Analog DC servo very big response delay

Hi to all,

I have done a retrofit of Mikron WF31CH running Bosch TR-10 230 amplifier and Siemens DC brushed servos 1HU 3070-0AF for X, Y axes and 1HU 3073-0AF.
This drive looks like have a very big delay in input command processing, despite being analog:

Drive manual: www.dropbox.com/sh/8owpaoax31ehf2x/AACji...older_nav_tracking=1
Is there any possibility to fix this? I have found in manual that drive has analog input low pass filter of 11.2 Ohms and 100 nF, which gives time constant of 1.12 ms and cuttoff frequency of 142 Hz. I'm thinking that it's possible to use smaller capacitor to change this, but I'm not sure that this will help much.
Also there is an option to exchange drive to microprocessor based with 8 KHz PWM, speed regulation of 1:10000 and 14 bit ADC, but drive developers can't give exact answer whether this situation will happen too, only experiment can help. I'm having same doubts as them, due to nature of brushed DC servos and DC tachos.

Have anyone encountered similar problem? I know there is a way to exchange with modern digital servos and PMSM motors which will give much better results, but this way is rather expensive.

Thanks!

Have you tuned your drives velocity loop? The nearly 10 ms delay from velocity command to actual motion seems a bit more like stiction combined with low velocity gain than drive bandwidth.

Tuning of this drive is a bit complicated, because it requires soldering capacitors and resistors somewhere on regulator board. It was pretuned for this motors on factory.
It does look like low servo gains on image with Z axis movement, when it tries to move back from overshoot, that takes almost 80 ms.

Looks like I need to find somewhere a RC decade box for the tuning. Once I get it I will comeback with updates.

Hello,

have you tuned the parameter FF1 ?

Sure, cruise ferror is close to 1-2 microns.

Hi again, I've got a time to get to the machine, but results are a bit disappointing.
Speed and current regulator inside the drive are pretty well tuned, increasing proportional gain for speed regulator, makes things only worse, changing integration term doesn't change much. I exchanged existing 220 nF capacitor with 100 nF, and it didn't change the delay.
Changing 100 nF filter capacitor with 22 nF haven't changed much too, but I still left it as it didn't made things worse.
Looks like the main problem with this machine and similar to it is static friction, even when using anti stiction lubes. I'm not sure that even exchanging with modern PMSM drives can solve this somehow.
Still question is open, whether something can be done with this machine.

Could someone please add FF3 for the jerk feedforward to the PID component in master branch? PCW mentioned that this could help in some thread, I would like to make a test of it.

There isn't much advantage in having jerk feedforward in the PID unless the motion controller has a matching output. That is likely to be the bigger part of the work required.

Maybe you could experiment with velocity-dependent P-gain? The "lincurve" component was designed with this sort of thing in mind.

I did a quick modification to PID component in servo sim to add this jerk derivative.

I will try to give it a go, as it possibly could remove some overshoots at movement phase 2 and 3(end of acceleration, start of deceleration) too, not only undershoot at start. But there is some overshoot at the end of move, and lincurve may help there.