deskthority

JP!'s recent finds had me taking another look at the SD boards that are sitting around here waiting to have something done with them.

This board, 69SD5, is the guts of a data entry terminal for a Burroughs TD-730/830. That is, it is more or less the same board as pr0ximity's in Burroughs TD016 "Data Preparation". Or the smaller sibling of another one here that still has its case, as discussed in Burroughs TD 830 Series Keyboard.

Specifically, it is the Latin American / Spanish variant. Of course it's the same PCB, SD-16201. And while pr0ximity's is 69SD5-8-1, this is 69SD5-13-1. The date codes are 7811 and 7923, respectively. So I imagine the 13 is the key cap variation. And perhaps firmware, although the one here doesn't have its AMI LSI controller at all.

Some of the keys look like they got scratched by some through-hole pins or something and it has one replacement key cap. Though there is one just like it with the same date code but complete and undamaged on eBay right now for €200. So there's a cache of them out there.

The earlier topic didn't take a close look at the circuit.

The keyboard interfaced to the cable through a 2x15 PCB edge connector. The top side has +5V on the left two pins and then connections for its 12(!) LEDs, of which 8 are populated here. And the final pin is the other side of the LEDs' several 220Ω resistors. The missing 4 LEDs are also missing their common resistor.

The interface to the keyboard proper is on the other side. Two pins of ground, the dozen outputs of two hex inverters, and a connection to the last pin (28) of the controller. The switch traces still have a bit of Chutes and Ladders that suggests not a matrix proper but two-of-n, like with SW series. And, indeed, almost all the switches are 4A3A. The beautiful baby blue command switches are a bit heavier and pulse only 4A8B.

The SHIFT LOCK is exactly that: the two SHIFT switches (61 and 48) and it (35) are all the same signal to the controller. These all have the dual outputs shorted together with a brief PCB trace.

Which leaves the strangest thing about the switch wiring. The RTAB (20) switch also has its outputs shorted. And it's a source level 4A3C and connected directly to one of the output inverters, skipping the controller. I cannot think of any reason to engineer it this way. Why would this function need special treatment? I am almost inclined to believe that the PCB was designed for something else where this was some kind of attention key and reused for this terminal. Can anyone else think of something?

I am still debating what the best way to build a converter is, because of the need to access both the switch "matrix" from the MCU slot and the LEDs from the edge connector.

In a spot check, some of the keys didn't work. Every one of the scan lines got at least one hit, so it isn't a mistake there. May need to scope those switches and see whether they got damaged somewhere along the way.

why are these keyboards so expensive? who is buying them to stack on a shelf in a museum.

gipetto wrote: 15 Jun 2020, 01:43 why are these keyboards so expensive? who is buying them to stack on a shelf in a museum.

Not sure what you mean but here but this particular module from that eBay listing has been for sale for a very long time at the price it sold for so maybe not exactly a deal. It seems there is less and less of this sort of stuff still around as a lot of it made it to the landfill long ago. They don't make stuff like this anymore and now there is more interest thanks to hobbyists who have an interest in bringing old electronics back to life. I personally have my own unique interests as far collecting. Certain boards that might command a hefty sum might be of little interest to me. If something like this were converted and put into a new case that would be something unique. Look at all of the modern custom keyboards built today. Lots of geeks enjoy building unique handcrafted peripherals. These can easily cost hundreds to build which might seem crazy to many people.

Is there anything better than old curly hand drawn traces? Love it.

I managed to find another of these keyboards. This is 69SD5-11-1, the French variant. The SD-16201 PCB is the same, so the wiring is essentially identical. This board has things that were missing from the Spanish variant above:

• The CLR EOL / EOP keycap.
• The AMI SW-20253 encoder.
• All twelve LEDs.

This gives a chance to confirm the details of the protocol. With all those LED signals on the input side of the edge connector, a Teensy 2.0++ is needed. And the CMOS needs -12V on S (it is actually probably fine with any low negative voltage). The output side of the edge connector is strobed parallel:

• There are, in fact, two direct keys: switch 20 (RTAB) on R and switch 63 (CHAR INS) on P.
• The next, N from pin 24 of the encoder, is, as usual, the strobe signal, idle high.
• The character data bits, D0-D7, are C-L from pins 15-22.
• The remaining signal, M from pin 23, is effectively a ninth data bit. It is synchronized with the strobe signal, but only set on a small number of key combinations: SHIFT CTRL and CHAR DEL are the ones I have identified so far.

