[sdiy] Re: [sdiy-interim] EPROM Armageddon?

[ASSI]

On Mittwoch, 28. März 2007 06:27, keith sterling wrote:
> I have searched the internet for info on this upcoming "EPROM Y2K",
> but haven't found any mention of it anywhere.  Maybe nobody knows?
> Does anyone have anything to add about the validity of this EPROM
> lifespan claim and is there anything that can be done about it?

Things aren't quite so simple and clear cut.  To relieve you of the 
immediate worry, there is no time-bomb or anything like that in the 
EPROM that checks the lifespan and blows up after exactly 20 years.

What Paul is probably talking about is called "retention time" or how 
long does an EPROM hold it's information.  The 20 years he's quoting is 
most likely a data sheet value for the (minimum) retention time.  The 
way you arrive at such values is via accelerated testing (standard in 
this case is storage of the programmed part at 525K for 24h) and 
extrapolating the test results to "use conditions" under the assumption 
of a certain maximum failure rate or total number of fails over the 
life time of a large population of parts.  It is common practise to 
safeband the extrapolation by an order of magnitude or more w.r.t. 
failure rate to insure against defects that might escape testing 
(so-called extrinsic fails) and the statistical uncertainties that 
comes from testing a relatively small number of parts to arrive at a 
figure for literally millions of them.  Retention is exponentially 
dependent on temperature (Arrhenius law) and use conditions for 
consumer parts are typically speced at 360K ambient.  So depending on 
the temperatures over lifetime you have about two to four times more 
retention time than the data sheet value just from the actual use 
conditions _on average_.  Retention also gets worse when an EPROM is 
cycled (reprogrammed) close to the maximum spec, but I would think that 
the parts in production units have seen just a single program cycle.

Remember that this is all lies, damn lies and statistics: none of these 
numbers tells you when any single unit is going to forget.  But 
considering the fail rate of the other components in a synth I'd say it 
is more likely to keel over from other defects than from EPROM 
retention failure.


Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk]>+

SD adaptation for Waldorf microQ V2.22R2:
http://Synth.Stromeko.net/Downloads.html#WaldorfSDada