Re: Re: EMC question

krnldmp

I'm going to cut this post all up and answer it out of the order you
posted it in.

On Tuesday 21 January 2003 11:45 pm, krnldmp wrote:
>    From: "krnldmp <krnldmp@...>" <krnldmp@...>
> Subject: Re: EMC question

> I'll be happy to figure this one out.

I like that attitude. I seem to remember a quote, "We learn from
experinece, if at all."

> Nothing says it should work with this hardware
> though, or my software alignment.

Yes, hardware can make a difference but it should not make to much of a
difference. IMO all of us who try to do most of the motion control and
signal generation in a PC have found variations between devices and have
had to modify software to provide for the variety of systems out there.
The wisdom a while back said don't use Cyrix processors. Now Via/Cyrix
processors are the most common in my basement.

> Sorry so light on the specifics.

One thing that I think that we need to know is the kind of stepper
driver/amp you're using. There has been a lot of discussion about the
nature of the signals required by it and the nature of the signals that
it gets. I've seen this first hand a couple of times. One amp board
maker assumed spikes from the EMC and the device he had made was jittery
as all get-out with the square waves.

Steve Stallings of PMD/PMDX fame recognized this problem and has built
some parport interface/breakout boards that will shape and regenerate
pulses.

> I was prepared to observe some
> limitations from using a Slow processor at 225MHz, but my trouble
> occurs as massive trashing of what should be a fixed period pulse
> train with only one axis driven at very slow speed... Indeed, any
> speed.

I don't think that the speed of this PC is the problem. I've got a very
old Gateway 166 overclocked to 200 here and it works fine for both
steppermod and freqmod. (I can just hear the chorus of detractors
singing, "Yea, but you've been tuning that damn program for years.")

Since the EMC is so configurable, it is possible to make quite a wide
range of errors in the melding of software, PC, and stepper stuff.

> Another or rogue process? How could that be, I thought this software
> was fairly ironed out.

A fellow on the emc list had what I think of as similar problems with a
PC not keeping up with reading encoders. (Both reading encoders and
computing stepping pulses are done on the rt side of the shared memory
buffer.) In theory, there isn't much that can get in the way of these
things happening. He fixed the problem with a different motherboard.

The motion process that is responsible for the pulse generation has a
really high priority. With the BDI 2.18 it is almost uninterruptable by
other processes because it is for the most part it's own OS running
between Linux and the processor. The task controller and interpreter
both run on the Linux side of things. This motion process could fail to
get an update from the task controller if it (motion) was such a hog that
Linux couldn't keep up. If this happened, you would see massive delays
in the refresh of the gui. You can test this by moving the mouse and
watching the delays in its update.

> I'm using steppermod.o (I think that's the name.) I've considered
> trying freqmod to see if I get a more predictable output, however more
> limited in speed.

Steppermod can jitter more than freqmod but I've seen it the other way
round as well. Good tuning of freqmod will make for some nice sounding
motion but steppermod works just about as well for my minimill.

> For the mention of it, the jitter may actually not be what should be
> considered jitter, at least not purely. It's difficult for me to tell
> because my scope doesn't want to trigger very well on the walking
> repetition rate, but it seems as though the output may be getting
> gated by something.

Right. Both steppermod and freqmod change pulse rate a thousand times a
second in a default install. I've put my scope on external trigger and
run it slow but the changes are still hard to see.

I think that the human ear may be a better judge of smooth. I'm always
listening to motors with a stethoscope. A long socket extension or steel
bar between your ear and the motor will also work. I wonder how it would
sound to hook up a pair of speakers in place of the stepper motor. Step
pulses in stereo!

Ray