It is said that the chip is very important. Today, I will take three minutes to show you the role of the main optoelectronic chip in coherent optical modules.

As shown in Figure 1, in the coherent optical module, at the transmitting end, the customer’s electrical signal passes through the digital signal processing function in the DSP chip to complete specific probability distribution and QAM mapping symbols, including probability distribution matching, FEC encoding, QAM mapping, etc., and then perform signal spectrum shaping and sender pre-compensation processes on the baseband digital QAM signal to compensate for the undesirable characteristics of the photoelectric device. After these DSP processing, it is sent to a four-channel high-speed DAC, which is converted into a high-speed analog bandwidth electrical signal through digital-to-analog conversion. After being amplified by the driver, it is loaded on the polarization multiplexed optical IQ modulator to complete the up-conversion, and the baseband electrical signal is loaded on the amplitude, phase and two orthogonal polarization states of the optical carrier, through the optical amplifier/attenuator inside the optical module output after optical power control is the generated high-speed optical signal.

At the receiving end, the high-speed optical signal undergoes coherent mixing and photoelectric conversion under the action of the local oscillator after ICR to obtain the baseband electrical signal, and then is sampled by the high-speed ADC to realize the digitization of the electrical signal, and then rely on the powerful DSP of the receiving end. The equalization ability compensates for the damage that the signal suffers in the circuit and the optical transceiver module. The receiving DSP processing flow includes resampling, optical front-end compensation, dispersion compensation, clock recovery, polarization demultiplexing adaptive equalization, frequency offset estimation, phase recovery, constellation diagram mapping and FEC decoding. Optical field linear IQ modulation and coherent detection enable theoretically the receiving end to recover all the optical field information of the optical signal. DSP processing can perfectly compensate the linear damage experienced by the optical signal, such as inter-channel skew, IQ amplitude and phase imbalance, and fiber color degree chromatic dispersion (CD) and polarization mode dispersion (PMD), frequency deviation and phase noise, etc., even through special algorithm design, the nonlinear effect of the system can be compensated or equalized to a certain extent, such as digital sub-carrier multiplexing can enhance the optical fiber non-linear tolerance. The high-performance SD-FEC algorithm in the 800G coherent DSP chip has better error correction capabilities than the HD-FEC in the early 100G DSP chip, further improving the OSNR tolerance of the signal, and ensuring error-free implementation in appropriate application scenarios code transmission.

Further, on the internal or external single disk of the optical module, through the Gearbox or Framer chip, the line-side electrical signal and the client-side electrical signal are connected to complete different functions, such as transponder and muxponder. Of course, some other auxiliary chips are needed in the optical module, such as power chip, clock chip, FPGA control chip and temperature monitoring chip. In addition, the DSP chip should also have the SerDes function to better interconnect with client-side signals.

In general, the core chips in coherent optical modules can be divided into two categories: optical chips, including dual-bias IQ modulation, lasers, coherent optical mixers, and balanced detectors; electrical chips, including modulator drivers, Trans-impedance amplifier (TIA), DSP chip. In terms of chip manufacturing technology, current electric drives are all based on Si or SiGe materials with CMOS standard process flow. The core part of DSP chips needs to use a more advanced 7nm node process due to power consumption and performance requirements.

The optical chips can be divided into two categories. One is based on III-V materials such as InP, including lasers, modulators, and detectors (GaAs). The second is based on Si, including Si-based modulators, coherent receivers and so on. Although in recent years, with the improvement of device integration and co-package requirements, SiP has become a hot academic and industry new favorite, but in essence, for coherent optical modules, InP vs. SiP is not a conclusive conclusion. In general, both have their own strengths, as shown in Table 1.

InP platform is better at making active devices, while SiP performs better in passive devices.  SiP has advantages in yield and cost, but may not be as good as InP in bandwidth and performance.  InP platform can integrate/package lasers with CDM and ICR, while SiP can integrate/co-package DSP and PIC without laser. SiP has many advantages, but there are two shortcomings that cannot be ignored: one is that it has a large coupling loss with the optical fiber, and the other is that it can not emit light (the LED distance practical optical communication application made by Dr. MIT is still early). Rather than say who replaces who, it is better to say that sometimes you may need to consider how to combine the advantages of the two, such as Si+X, hybrid integration, so that cost and performance are optimized as much as possible.