Coherent Transmission Modulators for 100G and 200G: Part II

By Milind Gokhale, Ph.D. on July 9, 2015 | Leave a Comment

In Part I we explored the principle of IQ phase modulation for coherent transmission modulator where the Mach Zehnder modulator design is used to convert coded symbols of digital information into combinations of phase and intensity of transmitted light. The total transmitted bit rate is determined by three transmission system parameters: baud rate (symbols/sec) X polarization (typically = 2) X coding (bits/symbol). The table summarizes popular combinations of these parameters for 100G and 200G. By adding more wavelengths per channel the total transmission capacity is increased beyond 400G/channel.

In this blog post, we discuss two new topics: (1) the push towards a higher data rate and (2) material platforms for building Mach Zehnder modulators to support that need.

Move To Higher Baud Rate

Baud rate also known as symbol rate is the rate at which symbols are transmitted and received. Note, that a symbol may contain multiple bits of information. The baud rates for 100G transmission vary from 28 to 32 Gbaud or 28 – 32 billion symbols/second with higher baud rate needed to support more aggressive error correction. There are two ways to increase the transmission capacity on the same optical fiber. One is to increase the bits/symbol by going to higher order modulation (HOM) from BPSK→QPSK→16-QAM→64-QAM. The other option is to increase the baud rate from 28Gbaud→45Gbaud→64Gbaud. Each option poses some technical and application challenges. The increased capacity with HOM comes at a penalty of shorter reach as the transmission distance is related to the phase and intensity distance between symbols. With higher baud rate, each component in the optical link has to operate at a higher rate, requiring more engineered components with signal bandwidth approaching 40GHz for 64Gbaud operation. Advances in new high speed transmission modulator and coherent detector have been demonstrated, but now the technology demonstrations need to be productized. Similarly the DSP speed and complexity increase is supported by Moore’s law where CMOS transistors get smaller and faster increasing the overall processing power at 1.8X every 2 years. Given that both these technology trends are well supported and the fact that higher baud rate does not drastically reduce the transmission distance or link budget, the bet is on higher baud rate for the future. OIF and industry standards forum has recently started work groups to standardize high bandwidth Tx and Rx interfaces for future coherent systems.

Today, HOM such as 16-QAM is deployed in metro and shorter datacenter interconnect links to maximize transmission capacity, while higher baud rate is making its presence felt in longer distance applications indicating that architects for fiber optic systems are exploiting both options. The macro long term trend favors high baud rate for a few key reasons: (1) higher speed links can transmit data over longer distance compared to HOM (2) it supports historic long term trends of increased transmission capacity with new DSP.

Material platforms for Coherent Transmission Modulator

The predominant material platform deployed today for coherent modulator is lithium niobate (LN). At 10G, over time as line cards were replaced with compact pluggable modules, lithium niobate modulators were replaced with a much smaller Indium Phosphide (InP) modulators. Since, 2008 a new material of choice namely, silicon has emerged as a contender to both LN and InP IQ modulators. The comparison of the different material platform for high speed coherent modulator is shown below.

$Pocket's effect, Stark effect, carrier refraction$

Lithium niobate is technically the most desired modulator platform, with linear electro optic effort, low loss, low Vpi and no need for cooling. However, the size of the LN modulator is too large for integration into compact pluggable modules such as CFP2-ACO. InP is well suited for both line card and pluggable modules due to its small size (<7mm) and high optical performance particularly of low Vpi and loss and BW > 40GHz which is critical for 64Gbaud symbol rate. Some disadvantages are the need for thermo electric cooling (TEC) and the integrated IQ chip size is still large in comparison to InP wafer size. The new Si based coherent modulator are based on standard CMOS process on much larger >8 inch wafers. The increased chip count per wafer, maturity of production technology of CMOS and a reasonable performance including uncooled operation has created strong interest in this technology. The one distinct disadvantage of Si vs LN or InP is a factor of 2-3 increase in Vpi that leads to higher drive voltage and increased power dissipation. The marketplace is likely to support all three platforms across the various applications.

In summary, similar to the doubling of transmission capacity by use of dual polarization 5 years ago, there is a new move to double the existing transmission rate by a factor of 2 by moving towards 64 GBaud. We also discussed the merits of the three main material platforms used to create coherent IQ modulators with both InP and LN showing great promise to meet requirements for current and new high speed coherent transmission systems.

Categories

Recent Blog Posts

Coherent Transmission Modulators for 100G and 200G: Part II

Leave a Reply Cancel reply

Related Posts

White Paper