Tom,
I checked out Dac0-3 as you said, and they all work fine, even with everything connected.
DACx=0 is about 0.05 V - is there a way to trim this?
Once I do a move command with kmotion, the output=0 when the axis is supposed to be at rest, however.
 
Also, once I do a move command, the console command DACx=xx no longer changes the DAC output - is this normal?
 
I also moved the Z axis around by hand, and the encoder behaves as it should. Its rotation is the opposite of the motor rotation, so I set its gain to -1.
 
The amp driving the motor is not very linear when the motor is under load - takes almost 5 volts just to get it turning.
However, I can still feel the motor shaft jump at the beginning of the move and at the end, and the voltmeter jumps, too. Commanding a 10,000 count move does not get a long 10V output on the DAC - it is still outputting the spikes I sent you earlier. I am not getting following errors, either.
So the DAC output is still not following the output graph.
Next?
 
Michael
 

Group: DynoMotion	Message: 4327	From: Michael Rosenfield	Date: 3/26/2012
Subject: Re: Servo question

Tom, That was the problem! Working fine now. However, I had axis1 also set to 4096, and it worked fine. Of course that axis has much quicker response. Anyway, I set them all to 2047.   New question: My Z axis is a very slow response - 5000 count/sec max velocity. I can't get any movement on the Bode plot, even with 5Hz limit and 2000 count move.  Anything else I can set on the Bode plot? Any guesses for gain settings? Currently, I have P=2.6 Any more than 0.0001I and I get some damped oscillation. FE is around 300 counts. Feed forward doesn't seem to make much difference.   Thanks,   Michael   Group: DynoMotion Message: 4328 From: Tom Kerekes Date: 3/26/2012 Subject: Re: Servo question Hi Michael,   If things are working correctly the axis should never try to exceed max output of 10V so it wouldn't make any difference.   I don't know, it is a bit of a catch22 as the axis must be working somewhat in order to measure how it is working.   Current (acceleration) mode has more phase lag than velocity mode because position is the 2nd integral of acceleration where position is only the 1st integral of velocity.  So try adding more phase lead down at the low frequency range.  Did you intend to include any plots?   Regards TK     Group: DynoMotion Message: 4337 From: Michael Rosenfield Date: 3/26/2012 Subject: Re: Servo question Attachments : z P5 move-4337.JPG z P5 fe-4337.JPG z filt-4337.JPG z bode-4337.JPG z bode sm-4337.JPG Tom, I did some tuning of the amp, and now I've got decent enough results to plot. Here they are. The Bode plot is difficult to decipher. I took a guess and added a lead lag filter with parameters of 1,1,10,10 but that caused massive driving of the servo to the endstop, so I had to remove it.  I played with the LPF, and found that 100 Hz provides better results than any lower cutoffs - I tried as low as 20 Hz. I've also tried adding bits of feed forward, I and D gain, and none of those seem to do much, until I add too much, and I get oscillations.   Michael Group: DynoMotion Message: 4339 From: Tom Kerekes Date: 3/26/2012 Subject: Re: Servo question [5 Attachments] Hi Michael,   Looks like there is a lot of stiction or the amplifier is not working right.  You can see the output increase to almost 500 (2.5V) before anything begins to move.  I don't think you will ever get good results until that is fixed.  You need to determine if this is stiction or a problem with the amplifier.   The Bode Plot is worthless.  A Time Domain plot will show why.   Change the LPF back to 500 Hz.   I think you need either a lead-lag filter (or D gain).  You will need to experiment with combinations of width, frequency, and P Gain.  I don't see why that ran to the end stop unless you had other inproper settings with it.      