1. Introduction

Due to the low power consumption and large bandwidth, microwave photonics has been extensively investigated for over 20 years [1,2]. Microwave photonic filter (MPF), as one of the most important microwave signal-processing implements, has attracted great interests from numerous researchers [3]. A good amount of solutions had been proposed [4–6]. Most of them were constructed with fiber-base devices, increasing the footprint and power consumption [7]. For broadband wireless access networks, it is desirable to adopt a flexible and scalable solution [7–9]. Recently, benefitting from the advantages of photonic integration, the integrated MPF (IMPF) implemented on different integrated platforms had been proposed and demonstrated. In [10], an IMPF including a laser diode (LD), a modulator, an optical filter and a photodetector (PD) was demonstrated based on indium phosphide substrate. On the other hand, in virtue of large index contrast and mature Complementary Metal-Oxide-Semiconductor (CMOS) fabrication compatibility, silicon photonics exhibit significant advantages. More recently, an integrated frequency-tunable microwave photonic bandpass filter based on silicon-on-insulator (SOI) platform was demonstrated [11], showing potential application in wireless and fiber communication.

Nowadays, with the rapid growth of data traffic, the access network architecture of the fifth generation (5G) wireless communication requires implementing massive multiple-input/multiple-output (MIMO) at base transceiver stations [12]. A multiple-IMPFs scheme is considered as an agile and applicable scheme for realizing the MIMO technology. However, there were no reports on a multiple-IMPFs scheme. In principle, this demand can be satisfied by designing and fabricating separate IMPFs on a same chip, which needs a great number of photonic devices and multiple transmission waveguides [7]. To reduce the complexity of a multi-channel system, the multiplexing technology is a promising solution for a compact and efficient multiple-IMPFs scheme, and the most straightforward one is the wavelength division multiplexing (WDM). However, it is not preferred for on-chip system, due to the requirement of multiple LDs. On the other hand, the mode division multiplexing (MDM) based on single wavelength exhibits advantages in terms of low cost and compatibility with other multiplexing technology, and it has been investigated extensively very recently [13]. Nevertheless, there is no previous report combining the MDM with a multiple-IMPFs scheme.

For the first time to our best knowledge, we introduce the MDM into microwave photonic system for simplifying the architecture, ensuring parallel processing and reducing power consumptions. A monolithically integrated parallel radio-frequency signal-processing unit (PRSU), enabling two independent lowpass IMPFs, is presented as a demonstration. The PRSU is an MDM photonic integrated circuit based on silicon platform, and it consists of grating couplers, two micro-ring modulators (MRMs), two germanium PDs (Ge PDs) and two pairs of micro-ring-based mode multiplexer/demultiplexer. Two lowpass IMPFs utilizing the first order and second order transverse electric modes (TE₁ and TE₂) share a common multimode bus waveguide. The multiplexer and demultiplexer also work as optical filters, which determine the lowpass radio-frequency (RF) filtering function. The bandwidth of the proposed IMPFs can be tuned by thermally controlling the ring based optical filter. The minimum bandwidth of the TE₁ and TE₂ based IMPFs are 3.7 and 3.8 GHz, with bandwidth tuning within a certain range.

2. Operation principle

Figure 1 illustrates the proposed PRSU, in which each IMPF is constituted by a MRM, an optical filter and a Ge PD. The optical filter for each IMPF is formed by a pair of cascaded add-drop micro-ring resonators (MRRs), which share a bus waveguide and also function as mode multiplexer and demultiplexer. More specifically, MRR-A and MRR-D work as filter and multiplexer/demultiplexer for TE₁ based IMPF, while MRR-B and MRR-C for TE₂ based one. An RF signal to be processed is first encoded in the optical domain through intensity modulation by MRM. Then, the optical signal is filtered by the MRRs pair. The resonance wavelengths of the MRRs pair can be tuned for alignment or misalignment by thermal-optic effect. Thus, the transfer function of optical filter is reconfigurable. Eventually, two optical signals are detected by two Ge PDs independently.

