Re: [avr-chat] AVRISP Upgrade - 1.0F to 2.04 in AVR studio 4.12

[Graham Davies]

Gary Douglas wrote:
> I have an AVRISP that ... I am trying to upgrade it from
> V1.0F to 2.04 with no success....
> The unit is failing verification after programming in all cases.   It
> does come back after, but continues to report a 1.0F version.

Here is a wild guess.  I'd appreciate any opinions on this.

Recently it has been discovered that AVR Butterflies become unprogrammable
in certain circumstances that are may relate to erratic power.  The symptom
is verification failures in AVRProg.  What is happening is that lock bits
become set, preventing writing to Flash in the application section.  Atmel
suspect that due to power fluctuations, the MCU is losing its way and ending
up in the part of the Flash loader (Atmel call it a bootloader) that
programs the lock bits.  They have re-released the Flash loader without this
capability.

Now, most AVR Flash loaders are derived in some way from application note
AVR109.  This is true of the Butterfly Flash loader.  It is also true of my
AVR ICE-Cube.  I am getting field failures of a fraction of a percent per
year that are due to lock bits becoming mysteriously set.  I thought I had
the fix when I caught wind of the Butterfly thing, but when I went back to
the source of my Flash loader, I see that I already thought of this and
disabled the programming of the lock bits that affect the application
section.  I've just tested the hell out of it and my modification is sound.

So, what I'm saying is:
a) Maybe the Flash loader of the AVRISP is derived from application note
AVR109.
b) Maybe the AVRISP is suffering from the same problem as the Butterfly and
my AVR ICE-Cube.
c) Maybe it isn't as simple as a stray instruction pointer wandering into
the wrong part of the code.
d) If you do manage to rescue your AVRISP, it would be really interesting to
know if any lock bits got set.

Graham.