> Hi, all,
>
> I designed a small device about 2 years ago, which can be operated in two
> different modes. Switching to mode 2 is only done here at the factory by
> pressing a key in a specific pattern (Press 10 seconds, then wait until
> light is off, then release key between 2 and 4 seconds, wait until light
> is
> on again, then press 4 times within 3 seconds, someting like that). The
> state of mode is stored in eeprom address 00, 0x00 for mode 1, 0x80 for
> mode
> 2.
>
> Now during the last 3 weeks about 10 customers called to tell, that the
> device is suddenly operating in mode 2, even if it was shipped in mode 1.
> The procedure to change mode was never made available to the public and
> futhermore there is no advantage at all for most customers to change mode=
..
>
> Question: Does anybody have information about the reliability of eeprom
> content?
> How is it possible that eeprom content changes (electric field, magnetic
> fields or or or)
>

I spend several weeks on this, finally pinning the problem on our Blessed
Microchip (sigh).

First: carefully search the archives for ANY EEPROM failures before the
PIC12F/16F/18Fxxxx series devices were released. I can count on less than
one finger's worth. And I designed the EEPROM into everything, making
configuration storage, calibration tables, everything.

Second: Microchip releases many PIC12F/PIC16F/PIC18F devices, and then
begins to make excuses for EEPROM failures. What happened was that when
Microchip reduced the die size on the PIC12/PIC16F/PIC18F devices, the
cell size was reduced to an almost unworkable size.

Having gone to an unworkable cell size, they can't go back and fix it. If
you search carefully, you will notice that Microchip recommends rewriting
(refreshing) the EEPROM cell array periodically. But I uncovered a BETTER
way:

Write 5 copies of your variables into the cell array. When you read them,
read all 5 copies into a buffer. (1) IF ALL ARE IDENTICAL, use the copy
and you are done; (2) if 3 or more of the 5 sets are identical, use one of
the 3 same copies as "good" as data, then rewrite all 5 copies of the
array. If less than 3 are identical, you are screwed... but guess what, I
NEVER saw that happen. What seems to happen is that one or more bits are
changed over a year's period, and this corrects them.

The advantage over Microchip's method is that there is a MAX rewrite
limit. My method only rewrites when there is a true error, so the limit is
very unlikely to be reached.

This method, that I call "Best of 5", always works for me.

Otherwise, drop the internal PIC EEPROM and use an external EEPROM, which
does not have super small cells.

--Bob A

































R



















, deciding

--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.