Connecting Mango Router ethernet to PS2 through capacitive coupling

[MRKane]

Of course the thread title could have also been called "the PS2 portable of Theseus" because that's kind of how it came about.

I've continued to try and perfect the PS2 portable I made for the BitBuilt competition back in 2020 (man doesn't that feel like a lifetime ago!) and I think I'm at the stage where I'm looking at getting a new carrier board made for it, hence I'm considering it to be a "new" portable (this also means I have, and have not, gone back against my oath to put down the iron and focus on other life stuff? I'm unsure here).

What's of immediate note is the project that MomSpagetti and I have been working on: to hard-wire the PS2 ethernet phy to phy using the smallest mobile router we could find, the GL-MT300N-V2 "Mango". Now while GMan's Pi3 inclusion in the PS2 Razer is a monumental step for modding, the method of capacitive coupling hasn't been replicable with different devices hosting different chipsets, and in the past people have used transformers scavenged from special RJ45 ports.

We needed a better way, and this is where someone could probably write a guide as I'm sure it'd be useful:

After weeks of pouring over documents the solution for this turned out to be contained in a document from Micrel (https://www.eeweb.com/wp-content/up...-coupling-ethernet-transciever-1321248893.pdf). Something the article was light on is that the ethernet bias for both sides needed to be correct, so for the PS2 this was 1.75v, and the Mango used the RX0 line instead of the 3.3v (not sure why). This has to do with the setup of the Broadcom (PS2) and Mediatek (Mango) chips and the expected bias. Simply put: if these aren't correct you'll frustratingly get listings, but nothing will load.

Spoiler: Reference pictures

Snippet of document schematic that was followed:

As a schematic from KiCad:

Also worth noting is the wiring of the ports for the Mango: The ethernet is mounted upside down. This diagram is from the underside of both units (we need more MSPaint around here!):

Now it's worth noting that I had the occasional stutter with my "antenna test setup" but I'm confident that this interference will be mitigated by shortening the leads. While this is fresh in my mind I'll also note that this setup probably won't be applicable for all systems as some chips apparently necessitate the use of magnetics or more elaborate phy interface circuits, but this is something that can be easily compressed into a small space.

Pertaining to variance in bias voltages: The numbers are pretty inflexible. I tested the PS2 with 1.25v, 2.5v, and 3.5v as provided by the PS2 board, and found that I was able to get listings but no games would load. I suspect that there's a slight tolerance to the ratings of capacitors and resistors here as I wasn't technically using the correct ratings, but couldn't say how durable the method is for large amounts of variance. Technically this is a double-biased system with both the PS2 and Mango having different bias voltages, so there should be a 100nF capacitor attached to the VDD lines as per the documentation from Micrel, but I'd assume that both systems had necessary smoothing in my tests.

Spoiler: Antenna Test Setup

In theory if I can get it working like this then it'll work great when everything is shortened down!

That task done I'm going to promptly forget all of this and move on with lifes next task!

 

[PvtDerp]

So just to wrap my head around things, mixing and matching different network transceivers with the Broadcom IC will not work unless it operates within the same voltage bias. If the Broadcom IC is 1.75v, would that mean that I am NOT able to drop in, say, a OrangePi or NanoPi that may operate on 3.3v? Or does this schematic basically rectify that and allow the use of any network transceiver operating on any voltage bias?
I am looking to use a SBC in my portable and I hope that it was as simple as dropping in 50Ohm resistors as well as the 100nF capacitors.
Please excuse me if this question is silly, I just feel a bit lost with the Micrel documentation.

 

[MRKane]

So just to wrap my head around things, mixing and matching different network transceivers with the Broadcom IC will not work unless it operates within the same voltage bias. If the Broadcom IC is 1.75v, would that mean that I am NOT able to drop in, say, a OrangePi or NanoPi that may operate on 3.3v? Or does this schematic basically rectify that and allow the use of any network transceiver operating on any voltage bias?
I am looking to use a SBC in my portable and I hope that it was as simple as dropping in 50Ohm resistors as well as the 100nF capacitors.
Please excuse me if this question is silly, I just feel a bit lost with the Micrel documentation.

The question isn't silly at all - this took me weeks to get my head around, and even longer to pick the Micrel documentation apart and create a working solution!

In theory to get the SBC you've got talking to your PS2 I'd first try the 33nF connection that GMan used (just in case it's actually an easy fix!), and failing that you'd need to figure out what the bias voltage for the PHY chip on your SBC is and use that for the pullup on the SBCs side of the circuit (in my schematic it's the 3V3 line from the Mango side). If you're lucky you could lookup the PHY chip it has and get a datasheet for what the expected value is, but otherwise you may have to go with a best guess, which is probably 3.3v knowing how SBCs are made, or seeing if you can look for traces that might go to different regulators, or simply trying all the different voltages on the board (but if you do this please start from the bottom up least you cook the chip!).

Spoiler: Mango nudes for examination

Now from my reading there are some systems that aren't friendly to capacitive coupling, and I've read a lot of numbers about them but it didn't make much sense and for these slightly more elaborate circuits may be necessary, so please don't hesitate to bug me about it if you have trouble getting it going why my memory is fresh!

 

[PvtDerp]

Ah Hah. Thanks for the clarification, This absolutely does clear up my concerns. Thank You!

Looks like I'm going to need to do some more testing on the bench first, before I put it into the portable.

 

[MRKane]

I've an update for this thread! So I really couldn't get this to work in my competition entry and I've spent the better part of a month trying to figure out why it wouldn't go. Recently I sat down and did a two-week deep dive as to why I just couldn't seem to get the two to talk to each other despite having success in the past. The solution turned out to be that the first time I'd accidentally used the pull down on the ethernet lines on the Mango (or is it pull up in this case?) to RX0 instead of 3.3v. Connecting as such results in things working pretty much flawlessly, and I have no idea why it works.