 15.15 GHz to 15.25 GHz of the time-frequency map, indicate a 3 dB SNR improvement in the overlap frequency range. The SNR improvement is attributed to the noncorrelation of noise components and the high amplitude-phase consistency of signal components between the adjacent channels. We also observed noise reduction regions along with the time-frequency curves of the signals, profiting from the phase noise elimination procedure utilizing the calibration signal, but with a limited frequency range. The limited reduction frequency range is attributed to the high phase noise level of the local oscillator lasers and the limited frequency-division multiplexing space for the calibration signal. We also attribute the observed slight rotational distortion in the constellation diagrams of Figs. 4(c-iii) and 4(d-iii) to the residual far-end phase noise. We believe that this problem can be overcome by replacing the local oscillator lasers with narrower linewidth ones, allocating larger frequency-division multiplexing space, or utilizing alternative multiplexing methods.

    To further verify the theoretic potential of our scheme in practical applications, we apply pulse compression to the reconstructed LFM and NLFM signals. The pulse compression curves are depicted in Figs. 4(a-iv) and 4(b-iv). For the LFM signal, the measured impulse response width (IRW), PSLR, and ISLR are 369 ps, 13.20 dB, and 9.75 dB, respectively, closely matching the theoretical values of 368.75 ps, 13.26 dB, and 9.76 dB. Similarly, for the NLFM signal, the measured IRW, PSLR, and ISLR are 531.75 ps, 38.12 dB, and 31.45 dB, respectively, matching the theoretical values of 531.75 ps, 40.09 dB, and 34.43 dB. Note that the NLFM signal achieves a 25 dB improvement in sidelobe suppression compared to the LFM signal, at the cost of broadening the main lobe by a factor of 1.45. For communication signals, we calculate the averaged error vector magnitudes (EVMs) and bit error rates (BERs) over ten 40-µs-long pulses with the corresponding constellation diagrams shown in Figs. 4(c-iii) and 4(d-iii). The calculated EVM and BER for the QPSK signal are 9.11% and 0, respectively. For the 16QAM signal, the calculated EVM is 9.91%, and the BER is 6.78×105. All the calculated BERs are well below the 7% overhead hard-decision forward error correction (HD-FEC) threshold of 3.8×103.

    B. Dual-Band Microwave Imaging

    To demonstrate the performance of our PFI-enabled reception scheme in practical radar application, the experiments of inverse synthetic-aperture radar (ISAR) imaging for two trihedral corner reflectors (TCRs) are carried out using both LFM signal and NLFM signal. The two TCRs are placed on the rotational platform with an initial range-projected distance of 45 cm and an initial cross-range distance of 37 cm. Note that the reconfigurability of our scheme is also demonstrated by performing microwave imaging in the Ku-band and C-band simultaneously. The experimental setup is shown in Fig. 2, where the signal generated in the transmitter unit is sent to the antennas. The spectrum relationship is depicted in Fig. 2(b), the Ku-band setup is the same as the setting in Fig. 2(a), and the reception frequency range of the added C-band is 6.4 GHz to 7.7 GHz with a calibration signal located at 6.25 GHz. The signal bandwidths of the generated RF signals are 2.4 GHz for the Ku-band and 1.3 GHz for the C-band with the corresponding carrier frequencies of 15.2 GHz and 7.05 GHz, respectively. In addition, the pulse width, pulse repetition interval (PRI), and the speed of the rotational platform are set as 100 μs, 500 μs, and 30 deg/s, respectively. Due to the limited working band of the antennas in the lab, the antennas for different bands are different and placed side by side, which may lead to slight relative position differences in the imaging results but would not affect the analysis of the results. The signal reconstruction in the digital domain follows the method illustrated in Section 2, except that the frequency-dependent responses of the antennas cannot be measured in a closed-loop response measurement. This dilemma can be overcome using the external calibration method, which is commonly employed in practical radar systems.