 

Fig. 1 Schematic and structure of the proposed parallel RF signal-processing unit. The inserted pictures are the details of the (a) MRM, (b) MDM optical filters and (c) Ge PD.

Download Full Size | PDF

The MRR based multiplexer/de-multiplexer had been previously investigated [14]. The structure and detailed parameters are shown in Fig. 1(b). The radius and gap of all MMRs are optimized to be 50 and 0.3 µm. For TE₁ and TE₂ cases, the widths of the straight waveguide are 1.1 and 1.7 µm, respectively. The waveguides with different widths are connected through the 150 µm long adiabatic taper. Numerical simulations are further performed here to study the reconfigurable transfer function of the MRRs pair. Taking TE₁ based optical filter (MRR-A and MRR-D) as a representative, the transfer function of single MRR can be first calculated. While the resonance wavelengths of the MRRs pair in series are aligned, the transfer function of whole optical filter can be obtained by multiplying two rings’ profiles. While the resonant wavelengths of the MRRs pair in series are detuned, the transfer function is rebuilt. Figures 2(a) and 2(b) show the simulated magnitude and phase responses of TE₁ and TE₂ optical filters when the MRRs pair is in alignment. The simulated bandwidths of the two filters are 7.7 and 7.3 GHz, respectively. Both optical filters can achieve phase shift from -π to π. The simulated transmission losses for TE₁ and TE₂ based optical filters are 4.2 and 6.4 dB. Figures 3(a) and 3(b) show the simulated magnitude responses of TE₁ and TE₂ based optical filters while the MRRs pair is in the state of detuning. When the detuning value changes from 1 to 10 GHz, the bandwidth of TE₁ based optical filter is correspondingly increased from 8 to 14.4 GHz due to the Vernier effect in the MRRs pair, and the result of TE₂ based optical filter is increased from 7.6 to 14.25 GHz. As shown in Fig. 3, the center wavelength of optical filter is also changed with the detuning value, and the power losses at the center wavelength are correspondingly increased from 4.3 to 9.5 dB for TE₁ based optical filter and 6.6 to 11.3 dB for TE₂ based optical filter. Even though power loss is inevitable for the mode conversion and wavelength detuning, the reconfigurable bandwidth of optical filters still has good performance.

 

Fig. 2 Simulated magnitude and phase responses of (a) TE₁ and (b) TE₂ based optical filters while the MRRs pair in alignment.

Download Full Size | PDF

 

Fig. 3 Simulated magnitude responses of (a) TE₁ and (b) TE₂ based optical filters while the MRRs pair with different detuning values Δ.

Download Full Size | PDF

3. Experimental result

The proposed PRSU is fabricated on 220 nm thick SOI platform using 0.18 μm CMOS lithography technology. The microscope image of the fabricated circuit is shown in Fig. 4(a), where the MRM1 and PD1 are used for the TE₁ link and MRM2 and PD2 for the TE₂ link.

 

Fig. 4 (a) Microscope image of the proposed PRSU, (b) the optical spectrum for TE₁ and TE₂ links.

Download Full Size | PDF

3.1 Optical spectrum

The spectrum of optical filter vastly influences the RF response of the IMPF. In order to fully characterize the proposed IMPFs, a reference structure of optical filter was fabricated on the same chip, and the experimental results are shown in Fig. 4(b). The insertion loss for TE₁ and TE₂ links are 3.5 and 6.9 dB, respectively. The extinction ratio is larger than 25 dB and the crosstalk is about −22.9 dB for TE₁ and −24.8 dB for TE₂, respectively. The spectrums of two filters are measured with different detuning values Δ, and the output optical powers (P) reach to the maximum (P_max) when Δ equals to zero. With the growth of Δ, the output power is decreased as shown in Fig. 3. Considering the difficulty on accurately acquiring the exact value of Δ, it is convenient to use the change of power ΔP (ΔP = P_max-P) to represent Δ. The results of two filters are shown with different ΔP in Fig. 5. The optical bandwidth of micro-ring limits the maximum bandwidth of the RF filters. In a real application, the filter bandwidth is also restricted by the optical loss caused by the detuning between two rings. Thus, with acceptable optical losses as shown in Fig. 5, the measured optical bandwidth are 8.5 to 24.5 GHz and 8.3 to 22 GHz for TE₁ and TE₂ links, respectively.

