Problaby one of those 6500-series microcontrollers, which were used in some disk drives... And it's old enoguh that the Beeb should have taken account of that design .Perhaps it was just so obscure that nobody over here had come across it (I certainly hadn't...) However, can you step it slowly enough (sensibly) that this is not a problem? If you snd, say, 12ms or 20ms-spaced step pulses, will it remain ready all the time? For pure bad deisgn, nothing beats the Shugart SA4000 hard disk, IMHO. On that thing, you can eitehr send the step pulses one at a time, waiting for hte 'on cylinder; signal to go active after each one. Or you can send a burdt sufficiently fast that the drive will buffer them, then move the head and assert 'on cylinder' when it's done. But if you send the burst of sstep pulses too _slowly_, the drive wil lget confusedand end up o nthe worng cyulinder. Ther reason (and I have the schematics...) is that there's on coiunter that is controleld both by the incoming puses and by the stepper motor derive clock. If the heads are moving when a step pulse comes in, there's a good change the control logic will be in the wrong state and will consider the incoming pulse to be a stepper motor clock pulse or vice versa, so the counter changes in the wrong direction... -tony

Except it wasn't an early design. OK, it was an early-ish wincheaster, but stepper motor controllers are much the same whatever you sue them for. The problems were known about and could have been avoided. -tony

I don;t know abotu the OEM manual becuase I don't have it, but I would assume it was in there. The service manual, which I do have, says : 'In order to be valid, step pulses must be at one of two rates. I nthe normal mode there will be 1.1 msec or more time delay between incoming step pulses. In the buffered mode, there will be less than 350 $\mu$sec time delay betwene them. If step pulses are incoming 600 $\mu$sec apart for exmaple, then every other pulse would be lost causing the heads to move only half the required number of tracks. The reason for this is that the stap count timer locatad at positon 1A which counts down to zero in apporoximately 500 $\mu$secwould output a low pulse. This pulse then clocks chip 5B blocking out incoming step pulses for another 500 $\mu$sec while cjip 1A once again counts down to zero generating a single track step and a SEEK COMPLETE signal. At this time another step pulse is allowed to enter, but the precious one was lost.' -tony

Well, the Beeb was gnerally well-designed and stuck to stnadards (apart guessis that they would have prefered you to buy their own drvie system, but werne't goiing to stop you using whatever you liked... My guess ithat buffered-seek drives were sufficiently uncommon that either they'd not come caross them or thought the problem would almost nver arise. [...] Full height drives were ocmmon on Beebs. The original Acorn-badged drive was a full-height 40 cylidner single-head thing. It was also unreliable. I never got to see inside one, so I don't know who made the mechanism, but it wasn't one of the common ones. I've seen plenty of other full-height drive systems used on beebs too, normally using Tandon TM100s (in one of their many versions) or similar. -tony