Re: [avr-gcc-list] Overclocking Microcontroller

[David Brown]

> Andrew M. Cullen wrote:
> > I've overclocked the ATMEGA32/64 at 20MHz with no problems. BUT... this
> > is only in a laboratory settings, you can't rely on this for any
> > commercial product.
>
> About 15 years ago I heard of someone overclocking a 20 MHz PIC to 60
> MHz by immersing it in liquid nitrogen. Somehow I don't think this is
> what the original poster is looking for though.
>
> John
>

I regularly overclock 40 MHz MPC561 chips to 90 MHz during board testing,
without any cooling whatsoever.

The general trick to overclocking (including for AVRs) is remember that
silicon runs faster at lower temperatures.  If a chip is rated at 16 MHz for
up to 85 C, then it can probably run at 20+ MHz at 25 C.  Similarly, the
higher the voltage, the faster the chips can run (within limits, of
course!).  Typical AVR's come in two variations - 16 MHz, 5V chips and 8
MHz, 3V chips.  These are in fact the same chips, and can in practice be
used at either range (although of course you have no guarentees - what you
do for your own lab cards and what you can sell to your customers are two
different things), they are just binned according to small variations during
testing.  So if you want to run your AVR faster, aim for 5.5V and low
temperatures.

The other important thing is to know what is likely to fail at higher
speeds.  Oscillators are one thing - you might get unstable clock references
when overclocking significantly, unless you use an external oscillator (as
distinct from just an external crystal).  The other is the flash - you will
get read errors when running too fast.  Finally, some peripherals (such as
ADCs) will have problems at higher speeds.  The core logic of the cpu, along
with the ram and digital peripherals like timers and uarts, should be good
for at least 40 MHz on an AVR.

David