jrok68 
So.. I happened to look at an old Atmega88 datasheet that shows a spec up to 24MHz at 5v(!) Yet the newer datasheet only specs to 20MHz. So does anyone know did the 24MHz part get cancelled, was it announced too early or just a good old fashioned typo.
I ask as I'd really like to clock one at 21.477MHz.
- James
theusch 
If you are just doing a one-off and not shipping, I wouldn't expect any problems. My guess is production problems certifying over the entire temperature and supply voltage range. Typical sore spots on overclocking are EEPROM and flash.
For a one-off, whatever works. Gonna use 5.5V Vcc? ;) (extrapolating the Speed Grades charts)
Tell more about 21.477 -- is this a "magic frequency" for some application area? SuperNintendo? Wide-screen TV video?
Lee
jrok68 Tell more about 21.477 -- is this a "magic frequency" for some application area?
It is exactly 6 x 3.579 MHz, which would help improve this .. http://www.jrok.com/hardware/wms/sound/
The original 6802/8 hardware ran on a 3.579MHz clock so using a multiple makes getting the perfect emulated cycle timing for a MC6800 a lot easier. At the moment with a 14.318 MHz clock you get a minimum of 32 AVR cycles per 6800 instruction execute, with the 21.477 MHz you get 48. I have uses for those extra 16 cycles ;)
- James
bobgardner 22.1188 is 3.686 x 6... might also be a good freq
theusch Interesting, and a nice use of the AVR's near-single-chip ability.
What kind of instructions do you have problems with @14.318 MHz ? (See if we can speed it up).
Or is it some add on?
Jens
microcarl 
I've run the Mega64 and Mega128 at the 21.something Megahertz magic frequency while doing some experiments. Everything seemed to function properly. Though, I did not use EEPROM.
Currently, I'm working on a closed loop servo for my mill that's running at 18.432MHz. The system runs the servo (read brushed DC motor) motor from 0 to +/-1800 RPM using 10 bit FAST PWM configured for "Locked Anti-Phase " operation. 1800 RPM roughly corresponds to 100 inches per minute using an 18 turn per inch lead pitch.
This servo system currently incorporates 6 interrupts, two of which are the Encoder A & B signals decoding full quadrature. It works pretty good, but frankly, I'm out of time. My fear is that the encoder interrupts, the PID update interrupt, the velocity capture interrupt, the ADC interrupt and USART interrupt are using so much time that there isn't any time left to do anything else.
As a result, I'm thinking about moving the encoder translation over to a Mega88/168 controller. My original goal was 200 inches per minute. But I now realize that may not be practical with this class of micro-controller. But...
I'm thinking about running the Mega88/168 at the 21.something magic frequency to achieve a bit higher encoder translation rate. As EEPROM won't be involved, I'm thinking that the roughly 15% increase in operating frequency won't affect the general operation of the Mega88/168 - based on the experiments I performed with the Mega64/128 controllers.
I plan on running the Mega88/168 experiments over this coming weekend, and the days that follow.
jrok68 What kind of instructions do you have problems with @14.318 MHz ? (See if we can speed it up).
Or is it some add on?
- James
Koshchi The original 6802/8 hardware ran on a 3.579MHz clock
kevin123 I have clocked an AVR at 27 MHz before. If I recall correctly, it was only a 16 MHz part. It was a little flaky at this speed, but it got the job done that I needed at the time. I don't remember exactly what I was doing, but I was basically building my own make shift test equipment of some sort.
To Steve
And a PC !!!
The good old 4.77MHz XT (3*4.77=14.318)
Jens
theusch If this is going to be a hobby posting where everyone is "at their own risk", I wouldn't anticipate any problems at 110% and indoor conditions. If you are going to sell the kits, maybe 5x would stay in the specs but still give you a few more cycles? But 17.895 doesn't appear to be a standard crystal. 21.477 is.
Lee
CountZero You could get 5x using a PLL, will add more chips though....
I volumes are low enough you could test each unit at a range of temperatures to see if they will work.
CountZero Or wait for the XMEGA with (if I read it correctly) 32MHz max frequency.
And the Xmega has a PLL so you can run from a 3.579 xtal or what is the cheapest.
But it's not a 5V part !!!!
Jens
