Re: [avrdude-dev] [bug #35590] avrdude make error when programmingATmega8 through USBasp

[Joerg Wunsch]

As Bill O'Neill wrote:

> So at this point, it seems that it may be partially a CPU speed
> problem.  Something in avrdude5.1x makes this a bit worse.  Probably
> requires a bit more processing power.  I have not done much testing
> under Linux as I have had no complaints from Linux users.

Well, I also experimented a bit with my "ponyser" programmer.  The
results basically indicate the kind of flakiness which I always
suspected for this crocky and cheap design.

I got hold of a Windows machine, and repeated the tests there as a
first step.  The hardware is approximately the same as the Linux
machine I've been using before (a little faster though), yet I
experienced random verification errors there, as you've also seen.
The errors could be fixed by introducing an addtional delay; even a
simple "-i 1" was already sufficient to get consistent programming
results.  The overall read or write time for the full 16 KiB was about
16 s.

In contrast, the Linux machine didn't require additional delays, but
read/write time there is about 6 s.

I then repeated it on my FreeBSD machine at home.  Without additional
delays, I cannot even make AVRDUDE recognize the device signature.
With -i2, it starts reading the signature, but reliable programming
results can only be obtained with a delay as long as -i10.  Almost
independent of the -i value used, programming time there is much
longer, about 30 s for a successful run (but still 26 s with -i2).

Now, compare this to using an AVRISPmkII: with option -B0.5 (2 MHz SCK
frequency, the ATmega16 is fused for 8 MHz CPU clock), programming
time for the same 16 KiB file is 1.5 s!

So all in all, the "ponyser" programmer type might be a solution for
some people, but it turns out being rather a problem than a solution
overall, and anyone interested in consistent (and fast) results is
well advised to invest a little into a more robust programmer.

Those who want to use the ponyser (and related) programmer are well
advised to experiment a little with the -i option until they get
reliable and consistent programming results.

> There is something odd about the Pentium D945 though.  It is the
> only one that has problems with the USBtinyISP.

That's indeed strange, and explains why it's not reproducible
anywhere else.

> In any case, we can put this to rest.

Agreed.
-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)