Devlin hardened keyboard

While I was recently at the lovely HACMan, I bought from them, for the tiny price of £10, a kiosk-style hardened keyboard/trackball combination, a Devlin with serial number D107922. (I say "a Devlin", because there's nothing on it approaching a model name.)

So, naturally, one of the first things to do is to disassemble it, because I'm me. The power is noted to be 5V, 0.3A, and "the power supplied to this product should be limited to 15VA MAX". (Note that this is precisely 10 times 5V * 0.3A.)

There are two covers on the back. The large black cover is held on by two PZ2-headed screws, and reveals two connectors, CON1 and CON2, each four pins: black, orange, red, brown. CON1 is "keyboard" and con2 is "trackball". I am reliably informed that these are PS2. There is also a two-pin 0.1" jumper header, marked J1, and a resocketable microcontroller(?) covered by a sticker marked "DEVLIN \n ElectronicsLtd \n\n DSL-600-1"[1]. (At least, that's what I think it says, the last line especially is somewhat flaking.) There is also another sticker, marked "DEVLIN \n 371127".

The other cover is a piece of steel held on by four bolts. The stacking on these is bolt held staticly in the front-plate, trackball module, washer, plate, lock-nut (with round side into open air). The nuts are 7mm across flats, implying that they are m5. The trackball module seems to be, somewhat suprisingly, a completely seperate module, made by a seperate company and simply mounted whole. The module is, according to it's sticker, from Penny-Gilescp, which seems to have been bought by / spun off as / whatever as traxsys. It is "D No" D509147/3 and "S No" 3211500. It has two populated connectors, P2 and P1, which have their pinouts semi-marked. P1 is for the buttons (which aren't part of the module), and P2 seems to be PS/2. P1 is M R L , with wide spacing. I think this is middle button, ground, right button, ground, left button, ground, going partly from the datasheet for the DS509409, which seems similar. P2 is marked NC, GND, +5V, ADB, CLK, DATA. This is somewhat interesting in-as-much-as the PS/2 mouse standard not having a line called ADB. In fact, it has only four lines -- GND, +5V, CLK, and DATA. It's quite possible that the ADB line is for some sort of test purpose. There's also an unpopulated 8-pin connector. Given it's location, and that the four pins on one side seem to be connected to each-other, I'm betting it's JTAG. The microcontroller seems to be an atmel 80c52x2-lc. A quick look finds this Philips datasheet, which tells me that 80c52x2 is an 8-bit microcontroller, with 256 bytes of RAM and 16KB (mask) ROM.

The trackball unit splits into PCB and ball units with three pz1 screws. A further examination of the unpopulated header shows that on the mostly-ground-plane side of the PCB, it is marked as four seperate headers, P1..4, each of which is a ground and a single signal pin, which I suspect to be the four signals from the optical encoders (X, Y * A, B, since they will be quadrature encoded).

The main back of the keyboard is held on by 16(!) counter-sunk PZ2 machine screws. M3, I think. You need to remove the grounding straps using a 6mm AF wrench, as well.

Removing this cover allows you to see the female connectors that mate to the male connectors on the trackball unit.

   Keyboard   Trackball
   CON4       P1
   N/C        (blank)
   LEFT       L
   N/C        (blank)
   RIGHT      R
   N/C        (blank)
   N/C        M
   (unmarked) P2
   CLK        CLK
   RTS        ADB
   VCC        +5V
   GND        GND
   N/C        NC

The mysterious "ADB" line appears to be the slighty less mysterious RTS, and we see that they haven't even attempted to show where the middle mouse button might be connected.

You can also find that under the not-cut-out in the back is CON5, EXTERNAL SWITCH, which is a four-pin header, and a label for assembly steps. Batch no is 9H29 (or possibly GH29). P. Flow is 19, instr 2 is stampted S, and Test 1 is a squiggle that might be a w or a sideways 3. AMISTAR, COMP. INS, INSP. 1, SW.ST.MASK, SHRTG.FIT, and KEYTOP are all blank. There's also text in the copper layer reading:


Further disassembly requires removing the 16 spacers for the 16 m3 bolts we removed earlier, and 6 6mm AF lock-nuts.

The front side of the board has a couple of features of note. The switches are lovely beasts, with a very satisfying clock, and a nicely designed plastic top which could keep keycaps straight and true. The extra-wide keys have multiple buttons under them, one on each side (left and right mouse button and enter), each marked SWNa and SWNb. The space bar has provision for three, a, b, and c, but b (the middle) is unpopulated.

There is also a fairly chunky PTC resistor, presumably to keep the keyboard from burning itself or something else if water ingress, blocked vents, etc, cause it to overheat.

  1. I'm using the C-like " \n " notation to mean "new line". ↩︎