Re: step pulse timing resolution

Matt Shaver wrote:

> > From: Art Fenerty <fenerty@...>
> > I do have a question though, how many pulses/second are actually
> required
> > for most people. A VXD is not very
> > hard to write for up to 8192 pulses per second, but get more
> difficult at
> > 40000 or so. I use about 4000
> > or so pulses per second on my machine, but how many pulses does the
> average
> > user require. If I'm going to do this, I might as well make it for
> most
> > users.
>
> Let's say I had a machine with 5tpi screws, 2:1 reduction from motor
> to
> screw, 200 step/rev motors and I wanted a basic unit of movement of
> .00025"
> (I'm half stepping the motor) with a max feed rate of 120"/min. That
> works
> out to 8kHz at 120"/min. I'd be happy with that for my stepper
> applications.
> I'd like the following output step rates:
>
> 8000 Hz
> 7999 Hz
> 7998 Hz
> <...>
> 3 Hz
> 2 Hz
> 1 Hz
> .9 Hz
> etc.
>
> Can this be done without locking up the GUI? If so, how?

Nope, can't be done by a CPU. 8000 Hz is a period of 125 uS, 7999 Hz is
a period
of 125.016 uS, or only a difference of 16 nS! You SURE can't interrupt
a Pentium - ANY pentium - every 16 nS!

I can't even do this with my chip unless I run it at 62.5 MHz! I COULD
do that,
but it isn't really necessary. Clearly, the jitter, etc. in a stepper
drive is likely
to be bigger than 16 nS. Especially in a chopper drive, the chopper
generally
runs at only 25-40 KHz. So, a 16 nS resolution in step pulses is
meaningless,
when the chopper is switching on and off on 40 uS cycles.

I CAN, however, give a 100 nS step pulse resolution. So, your velocity
choices are 125 uS = 8000 steps/sec, and then 125.1 uS = 7993.6 Hz
This is a change of only 0.08 % ! This is a change of about 6 Hz, up at

that end of the hearing range, you'd hardly be able to hear the pitch
change,
and our ears are VERY sensitive to pitch. The motor is REALLY not going

to have much trouble making the jump of .08% in speed. AND, that is the

WORST velocity jump there will be (in the static sense). When large
accelerations are required, the velocity command out of the CPU will
not go through every speed value monotonically, it will jump in whatever

size velocity steps are needed to accomplish the acceleration. But, the

jumps should still be quite small.

This should all still work fine with the CPU only needing to monitor the

position every milli-second or so, as is done for servos.

Jon