General > General Technical Chat
Big Test MFG says our Ethernet Chip violates the IEEE standard - are they right?
<< < (4/8) > >>
metrologist:
So I see mention of MDI and MDIX and what I see are two twisted pairs 8P4C, pins 1236. I also see token ring configuration using pins 3456. I suppose the latter would not be compatible with modern 10/100/1000 networks.
T3sl4co1l:
Ask them to test some PCs with the suspect cable and confirm your device is specifically failing.

Though if your interface is missing those nice features (AutoMDIX), the cable fault might be hidden because almost everything does support it.  The question then would be, why doesn't yours?  So, a possibility there, or have them try other cables, or more closely replicate your configuration in case it really is something.

Tim
PaulAm:
And maybe, since they used a cable with an RJ11 connector, it was just an old phone cord.

Who would plug in an RJ11 into an RJ45 jack and expect it to work?

Even when I was running wires in the dim past, I only used  Cat3 for phone lines and everything else was cat5.  And that's gotta be at least 15 years ago.

And I hear you say "phone line?  what's a phone line?"
metrologist:
If I can get access to an instrument, I'd probably make or find a cable straight wired pins 1236 and try it through the LAN or direct to NIC. If both connections work then we don't have an issue on our end.
ve7xen:

--- Quote from: metrologist on October 20, 2021, 08:22:20 am ---I don't have access to the specification but it seems reasonable that it would state backward compatibility, so I'd expect the system to fall back to 10Base-T, and if our implementation does not do that then I'd be inclined to agree with the customer. The interface supports gigabit, but would only be used for sending settings, like for a function generator. We even supply PC software with a simple interface for remote operation - as simple as entering the IP address.

--- End quote ---

That's not the case. Autonegotiation is completely unaware of cabling limitations, and doesn't compensate for them. The two devices will negotiate their highest-common-denominator and attempt to establish link. If pairs are missing, and that HCD is 1000base-T (or 10GBASE-T), then link will simply fail to establish and the autoneg process will repeat ad infinitum. Some clever PHYs will retry autoneg without offering 1000base-T after repeated failed attempts, and succeed at 100base-TX link, but this isn't required (or even mentioned) by the spec, and I can confirm it is not implemented in many popular PHYs/drivers. Typically link will just fail if there are missing or broken pairs. If the cable is complete but out of spec (e.g. using Cat3 for 1000base-T), link may or may not come up, and may or may not work at an acceptable loss rate, it's simply operating out of spec, similar to going beyond 100m.

tl;dr you should not expect a 2-pair cable to work at all between two 1000base-T+ interfaces, unless you manually force them 100mbps.

And as tooki points out, the 10base-T PMA isn't required to be implemented, and modern equipment is starting to omit it, though plenty of 10GBASE-T stuff does.

Auto-MDIX, if implemented on one side, should work even if it's not supported by the link partner, on any PMA including 10base-T (it happens prior to autonegotiation). It's actually optional for 1000base-T PHYs, but is almost always implemented; it's required as of 10GBASE-T.
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod