[clue-tech] making ethernet autonegotiation not work

[David L. Anselmi]

Jim Ockers wrote:
[...]
> Basically we have ruled out trying to solve this problem with different
> NICs (the link partners are industrial single board computers with
> built-on onboard network interfaces), hubs (because we need the switch
> signal processing), and software configuration on the link partners
> (already tried that).  Changing the cabling and connectors is not an
> option - that's what's driving this whole investigation.

So you can find a different SBC that either has hardware settings for 
NIC speed, or that is configurable via software.  I don't think that's 
any harder a requirement than what you have for switch environment.

> Basically the only thing I can think of that would make the ethernet
> autonegotiation not work is to somehow defeat the link pulses so
> that everything falls back to 10BaseT half duplex, the lowest common
> denominator.

I don't think you're going to manage that, because the same pulses are 
used for link presence.  You'd need something to rewrite the 
negotiation.  If you're putting a box in the line can you put in 
something that acts as a repeater?

As far as filtering, what's the frequency spectrum of the NLP signal? 
If it is higher than the frequency of the ethernet signal (350MHz or 
something) then you can filter it.  But I'd guess it isn't high enough 
that you can filter it without distorting the signal you want.

Just guessing, it's been a while since I did signal processing.

Dave