DIYbanter - View Single Post

cavelamb

Joe Pfeiffer wrote:
cavelamb writes:

Joe Pfeiffer wrote:
writes:

On Mon, 22 Mar 2010 23:51:54 -0500, cavelamb
wrote:

Joe Pfeiffer wrote:
cavelamb writes:

Joe Pfeiffer wrote:
And yet, when it comes down to it, the electronics and the code are the
*only* parts in the car that we can't take apart and look at.
Well, actually, you can.
But it's a lot more involved than examining mechanical parts.
Microphotography to read a ROM is a little involved, yes....

And, perhaps an important question...

How can you tell when software is getting worn out?
Jokes about bit rot aside, it doesn't.
Are you sure?

What does it take to cause a bit to flip?
Say, in ten years?
In firmware, it is possible. But the question was about software. And
even in firmware (like a ROM) it is extremely unlikely in 10 years -
or even 20. In a "burned rom" it would require voltage being applied
where voltage should not be to burn a 1 to a 0 or short a 0 to a 1
In fairness, in this context when we say 'software' we're really talking
about firmware, and it's more likely to be flash than burned-in or
mask-programmed. But, a quick look over at Freescale shows data
retention guarantees of well over 10 years (they actually guarantee over
100 years, with derating for temperature).

Also, of course, if the code were available to us, we'd be able to
replace a failed EEPROM long after the manufacturer went under.

Thank you, Joe.

But actually, nobody "guarantees" retention.
What they guarantee is a statistical failure rate.

An error every so-many operations...

I was actually pretty surprised to find that Freescale does in fact
quote a minimum data retention time, with a derating formula. No
statistical failure rate to be found in their data sheets.

I see. Yes, there would be a minimum retention time.

I was thinking MTBF.

--

Richard Lamb
http://www.home.earthlink.net/~cavelamb/