Breaking News

An Oversampling Technique for Serial Protocols

1 saves
View Comments: Oldest First | Newest First | Threaded View
User Rank
Oversampling CDR
michal.kubicek   10/11/2013 2:57:50 AM
We did some FPGA-based research and implementation of oversampling CDR in FPGAs few years ago. The target application was a special device for a local company. We successfully used Cyclone-II as a 4-channel transceiver in this application and later implemented the same algorithms into another product based on Spartan-3 FPGA.

The algorithms were initially based on Xilinx xapp224 but we had to improve them to get reliable data recovery (at the end with very little hardware overhead). With the improvement we were able to compete with traditional PLL-based CDRs. During the research we evaluated FPGA transceivers in oversampling mode as well, see this paper:

Our conclusion is that it is efficient and reliable to use oversampling data recovery. The asynchronous (blind) oversampling data recovery is extremely efficient in multichannel (multilane) application, as you need only single clock management block (eg. PLL), which even need not to be synchronized to the data. However, the limiting factor of today FPGAs is the sampler itself - if you use general purpose IOs. Naturally, In ASIC this is not an issue.

Our evaluation of FPGA transceivers (Virtex-5 GTP) was only shallow. The oversampling feature is actually supported in the transceiver core wizard so it was very straightforward to implement it, as the data recovery engine is already part of the IP. The results were somewhat worse than we expected, which might be caused by a poor data extraction algorithm. However, we did not try to fix it by using our data recovery engine implemented within FPGA.

Feel free to email me:

Most Recent Comments
Top Comments of the Week
August Cartoon Caption Winner!
August Cartoon Caption Winner!
"All the King's horses and all the KIng's men gave up on Humpty, so they handed the problem off to Engineering."
Like Us on Facebook Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
EE Times on Twitter
EE Times Twitter Feed
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.
Flash Poll