    The spectra of the reconstructed dual-band radar echoes of the LFM signal and the NLFM signal are shown in Fig. 5(a-i) and Fig. 5(b-i), respectively. The high consistencies of the spectral power in the overlap regions of the Ku-band indicate the effectiveness of our compensation method in practical application. The periodic notches in the spectra are induced by the destructive interference between the echoes from the two TCRs. The one-dimensional (1D) range profiles are obtained by applying pulse compression to the echo pulses as shown in Figs. 5(a-ii) and 5(b-ii). For the LFM signals, a 3 dB width of 5.71 cm for the Ku-band and 10.88 cm for the C-band are achieved, both consistent with the theoretical values. The PSLR measured in the Ku-band result is 14.39 dB, while the measured PSLR for the C-band is unreliable due to the pronounced environmental scattering interference. We attribute the interference in the C-band to the limited work band of microwave-absorbing material designed for the Ku-band and the stronger scattering characteristics of other devices in the lab within the C-band. Similarly, for the NLFM signals, a 3 dB width of 8.55 cm along with a PSLR of 32.05 dB for the Ku-band and a 3 dB width of 15.30 cm for the C-band are achieved. Two point-targets can be identified in the 2D images and the low sidelobe advantage of the NLFM signal is intuitively embodied as shown in Figs. 6(b-iv) and 6(b-v), which is calculated in a coherent integration time of 0.15 s (i.e., 300 pulses). Note that a zero-Doppler-filtering is performed in the range-Doppler domain for C-band echo data to exclude interference from stationary targets. In addition, the azimuth slices of one TCR are depicted in Figs. 5(a-iii) and 5(b-iii). For the LFM signals, a 3 dB width of 11.27 cm for the Ku-band and 23.94 cm for the C-band are achieved in the azimuth direction. Similarly, a 3 dB width of 11.43 cm for the Ku-band and 24.02 cm for the C-band are achieved for the NLFM signals in the azimuth direction. The azimuth resolutions obtained are consistent with the theoretical values calculated by the equation ρa=λ/2(w×Δt), where λ is the center wavelength of the transmitting RF signals, w is the angular speed of the rotational platform, and Δt is the integration time.

    The experiment results of the dual-band microwave imaging of two TCRs. (a) LFM signal experiment results. (i) The spectra of the compensated signals of three channels. (ii) The range profiles of the two bands. (iii) The azimuth slice of one TCR. (iv) The ISAR image of C-band. (v) The ISAR image of Ku-band. (b) NLFM signal experiment results. (i) The spectra of the compensated signals of three channels. (ii) The range profiles of the two bands. (iii) The azimuth slice of one TCR. (iv) The ISAR image of C-band. (v) The ISAR image of Ku-band.

    Figure 5.The experiment results of the dual-band microwave imaging of two TCRs. (a) LFM signal experiment results. (i) The spectra of the compensated signals of three channels. (ii) The range profiles of the two bands. (iii) The azimuth slice of one TCR. (iv) The ISAR image of C-band. (v) The ISAR image of Ku-band. (b) NLFM signal experiment results. (i) The spectra of the compensated signals of three channels. (ii) The range profiles of the two bands. (iii) The azimuth slice of one TCR. (iv) The ISAR image of C-band. (v) The ISAR image of Ku-band.

    The results of the image interference suppression in the second channel. (a) The spectrum slices of the received LFM echo in the second channel with (red) and without (blue) image suppression after a time-frequency 2D filtering. (b) The time-frequency diagram of the LFM echoes with (left) and without (right) image suppression. (c) The time-frequency diagram of the NLFM echoes with (left) and without (right) image suppression.

    Figure 6.The results of the image interference suppression in the second channel. (a) The spectrum slices of the received LFM echo in the second channel with (red) and without (blue) image suppression after a time-frequency 2D filtering. (b) The time-frequency diagram of the LFM echoes with (left) and without (right) image suppression. (c) The time-frequency diagram of the NLFM echoes with (left) and without (right) image suppression.

