iss wrote:Additionally there is no clearly stated info how long is the data retention time of these chips - different sources claim 10-20-40 years (and even >200 years but for EEPROM's).
Most accurate estimate: "decades". I have some datasheets for UV EPROMs that state 10-year data retention. I think that's rather short. More likely to cover manufacturer's ass in case EPROM looses its data even sooner. I have some Flash EPROM datasheet that states "Greater than 100 years Data Retention". While it's doubtful an EPROM in plastic housing (the device
itself) would remain in working order that long. Let alone its contents. More so since the underlying technology is similar to that of UV EPROMs.
And there are many factors involved. Temperature (and -variations), storage conditions, how often it's powered up, programming algorithm used, technology node (size of the smallest structures), type of window cover used for UV EPROMs, etc etc. I don't remember seeing a <20 year old EPROM loose its data while otherwise being in working order. Likewise, the few EPROMs I've seen where bits had fallen over were nearing 30 years or older.
I suspect manufacturers themselves don't really know. Nor am I aware of any scientific research concerning EPROMs losing their data in the field (as opposed to accelerated life testing which may give a good
indication, but no more than that). Expect 20 year minimum for a properly programmed device. In any case: much much longer than commercial lifespan of equipment it's used in. Might be interesting how this goes in the years ahead, as in-circuit programmable stuff (uC's etc) is in 'everything' these days. Even dumb gear like say, flashlights. And EPROMs in machines like the Oric & many others from the early 80's are >30 years age now. So I suspect "EPROM lost its contents" to become an increasingly common cause of failure in the years ahead. Well that and electrolytic capacitors...
The good thing is they can be reprogrammed, so our old machines will survive
.
That only helps for owners who can diagnose & fix the problem themselves. For
most owners who:
- Don't have the knowledge or equipment to diagnose the problem
- Dare not open a case
- Or unsolder an IC
- Can't find a (pre-programmed) replacement IC
- Don't own an EPROM programmer (or one that works with their current PC etc)
- Don't know anyone in their area with these things (or shipping costs too high)
In all the above cases, EPROM lost its contents means: device = bricked. Or at best, looses functionality. That only an EPROM failed, helps nothing when its owner doesn't know whether it's ROM, RAM, CPU, ULA, video circuit, crack in a pcb track, or power supply. Or how to fix.
NekoNoNiaow wrote:Yup, whenever you obtain a cartridge, one should immediately dump it properly for preservation:
Depends. Mask ROMs are extremely reliable especially if they spend 99% of the time sitting in a box. Usually surviving much longer than the hdd's, USB sticks etc that people store their ROM dumps on. Yeah I know... backups, off-site copies, upload to some online archive...
If it's an EPROM, then it depends on age & rarity. No existing dump: yes PLEASE, dump ASAP! If it's a common title, and your dump comes from an EPROM that was programmed 28 years ago, who's to say any differences from existing dump is in fact another version. Might as well be some corrupted bits.
Which is one reason I'm very careful with my personal archive of system ROMs (and to a lesser degree, game ROMs). Not so much because they're hard to find, but because I
personally dumped them back in the day when all those ROMs and EPROMs were still 'fresh'. And kept multiple checksums for those dumps over all those years. Bring in a dump for a machine that I already have in my dumps, and (unless
proven otherwise) I'll assume my existing copy to be more trustworthy than the new one. Or in the case of say, a Sinclair ZX81 or ZX Spectrum, I have 5+
actual machines to check against. Not some online hash database (though I understand & respect the purpose of databases like TOSEC & co).