Fully integrated on-chip THz transmitter for a THz FDM high-speed WLAN
The vision of a fully integrated on-chip photonic-wireless THz transmitter is shown in Fig. 2, where the THz frequency division multiplexed (FDM) high-speed wireless local area network (WLAN) is used as an exemplary application. The THz FDM high-speed WLAN system consists of a shared chip-based optical frequency comb and several integrated THz photonic-wireless transmitters. The frequency comb is distributed to each THz transmitter for injection locking dual-DFB lasers respectively. The THz transmitter consists of dual-DFB lasers, data modulator, and UTC-PD integrated THz antenna. The two free-running DFB lasers become phase coherent after injection locked to the frequency comb. One of the CW tones is modulated with a high-speed data signal and the other acts as a LO, and after the recombining they are launched into the UTC-PD for photomixing, and finally emitted as a THz signal through the THz antenna. By tuning the wavelengths of the DFB lasers in each THz transmitter, they can be injection locked to different tones of the frequency comb with different spacing and generate a different beating THz carrier to support THz FDM high-speed WLAN. Such a THz FDM high-speed WLAN will be desirable for indoor communications with high cybersecurity, since each user can have a distinct frequency channel.
Dual-DFB PIC injection-locked to OFCG for high-purity THz generation
As shown in Fig. 3(a), on the monolithically integrated dual-DFB PIC, the two tunable DFB lasers are coupled to the 3-dB multimode interference (MMI) coupler. The wavelength and intensity of each DFB tone are controlled with the thermal tuning current for integrated heater and injection current. The GSG transmission coplanar waveguide (CPW) supports RF current injection. After the MMI coupler, the superimposed optical dual-DFB emission is accessible via the spot-size converter (SSC) on the cleaved facet with anti-reflection (AR) coating. The on-chip PIN-PD allows for on-chip optical heterodyning generation up to 40 GHz that is far below the THz range. The operation frequency of PIN-PD and GSG are not within the THz range and thus not studied in this paper. The wavelength spacing between the two free-tuning DFB modes is variable within the range of 0-10.7 nm, corresponding to 0-1.4 THz, in the telecom C-band. To reduce the phase fluctuations between the two free-running tones, an off-the-shelf 9.951-GHz MLL-based OFCG is employed to injection-lock the two modes simultaneously to keep them correlated. The dual-DFB lasers are set to 1555.575 nm and 1558.975 nm, injection locked to two selected modes of the coherent OFC.
Fig. 3(a) also shows the overall experimental setup for characterizing the OFCG-locked dual-tone laser. Before sending the two coherent tones generated by the OFCG-locked dual-DFB PIC into the UTC-PD, an optical band-pass filter (OBPF), a polarization controller (PC), a polarizer, a polarization maintaining variable optical attenuator (VOA) are employed to control the polarization alignment and optical power of the input signal. At the output of the UTC-PD, the THz signal O/E-converted from the dual-tone laser emission with a frequency of 408 GHz is generated and emitted into a 10.7-m wireless link. A pair of THz lenses with 100-mm diameter and 200-mm focus length is used at the transmitter and receiver sides to collimate the THz beam. The 408-GHz THz carrier signal is then down-converted to the IF of 10 GHz with a SBD mixer, with the 12-time electrical tone being 398 GHz (corresponding to a fundamental rate about 33.17 GHz). The IF signal is collected with the ESA.
The typical behavior of the DFB laser is presented in Fig. 3(b), where the intensity and wavelength continuously increase as the injection current raises. The intensity and wavelength can also be tuned with the heater current. DFB-1 covers 1552-1556 nm and DFB-2 covers 1556-1560 nm. In the pure single-wavelength operation, the optical signal-to-noise ratio (OSNR) is around 60 dB. Fig. 3(c) shows the OFC spectra and the optically injection-locked two-tone emission collected at port 1 and port 3 of the optical circulator, respectively. From the OFC spectra, it can be seen that the OSNR of the tone of OFC is around 40-50 dB. When both lasers are aligned to the OFC and properly biased to have the 3.3-nm (408-GHz) wavelength separation, the OSNR of the two selected OFC tones increases to 50-60 dB. Therefore, the optical injection locking method can increase the OSNR of the tones by around 10 dB in respect to the original OFC. To enhance the side-mode-suppression ratio, the unwanted adjacent modes could be suppressed by using a sharp optical filter or a comb with wider spacing.
The IF electrical spectrum are shown in Fig. 3(d). The blue trace shows the down-converted beat-note when the two DFB lasers are free running. The red trace shows the down-converted beat-note when the lasers are properly biased and aligned to the OFC. The injection power of the master comb laser is 10 dBm, measured with an optical power meter at port-1 of the optical circulator. This locked beat-note shows a Hz-level 3-dB linewidth (see the inset) while the free-running beat-note shows a sub-GHz linewidth. In the free-running operation, the dual-DFB lasers are not correlated and thus generating a huge amount of phase perturbation, resulting in the large linewidth as well as the frequency drifting that is associated with the long-term frequency stability. The single sideband (SSB) phase noise power spectral density of the down-converted synthesized signal is measured experimentally for the dual-DFB lasers separated by 408 GHz (3.3 nm), as shown in Fig. 3(e). It demonstrates the effect of injection power level. In this range, a higher injection power level offers further phase noise reduction. The phase noise of the heterodyne signal is reduced to the level of less than -100 dBc/Hz at >1 MHz offset when the injection power is larger than 9 dBm.
