Step response doesn't stop

[pblase]

I'm not up to speed on servo systems and could use some help in getting the loop set up.
The system is as follows: an H2W Tech BLDM-B02 https://www.h2wtech.com/product/brushle ... r/BLDM-B02 linear motor driven by an AMC AB30A200AC servo amp https://www.a-m-c.com/products/?page=pr ... AB30A200AC, using +/-10V inputs and Renislaw LM10 magnetic encoders https://www.rls.si/eng/lm10-linear-magn ... der-system for feedback. This ensemble is driven by a KFLOP/KANALOG through the DAC output using the encoder feedback (differential RS-422). I can move the stage manually and see the position counter change properly on the configuration screen nicely.

The task is simple: move the stage from point A to point B. Right now things are pretty much at the default setting, although I had to increase the output limit significantly to get the acceleration I need. Using the Step Response screen, I can feed it a position step and the stage takes off nicely. However, it doesn't stop when it reaches the commanded position! It doesn't even slow down. Obviously, I'm missing something, but haven't a clue as to what it could be. Any suggestions would be appreciated.

[TomKerekes]

Hi pblase,

Possibly the motor is driving the wrong direction? See FAQ.

Post a Step Response Plot of Command, Position, Output and we could tell you if this is the issue.

[pblase]

You're right, it was. I had to flip the "invert" switch on the driver. Might I suggest adding an "invert output" switch capability?

Now I can't get this thing to stop oscillating. I have the following settings:
100 Hz low-pass filter 2nd order filter on filter 2
P = .1
I = 1e-5
D = 0
Gain = 1 on the Configuration page.
That works, but it's really slow. I tried applying the Ziegler-Nichols method, but any higher integration factor and the stage starts oscillating out of control. Any higher P and there's tremendous overshoot. I'm not sure that Ziegler-Nichols works with this loop.

Question: is the Servo-Loop drawing correct? Usually, the "D" factor is taken from the error term, along with the P and I signals, not from the position feedback.

Also, should I use the deadband or the backlash to get rid of the step on the amplifier input? I adjusted the gain on the servo amp so that I get minimal motion right at about a half volt of input, but there's a lot of slop there. With the integration on and the stage sitting still, about every 2 or 3 seconds it will move to the other side of where it thinks it's supposed to be, overshooting and than waiting for the integration to catch up.

[TomKerekes]

Hi pblase,

Now I can't get this thing to stop oscillating. I have the following settings:
100 Hz low-pass filter 2nd order filter on filter 2

That seems low. Consider trying 500-1000Hz Q=1.4

P = .1
I = 1e-5
D = 0
Gain = 1 on the Configuration page.
That works, but it's really slow. I tried applying the Ziegler-Nichols method, but any higher integration factor and the stage starts oscillating out of control. Any higher P and there's tremendous overshoot. I'm not sure that Ziegler-Nichols works with this loop.

Is the amplifier in encoder velocity mode? Or Torque/Current mode? You might read this wiki article.

Please post screen shots of your settings and the response if you would like us to help.

Question: is the Servo-Loop drawing correct? Usually, the "D" factor is taken from the error term, along with the P and I signals, not from the position feedback.

The Servo-Loop Drawing is correct. We feel that method has some advantages and is the same as some other major motion controllers work. You can make it the alternate way by adding velocity feed forward.

Also, should I use the deadband or the backlash to get rid of the step on the amplifier input? I adjusted the gain on the servo amp so that I get minimal motion right at about a half volt of input, but there's a lot of slop there. With the integration on and the stage sitting still, about every 2 or 3 seconds it will move to the other side of where it thinks it's supposed to be, overshooting and than waiting for the integration to catch up.

Do you know why there is so much deadband in the amplifier? or is this due to stiction? You might want to wait until the loop is properly tuned before worrying about it. You might also see this wiki article.

[pblase]

Here are screenshots of my setup:
Configuration

IIR Filters

Step Response

Setup

The deadband is due to stiction in the bearing carriages.

[TomKerekes]

Hi pblase,

Unstable.

It seems your amplifier is in torque/acceleration mode as the Position (red) is 180 degrees (2nd derivative) out of phase from the command (green).

Did you read the previous link explaining why D Gain is needed in this case? It would be helpful if you explained what you have read, what you understand, and what you have tried.

Note it is usually best to test "Move" not "Step". Especially steps of thousands of counts that put the amplifier in saturation, which wouldn't occur in normal operation.

[pblase]

"It seems your amplifier is in torque/acceleration mode as the Position...." Yes, it is. I'm guessing that encoder-velocity would be better? I'll have to figure out how to run the signals from the encoders to the driver, since they're going to the KANALOG right now - and they're differential. Can I tap off of the signals after the KANALOG differential-to-single-ended decoders?

I could try it will the Hall-Velocity mode, but the manual says that that doesn't work well at slow speeds.

If I leave it in torque/acceleration mode, what would you recommend for a starting D-gain value?

[TomKerekes]

Hi pblase,

I'm guessing that encoder-velocity would be better?

I don't think so. Adding D Gain in KFLOP should do the same or better.

I could try it will the Hall-Velocity mode, but the manual says that that doesn't work well at slow speeds.

Halls have much too low of resolution to get instantaneous velocity.

If I leave it in torque/acceleration mode, what would you recommend for a starting D-gain value?

D-Gain values are typically 1 ~ 2 orders of magnitude larger than the P Gain.