I. Introduction
CHANNELIZED receivers are key building blocks of modern radio frequency (RF) systems including modern electronic warfare systems, deep space tracking and telecommunications [
1,
2,
3,
4], as they enable the detection of wideband RF spectra with digital-compatible devices that have much lower bandwidths (such as generic analog-to-digital converters at hundreds of MHz). However, multi-format, multi-frequency and wideband signals pose challenges to present electronic processing systems. While electronic RF channelizers (generally formed by filter banks) are subject to the bandwidth bottleneck, photonic approaches are promising since they can offer ultra-large bandwidths, low transmission loss and strong immunity to electromagnetic interference.
Photonic RF channelizers can generally be divided into three categories. The first category depends on the spectral-to-spatial conversion performed by a diffraction grating or an integrated Fresnel lens, yielding multiple parallel channels containing different frequency components in free space [
5,
6]. This method, while advantageous in terms of instantaneous bandwidth, poses limitations in the system’s overall footprint and resolution. The second category is more compact, which employs the Vernier effect between a multi-wavelength source and a periodic filter array to achieve high-resolution wideband RF spectral channelization. Several types of devices and platforms have been utilized for this purpose, such as discrete laser arrays [
7], parametric processes in nonlinear fiber [
8], cascaded electro-optic modulators [
9] and so on. However, the above approaches are restricted in channel number, spectral resolution, and compatibility for monolithic integration. The last solution is based on a frequency-to-time mapping, where the RF spectra are mapped to the time domain by using wavelength scanning or frequency shifting [
10]. Each wavelength is marked with a specific channel at the corresponding time slot, and only one photodetector (PD) is required to obtain the RF signal at a fast scan rate. Yet those approaches face limitations in one form or another — the scanning frequency step determines final channel number and operating bandwidth, even the serial detected results are non-contiguous.
Recently, microcombs, especially CMOS-compatible microcombs [
11], have shown unique advantages over conventional mode-locked fiber combs [
12] and electro-optical combs [
9] as they provide massively coherent wavelength channels at the chip-scale size and have proven to be widely used in microwave photonics [
13,
14,
15].
Here, we first leverage the polarization division of integrated photonics for RF channelization, and report a dual-polarization photonic RF channelizer, achieved with two MRRs with slightly different FSRs at ~49 GHz. The first MRR is used to generate optical comb lines (over 80 in the C band and 20 were used in this work), while the second acts as dual narrowband notch filters in two polarization states to slice the RF spectrum. The high Q passive MRR offers narrow resonance linewidths of 144 MHz (TE) and 163 MHz (TM), enabling high-resolution RF channelization and thus lowered requirements of subsequential analog-to-digital converters (ADCs) for digital processing; in addition, with tailored FSR mismatch between the active and passive MRRs, the RF channelization steps (~163 MHz for TE and ~117 MHz for TM) are closer to the slicing resolution, leading to an improved channel crosstalk of ~12 dB. Most importantly, the use of dual polarization modes doubled the number of channels (40 in total, 20 for each polarization state) and instantaneous bandwidth (3.1 GHz for TE and 2.2 GHz for TM in this work) in contrast to those using a single polarization mode, with wideband operation verified via thermal tuning of the passive MRR. This approach explores the polarization division of optics using integrated devices, further demonstrating the potentials of photonic channelizers for wideband RF signal processing.
II. Principle
Figure 1 depicts the schematic of the 40-channel dual-polarization RF channelizer that consists of three modules. The first module achieves microcomb generation and flattening, where an active MRR is pumped by a continuous-wave (CW) laser and amplified by an erbium-doped fiber amplifier (EDFA) to excite intracavity parametric oscillations. The MRR features high Q-factor, high nonlinear coefficients and designed anomalous dispersion, providing sufficient parametric gain to generate a Kerr frequency comb. Here, we utilize the soliton crystal combs and perform spectral shaping by a commercial WaveShaper to achieve equalized channel power.
Assuming the pre-shaped lines are generated with a spacing of
δOFC, the optical frequency of the
kth (
k = 1, 2, 3, …, 20) comb line can be written as
where
fOFC(1) is the frequency of the first comb line.
In the second module, the flattened combs are fed into an electro-optic phase modulator, where the broadband RF spectra are multicast onto each wavelength channel. Next, the copied RF signals are sliced into segments by a dual-polarization passive MRR with FSRs of δMRR-TE and δMRR-TM for TE and TM polarization, respectively, where the channel resolution is determined by the 3 dB bandwidth of the TE- and TM-polarization resonances, denoted as ∆fTE and ∆fTM.
The
kth centre frequency of the passive MRR’s resonance follows
where
fMRR-TE(1) and
fMRR-TE(1) are the frequency of the first filtering transmission lines of two polarizations,
δMRR-TE and
δMRR-TM denote the FSRs of passive MRR at TE- and TM-polarization.
