[Note from MC marked "Feel free to publish".]
[Attachment is alongside this file.]


Date: Tue, 8 Mar 2016 08:39:41 +0000
Subject: REV2 resets
From: Marko Cosic
To: Damon Hart-Davis

The attached file shows how many resets there were, per REV2, per day.

"Resets" are self inflicted by the board itself. "Catatonic" is where we
deliberately reset the REV2 because it gets into a state where it ceases to
respond.


On 23/02 the code was as follows:

1/2 - BETA3
3/4 - OLD
5/6 - OLD
7/8 - MAINLINE
9/10 - OLD
11/12 - OLD
14/15 - OLD
16/17 - OLD
18/19 - OLD
20/21 - OLD
22/23 - OLD
25/26 - MAINLINE


On 07/03 the code was as follows:

1/2 - BETA4
3/4 - BETA4
5/6 - BETA4
7/8 - BETA4
9/10 - BETA4
11/12 - OLD
14/15 - BETA4
16/17 - BETA4
18/19 - OLD
20/21 - OLD
22/23 - BETA4
25/26 - MAINLINE


There is no typo on the number of resets on the Flat 16/17 REV2, and it did
not work for the start of the day then suddenly enter a constant reset
loop, it just averaged a reset every 20-40 seconds and has been doing so
for many moons.

I *did not* replace the REV2 in 16/17 to get the zero resets figure: I just
reprogrammed it.

This fits with what we have observed before: The REV2s will fail on you but
if burn the flash again and they work just fine. (e.g. Flat 25/26)

They are powered from the USB bus of a Raspberry Pi via an FTDI lead and
are operating in a dry environment at 10-40C.

Compared with most installations they will be being hammered in terms of
the number of messages received and the number of messages they're asked to
send though!


Our office REV2s now have 2.7V BOD and we will be watching these for resets
and flash corruption.


Future REV2s on site will get 2.7V BOD: a reset is much less bad than flash
corruption.


Only REV2 board has completely failed and is unprogrammable. If you handle
it, and specifically massage the 328P, then this makes it programmable
again. Probably I would hard-solder the 328P chips in place in future and
use clips to programme them.