The 100G networking architecture gradually becomes the mainstream in today’s data center network deployment, driving the boom of the 100G transceiver market. In the long run, there is still a need for data center upgrading. It is generally believed in the industry that the network will evolve towards 400G network architecture and will drive the market demand and technical innovation of 400G optical module. Meanwhile, silicon photonics, the new idea that would use the principles of optic fiber inside the devices, has been chasen by the industry’s biggest players, including Intel, IBM and HP.
Lasers are the core devices of optical transceivers, which injecting current into semiconductor materials and injecting laser light through the photon oscillations and gains in the resonator. Accounting for 60% of the cost of the transceiver module, the laser is closely related to the transmission distance of the transceiver. Typical laser types in the market include VCSEL, FP, DFB, DML and EML. The table below shows their wavelengths, working pattern and applications.
|EML||1310nm/1550nm||External Modulation; Electro Absorption Modulation||40KM|
VCSEL laser features small size, high coupling rate, low power consumption, and low price, thus favored by the 100G-SR4 QSFP28 optical transceiver which is mainly deployed in the 100G multimode fiber network.
The EML has less wavelength dispersion and a stable wavelength at high-speed operation. The frequency response of the EML depends on the capacitance of the EAM section and can achieve high operating speeds even above 40 GHz. It is often adopted by 100G-ER4 and 100G-LR4 QSFP28 transceivers which are designed for applications over single-mode fiber (SMF) with transmission distances of up to 10km.
DMLs are mainly used for relatively lower speeds (≤25Gbps), and shorter reaches (2-10km) in telecom and datacom applications, due to constraints such as larger chromatic dispersion, lower frequency response, and relatively low extinction ratio, compared to EMLs. 100G-CWDM4 QSFP28 optics for CWDM programs over 2km frequently adopt the DML laser.
As for the 100G-PSM4 QSFP28 transceiver, there is a new breakthrough in chip technology as Intel has realized volume shipments of 100 Gbps PSM4 optical transceivers that leverage the silicon photonics technology. Moreover, the competitive cost advantage makes the transceiver occupy 80% of the market share of PSM4 products.
At present, the technology route of optical integration commercial products is mainly divided into InP and Si. Lasers like DFB, DML, EML belong to InP which is relatively mature in technology, but has high cost and is incompatible with CMOS process, and its substrate material only doubled every 2.6 years. The Si silicon photonic device can be integrated on a large scale by passive optoelectronic devices and integrated circuits with the CMOS process. It is featured by the high-density and the substrate material doubles every year.
At present, the 100G transceiver has seen the bright future in silicon photonics with the debut of the Intel silicon photonics transceiver, yet encountered some challenges in the development.
The approach to integrate photonics with silicon based microelectronics should be achieved. Silicon is an indirect bandgap semiconductor, unable to emit light efficiently, thus the independent light is required, which, however does not comply to the Moore's law. The more coupled integration, the higher the cost, which eventually offsets the cost advantages of silicon materials and process integration.
In addition, the silicon photonics transceiver is hard to package, with low yield. The packaging of silicon photonic interface is in the initial stage, with the conundrum lying in the packaging of optical interface formed by opto-electronic chip and optical fiber array, which requires high precision but shows low efficiency. The current technology is difficult to achieve high quality, low cost packaging. What’s worse, product yield hinders the mass-scale production of silicon photonics transceiver.
The silicon photonics is an irresistible trend in 100G optical transceiver industry in the long run, and may make inroads into the 400G transceiver market.
In the face of the demand of high-speed 400G network, although the multi-channel coherent technology proposed lowers the requirement of laser chip, the overall cost is higher. However, the single-channel technology has much higher requirements on chips. In the case of 100G optical module, the traditional laser is close to the bandwidth limit, and the only possible EML laser has a relatively high cost. Given this, the silicon photonics may become the mainstream in the 400G era if the above puzzles can be addressed.
Although silicon photonics transceiver optics may become the mainstream in the era of 400G optical module, 100G network still prevails at the moment. 100G QSFP28 optical module laser chip is still dominated by VCSEL, EML and DML. 100G Ethernet data center deployment has drawn much attention to silicon photonics; many of the 100G optical transceiver modules are made with silicon photonics technology.