Re: [CAD_CAM_EDM_DRO] Servo Error, - Was backlash.

Mark Vaughan wrote:

>Sorry Scott.
>
>I try to make descriptions easy, but often don’t succeed.
>
>Jon is pretty switched on as well, and way more experienced at some of this
>than I.
>
>
>
>Interesting Jon. I guess I did not understand what
>Mark was talking about when I first read his post.
>Thanks for making it clear to me.
>I am currently set up with my servos for 16000 counts
>per inch of travel. If you figure it out, 128 counts
>is a whopping .008"!!! That is a lot of error and still
>have have the drive say everything is OK. But that
>is following error. When the motor stops, the drive
>should "hunt" for the actual position. Is there a case
>when this is not true and the drive just stays where it
>is?
>
>
>

Apparently, under some conditions, yes. I still don't understand the
exact conditions,
but I have a friend/client here who has a machine with Gecko 320 drives,
the Camtronics
motors made by Ametek, and a 36 V switching power supply. The motors
drive the
machine's leadscrews 1:1. These may be Acme, not ballscrews, now I can't
remember.
Anyway, with the 250 cycle/rev encoders, or 1000 counts/rev x 5 TPI that
has an
encoder resolution of 5000 counts/inch, or .0002", which is a bit coarse
for a CNC
machine. I can twist the handles up to 20 degrees or so, and the Gecko
drives will
not return the motor to the servo null point. Mariss at Gecko says this
is impossible,
but this machine does it on all 3 axes. It acts much like the PID's I
term is way too
low. My thinking is that this has something to do with the DC supply
voltage and
maybe the motor's winding resistance. The way PWM works, the DC supply
voltage
is just like a gain knob. Double the voltage, and it has exactly the
same effect as
turning the P gain up twice as high.

>
>Yes. Whilst this isn’t quite correct, think of it that the drive doesn’t put
>enough power in to overcome friction for that last little bit of follow up.
>
>The P parameter sort of looks for the error and applies power proportional
>to that error. The more the error, the more power and faster it goes towards
>position. But if the error is small, there isn’t enough power. Think of
>pushing a chair across a floor, you can apply a little pressure and nothing
>happens, so if you wanted it to move a short distance, you wouldn’t apply a
>lot of force, you’d probably give it a short sharp nudge, it it still didn’t
>move, you’d wack it a bit harder until it does. If you wack it too hard it
>will overshoot. Think of the I as correcting this. Some cheap servo drives
>are just P and D they have no I parameter.
>
>
>
>

Yes, exactly.

>
>Why would the servo controller not move the motor
>to the command postion within say several counts?,even
>if it is over or underdamped. I do not understand how
>tuning would change this.
>

Tuning changes EVERYTHING, believe me. But, on the Gecko, you don't
have adjustment on the I term. You do have adjustments on P and D.
(The last one is current limit.)

> I thought tuning just changed
>how smoothly the servo went to command postion. I never
>thought it would prevent the motor from getting there.
>
>

Turn the P down to minimum, and it probably will not move at all!

>When I am single stepping the motor,
>my dial test indicator shows clean solid .001" steps.
>Is it possible that my motors are not powerful enough
>and the drag on the table just prevents the servo
>from getting to it's command position?
>
>

it NEVER gets to it's commanded position! If the system is "powerful
enough" and
the gain can be turned up high enough, then it will get really close, or
even start
overshooting and bounce back and forth over the commanded pos. But, it might
not, and with the Gecko 320/340, you never really know how close it is
at any time,
other than less than 128 counts.

>The easiest method to know whether there is a following error is to take the
>encoder signal out to another counter, perhaps connect it to the mpg input
>for mach. Then if mach sends say 100 pulses, you would expect to see 100
>pulses worth of movement on the encoder. Some servos actually have an output
>that shows the difference between the correct position and the actual
>position, sometimes this is a digital output, sometimes analogue.
>
>

The Gecko 320/340 has an analog testpoint for this. The datasheet says
40 mV per encoder
count.

>
>

Jon