The Great Gigabit Backplane Shootout - Question
#13
We see no evidence that the cost of optical links in the backplane can be cost competitive with copper links. The electronics for copper links continue to drop in price as costs come down. Optical transceivers are several orders of magnitude more expensive. In addition, optical links require a great deal more power, and are more difficult to manufacture. In fact, the few companies that have attempted optical backplanes have converted back to electrical. It is clear that electrical backplanes will continue to increase in speed for the foreseeable future. It will be many years before optical backplanes will be a viable alternative to electrical ones.
Every generation, there seems to be a question about whether the backplane can handle the higher speeds. To date, we have always solved this problem. It is likely that the backplane and IC industry will continue to solve this problem. It is a matter of whether the optics suppliers can drive down their costs faster or further that the IC suppliers can solve their signal integrity issues.
As mentioned in the comment section, SerDes, in either discrete or integrated form, is required for either copper or optical backplane.
SERDES technology actually can be considered as an adjunct technology for optical backplanes. You still have to convert from electrical to optical domain. Most of the existing optical backplanes are parallel optics, with each fiber running at 2.5-3.125GB/s rates. The SERDES function is still needed here, and depending on how the backplane is constructed, the electrical link can be short to long, so equalization is still a requirement. Until optical backplanes start migrating to 10GB/s per fiber, multi-level coding is probably not a serious contender.
Absolutely not.
There is a saying in the electronics industry that if a device can be designed in CMOS, it will be. In the communications world the saying is that if it can done on copper it will be. Over the years the Ethernet world has proven this again and again. Optics usually takes an early market lead and then as copper solutions roll out the pendulum swings back (for short runs) as people weigh the price / performance options. Optical backplanes will always have their place in high-end applications; however, for the next three product cycles (5Gb/s, 10Gb/s, and 20Gb/s) copper alternatives are clearly viable. This should extend the life of copper backplanes for at least ten years. OEM's want simple, low cost, connections for their boards using existing material sets and connectors. If IC companies can deliver the type of parts OEM's want in this timeframe, optical backplanes will have to wait for "next year."
Optical backplanes are still orders of magnitude more expensive compared to copper backplanes. SERDES will continue to be deployed as the primary mechanism for copper backplanes. In any case, SERDES chips are required to implement optical backplane. The only difference is that the pre-emphasis is turned off.
We believe that the SerDes technology will continue to be the technology of choice because of high manufacturing costs and low yield of the optical backplanes.
No, cost of optics still makes optical backplanes prohibitive and you still need a SERDES in behind the optics anyways.
What optical backplanes?