Re: Problem in Mar 15 build?

[Fred Proctor <proctor-at-cme.nist.gov>]

Jon Elson wrote:

> Anyway, I was able to jog around for a while, and experiment with the
> backlash compensation.  But, there was a very strange phenomenon!
> You've managed to make a beautifully tuned servo machine sound like
> a stepper-driven one!

One thing we added since the 20-Dec-1999 release was the quantization of
the output to the nearest input step. This sounds like it might be the
culprit. Jon, you can test it and see, by editing
emc/src/emcmot/emcmot.c and looking for the "ADJUST OUTPUT" comment near
line 2448. #ifdef out the quantization, like this:

	  /* ADJUST OUTPUT: */

	  /*
	    this computes outJointPos[] from jointPos[], by adding backlash
	    compensation and rounding result to nearest input unit
	  */
	  outJointPos[axis] = jointPos[axis] + bcompincr[axis];
#if 0
	  numres = outJointPos[axis] * emcmotStatus->inputScale[axis];
	  if (numres < 0.0) {
	    outJointPos[axis] = inverseInputScale[axis] * ((int) (numres -
0.5));
	  }
	  else {
	    outJointPos[axis] = inverseInputScale[axis] * ((int) (numres +
0.5));
	  }
#endif

and do an incremental compile of this, via:

cd /usr/local/nist/emc/src
make PLAT=rtlinux_09J all

You should see emcmot.c being recompiled, and stgmod.o (among others)
being built. Re-run the controller and see if this helps.

Another thing we did since the 20-Dec-1999 release was go to pure
periodic scheduling, rather than oneshot mode scheduling. This reduces
the timing jitter, and I don't see how this could make things worse. One
never knows. Let me know what happens and I'll try loading the
20-Dec-1999 release on my Bridgeport and see if I can tell the
difference. I didn't notice any problems with the new release, but then
again I didn't hear my car's wheel bearing falling apart until my wife
shouted to me over the noise.

--Fred