Experimental demonstration of 131 Gbit/s THz wireless transmission over 10.7 m
At the transmitter, two continuous waves (CWs) at 1555.675-nm and 1558.975-nm are generated by the aforementioned dual-DFB laser chip, as shown in the inset of Fig. 4(a). For a phase-stabilized beat note, an off-the-shelf MLL, which emits a 9.951-GHz-spacing OFC, is used to injection-lock the two CW modes, making them correlated in terms of frequency and phase. In the experimental demonstration, the spacing between the two DFB lasers is also tuned to be 408 GHz in order to generate the carrier frequency. The overall experimental system setup is shown in Fig. 4. A de-multiplexer separates the two coherent tones generated in the dual-DFB laser chip with 408 GHz spacing. One tone is used as an optical local oscillator (LO) for heterodyne mixing in order to generate the THz wave. The other tone is used as modulation carrier and it is launched into the in-phase (I) and quadrature (Q) arms of an optical modulator (IQM).
A two-channel 64-GSa/s arbitrary waveform generator (AWG) is used to generate the IQ-OFDM signal. The length of the inverse fast Fourier transform (IFFT) and cyclic prefix (CP) of the IQ-OFDM signal are set to 1024 and 16, respectively, and the first subcarrier is set to null. The binary sequence used to generate the OFDM symbols is a random sequence generated from MATLAB software. The modulated 16-QAM-OFDM optical signals after the IQM are amplified by an erbium doped fiber amplifier (EDFA) followed by an OBPF. Here, a VOA is used to control the power of the optical signal before combining the optical LO, to keep the power ratio balanced between the optical LO and signal for the highest photo-mixing efficiency in the UTC-PD .
The baseband signal and the optical LO are polarization aligned, and then combined before launching into the broadband UTC-PD. A polarization maintaining (Pol. M) VOA is used to control the optical power launched into the UTC-PD. The optical spectrum of the combined signal and LO is shown in the inset of Fig. 4(b). At the output of the UTC-PD, a THz signal with carrier frequency centered at 408 GHz is generated and emitted into a 10.7-m line-of-sight (LOS) wireless link, as shown in Fig. 4(d). A pair of THz lenses with a 100-mm diameter and 200-mm focus length are used to collimate the THz beam. At the receiver, the THz signal is down-converted to an IF by employing a sub-harmonic Schottky mixer operating in the 0.3-0.5 THz band, driven by a 12-time (×12) frequency multiplied electrical LO. The electrical LO is tuned to be 32 GHz, resulting in a corresponding IF carrier frequency of 24 GHz. The IF signal is amplified by an RF amplifier with 45 GHz bandwidth and then converted to digital samples in a 160 GSa/s real time digital sampling oscilloscope (DSO) with 63 GHz analog bandwidth. The digital signals are processed and analyzed offline with a digital signal processing (DSP) routine.
The structure of the DSP routine is shown in Fig. 4(c). The channel equalization is composed of linear equalization (LE), phase noise compensation (PNC) and nonlinear equalization (NLE). First, the signal after the FFT module passes through the pilot based one-tap LE, which is used to compensate the system linear response and to reduce the system additive noise influence. After LE, the signal is equalized with a least-squares method-based PNC to reduce the impairment from phase noise. After the PNC, the Volterra series nonlinear model is used for estimating the nonlinear impairment, which considers the 2nd-order and the 3rd-order distortion terms (see details in Methods and Supplementary information).
The single-channel THz signal is evaluated after the 10.7-m wireless transmission. As shown in Fig. 5(a), the BER performance for four cases of different DSP modules combined (w/o equalization, LE, LE+PNC, LE+PNC+NLE) have been measured versus the optical power launched into the UTC-PD. For the case of nonlinear DSP (LE+PNC+NLE) employed, a BER below low-density parity-check convolutional codes (LDPC-CC) forward error correction (FEC) threshold (2.7 e-2, 20%-OH, the pre-FEC BER was calculated from the given Q factor in dB as (1/2)erfc(105.7dB/20/√2)) [20, 21] is successfully achieved. The 16-QAM-OFDM has a total bandwidth of 44.43 GHz, which corresponds to a gross bit rate of 157.46 Gbit/s (subtracting the pilot overhead) and a net rate of 131.21 Gbit/s after subtracting the FEC overhead. The capacity calculated by the generalized mutual information (GMI)  is also presented, and the capacity at 14 dBm optical power is 134.56 Gbit/s, which has ~2.5% variation with post-FEC capacity. The corresponding signal constellations captured at an optical power of 14 dBm with different DSP modules are shown in Fig. 5(b)-(e) respectively. The electrical spectra of the 44.43 GHz OFDM signal both before and after down conversion and filtering are shown in Fig. 5(f)-(g) respectively. The performance of the system here is limited by the SNR of the received signal, as shown in Fig. 5(g), a mean SNR of 13.45 dB is achieved.