3
\$\begingroup\$

I'm trying to evaluate the quality of an 8b/10b encoded data stream (Gigabit Ethernet, 1.25 gigabits/sec). Without doing clock recovery from the data stream, I want to measure jitter (et al) with a plain oscilliscope. Is there any 8b data stream I can send which ends up as a regular square wave in 10b format? I don't care it's divided or uneven duty cycle, as long as it's a regular grouping of 0's and 1's.

\$\endgroup\$
6
  • \$\begingroup\$ Wow. Are there really oscilloscopes out there that can visualize a 10 GHz signal? That's pretty impressing. \$\endgroup\$ – Roland Illig May 17 '11 at 20:38
  • \$\begingroup\$ oops, wrote 1.25 gigabytes when I meant 1.25 gigabits. \$\endgroup\$ – GlobalReset May 17 '11 at 21:33
  • \$\begingroup\$ But yeah, there are scopes that can do a really good job on 10Gig signaling. Anyone doing the latest PCI or 10Gig ethernet stuff needs them. \$\endgroup\$ – GlobalReset May 17 '11 at 21:35
  • \$\begingroup\$ possible duplicate of Alternating patterns in 8b/10b encoding \$\endgroup\$ – stevenvh May 18 '11 at 5:57
  • \$\begingroup\$ I doubt it because the running disparity calcs will keep selecting between different codes. Is there any testmode you can put the device into that you could use? \$\endgroup\$ – Marty May 18 '11 at 6:42
3
\$\begingroup\$

Alternating pattern for 8b10b encoding is D10.2 symbol. Check out table in wikipedia.

However this may not be what you want as for example gigabit Ethernet of the 1000BASE-T flavor uses PAM-5 modulation with data scrambling instead of NRZ with 8b10b encoding (again Wikipedia is helpful).

Also end result will depend in which protocol layer you are injecting test data as different protocol layers will add/remove/modify data.

\$\endgroup\$
3
  • \$\begingroup\$ My setup is a GMII interface to a PCS/PMA wrapper that directly drives a transceiver. Output is 1000Base-TX (for driving an optical SFP). \$\endgroup\$ – GlobalReset May 18 '11 at 14:01
  • \$\begingroup\$ you should be ok then by driving d10.2 (010 01010) \$\endgroup\$ – mazurnification May 18 '11 at 17:24
  • \$\begingroup\$ You're right. I got distracted by your mention of PAM-5 and missed that you gave me the data word I needed. Thanks for your help. \$\endgroup\$ – GlobalReset May 18 '11 at 21:22
0
\$\begingroup\$

I think that I've found that the 10b words are selected specifically to avoid this kind of pattern. I looked at the 8b/10b tables for an hour or so before I gave up and determined it was going to have a very similar solution as the MU puzzle.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ It is possible to achieve that kind of signal - it is right there in the table D10.2 D10=01010 and D.X.2=010 (EDCBA=01010 and HGF=010) and this is the same for both run disparities. \$\endgroup\$ – mazurnification May 18 '11 at 17:18
  • \$\begingroup\$ oh man, you're right. I don't know how many times I scanned the table (literally dozens) and missed that. Should've used a regular expression. \$\endgroup\$ – GlobalReset May 18 '11 at 21:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.