matthiasm wrote:
> On May 24, 2007, at 2:07 PM, Marc R.J. Brevoort wrote:
>
>> Hm... In this case I'm running Linux 2.6.20. I'll be needing 1/1000 or
>> better accuracy. Any suggestions on this?
>
> Have you though about realtime operating systems?
..or doing whatever it is in hardware or as a kernel driver.
If you need an accurate time trigger for 1000'ths of a second,
I think Matthias is right, you'd only be guaranteed accuracy
from a real time system or from a kernel driver.
None of the others
> will guarantee you a resolution of 1/1000th of a second. They need
> much more time just to switch between tasks.
I gave rtlinux a real test for accuracy some years ago (2001)..
wrote a small installable C driver triggering outputs on a
parallel port and monitored with a scope:
http://seriss.com/people/erco/rtlinux/
Based on those tests, I was quite convinced it was really
and truly working as advertised.. the OS was lower priority
than my driver app. Even while accessing slow devices like
floppy disk drives.. the OS's ability to manipulate the drive
was usurped by my app.
Same tests on a non-rt system were pathetic. Glitches everywhere,
and the signal would go off completely whenever I moved the mouse,
or did anything with the disks.
In my case I was driving a stepper motor, so glitches were
not allowed. And it gave a nice steady squarewave with
non-accumulative error.
And this was with a slow-assed 120MHz box.
_______________________________________________
fltk mailing list
[email protected]
http://lists.easysw.com/mailman/listinfo/fltk
