Those are all lies. You are just making shit up.
To the makers of this ethernet thing:
So what single chip ethernet switch IC is this using? There ain't many. Marvel won't usually deal with small companies outside of asia. Broadcom, well, ask any of the Rasp Pi guys for a complete datasheet on their IC's. I did some of the crypto accelerators for Bluesteel after Broadcom bought them. They were not fun to deal with...
So that leaves the Micrel based or Vitesse, unless you spin your own in an FPGA, but that's probably suck as to thruput and latency unless you used a silly expensive part.
And what ARM are you using? I see the JLINK ICE headers. Are they M or A core ARMS's? I do a lot of design with both; the A cores are sometimes available in POP packaging - where the DDRx is mounted on top of the uC itself... Snapdragon 8xx is one, tho again they will not sell you IC's unless you're Samsung and buying 100K or so a quarter.
So can you show this on a report from something like the Sifos analyzer?
I'd be interested in seeing that...
I deal with some silly fast ethernet designs - silly expensive, very low latency. We've had our stuff tested by firms that have gear like the Sifos for PHY and stuff like Teledyne/LeCroy for protocol analysis. I'd be curious to see any report you'd have that's similar.
I'm guessing the input is just a standard 802.3ab using PAM-5 and typical 1000-T encoding... What I'm not quite sure of is that in maintaining the encoding for this
... how can this be any different then any other 802.3ab compliant device?
And what's with the isolation between the in and output? What does that really do as far as any measurable performance of the actual ethernet switching? The galvanic isolation proved by the transformers - transformers with fairly decent CMR - should be sufficient. There's no galvanic connection to either link partner...
I've had to design isolation for things like USB for medical gear - compliant to IEC 60601 for patient barrier - here's a pic
We tried using the ADUM Icoupler's with power isolation but it put out more RF (1MHz to about 150MHz if I recall) that the Welch Allyn test lab mentioned it was the most radiated EMI they'd ever seen.... so we bagged that and went with what you see there.
The USB isolator that Lady ADA sells is also based on those AD ICouplers. And yep - they're also hideous for radiated too...
So any info and ethernet PHY/Protocol test results that show a distinct improvement in link partner performance/resilience would be appreciated...