[sdiy] [OT] puzzling EPROM problem

[Magnus Danielson]

From: "Trevor Page" <trevor at resonance.fsnet.co.uk>
Subject: [sdiy] [OT] puzzling EPROM problem
Date: Thu, 20 Sep 2001 22:22:26 +0100

> Hi list,

Hi Trevor,

> I've got a funny EPROM problem here. Sorry for the OT, but this is probably
> the best place to ask this sort of question.

Indeed. However, as I read through it I also think it applies to ower
older synth friends. Some of them might actually experience similar things.

> I'm repairing two Martin Imagescan projectors (dirty great big heavy things
> that spin strange images all over the place in nightclubs!!!). Without going
> too much into detail, the projectors were failing to initialise properly.
> 
> They use an Atmel CPU and have an EPROM (27256) containing programme
> information. The EPROM versions in these projectors was V3.0. Having spent
> ages on these things and coming to a complete loss as to what the cause of
> the fault was, for a laugh I tried reprogramming one of the EPROMs with the
> latest image, V5.0.
> 
> That caused the projector to work perfectly.

Ah. Interesting.

> Before reprogramming the EPROM, I verified it's contents against a V3.0 mask
> that I downloaded also, and it verified correctly. So it's not as if the
> EPROM contained bad data in the first place. V5 may well have bug fixes and
> so on over V3 - but it's strange how updating the firmware like this would
> make a suddenly disfunctional machine work again.

Naturally, you should try to load the old firmware once again just to
see if the errors comes back or not. I don't think they will thought.

> When an EPROM is programmed, or with the passing of time, is it at all
> possible for certain bits, or 'fuses', to be only partially 'blown' so that
> although the logic levels of those bits are such that they satisfy the EPROM
> programmer that they're correct, these particular bits read out at funny
> voltages when the EPROM is in use, and hence the CPU reads them wrong? But
> then again, the outputs are surely buffered, aren't they.

Let's forget about fuses. Those exist in PROMs but not EPROMs.

EPROMs operate with a floating gate. This means that one has a little
piece of deposited aluminium in the siliconoxide. When programing one
deposit a charge onto it. The isolation is so great that it takes the
energy of an UV lamp for the electrons to jump back. There is however
a slow leakage mechanism, so the deposited charge will very slowly
evaporate.

> The EPROMs are being accessed pretty fast. The CPU is on a 24MHz xtal. If
> this means anything. They're 27256-10's.

It could be, but not necessarilly have to be so, that the slow access
of these reads correctly where as a quick access can fail.

> One other possibility is that the CPU is detecting a certain fault condition
> from one of the sensors, and whereas V3 would just hang up in the event of
> this happening, V5.0 will carry on regardless. But there are really very few
> inputs to the CPU - DIP switches for mode settings, a microswitch and an
> optosensor - all of which work fine.

Try loading the old firmware back again.

> Any thoughts? It's bad enough not being able to repair something, but it's
> worse having got something working and not knowing why...

Indeed. I hope you got some dash of insight which can come in handy.

> Trev (sitting in a dark room, with psychaedelic images floating around the
> walls and ceiling... :-)

Trev, please come down...  ;O)

Cheers,
Magnus - about to hit the bed