    C. Scalability Demonstration

    In the final experiment, we increased the interleaving channel counts to three to validate the scalability of our PFI scheme. Except for the two aspects illustrated below, the rest of the experimental settings remain the same as the dual-band imaging experiment. First, in the transmitter unit, only the bottom link is retained to generate the Ku-band broadband signal. The frequency range of the AWG-generated baseband signal spans from 0.6 GHz to 4.6 GHz and is upconverted to 13.2 GHz to 17.2 GHz by mixing it with a 17.8 GHz sinewave, with only the negative first-order sideband reserved. The generated RF signal is then amplified to 0 dBm and transmitted into free space by the transmitting antenna. Second, in the receiver unit, the relative spectral positions of the OLOs and reception channels are depicted in Fig. 2(c). The two tones of the calibration signal are set at 13.05 GHz and 17.35 GHz, and the former can be received simultaneously by both CH1 and CH2. In CH2, a simple time-frequency peak filtering algorithm is utilized to separate the calibration signal component from the incoming signal component.

    As then the number of interleaving channels increases to three with a total acquisition bandwidth of 4 GHz, the image interference problem in CH2 occurs due to the non-IQ mixing of the optical CR used. To overcome this problem, the image interference cancellation algorithm discussed at the beginning of this section is implemented after the frequency-dependent responses are compensated. The performance of our image cancellation algorithm is shown in Fig. 6. By applying time-frequency domain filtering to the LFM signal echoes, we calculated a 23.09 dB image suppression ratio, which also indicates the accurate compensation ability of our methods. The time-frequency diagrams in Fig. 6(b) and Fig. 6(c) illustrate the image suppression performance for the LFM echoes and the NLFM echoes, respectively. Note that the pulse compression technique can further eliminate the impairments induced by image interference.

    Figures 7(a-i) and 7(b-i) show the spectra of the reconstructed signals from the three interleaving channels, where good channel consistency and the absence of noticeable image interference are observed. To illustrate the effectiveness of our scheme in radar resolution enhancement, we apply pulse compression to both the reconstructed signal and the compensated signal slice in CH1. The pulse-compression curves are shown in Figs. 7(a-ii) and 7(b-ii). For the reconstructed LFM signal, a 3 dB width of 3.29 cm is achieved compared to a 3 dB width of 10.29 cm for the signal slice in CH1, corresponding to a 3.13× improvement in resolution. For the reconstructed NLFM signal, a 3 dB width of 4.70 cm is achieved, while only a 3 dB width of 12.56 cm is reached for the signal slice in CH1.

    The experiment results of the system’s scalability demonstration. (a) LFM signal experiment results. (i) The power spectra of the compensated signals of three channels. (ii) The range profiles after pulse compression utilizing the reconstructed broadband signal and the signal slice in first channel. (iii) The ISAR image of the signal slice in first channel. (iv) The ISAR image of the reconstructed signal. (b) NLFM signal experiment results. (i) The power spectra of the compensated signals of three channels. (ii) The range profiles after pulse compression utilizing the reconstructed broadband signal and the signal slice in first channel. (iii) The ISAR image of the signal slice in first channel. (iv) The ISAR image of the reconstructed signal.

    Figure 7.The experiment results of the system’s scalability demonstration. (a) LFM signal experiment results. (i) The power spectra of the compensated signals of three channels. (ii) The range profiles after pulse compression utilizing the reconstructed broadband signal and the signal slice in first channel. (iii) The ISAR image of the signal slice in first channel. (iv) The ISAR image of the reconstructed signal. (b) NLFM signal experiment results. (i) The power spectra of the compensated signals of three channels. (ii) The range profiles after pulse compression utilizing the reconstructed broadband signal and the signal slice in first channel. (iii) The ISAR image of the signal slice in first channel. (iv) The ISAR image of the reconstructed signal.

    4. SUMMARY

    A photonic-frequency-interleaving (PFI)-enabled broadband receiver scheme that exploits multiple free-running lasers as multi-wavelength optical local oscillators (OLOs) is proposed and demonstrated. Leveraging the broadband operation capability of photonics devices and the insensitivity of the frequency-interleaving technique to clock jitter and channel mismatch, the PFI-enabled reception scheme promises the potential to overcome the accuracy and bandwidth tradeoff in the conventional time-interleaving reception paradigm. By employing independent tunable lasers as OLOs, our approach eliminates the need for optical frequency combs, optical filters, and optical amplifiers, which are highly relied on in comb-based PFI schemes. This greatly simplifies the system implementation while enhancing the system’s reconfigurability, scalability, and integration potential. For the precise reconstruction of the incoming signal, a compensation method utilizing the premeasured system’s frequency responses and a multiplexed calibration signal is proposed and evaluated in a two-channel interleaving experiment for multi-format broadband signals. The reconfigurability and practical performance of our scheme in radar application are validated in a dual-band ISAR experiment. Finally, the scalability of our scheme is further demonstrated by performing a three-channel interleaving ISAR experiment with a total acquisition bandwidth increased to 4 GHz. The outstanding performance validates the practicability of our PFI-enabled reception scheme, which paves the way for the use of PFI in next-generation RF receivers in sensing, communication, and electronic countermeasure applications.

