How to determine IBM USB 6 pin header pinout?

I have two IBM Power 7 machines - Power720 and Power740 (8202-E4B and 8205-E6B) and both have an internal USB 6 pin header (labelled J3_TAPEUSB_SIGNAL). These are for connection to a single device (either an LTO drive, an RDX drive, etc.).

The IBM cable is either unobtainable or the price is insane, so I decided to just make one myself.

I am going to mount an RDX drive (dock). I already got the parts (the RDX, a USB Type B cable and a Dupont style 2x3 connector). Now the problem is, how do I determine the header pinout (using a multimeter, etc.) without frying the motherboard, port or any device attached to it?

I searched all over the place for IBM-related pinout information to no avail. So this is basically a hands-on job, hopefully without toasting anything.

Any help or pointers appreciated.

EDIT:

These are resistance and voltage readings I took as per @KH's recommendations:

As for other components/devices nearby, this is it (bottom right SAS backplane cable removed for clarity):

I tested impedance on the front and back USB Type A ports (receptacles) and got 1.2 Ohms on pin 4 (which can maybe match pin 3 top right on the header). All others infinity.

All readings done against the chassis.

• You can start by attempting to find the ground pin. Power down the mobo obviously, and meter voltage (to confirm 0) and resistance to ground on all 6 pins. Hopefully one should have much lower impedance. For +5V, you can power the motherboard and meter voltage on all 6 pins, and the remaining 4 pins should be 2 data sets. Inspect thoroughly to figure out which of those 4 are paired with each other if you can, and hopefully someone around here has a method of determining data wire polarity. Devices may not detect if they're reversed, I'm just not sure of potential for damage. – K H Jan 17 at 3:19
• Once you know which are power and ground you may be able to just connect the data wires both ways to see which one works. One of the answers to this question does not imply that devices will be damaged, but that is hardly a guarantee. I assume your measurement tools are limited to a multimeter? – K H Jan 17 at 3:20
• do you know the part number for the cable? – jsotola Jan 17 at 4:56
• ebay.ca/itm/… – Sunnyskyguy EE75 Jan 17 at 5:58
• @SunnyskyguyEE75 That photo shows a cable assembly that definitely doesn't fit in this connector (just count the wires...) – Marcus Müller Jan 17 at 8:28

Going with the conventional numbering

┌── ──┐
│5 3 1│
│6 4 2│
└─────┘


From the measurements. 1 and 4 appear to be ground, 2 is 5V/VUSB, and 3 is the chassis. That also matches the components, and the track widths - the DRC has made the power and ground lines thicker than the chassis ground (Which the DRC probably doesn't know is a power net). The component south of pin 2 is a suppression choke, and a little capacitor between chassis ground and ground to help EMI, but double check your measurements before you hook it up.

That just leaves the thin signal tracks on pins 5 and 6 for D+ and D-. Fortunately nothing terrible happens if you have these swapped - just things don't work. Given sod's law you'll probably get this wrong first time round...

The only USB cable for that motherboard, that I can find pictures of on the web, is an IBM 46K4646. That has an 8 pin header - but with the pinout

1 brown
2 red
3 shield
4 black
5 green
6 white
7 nc
8 nc


That looks like a plausible match for your connector, so as a first guess I'd go with green (D+) to pin 5 and white (D-) to pin 6.

so long as you don't short 5V to ground you should be fine. The USB spec. says that host controllers have to be tolerant of D+ and D- being shorted to +5V (originally for ever, but now for minutes) and ground.

• Thanks James. Very thorough explanation! I noticed the same OP_USB 8 pin connector upper left that leads to the front USB for which I thought would be the exact same pinout for this connector, but with 6 instead of 8 pins. The connector key is also 180 degrees on the motherboard, and, save for the 2 extra non-use pins, basically should be the same. I also noticed my readings were off (miscalibrated) by about 1.2Ohms, which basically renders pins 1 and 4 at 0Ohms (and pin 3 at about 1.8Ohms), matching exactly what you said. – Molasar Jan 21 at 0:42
• This connector here: electronics.stackexchange.com/questions/417763/… – Molasar Jan 21 at 0:51