The detailed mechanism of the dual polarization RF photonic channelizer is described in right part of
Figure 1. For each channel, the microcomb is phase-modulated by RF signals to produce counter-phase sidebands whose offset angle from the TE polarization is marked as
θ (
Figure 1(i)). The components of modulated signal falling in TE- and TM-planes are filtered by the orthogonally-polarized notch filters (i.e., the TE- and TM-resonances of the passive MRR), respectively (
Figure 1(ii)), which break the balance between the inverted sidebands and enable the phase-to-intensity modulation conversion. Finally, the dual polarization modes are separated by a polarization beam splitter (PBS), followed by wavelength-division demultiplexers to separate the wavelength channels for photodetection. After parallel detection, the channelized RF signals are each centered at
fRF-TE(
k) (or
fRF-TM(
k)) with a spectral bandwidth of ∆
fTE (or ∆
fTM), within the operation bandwidth of generic ADCs (
Figure 1(iii)).
Therefore, the progressive RF centre frequencies of the RF spectra on the TE- and TM-channel, respectively, are given by
where
fRF-TE(k) and
fRF-TM(k) are the
kth channelized RF centre frequencies of TE- and TM-channel, [
fMRR-TE(1) –
fOFC(1)] and [
fMRR-TM(1) –
fOFC(1)] denote the relative spacing between the first comb and the adjacent dual polarization resonances, namely, the offset of the channelized RF frequency. And (
δMRR-TE –
δOFC) and (
δMRR-TM –
δOFC) represent the channelized RF frequency step between adjacent wavelength channels for TE-and TM-channel, respectively.
We further analyzed the operation of the dual-polarization passive MRR using the Jones matrix [
16], and the through-port transmission can be written by
where
TTE and
TTM are the through-port transfer functions of the passive MRR given by
where
t is the transmission coefficient between the bus waveguide and the passive MRR,
a is the round-trip transmission factor,
ϕTE = 2π
L ×
neff_TE /
λ and
ϕTM = 2π
L ×
neff_TM /
λ are the single-pass phase shifts of TE and TM modes, respectively,
L denotes the round-trip length,
neff_TE and
neff_TM denote the effective indices of TE and TM modes, and
λ represents the wavelength.
The phase-modulated optical signal can be given as
where
E0 denotes the modulated optical signal,
θ is the polarization angle relative to the TE-axis, so the output field of the passive MRR can be written as
According to the above equation, the optical power of TE- and TM-polarized optical signals are proportional to cos
2θ and sin
2θ, respectively. Hence the extinction ratio between the channelized RF spectral segments of TE- and TM-resonances is given by
ER(
θ) can be continuously adjusted by changing
θ, limited only by the performance of the polarization controller. Moreover, cot
2θ can infinitely approach 0 or infinity as
θ approaches 90° or 0°, thus theoretically an ultra-large dynamic tuning range of the extinction ratio for dual polarization states can be expected. Specifically, when
θ = 45°, the amplitude of TE and TM polarization is equal.
III. Experimental results
Both active and passive MRRs are fabricated on a high-index doped silica glass platform through a CMOS compatible fabrication process [
17]. The two MRRs featured similar characteristics with radii of ~592 μm, corresponding to a FSR of ~0.4 nm (~49 GHz), with Q factors over one million. To demonstrate the analysis above regarding the dual-polarization passive MRR, we measured its transmission spectra with different polarization angles
θ. As
θ varies from 0° to 90°, the transmission spectra evolution of the passive MRR (TE and TM resonances) and extinction ratios between the orthogonally polarized resonances are shown in
Figure 2(a-b). We note that the polarization angle
θ is deferred by comparing measured and simulated results. As for the extinction ratio of dual modes (TE2-to-TM1 and TE2-to-TM2), the former varies from 34 dB to -35 dB, which indicates a continuously tunable extinction ratio of over 69 dB. Note that we actually used the through-port transmission of the passive MRR to achieve the channelizer, although here we plotted the drop-port transmission spectra to reveal the relationships between
θ and the extinction ratio between the orthogonally polarized resonances.
The measured drop-port transmission spectra at dual polarization of the passive MRR and 20 microcomb lines are shown in
Figure 2(c). The microcombs are pre-shaped by an optical programmable processor. Proper polarization state allows both TE and TM mode resonances to be observed simultaneously, as the blue lines mark, while the orange lines denote the flattened combs. The zoom-in views show the details near the second and the 20th comb lines. We note that the combs’ linewidths are much smaller (potentially at kHz level) than those shown in the figure (limited by the resolution of the optical spectrum analyzer at 0.02 nm). As
Figure 2(d) depicts, the full-width at half-maximum (FWHM) of TE and TM resonance are ~144 MHz and ~163 MHz, respectively, corresponding to a high Q factor over 1.2 million, and a 20 dB bandwidth of ~ 1.2 GHz, which leads a high RF spectral resolution for the channelizer and thus reduced bandwidth requirements to ADCs (<200 MHz).