It is mostly ASCII:

• The number keys and their shifted forms are regular 7-bit ASCII.
• So too most of the keys for punctuation.
• The letters only send the low five bits, that is, the CTRL form. So 1 for A.
• SHIFT does not have any effect on letters.
• So too [, although { is available shifted as expected.
• Command keys send codes with the eight bit set.

I cannot find anything in the documentation to suggest that the Data Preparation Keyboard was uppercase-only, but that was not unheard of around then. It does describe the somewhat unusual CTRL function. It is not a shift as we expect today. You press (and release) it, the CTRL LED lights up, and the next key has CTRL applied. Or you press SHIFT CTRL and CTRL stays lit and functioning until you do that again.

It is easy enough for the terminal to get back printing ASCII: if the code is less than 32, add 64 (or the bitwise equivalent, if that's cheaper). I am not sure whether it ever doesn't do that, such as to implement that CTRL form. If so, I'm not sure why the current setup was deemed easier than bit-anding real ASCII with 31 to do that when necessary.

The way these MOS encoder chips work, CTRL and SHIFT are not scanned as part of the matrix (or, in this case, two-of-n). Rather they are input directly. And indeed, both SHIFT keys and SHIFT LOCK all connect to pin 14. Pin 13, which would be expected to control CTRL, is unconnected -- there are no traces to it on either side of the board. These shift key signals are active low and often have pull-ups outside the chip. So I wonder what it does when not connected and whether the encoder is programmed as though CTRL were always pressed. It doesn't matter, really, since I believe the OEM process was to specify every code for each of the four shift combinations and just put this into ROM.

The four new LEDs do not work yet. That is because there is a 220Ω resistor needed in position 24. It can be seen as missing on the far left of the photo (top right of the keyboard). The solder isn't pristine, though, so I am not sure whether it was removed or never successfully added. I will probably add one, on the grounds that this is restoring functionality, not altering it per se.

This second keyboard has the same problem as the first one: about a third of the switches don't work. This is really quite surprising, since the SD-series is famously reliable. There is no sign of anyone messing with how the sensors are soldered in, which is delicate. And the switches (mostly 4A3A) all seem physically intact with no signs of corrosion.

Since the voltage from USB was near the low-end of the rating from the datasheet, I tried directly powering with something solidly at 5V but that did not make any difference. There is a big electrolytic capacitor across the power line, but it does not show any signs of swelling or leaking.

Most of the broken keys do not ground either of their outputs. A few are missing only one of the two.

Hooking up to a scope does not reveal much, except that a couple of the bad keys do have a small ripple when pressed.

A working key will stay low as long as pressed -- it is the encoder the loses up transitions. Here are Q (2 + 4) and W (2 + 3 -- zoomed 4x). Here is A (3 + 4), where something happens but not all the way to ground and only for a few hundred microseconds. So it really does seem like the Hall Effect circuit is damaged. Could it be something about how these keyboards got from Burroughs terminals into the surplus market?

Do you reckon you could handle this fella?



I'm too thick at electronics / protocol logic to really understand what you're doing, but I see a glimmer of hope for the original gangster Round 5 keyboard in these Micro Switch conversions of yours!

Muirium wrote: 24 May 2021, 13:30 Do you reckon you could handle this fella?

See here, which if not the same is very close.

It's at the borderline for what can be made into a USB keyboard without rewiring: it sends key down codes (not ASCII and not affected by shifts) via strobed parallel, with the shifts output separately. So you can't tell when an ordinary key is released and you can't add / change shifts from CTRL and SHIFT.

@Muirium

I think both of us can do this conversion: viewtopic.php?f=7&t=24822
Looks like the same set up, although your board probably have a completely different protocol, that should not be of any consequence because you have the same controller I think (intel 8048) and the same multiplexer, so its the same where it counts.