Re: stepper pulse rate idea

[Dean Hedin <dhedin-at-ieee.org>]

> Stepper timing in software on EMC is already done at somewhere between
> 50 uS and
> 16 uS, faster if your CPU can handle the interrupts.  So, there is NO
> advantage if
> adding another CPU and CPU-CPU communications overhead to get faster
> interrupts.

There is a faster way if your willing to give up checking limit switches 
between motions.

Rather than relying on a periodic interrupt, you instead fill  very large 
buffer(s) with the pre-calculated step sequence.   You could then use the x86 
REP OUTSTR instruction to blast the data out to the port.   You can then get 
1us bus resolution on the i/o pins.

Maybe you could use freqmod.o to layout and debug your g-code.  When it all 
works you could load the repoutst.o  module for fast production work.

Obviously, this works in open loop mode only.  A hybrid driver could also be 
made that incorporates both methods i.e, used freqmod method for homing 
etc...