80-Bus News

A Strange Drive Fault

Thinking about faults you don’t see, try this one. A brand new Galaxy straight out of its box was being fitted with the Winchester Technology colour card for Prestel compatibility. There it was on the bench with its lid off, monitor alongside, being tested. At this stage, had we put the lid back on, then, of course, a certain immutable law would have guaranteed that the colour card would not to work. Anyway, everything worked fine except, try as we might, we could not get drive B: to give us anything else but disk read errors. Of course the obvious ocurred to us – Gemini – God bless their little cotton socks – had supplied a duff drive. It could happen, especially if it was made on a Friday and if someone had had a particularly liquid lunch that day. So we put in a new drive, and guess what, it didn’t work either. Well I have this little theory, it’s been printed in these annuls before, but repetition is good for the soul, so here goes:

One dead one is tough
Two dead ones is coincedence
Three dead ones and something else is wrong

So after the third drive failure, the drive card was changed – no difference. So the colour card was removed – no difference. So the processor card was changed (for no good reason, except desparation) – no difference. We even thought about changing the box, but even Graham at his most pessimistic could not advance any valid theory as to how the box could affect drive B:. So what was it? Well I gave you the clue at the beginning, have you spotted it. “The lid was off and the monitor was along side.” Now monitors have fairly powerful line timebase stages (after all that’s how the man in the TV detector van knows you have the box on but no TV licence), and line timebases chuck out an awful lot of low frequency RF muck, about the right sort of frequencies disk drives are wanting to look at, so guess what. Move the monitor, put the steel lid on the Galaxy, and that was the answer to another wasted afternoon.

Keyboard reliability

And so on to another tale of woe concerning, this time, the Rotec function keyboards used on the Gemini Galaxy machines. Apart from my not liking the layout of the cursor keys, a certain amount of trouble has been experienced with the earlier ones. To put it baldly, the early keyboards are unreliable.

One of the causes of unreliability is the egress of muck, and the failures all stem from the way in which the keyboard works. Now I haven’t seen the circuits but it is obvious that the keyboard uses capacitive keys. The pcb has areas of track in the form of two pads under each key, which are separated by about 2 – 3mm. When a key is pushed a metal pad on the bottom of the key is pushed down so that the pad on the key covers the two pads on the pcb. The pcb is heavily lacquered so that there can be no electrical connection between the pads when the key is pushed. The coupling between the two pads is therefore affected capacitively, the keypad joining the two pads on the pcb with a capacitance of a few picoFarads. I understand that the key sense lines have 2pF capacitors in series so that the few additional pF’s caused by the key closing cause a change of between 0 and something a bit less than 2pF. The 2pF capacitors are there to minimize the reduction in capacitance caused by muck under the keys, but read on. Now I guess the on-board processor (yes the keyboard has a processor all to itself) must be scanning the pads by sending sharp pulses across the keys to check for coupling between them. When it finds a couple of pads joined capacitively, it does the rest, works out what key it