Electronics > Projects, Designs, and Technical Stuff
Ethernet Auto-negotiation & Signal integrity
(1/1)
michaelym:
So I understand that "Auto-negotiation" is based on the capability of the end devices. So for example, if they both can do GBE the link will be GBE. What happens if there are signal integrity issues on the line between them. Do they just "struggle" with many lost packets or do they automatically reduce to 100.
tooki:
Doubtful. GbE uses all 4 pairs in the cable, as opposed to 100Base-TX which uses only two. In essence, it’s two 500Mbps links aggregated. So chances are that degraded half-gigabit times two is better than dropping to a single 100Mbps, at least if the links half work.

(Just to be clear, this is pure speculation on my part. I haven’t done any testing or anything.)
Vovk_Z:

--- Quote from: michaelym on March 12, 2020, 04:27:25 am --- What happens if there are signal integrity issues on the line between them. Do they just "struggle" with many lost packets or do they automatically reduce to 100.

--- End quote ---
I know very little about wired/wireless Ethernet technology, but from what I know it tend to work as GBE with a big loss of packets (up to 100% loss).
Auto-negotiation is more about device capabilities (exact as you stated), but not about channel capability. It quarantining to work 10/100 MBE devices with 1/10 GBE devices together in different variations, but it is not about physical channel testing.

I don't know, may be there is "improved" Auto-Negotiation Technology where each device can test it's physical connection to other device, and "pretend" lower-speed device. But I haven't heard about that.
ve7xen:
Autonegotiation is, as you suspect, capability negotiation and not training / based on the signal integrity conditions whatsoever. In fact, since autoneg occurs over the two 'normal' pairs (used in 10base-T and 100base-TX), it's possible to successfully negotiate 1000base-T even if the other two pairs are completely missing, and the link will be completely nonfunctional. Some PHYs do include cable tests / TDR measurements, but this isn't generally done routinely before link-up, and isn't standardized or part of the autonegotiation process.


--- Quote ---In essence, it’s two 500Mbps links aggregated. So chances are that degraded half-gigabit times two is better than dropping to a single 100Mbps, at least if the links half work.
--- End quote ---
Not quite, all 4 pairs are used bidirectionally in 1000base-T; it's actually 250Mbps x 4 (same 125Mbaud as 100base-TX). Also the line coding splits each byte across all 4 channels, so if you lose any of them, the link won't pass any data at all. Even if it did, higher level network protocols would tolerate 50% packet loss extremely poorly, and real-world performance would be much, much, much worse than a properly working 100mbps link (it would be essentially unusable).

Due to improvements in line coding, the use of bidirectional channels, and adding more pairs, the signal quality requirements for 100base-TX and 1000base-T are basically the same, so as long as the extra pairs aren't damaged, it shouldn't generally have a different error rate than 100base-TX on the same cable.
Navigation
Message Index
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod