Power supply connector pin crimping, and 8-inch FDD power connectors (another chapter in the Quay 900 saga)

Eric Smith spacewar at gmail.com
Thu Aug 20 22:16:07 CDT 2015


Last week I wanted to test some half-height eight-inch double-sided
drives (NEC, Mistubishi, and Qume) on the Quay 900.  I cabled a
Mitsubishi drive in place of the original CDC/MPI 9406 77618022
drives, and the machine apparently wouldn't reset properly, since I
wasn't getting the prompt from my ROM monitor either on power-up or by
manually resetting it.  I poked around a bit and discovered that the
+5V DC supply was at about 0.7V.  I disconnected the Mitsubishi, and
it still didn't work, and the +5V was still at 0.7V. Uh oh, what did I
break?

After a lot of pulling of hair, gnashing of teeth, and sacrificing a
chicken at midnight, I discovered that +5V pin in the connector that
plugs onto the switching power supply was not crimped properly. It was
partially crimped, but the wire was just loose.  The cable connector
is an AMP (now TE) 87025-7 "Ampmodu" 0.156-inch pitch shell which
accepts 102103-3 rectangular crimp receptacles.  The shell has 20
positions, of which they inserted a keying plug in position 1, and
only use contacts in some of the even positions from 2 though 20,
because the header on the power supply PCB only has every other pin
loaded.  TE no longer makes the 87025-7, but they still make the
87025-8, which is apparently the same thing without the part number
being stamped on the housing. I don't need another housing though,
just a pin, because without the right extraction tool I hadn't been
able to get the old pin out without mangling it a fair bit.

Mouser and Digikey sell the pins in small quantity for $0.50 each,
which seems absurdly high for a crimp pin with only tin plating.
(There's another part number for a gold contact, but distributors
don't stock it.)

Just for the hell of it, I looked up the TE manual (hand) crimping
tool designed for this pin, p/n 90274-2.  It sells for over $6500.
There is a used one on eBay for $75, but I've had bad experience
buying used crimping tools.  The only crimping tool I have on hand is
designed for terminals with a round shell that just have to be crimped
flat, vs. for terminals with V-shaped edges that have to be folded
back in, as is typical of Molex pins and the like. I decided to order
an inexpensive ratcheting crimping tool from an Amazon seller. It's an
Iwiss SN-28B, also sold under the Estone and other brands. The Iwiss
was $19.  I couldn't tell from the photos whether it would be
suitable.

It turns out that it worked perfectly for the TE pins. It has two
pairs of dies stacked with one pair having a larger profile, so it
does crimp both the conductor and the insulator at the same time,
which I wasn't expecting for a sub-$20 tool.

That got the machine working again, and I verified that the CDC/MPI
drives are still working, or at least working as well as they were
before. I'm still seeing a lot of unreliability when using
double-density on the highest-numbered tracks (closest to spindle).
Could be the wrong amount of precomp, or the low-quality data
separator design. Since one of the two MPI drives gets more errors
than the other, there may be some issue with drive alignment or drive
electronics adjustment as well.

I unplugged the MPI drives and plugged in the Mitsubishi. Once again
the machine wouldn't reset properly.

It turns out that even though this bizarre variant of the 9406 uses
the Shugart pinout for the data connector instead of the MPI pinout,
and uses the same DC power connector as the Shugart, instead of the
header used in normal MPI 9406 drives, the DC power connector pinout
for the MPI does NOT match the Shugart DC pinout, as also used by the
various half-height drives I want to try. I'm becoming less and less
impressed with these MPI drives as I learn more about them.

With the Mitsubishi cabled up to the Quay, but using a separate DC
power supply with the correct pinout, I was able to verify that the
Mitsubishi drive actually works fine.  When used with the Quay FDC, it
does need some retries for double-density on the inner tracks, like
the MPIs, but it doesn't need as many retries as either of the MPI
drives.


More information about the cctalk mailing list