80-Bus News


November-December 1983, Volume 2, Issue 6

Page 10 of 67


is and generates the appropriate ASCII code and a strobe. Fine, it sounds good in theory, and given good close proximity of the pad on the key and the pads on the peb, it works.

How does muck affect the keyboard? Where does it go wrong? The lacquer on the peb is quite thick as far as pcb lacquers go, but is still only about 0.05mm. I haven’t worked out the capacitance involved, but the capacitance is proportional to the area of the pads, and inversely proportional to the distance between them. This is also multiplied by the permittivity (dielectric constant) of the lacquer, which I will assume to be about 3 – 4 (air has a dielectric constant of 1). Muck under the keys holds the keypad off the pad, and the chunks of stuff I’ve found under the keypads is at least as thick as the laquer. So the muck halves the coupling capacitance for a start, next, because the keypad no longer makes contact with the lacquer (except for a small area) the dielectric is now air and not the lacquer, let’s be generous and assume that this only reduces the permittivity by half. The capacitive coupling between the keys and the pads is now only a quarter of what it was, so the capacitive change detected by the CPU is now between O and some rather less than the 2pF (minus) which was present before.

Now this has pugzled Rotec, as the keyboards were designed with a certain amount of muck in mind, after all that’s why the sense lines have 2pF in series, to minimize the affects of the muck. After much poking around, they have discovered that a whole batch of keyboards were made with a 10OnF ceramic capacitor which was marked as 10% tolerence component (to their spec) which was in fact a wrongly marked -40/+100% component intended for use as a decoupling capacitor. This of course degraded the sharpness of the scanning pulses, with the predictable result that when the coupling capacitance was reduced by muck, then the key failed to work.

Two further things compound the liability of key failure through muck egress, the first batch of keyboards made no attempt to keep muck out, the next and susbsequent batches have rubber strips round the edges between the peb and the keyframe to stop muck getting in, and between the keys to stop any muck that does get in from moving around (and also to cut down noise). All well and good, except there are 16 unfilled keyholes on the top of the keyboard frame covered with sticky tape (not really tape, it’s quite thick). Sad to say, this can peel off after a few months use leaving half inch square holes on the top of the keyboard for the odd passing brick to drop through. Secondly, the pads on the bottoms of the keys seem to be slightly sticky and any muck that gets under a key promptly attaches itself to the keypad and stays put.

So onto the cure. Sling the thing back at your Gemini dealer, who in turn will sling it back at Rotec, via Gemini, and shout at them. Actually, I understand Rotec have been very good over this and all faulty keyboards returned have been sorted out. That of course is the legal recourse and you would be entitled to do just that, (provided that it is till under warranty) unfortunately, it’s not a lot of help if you have just written a couple of hours rubbish using PEN (not having backed up of course) and then discover a passing brick has got under the CONTROL key, and all PEN does is print 222222 instead of breaking out of the insert mode (*Z, for those who don’t know PEN). That’s what happened to me once, which is the reason for this bit. The well tried cure is to turn the keyboard on its end and ‘thumping it one’ (known in the trade as a technical tap) in the hope that the muck will get dislodged and the key will become functional. Needless to say this didn’t work. So in my wisdom I decided to take the keyboard apart. When I’d fixed the keyboard, I found that the action of removing or replacing the keyboard had crashed the video card, and using RESET to clear it lost about half of what I had written.

This is an OCR’d version of the scanned page and likely contains recognition errors.

Page 10 of 67