Argh. ATmega128 users, please help?

Allrighty... some progress.

Richard H. wrote:

> Now, no knowing why changing CKOPT didn't fix your other ailments.
> Perhaps it caused some damage?

I put an 8MHz xtal on another board, and that brought the swing up to
about 824mV.
With a colorburst xtal, that went up to about 900mV. I figured this was
probably safe to attempt another write operation.

Per a couple of application notes I read from programmer vendors, I put
a 100pF cap between TCK and ground.

Next I tried writing 0x89 to the high fuse byte, and whomp! the swing
when I put back the 16MHz xtal is from 760mV to 4.32V (3.56V) which is
great. I was hoping that this would fix all my programming woes, but it
doesn't.

I tried writing hfuse, lfuse, efuse individually, in that order. Side
effects!!

Write hfuse = 0x89 -> changes efuse from 0xff to 0xfc
Write lfuse = 0xff -> changes hfuse from 0x89 to 0x20
Write efuse = 0xff -> apparently no side effect.

Incidentally, I discovered that the undocumented bits in the extended
fuse byte seem to have a meaning related to clocking! Specifically,
0xFC results in the chip running _incredibly_ slow; while (1) { LED =
^LED; } on 16MHz clock results in blinking at about 4Hz!

Dave Hansen wrote:

> If you're using an STK-200 type programmer, you might try changing the
> buffer.  That generally solved any programming problems we had.

No, I'm using a USB JTAG-ICE (Olimex's). I tried it with the Atmel
JTAG-ICE at work, same results.

GRRRRRRR!!!!!!!