The channelized RF frequencies
fRF(k) of dual polarization are derived, corresponding to the spacing between the comb and adjacent TE and TM resonances, showing an upward trend from red to blue in
Figure 2(e). Due to the insufficient resolution of the spectrometer, here the peak wavelength of each comb is calculated using the inherent FSR of the MRR. The fitted results indicate that the channelized RF frequencies increase at 160 MHz (TE) and 110 MHz (TM) per channel, eventually achieving instantaneous operating bandwidths of 3.04 GHz (TE) and 2.1 GHz (TM), 5.14 GHz in total. We note that a wider instantaneous bandwidth can be achieved with more wavelength channels, for example, with 80 channels the instantaneous bandwidths can be increased by 4 times to over 20 GHz, sufficient for general RF applications.
Next, the frequency responses of the 40 channels under dual polarization are verified by VNA in the RF domain, as shown in
Figure 3. The 3dB bandwidth of RF channels, or the achieved RF channelizing resolution, are ~ 115 MHz. We note that due to imperfections of devices across a broad optical bandwidth, the measured RF channels can have power fluctuations, although they can be equalized straightforwardly by adjust the optical power of each comb line during the flattening process. As shown in
Figure 3(c), the channelized RF frequencies (i.e., the center frequencies of each RF channel) indicate instantaneous operation bandwidths of 3.1 GHz and 2.2 GHz for TE and TM polarization, respectively, with RF channelization steps of 163 MHz (TE) and 117 MHz (TM), which closely matches with the results in
Figure 2(e). We note that, due to the relatively close match between the channelized RF frequencies’ step and the resolution, the crosstalk between adjacent RF channels is further reduced (~12 dB, as shown in
Figure 3(b)), in contrast to our previous work [
15]. To further reduce the crosstalk, several solutions can be adopted [
15]: (i) tailored passive MRR’s FSR via accurate design and nanofabrication allows matching between the RF channelization step and resolution; (ii) passive optical filters with higher roll-offs and flat passbands, which can be achieved by high-order cascaded MRRs [
18].
To further verify the tunability of the proposed RF channelizer for spectral analysis at diverse RF bands (such as frequency-upconverted baseband RF signals), we thermally tuned the passive MRR to change the offset or spectral interval between a comb line and its adjacent passive resonances.
Figure 4 shows the measured RF transmission spectra. The thermal tuning efficiencies of TE and TM channels are -1.59 GHz/°C and 1.65 GHz/°C, respectively, shown as in
Figure 4(b), indicating a wide tuning range up to >60 GHz with ~40 °C temperature variation, well within the capability of external or on-chip heaters and sufficient for wideband RF applications. Furthermore, to verify the flexibility in the extinction ratio between TE- and TM-polarized channels, which can be used to equalize the channels’ power, we continuously adjusted the polarization angle
θ via a polarization controller. By varying from 0 to 90 degree, the extinction ratio between the TE- and TM- RF channels varies from -25.1 to 25.4 dB, as shown in
Figure 4(c). This result indicates that the TE- and TM- RF channels can be equalized or switched on/off, bringing additional flexibilities for post sub-bands receiving and processing.
We note that the instantaneous RF operation bandwidth is given by the product of channel number and the channelizing resolution, as shown in
Figure 2(e) and
Figure 3(c), thus it can be further increased by: a) increasing the number of channels within available optical bands — this can be achieved by using active/passive MRRs with wider optical bandwidths and smaller FSRs, although we note that smaller FSRs introduces tradeoffs with the Nyquist bandwidth (i.e., half of the MRR’s FSR); b) decreasing the channelizing resolution — this can be achieved by using passive MRRs with lower Q factor (ideally high-order MRRs with wider passbands), although this imposes higher requirements on the performance of ADCs, thus needs to be chose upon the practical systems’ requirements of the instantaneous bandwidth, cost and size; c) the optical spectral interval between adjacent resonances of the passive MRR/the FSR of the comb source — this can be enlarged by using MRRs with higher FSRs [
19,
20], nonetheless, this brings about tradeoffs that larger FSRs lead to less channels within a certain optical bandwidth. This work will benefit from recent advances made in microcombs generally [
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38,
39,
40,
41,
42,
43] and microwave and other high bandwidth applications of microcombs. [
44,
45,
46,
47,
48,
49,
50,
51,
52,
53,
54,
55,
56,
57,
58,
59,
60,
61,
62,
63,
64,
65,
66,
67,
68,
69,
70,
71,
72,
73]