    References

    [1] M. Skolnik. An introduction and overview of radar. Radar Handbook, 3, 1.1-1.24(2008).

    [2] M. Agiwal, A. Roy, N. Saxena. Next generation 5G wireless networks: a comprehensive survey. Commun. Surveys Tuts., 18, 1617-1655(2016).

    [3] S. Pan, J. Yao. Photonics-based broadband microwave measurement. J. Lightwave Technol., 35, 3498-3513(2017).

    [4] X. Zou, B. Lu, W. Pan. Photonics for microwave measurements. Laser Photon. Rev., 10, 711-734(2016).

    [5] L. Zheng, M. Lops, Y. C. Eldar. Radar and communication coexistence: an overview: a review of recent methods. IEEE Signal Process Mag., 36, 85-99(2019).

    [6] G. Serafino, F. Scotti, L. Lembo. Toward a new generation of radar systems based on microwave photonic technologies. J. Lightwave Technol., 37, 643-650(2019).

    [7] P. Ghelfi, F. Laghezza, F. Scotti. A fully photonics-based coherent radar system. Nature, 507, 341-345(2014).

    [8] C. Deakin, Z. Liu. Dual frequency comb assisted analog-to-digital conversion. Opt. Lett., 45, 173-176(2020).

    [9] J. Capmany, D. Novak. Microwave photonics combines two worlds. Nat. Photonics, 1, 319-330(2007).

    [10] J. Yao. Microwave photonics. J. Lightwave Technol., 27, 314-335(2009).

    [11] S. Pan, Y. Zhang. Microwave photonic radars. J. Lightwave Technol., 38, 5450-5484(2020).

    [12] A. Khilo, S. J. Spector, M. E. Grein. Photonic ADC: overcoming the bottleneck of electronic jitter. Opt. Express, 20, 4454-4469(2012).

    [13] X. Ye, F. Zhang, Y. Yang. Photonics-based radar with balanced I/Q de-chirping for interference-suppressed high-resolution detection and imaging. Photon. Res., 7, 265-272(2019).

    [14] B. Jalali, A. S. Bhushan, F. Coppinger. Photonic time-stretch: a potential solution for ultrafast A/D conversion. International Topical Meeting on Microwave Photonics, 197-198(1998).

    [15] T. R. Clark, J. U. Kang, R. D. Esman. Performance of a time- and wavelength-interleaved photonic sampler for analog-digital conversion. IEEE Photon. Technol. Lett., 11, 1168-1170(1999).

    [16] W. Han, Z. Liu, X. Xu. Photonic RF channelization applications of microcombs. IEEE J. Sel. Top. Quantum Electron., 30, 7600417(2024).

    [17] N. K. Fontaine, R. P. Scott, L. Zhou. Real-time full-field arbitrary optical waveform measurement. Nat. Photonics, 4, 248-254(2010).

    [18] G. Gao, L. Lei. Photonics-based broadband RF spectrum measurement with sliced coherent detection and spectrum stitching technique. IEEE Photon. J., 9, 5503111(2017).

    [19] J. Yang, R. Li, Y. Dai. Wide-band RF receiver based on dual-OFC-based photonic channelization and spectrum stitching technique. Opt. Express, 27, 33194-33204(2019).

    [20] A. S. Bhushan, P. V. Kelkar, B. Jalali. 130-GSa/s photonic analog-to-digital converter with time stretch preprocessor. IEEE Photon. Technol. Lett., 14, 684-686(2002).

