PID Tuning Problems

[ltoy]

I have been trying to tune the PID loop on my Skyko cards for the last 3 night with no luck. I am hoping someone can shed some light on where I may be going wrong.

First Let me describe the setup

The Server Moto Is a Baldor 500 oz/in 100 Volt 1.85 amp motor.
The Encoders are also Baldor and they are 500L Encoders
(I believe my steps per rev are 12000)
The Amplifier is a Servo Dynamics SD1525
The Input to the amp is from a Skyko Pixie-100
The input toe the Pixie is from a PMDX 122 Break out board
The Break out board is drive by mach3

Now I am trying to tune the pid loop in the skyko card and this is my problem
If I increase the Kp factor to anything over 1 I can not get the motor to stop occillating. and If I leave all of the factors at 1 the motor has a huge amount of over shot (like .250). I am looking for the next step to getting this system running right.

[sanjiv]

Dear Sir
what is the difference between servo & Stepper? Witch control is easy?
I am sorry because you are expert in this field as i feel by reading your problem. i am trying to built a CNC router from junk. can you help me in ELECTRONICS & Software?

[ltoy]

I am running the Sd1525 Amp in Torque Mode. Do you think I should run in velocity Mode? What would the advantage / disadvantage be?

[HuFlungDung]

I think your steps per revolution is out of wack. In quadrature, the most it could be is 4x the resolution of the encoder, which would make 2000 ppr. Unless you made a typo, I think this is going to have an effect on your tuning

My 2 cents: leave KI at zero or 1, KP between 1 and 100 and KD between 1000 to 2000. YMMV. I have no idea how 'portable' the PID settings are between systems, but this is what I have used for several servo setups on various retrofits.

[ltoy]

Hi,

You are correct on the steps. I believe the steps on the motor are 2000 however the steps I have set in Mach3 are 12000 (6 tpi ball screw). I have done some extensive experimentaion with the Kp and Kd. If I leave Kp at 1 and run Kd at around 100 it seames oscillates bad However If i put some drag on the pully of the motor it runs smooth. It seames like the system does not have enough load for the values. Also a note on the over shot. It only overshots when I try and do a go1 x 4 y0 F100 if I use a f value of say 10 it does not seame to have the over shot.

I ran the system before the upgrade and the motors ran smooth from 1 to 120 IPM rapid with no problems. So I know the motors and amplifiers are capable.

I will try the settings that you suggested at lunch and let you know.

[ltoy]

I just tried the above settings. Started at 100 for A Kp and a Kd of 1000
kept increasing Kd to try and get the oscillation to stop. I went up in steps of 100 to 2000 with just about no change other than the motor oscillation faster. So then I made jumps of 500 all the way up to 9000 with no real change.

[ltoy]

Any other Ideas on how to get this system working right?

[HuFlungDung]

I don't know. Start over?

Did you first of all eliminate all motor drift?

Are your encoders attached directly to the servomotor or are they mounted elsewhere. Is there any suspicion of backlash between the motor and encoder?

What do the documents that came with your controller suggest for a tuning procedure, and range of values?

[ltoy]

Hi All,

I am still having problems tuning this system. I have verified that with no input on the Amps there is no motor drift. I also verified that with the Pixie 100 Connected and outputing 0 volts there is no drift. I am unable to tune the pixie 100 to get any near a good graph. I have even experimented with the gain to see if that helped. It did not. it seames that the machine moves too smoothly. If I add resistance to the ball scred (ie hold the pully while the motor is turning) i get much better graphs. The ball screws are tight and the servos a hooked to them via a belt with no play. Any one else ever run into this type of problem?

[KTP]

Can you post a screenshot of your graph from the Pixie com software?

Most users will end up with a Kp setting of 5 to 20, a Kd of 300 to 1000 and a KI of 1 to 40.

I haven't had any motor use a Kp of 100...if you are using that high of a value for your P term it is no wonder you are oscillating. The Kd term is the damping term, this can vary depending on how much mass your motor is moving around (ie, the Kd term might be different for a motor sitting on a bench vs a motor mounted to a 500lb table).

Have you tried working from a stable graph with values such as Kp=4, Kd =300, Ki=1 and then gradually play with the numbers (increasing Kp by 1 at a time while also increasing Kd by 50 at a time) until you get a sharp response with minimal overshoot? I haven't found a single motor/amplifier combo that I could not get a good graph and response from with the Pixie P100. The only problem I ever encountered was with some low pole count brushless motors and trapezoidal amplifiers..but this is more the fault of the amplifier not being sinusoidal and not a problem with the Pixie

[bbrreid]

I cant even get near tuning my skyko card even with the drive disconnected, the encoder disconnected and only the parrallel cable from the PC connected I still get the RED flashing light on the pixi 100 as soon as I press the jog key any ideas please

[KTP]

Originally Posted by bbrreid I cant even get near tuning my skyko card even with the drive disconnected, the encoder disconnected and only the parrallel cable from the PC connected I still get the RED flashing light on the pixi 100 as soon as I press the jog key any ideas please

...

You don't have the encoder connected and you are wondering why you can't tune the Pixie card?

The Pixie flashes the red light when the encoder count it is reading is off of the commanded position (from the steps input from the parallel port) by more than a certain number (this number can be set in the Pixiecom software, default value = 100 steps). If the encoder is not connected to the Pixie and/or if the motor and Pixie are not correctly hooked up to an amplifier and tuned, then of course the red light will flash on the Pixie as soon as you send 101 steps over the parallel port.