Regards TK         Group: DynoMotion Message: 4341 From: Michael Rosenfield Date: 3/26/2012 Subject: Re: Servo question Tom, There is a lot of stiction. Unfortunately, the z axis of this mill was not designed to be motor-driven, and I had to come up with a drive system to retrofit on it. A ball screw was out of the question, so I have a 68:1 gear motor driving the quill through a 2:1 toothed belt to the quill handle. There is a spur gear driving the rack built into the quill ram. This quill will never allow anything but 2 1/2 axis machining, so the response isn't critical, as long as it gets to the position accurately. This also means it will never be driving a cutting load very fast - 26 IPM tops. I also have an actuator on the quill clamp, that I must drive after a z move, to lock the quill in position for milling. I looked into putting an encoder or tach on the motor for velocity feedback, but the motor speed just isn't high enough to generate much of a velocity signal - I am hoping this thing will work well enough to drill holes and set the cutting depth of an endmill. What do you think?   Michael   Group: DynoMotion Message: 4343 From: Tom Kerekes Date: 3/27/2012 Subject: Re: Servo question Hi Michael,   We can try.  What is the resolution (counts/inch) ?   If you don't care about performance a little Integrator gain can usually be used to creep into an accurate final position.   But basically try those suggestions in my previous email.   Regards   TK Group: DynoMotion Message: 4344 From: Michael Rosenfield Date: 3/27/2012 Subject: Re: Servo question The encoder on the quill is 1666 counts/inch. Quill speed varies between 1 - 26 IPM for drilling/plunge cutting, and 180 IPM for rapid move. Motor speed for this travel is 17 - 300 RPM, and 2000 rpm. When I had the lead lag filter, and the motor drove the quill past the end stop, the only thing I changed was adding that filter. When I removed it, I got the plots I already posted.   Michael   Group: DynoMotion Message: 4345 From: TK Date: 3/27/2012 Subject: Re: Servo question Hmmm. The resolution is much lower than I understood from all the gearing you described in your last email. I guess I don't understand the mechanical arrangement.  You need to find an arrangement that is stable with the lead/lag filter in. You can't just put it in and not expect to change other things. You should always be able to reduce the P gain enough to make it stable (it may not move well but it should be stable).  I also don't understand why it runs into the end stop. I can understand violent oscillation if the gain is too high, but not just driving one direction. Generate a plot of this problem so we can see how it is possible. Add in the lead/lag compensator with the axis disabled. Then use the Move button to enable the axis, do a small move, and capture the response.  TK  On Mar 27, 2012, at 12:22 AM, Michael Rosenfield <mrosenfield@...> wrote:   The encoder on the quill is 1666 counts/inch. Quill speed varies between 1 - 26 IPM for drilling/plunge cutting, and 180 IPM for rapid move. Motor speed for this travel is 17 - 300 RPM, and 2000 rpm. When I had the lead lag filter, and the motor drove the quill past the end stop, the only thing I changed was adding that filter. When I removed it, I got the plots I already posted.   Michael   Group: DynoMotion Message: 4395 From: Michael Rosenfield Date: 3/27/2012 Subject: Re: Servo question Attachments : z 12move-4395.JPG z 12 move sm-4395.JPG z 12 filt-4395.JPG z 12 fe-4395.JPG z 12 fe sm-4395.JPG Tom, Finally got back to the mill. I have attached photos of the mechanical setup, and the latest tunings. The system just appears too slow to react to the driver used to generate Bode plots, so I'm guessing that tool is not available for this axis. The quill encoder is driven from a friction wheel on the side of  the quill guide. No other way to set it up with this machine, that I could think of. The motor has an integral gearbox. Have no problems with stability with the current settings. Don't know what was going on yesterday. I did both a 100-count move and a 1000-count move.   Michael   Group: DynoMotion Message: 4399 From: Tom Kerekes Date: 3/27/2012 Subject: Re: Servo question [9 Attachments] Hi Michael,   That looks like it is meeting your requirements.  But it does take a while (1~1/2 second to settle in to the last 1 or 2 counts).   The only think I might suggest is to make the lead/lag wider to add more phase margin and damping so maybe the P and I gains can go a bit higher.   Regarding the motion profile:  With a 1000 count  (2/3 inch?) move it looks like you are only accerating up to about 100ipm.  A longer move would go higher.  Your Jerk is set very high (4e6/1e4 = 400) so full current and torque is attempting to be applied in 1/400th of a second.  Reducing Jerk and increasing Acceleration may give a faster yet smoother motion.   Regards TK