    [21] N. Qian, L. Yu, J. Chen. Influence of the demultiplexer on channel-interleaved photonic analog-to-digital converters. IEEE Photon. J., 12, 5502110(2020).

    [22] G. Yang, W. Zou, L. Yu. Influence of the sampling clock pulse shape mismatch on channel-interleaved photonic analog-to-digital conversion. Opt. Lett., 43, 3530-3533(2018).

    [23] S. Xu, X. Zou, B. Ma. Deep-learning-powered photonic analog-to-digital conversion. Light Sci. Appl., 8, 66(2019).

    [24] Y. Dai, K. Xu, X. Xie. Broadband photonic radio frequency (RF) channelization based on coherent optical frequency combs and polarization I/Q demodulation. Sci. China Technol. Sci., 56, 621-628(2013).

    [25] J. Ding, Y. Wu, H. Yang. Wideband image-reject RF channelization based on soliton microcombs. APL Photon., 8, 090801(2023).

    [26] X. Xu, M. Tan, J. Wu. Broadband photonic RF channelizer with 92 channels based on a soliton crystal microcomb. J. Lightwave Technol., 38, 5116-5121(2020).

    [27] D. Onori, P. Ghelfi, J. Azaña. A 0–40 GHz RF tunable receiver based on photonic direct conversion and digital feed-forward lasers noise cancellation. J. Lightwave Technol., 36, 4423-4429(2018).

    [28] X. Xu, J. Wu, T. G. Nguyen. Broadband RF channelizer based on an integrated optical frequency Kerr comb source. J. Lightwave Technol., 36, 4519-4526(2018).

    [29] C. Deakin, Z. Liu. Frequency interleaving dual comb photonic ADC with 7 bits ENOB up to 40 GHz. Conference on Lasers and Electro-Optics (CLEO), 1-2(2022).

    [30] D. Fang, A. Zazzi, J. Müller. Optical arbitrary waveform measurement using silicon photonic slicing filters. J. Lightwave Technol., 40, 1705-1717(2021).

    [31] D. Drayss, D. Fang, C. Füllner. Non-sliced optical arbitrary waveform measurement (OAWM) using soliton microcombs. Optica, 10, 888-896(2023).

    [32] K. Igarashi, Y. Kawabata, N. Urakawa. Measuring complex field waveforms of quadrature amplitude modulation optical signals using a spectrally slicing-and-synthesizing coherent optical spectrum analyzer. Opt. Express, 28, 21560-21570(2020).

    [33] H. Xue, C. Song, Z. Zheng. Segment reception of broadband millimeter-wave for reducing ADC bandwidth requirements. J. Lightwave Technol., 42, 7619-7627(2024).

    [34] P. Vaidyanathan. Quadrature mirror filter banks, M-band extensions and perfect-reconstruction techniques. IEEE ASSP Mag., 4, 4-20(1987).

    [35] G. Ding, C. Dehollain, M. Declercq. Frequency-interleaving technique for high-speed A/D conversion. International Symposium on Circuits and Systems (ISCAS)(2003).

    [36] N. K. Fontaine. Spectrally-sliced coherent receivers for THz bandwidth optical communications. IEEE 27th Convention of Electrical and Electronics Engineers in Israel, 1-4(2012).

    [37] H. Othman, X. Ouyang, C. Antony. Spectrally-sliced coherent receiver utilizing a gain-switched optical frequency comb. J. Lightwave Technol., 41, 5262-5274(2023).

    [38] L. Chang, S. Liu, J. E. Bowers. Integrated optical frequency comb technologies. Nat. Photonics, 16, 95-108(2022).

    [39] Q. Xie, H. Zeng, Z. Mo. A two-step optimization framework for low sidelobe NLFM waveform using Fourier series. IEEE Geosci. Remote Sens. Lett., 19, 4020905(2022).

    Jianwei Liu, Ruixuan Wang, Jiyao Yang, Weichao Ma, Henan Zeng, Chenyu Liu, Wen Jiang, Xiangpeng Zhang, Qinyu Xie, Wangzhe Li, "Photonic-frequency-interleaving-enabled broadband receiver with high reconfigurability and scalability," Photonics Res. 13, 395 (2025)
    Download Citation