Ampex 219

[MMcM]

This is a keyboard from an Ampex 219 terminal from the mid '80s.

Ampex-219-case.jpg (1.16 MiB) Viewed 1596 times

Ampex-219-back.jpg (1015.63 KiB) Viewed 1596 times

The key layout is almost identical to the Ampex 230, discussed here before and photographed in the wiki and on Bitsavers. So much so that you could almost believe that one of them has had the keys put back on wrong.

Moreover, the PCB is almost the same, except for the placement of the buzzer.

Ampex-219-top.jpg (1.27 MiB) Viewed 1596 times

Really the biggest difference is that the 219 has Stackpole switch grids with double shot caps and the 230 has NMB Hi-Tek 725 series linear switches with dye-sub caps.

Ampex-219-bottom.jpg (1.17 MiB) Viewed 1596 times

(Compare 230 top and bottom from Bitsavers.)

The caps are pretty thin and the switch travel isn't that deep. Plus this board has the usual corner cracks. Therefore it isn't all that great a keyboard to type on.

The 4P4C RJ plug has the same pinout as given for the Ampex 210 on kbdbabel, so I assume the whole series has pretty much the same protocol, maybe even the same codes. (The 210 lacks function keys.)

• chassis ground
• Vin
• data
• signal ground

Apparently some spec said 11VDC for the voltage, but it just goes to a 5V regulator, so most anything will do. I've got 12V.

The remarkable thing about the components on the PCB is that the MCS-48 is the only digital element. There is no 74 series TTL.

Even though there is only one data line, it is bidirectional, having two open-collector circuits, using just NPN transistors, separated by a diode. Which, of course, means that the protocol has to always know which side is supposed to be pulling the data line down.

A 4.9125MHz crystal might suggest that someone intended to use a standard baud rate, but that doesn't actually seem to be so.

It is bit-serial, but with an odd frame size.

• start bit
• three bits of shift state: shift, ctrl, func
• eight bits of key code
• stop bit

Nothing is sent on key up. In the other direction, the terminal has to ack each frame. I found it more reliable to just strobe periodically in case something got lost. Assuming that it's possible for the terminal to turn off the clicker and sound the buzzer, there are probably some longer sequences of bit changes in a short period of time, but I didn't work them out yet.

The bits are around 725µs long, slightly faster than 1200 baud (833µs). Slow enough that there is no real need for anything more complicating than reading it directly in a loop, which a quick QMK implementation does.

I used a Teensy 2.0 because I recently restocked those, but I believe the FETs would allow a 3V microcontroller to work as well.

Ampex-219-converter.jpg (447.78 KiB) Viewed 1596 times