Re: regarding future plans for EMC

[Fred Herenius <fred-at-twonky.xs4all.nl>]

Ray Henry wrote:
> I probably don't understand Fred H's mix of hardware and software very
> well but I think that he is saying that parts of the EMC could be
> rewritten and compiled in such a way that they reside in hardware
> rather than in memory.  That steppermod, freqmod, (or his
> steptablemod), or for that matter stgmod could be committed to an
> outboard device which would handle signals normally sent by other files
> like minimilltask and do the work of connecting to motor amps.

That is indeed what I was saying. And depending on how much time and
money someone is willing to commit this can be incrementally improved
(which is what I am doing right now). PCs are good for parsing rs274ngc,
numbercrunching, handling user interaction. Not so good at providing good
timing resolution for pulseshaping. Offloading the pulseshaping (nothing
more than toggling a number of IO lines at specified points in time) to a
small external board is IMO an easy and cheap solution (competitive even,
for those reading Dilbert ;). You have your highlevel task that says
something like

- at time X, generate this type of pulsetrain at this frequency and for
  this duration.
- s/X/Y/g and repeat until bored

Send these commands to the board over a link of your choice. I use a simple
parport i2c interface, but this can be changed easily. The board is able to
buffer a few of these commands, so you don't even need good realtime
response on the PC side. And what's cute about i2c, you can connect several
devices to the same bus.
Currently I use a few at90s1200-at-12MHz for this (had some spares). Together
with sense resistors + opamps + comparators these do the above and also
handle the PWMing. All parts (minus avr + xtal) salvaged from old PSUs :)
Oh, I tell a lie, I had to go out and buy more BUZ11s, as these ran out.

I have to use more than one at90s1200 (currently one for each axis) because
those beasties don't have enough IO lines, but in principle a single big
avr would also work. And some of the bigger ones have pwm channels to boot,
so no need to do that in software anymore :)

And ofcourse the DSP based solution on the drawingboard, but I already
posted some on that.

Executive summary:
Use of cost competitive external hardware enables improvement of realtime
characteristics of stepper motor control.

Fred