 

Fig. 5 Measured magnitude responses of (a) TE₁ and (b) TE₂ based optical filters with different ΔP.

Download Full Size | PDF

3.2 RF responses

The performance of the utilized MRM and PD were characterized independently in previous work [15]. The 3-dB bandwidth of the MRM and PD are 13 and 43 GHz, respectively. A full electrical to electrical (E/E) measurement is performed here to investigate the basic functionality of single IMPF, through applying an electrical RF signal onto the MRM and measuring the output of PD. Limited by the two-port vector network analyzer (VNA, Anritsu MS4647B), the two IMPFs cannot be tested at the same time. The experimental setup is shown in Fig. 6. The Impedance Standard Substrate (Cascade, 101-190C) is used to calibrate the influence of cables and microprobe (Cascade, I67-GSG-150). Both MRM and PD are supplied −3V bias voltage. The narrow-linewidth tunable laser (Alnair labs, TLG-200) generates continuous wave (CW) at 1552.3 nm with 15 dBm optical power. The signal wavelength is optimized for the modulated depth and insertion loss of MRM [16]. A 3 dB coupler is used to halving the optical power for two IMPFs.

 

Fig. 6 The E/E experimental setup.

Download Full Size | PDF

When the signal wavelength is aligned to the center wavelength of the optical filter, the photocurrents of PDs reach the maximum value and the RF response of the IMPF achieves the minimum bandwidth. The frequency responses of TE₁ and TE₂ IMPFs are shown in Figs. 7(a) and 7(b), respectively. The minimum bandwidth of TE₁ and TE₂ IMPF are 3.7 and 3.8 GHz, and the continuously tuning of bandwidth can be achieved from 3.7 to 8 GHz for TE₁ link and 3.8 to 7.6 GHz for TE₂ link. RF losses are caused by the attenuation of optical signals and the modulation penalty of RF signals. There are multiple factors inducing the difference between the two RF responses. First, it is hard to acquire the accurate detuning value of the MRRs pair for each RF response, since the spectrum of optical filter cannot be measured simultaneously in this system. Then, when the optical filter is detuned, its center wavelength is slightly changed accordingly, as shown in Fig. 5. Consequently, the wavelength mismatch between the input signal and the optical filter cannot be kept to 0 in this process, resulting the RF response changing from lowpass to bandpass. This mismatch leads to the difference between RF responses of TE₁ and TE₂ based IMPFs. On the other hand, optical losses caused by the mode multiplexing lead to the amplitude difference of RF responses.

 

Fig. 7 Frequency responses of (a) TE₁ and (b) TE₂ IMPFs.

Download Full Size | PDF

In the demonstration, two RF signals carried on a same wavelength share a common optical channel but are independently filtered by two IMPFs with low crosstalk. Due to the intensity modulation, each IMPF has only a half bandwidth of the corresponding optical filter. Considering the optical loss caused by MRRs detuning and the PD sensitivity, the maximum bandwidth of IMPFs is less than 10 GHz. On the other hand, the maximum extinction of the IMPF is about 12 dB and mainly effected by the optical extinct ratio of the optical filter.

To be noted, it is a conceptual experimental demonstration, and more comprehensive investigation in terms of dynamic range and group delay of the proposed IMPFs are not well evaluated due to the limited experimental facilities. The performance of each IMPF can be specially optimized, by using high linear modulators [17], high quality factor MRRs [18] and high power PDs [19].

4. Conclusion

For the first time to our best knowledge, we propose and demonstrate a monolithically integrated parallel RF signal-processing unit based on mode multiplexing. This compact and low-cost microwave photonic on-chip system achieves full integration of necessary components except the laser source. The MDM in the proposed system alleviates the requirement on laser number while enabling two independent lowpass IMPFs. The reconfigurable transfer function for two IMPFs is experimentally proved. The demonstrated lowpass IMPFs show significant data-processing capacity enhancement and great potential application in microwave system.