Re: Black box

Ok, Mariss, for one of the dumb guys you seem real bright:-)

I understand your explanation. What bothers me though, is that you still
have a critical time requirement on the PC to deliver the changes to the
counters. You set the counter to output some number of steps at a
startup rate, then you must get back to it with the next value before
it finishes that one. I guess that's why some of the discussion on
buffers.

I really want NO critical time constraint on the PC. With a smart box,
I send it off on a move, and the PC can go out to lunch, that move will
accelerate, count to its end and decelerate exactly as programmed. Your
approach lightens part of the load on the PC, but still keeps it in a
tight time loop.

I suspect the cost of the box is really unrelated to its smarts. Clearly
this work can be done with a PIC, because Kerr sells a box that includes
a servo amp in it for just under $200. I suspect most of the cost is in
making a board, adding connectors, etc, and putting it in a box. The
exact compliment of chips in the box probably doesn't change the price
much.

I re-read the PicServo manual last night. It has a mode where you send
it up to 96 points, and choose a time interval between points. It starts
working away to make each point at the required time, and the PC can
start loading more values into the point buffer. This would ne a fine
way to handle circular interp, or complex spline curves, etc. The PC
here does the trajectory calc, but it is not time critical- it could be
done in batch, as long as once started it keeps the buffer full.
According to the numbers in the Pic manual that is a very easy load on
any PC.

For $200 the PicServo does exactly what I want, but it has the servo amp
built in and only drives 4amps. If Kerr will separate the control logic
from the servo amp it would be my choice.

ron ginger