EPROM data life?

[SiriusHardware]

Does anyone have personal experience of how long the programmed contents of a windowed Eprom (ie, 27C256, etc) are likely to survive intact?

For several years in the early noughties I worked for a repair company with an interesting approach to repairing PCBs without any diagrams or even the chassis / equipment they fitted into. The approach, broadly, was to test everything that could be tested and change everything else.

This (expensive) approach was aimed at items like control PCBs from eye-wateringly expensive industrial robots, CNC machines and so on - the sort of things which would cost their owners thousands of pounds a day for as long as they weren't working.

With microprocessor cards / boards one of the things we did was the read out the contents of the eprom, make a backup, and then write the contents back to it, thus refreshing the programmed information for another 30-40 years or so.

I mention this now because it occurs to me that a lot of vintage micro equipment which used eproms for information storage may be coming up to the age where 'eprom fade' could begin to be an issue. (One example might be where you've got a BBC model B with the DFS contained in an eprom).

It's not even always obvious that the devices are EPROMs, because they may be the cheaper windowless one-time-programmable variety, which, nonetheless, use the same technology to store the data as the windowed type does and will be equally subject to 'data fade' in the long term.

[Refugee]

I used to work in a place that did the same with very expensive printers. We did have some circuit diagrams.
We used to copy everything to hard disc and then back onto sets of chips we kept on our benches.
We copied them to blanks as required.
I got one with the MCU blown so that there was a patch of charcoal over where the CPU was.
I decided to try it.
The data copied from the memory so i ordered a new blank chip.
I programed it and was rewarded when It burst into life again as if nothing had happened to it.
They can also be tough too.

[GMB]

EPROM life often depends on how well they were programmed.
The later types had fast-programming methods which tended to be manufacturer specific so scope for incorrect programming.

You are right that the early micros are now approaching the death zone. I have encountered failed EPROMs. But the thing that also does them in is any kind of abuse as although an electrostatic discharge might not always damage the actual circuits it can knock a lot of years off the life of the data.

[jjl]

This 2004 Microchip datasheet for their 27C256 claims > 200 years data retention.

http://ww1.microchip.com/downloads/e...doc/11001n.pdf

This 1993 National Semiconductor sheet makes no mention of data retention duration.

http://pdf1.alldatasheet.com/datashe...SC/27C256.html

I'm sure that I remember reading in the early '80s that NMOS EPROMs such as the 2716 and 2732 could be expected to retain data for 10 years.

I guess that silicon process improvements allow Microchip to make the > 200 year claim.

John

[emeritus]

Back in the 1970's when memory was expensive, a former colleage mentioned that, in his new firm, someone had discovered that non-windowed EPROMs could be erased by "cooking" them in a microwave oven for a few minutes.

[Refugee]

Microwaves can also be used to erase CDs

[cheerfulcharlie]

Kept a PC from around 1998 that had a BIOS eprom in it. Use to run it up every few months or so. Been too occupied to bother in the last year, switched it on a few weeks ago and the BIOS had died so it had to go up the tip.

[Refugee]

That would have most likely been the battery backed part of the BIOS.
It could have been fixed with a 3 volt battery from a pound shop.

[GMB]

This 2004 Microchip datasheet for their 27C256 claims > 200 years data retention.

I can't help but wonder if that is a typo.

We used to buy huge quantities of EPROMs from all the manufacturers. In those days they quoted life of 10, 15 and eventually 20 years.

I think it is doubtful that the chip's supporting circuits will last 200 years. We have been round this loop before. The fuse-link ROMs might have been expected to last for ever, I mean you don't expect a dead fuse to burst into life do you? Well they do, thanks to solid-state diffusion.

But I am prepared to eat my words if Microchip show me a working EPROM they made 200 years ago.

[ricard]

Originally Posted by Refugee

Microwaves can also be used to erase CDs

... Permanently...

[emeritus]

My previous post should perhaps have said " ... erased and reused like a windowed EPROM... ".

[G6Tanuki]

Keeping the windowed type in the dark is a good start to prolonging their life: the contents are erased by (UV) light!.

As to the overall durability, I rather suspect that this will depend on the individual manufacturing technologies used, just as we have AF117 transistors and Rifa capacitors failing whereas others go on for decades.

[I recall 'issues'with some of the early dynamic-RAM chips which were found to suffer occasional corruption because of Alpha-particle emission from the encapsulation resin!]

--G6Tanuki

[Dave Moll]

It may also depend on how well the window has been covered - thus keeping out the UV that gets as far as the exterior of the chip.

[julie_m]

Originally Posted by emeritus

Back in the 1970's when memory was expensive, a former colleage mentioned that, in his new firm, someone had discovered that non-windowed EPROMs could be erased by "cooking" them in a microwave oven for a few minutes.

Any sufficiently-energetic photon will dislodge electrons from a negatively-charged surface. The humble 1N4007, in reverse bias, works quite well as a detector of X-rays.

Also, watch what you use to cover the window on an EPROM -- some materials which are opaque to visible light are transparent to UV.

[lesmw0sec]

The original thing around the EPROMs were that you did the prototyping using them and then created a permanent ROM for production. The EPROMS were so successful that few did this. I have a nice boat-anchor which once used to be a Wayne-Kerr universal bridge as a result of data fading away!

[cheerfulcharlie]

Originally Posted by Refugee

That would have most likely been the battery backed part of the BIOS.
It could have been fixed with a 3 volt battery from a pound shop.

I thought the battery only kept the changes alive. ie clock settings & bios updates?

[electroanorak]

My old Strand MMS lighting desk was made in 1974. It uses the old 1702 EPROMS. They are all still fine.

Dan.

[Refugee]

Are you sure the BIOS chip had not just come part way out of the socket?

[SiriusHardware]

Originally Posted by electroanorak

My old Strand MMS lighting desk was made in 1974. It uses the old 1702 EPROMS. They are all still fine.

Dan.

Are they EPROMs or Bipolar PROMs? If they are really EPROMs I would be frantically trying to find something to read them with while the device still works. It's just getting dangerously close to the critical age now.

I have a Science Of Cambridge MK14 (1977) but its monitor (bios, you might say) is held in two bipolar PROMs, the sort in which physical fuses were blown in order to set the data.

I've always thought that was totally safe, until I read the post above

[jjl]

The 1702 was the first commercial EPROM type; 256 x 8 bit bytes, introduced by Intel in 1971.

John