Populating empty USB sockets on a PIIX3 motherboard (Micronics W6-LI, 440FX) \ VOGONS

Populating empty USB sockets on a PIIX3 motherboard (Micronics W6-LI, 440FX)

Topic actions

Post a reply

First post, by RussD

Posted on 2022-08-25, 18:54

RussD Offline

Rank Newbie

Rank: Newbie
Posts: 46
Joined: 2022-06-28, 08:04

Many PIIX3 based motherboards were designed with USB support in mind, but shipped without the necessary sockets and passives. Even though the USB support provided by the PIIX3 is pretty minimal (only 12mbps, most BIOS's don't provide any support ...no hit f2 to enter setup, navigate boot menus, etc..). But it's still a fun project and it's something handy to have. There's basically 4 areas to look at. The socket itself, the filtering and fusing on the power rail, the passives connecting the signal lines, and any necessary passives (termination, etc) for the USBCLK signal. A PIIX3 identified by the marking "Intel" "PCIset" "SB82371SB".

Anyway, here's what you might typically see around the USB socket:

The socket itself is bog standard. A wide variety of connectors will do, I chose a TE 5787745-4. Some

The little three pin components, L10 and L11, are LC EMI filters:

Filename

gnome-shell-screenshot-3k3jm3.png

File size

3.4 KiB

Views

308 views

File license

Public domain

So power goes in one one end, comes out on the other and the center is grounded. In a pinch, you could probably just bridge pin 1 to pin 3. In this case I used DSS1NB32A331Q91A which has a 330pF capacitance. F2 is the shared polyfuse for both sockets. I chose a Littelfuse RUEF250 with a 2.5A hold current and a 5A trip current. L12 and L13 are filters for the ground signals from the USB ports. This is pretty uncommon. These can probably be shorted, in my case I used 220 ohm ferrite beads. On the backside we have some more components:

Since the LC filter provides a small value capacitance for high frequency components, these are either the other half of a Pi filter, or meant for bulk capacitance, but the footprint is a little small for that. Probably not too critical either way as it's just for full speed USB devices and there's probably a bunch of bulk decoupling in this area anyway. I went with a couple of 560pF capacitors, but I can certainly see going with something larger.

Because motherboards of this era are typically wave soldered, all the unused through holes will be populated with solder. If you are determined enough, you can get this out with a combination of wicking, solder sucking, fresh solder, flux, wicking, and a lot of frustration. If you have a hot air station, using that on one side while using a solder sucker from the other side is magical. Anyway, components ready to go:

Bit of soldering later:

And that takes are of the socket and power filtering/decoupling.

Reply 1 of 2, by RussD

Posted on 2022-08-25, 19:20

RussD Offline

Rank Newbie

Rank: Newbie
Posts: 46
Joined: 2022-06-28, 08:04

The next step are the passives on the data lines. The PIIX3 datasheet indicates that there should be a 15k pulldown on the USB port side, a 27 ohm series terminator near the PIIX3, and a 47pF filter capacitor on the PIIX3 side of the series terminator. Tracing out the signals leads to this collection of unpopulated components on the back of the motherboard near the PIIX3:

It's pretty clear here by probing or even just taking a look at the traces which are which. They are 0805 or maybe 1206 footprints but rather than ordering some fresh components, I just used what's in my component drawers which include 0603's and 0402's. They look a bit wonky, but hey, it's the back of the board.

And the last bit is the 48MHz USBCLK. Depending on how this is hooked up, there may be some termination passives to populate. Probing around should let you know what you are dealing with. In my case, R101 and C63 form an AC terminator adjacent to where the clock signal hooks up to the PIIX3:

There's no hard and fast rule on these. There's a PCICLK nearby that's also terminated with an AC terminator with the resistor being 330 ohms. The capacitor is impossible to measure without depopulating it. There's a few rules of thumb out there for picking a capacitor value using the signal rise time and the trace impedance. The signal rise time is 1.2ns, and the impedance is whatever my best guess is. The exact value isn't super critical here. It's only 48MHz after all. I went with 47pF because I already had them out, and 270 ohm on the resistor because it's what I had in 0603.

Good to go. BIOS support for programming the chipset may be necessary and of course appropriate OS drivers are necessary as well. Without BIOS support it may be possible to poke the right register values early in the boot sequence to get things going. In my case, I needed to update my BIOS revision to get things working.