Discussion About 10GBASE-T Improvement

0

Error Correction to Improve Performance

The Japan Institute of Industrial Technology and Hitachi jointly proposed a scheme combining orthogonal frequency division multiplexing and PAM modulation. Among the remaining 8-value to 12-value multi-value PAM modulation proposals, this is a much more vocal scheme. Comparing the proposals of different companies, the difference in modulation method is not very big. The main difference is that the forward error correction method adopted is different.

 

For PAM modulation with a large number of values, it is easy to cause transmission errors because the value of the required signal-to-noise ratio is increased. In addition, since a symbol carries more information, even if a transmission error occurs, a large amount of data will be lost. Therefore, using an appropriate forward error correction method can make up for the shortcomings of multi-value PAM modulation with a large number of values, and obtain the same effect (coding gain) as using a modulation method with a small required signal-to-noise ratio.

 

Most communications use Trellis and Solomon codes. Especially the Trellis code and the modified low-density parity check have a large coding gain so that high-quality communication with extremely low transmission errors of 10GBASE-T can be realized. Along this technical direction, the IEEE802.3an working group received the Broadcom The company’s proposal to combine Trellis code and Solomon code, Mr. Sailesh Rao of American Ample Communications Company (who worked at Intel Corporation. When the proposal was proposed) proposed a low-density parity check, and Solar Flare’s proposal to use Trellis code. Supposedly, there is a tendency to favor low-density parity proposals.

Lower Latency Solution 

There are only 2 approaches to the proposal for latency under 1 μs. However, there are restrictions on the choice of FEC. In general, the method with high coding gain is more complicated in data processing, so the delay will increase. For the server and network computer environment, the large delay will affect the processing function of the server and grid computing. Therefore, the IEEE802.3an working group proposes that the delay should be less than 1ms, and only the solutions of SUN Microsystems and IBM of the United States meet this condition. The FEC method that combines the Trellis code and the Solomon code together has a large delay and cannot meet the requirements.

 

Among the remaining methods, some have larger delays. Mr. Guan Kemin, director of the first custom LSI business department of Japan’s NEC Electronics Corporation, who has participated in the IEEE802.3an working group since the beginning, believes: “The current work is in the LDPC method with large coding gain but also large delay, and the coding gain is small but the coding gain is small. Since the processing method is simple and the delay is also small, the Trellis code method is better to choose which method is used alone.”

 

Combining 8-value or 12-value PAM modulation with LDPC, or 10-value PAM modulation with Trellis codes is the most likely approach to 10GBASE-T based on the results discussed so far. But Mr. Guan Kemin said: “The center of discussion has shifted from which method to choose from the above-mentioned methods, to whether to use precoding or pre-correction supplementary coding and waveform processing at the transmitting end.

 

LDPC is a low-density parity-check code, which is a kind of forward error correction method. This method was developed in 1962 but was not practically applied until 1998 due to the large amount of computation. Through improvements in recent years, a method of approaching the “Shannon limit” of 0.1dB has been developed. It can be used to make electronic watermarks or generate metacontents, etc.

LEAVE A REPLY

Please enter your comment!
Please enter your name here