Re: EMC problem/question

On 5 Aug, Jon Elson wrote:
>
> I'll bet your step resolution is .0005", and you are losing exactly that much
> every time the machine reverses. The first thing is to check the specs on
> the Camtronics driver for the setup and hold time on the direction line
> relative to the step line. My guess is that the driver needs a certain time
> for the direction line to be stable either before the step pulse or after the
> step pulse. I suspect that EMC is violating this requirement in one
> direction, thereby causing a step to occur in the wrong direction.
>
> One possible way to fix this might be to invert the sense of the step pulse.
> This can't be changed in the .ini file, so your choices are to possibly
> reverse it by the way the connection is made at the driver, if it uses
> opto-isolators at its input. Otherwise, you can hack the software that
> drives the step/dir bits to the parallel port or add an inverter chip at
> the output of the parallel port (on the step signals, only).

This makes too much sense. :-) I built a small opto-isolation board
which goes between the computer and driver boards. Thinking through
this circuit I figure it IS inverting the signals. SO, I pulled it out
and tried another test. The directions, both X and Y, reversed so the
opto-isolation does invert the signals (easily changed in the *.ini).
And with the isolation circuit taken out no off-set occured, dead on.
PROBLEM SOLVED, or at least isolated. :-)

I'd like to put the isolation circuit back in so I may work on the
problem more. Seems like a simple do nothing loop at the right place
in EMC would do the trick. The problem is where to put it. :-(

> If this doesn't seem to be the problem, you should try steppermod (if you are
> using freqmod) or vice versa.
>
I tried to run freqmod but it wouldn't function. EMC fires up but the
motors don't move (just buzz as if getting pulses too fast). What
parameters are configured in *.ini besides "freqmod.o"? I briefly
looked at linuxcnc and nist but didn't find the info right off. Can
you, or anyone, direct me to a description? This could be a better fix
than patching steppermod and no doubt more useful to know.

> One other possiblity is the backlash compensation. There was a problem
> at one time, which could be in your version, where the extra steps added to
> correct for backlash were coming out too fast, causing lost steps. If you
> have a nonzero value there, try setting it to zero and see if the problem
> goes away.
>
I have no backlash comp set but I'll keep this in mind when I start to
set those.

Thank you greatly for the incredible insight.

Hugh Currin
Klamath Falls, OR