initial setup before srvo tuning

More
18 Feb 2021 13:33 #199250 by bevins
Is it deadband you ned to stop creep? I dont know if this is servo creep but when I
enable the drives the Y axis starts moving until it reaches f-error limit and throws joint following error.

I need to stop this before I can start tuning. Is there a trick? X and Z axis is ok, and the encoder scale is correct also.

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 14:16 #199255 by tommylight
Only one axis ?
Things to check:
-tacho feedbak/polarity/direction
-encoder wiring/polarity/direction
-pid setup in hal - feedback might be disconnected
-output polarity to drive input
Disconnect the servo input and enable the drive while the input is not connected to anything, it should creep slowly.
When connected and controlled by LinuxCNC, no creep should occur under any circumstances
Deadband will not help.

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 14:36 - 18 Feb 2021 14:38 #199259 by bevins

Only one axis ?
Things to check:
-tacho feedbak/polarity/direction
-encoder wiring/polarity/direction
-pid setup in hal - feedback might be disconnected
-output polarity to drive input
Disconnect the servo input and enable the drive while the input is not connected to anything, it should creep slowly.
When connected and controlled by LinuxCNC, no creep should occur under any circumstances
Deadband will not help.


Only one axis is doing it,
encoder is connected and working and in the correct direction.
If I disconnect the servo input it doesnt do it.
I'll check hal FB connection but I think the servo would runaway alot faster if this was the case.

I am just trying to get it to a point where I can start tuning the servos. It's annoying....
I'm stumped.
Last edit: 18 Feb 2021 14:38 by bevins.

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 15:02 #199261 by tommylight

If I disconnect the servo input it doesnt do it.

It should, so check the tacho, namely the brushes if they are clean and in good condition.
Also check the drive pots, maybe someone messed with them ...

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 16:31 #199289 by Hakan
That creep thing just happened to me too. It was before feedback was engaged (P=I=D=0).
There is a deadband setting in the servo drive and I set it to 50 mV. The servo drive displayed -30 mV
when it was supposed to be 0 mV. No more creep. I suppose I can remove that setting when feedback
is engaged,

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 21:48 #199352 by bevins
It is creeping with it unplugged.

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 22:37 #199355 by PCW
Sounds like you don't have the drive enable working
the drive should be disabled until LinuxCNC closes the loop
And perhaps more importantly when LinuxCNC detects a problem
like a following error or a watchdog bite

Normally you dont want any analog deadband

Please Log in or Create an account to join the conversation.

More
19 Feb 2021 01:33 #199364 by Michael
What drives and mode of operation? Have you tuned the drive yet?

Please Log in or Create an account to join the conversation.

More
19 Feb 2021 07:09 #199375 by Hakan
In my case it is well understood why it happened. It was the very first test with the servo
connected to linuxcnc and I just wanted to test that it at all reacts to commands without
needing to worry about reactions from the feedback system. Already here the thought might be wrong.
The encoders are connected, but P,I and D are all set to zero. So no feedback used.
Instead FF1 has got the right value, feed forward velocity value and we all know how to calculate that value.
When hitting the "machine on" button, servo enable goes high and the 7i83 outputs 0.0 V as measured by the dmm.
It is just that the driver has a small 30 mV offset in the input so it takes it as 30 mV input and starts to move slowly.
No feedback is engaged in linuxcnc so nothing will tell it to stop moving. And there we have the creep.

What I did was to set an analog deadband which stopped the creep. I could verify that the servo moved and
that it went in the wrong direction so I changed that. After that I started to tune the servo, gave P a value
and by that the analog deadband value I had set in the servo drive wasn't needed (or wanted) so was removed.
Instead an offset value could be set to compensate for that 30 mV reading. Kind of stupid, it would be better if it actually
understood that 0.0V actually is 0.0V.

tldr; Can/will happen when no feedback is engaged and the servo drive doesn't read exactly 0V when 0V is output

Please Log in or Create an account to join the conversation.

More
19 Feb 2021 12:17 - 19 Feb 2021 12:18 #199409 by bevins

Sounds like you don't have the drive enable working
the drive should be disabled until LinuxCNC closes the loop
And perhaps more importantly when LinuxCNC detects a problem
like a following error or a watchdog bite

Normally you dont want any analog deadband


I will check if the drive is enabled. I am pretty sure it is. But I can check. It is a retrofit so I am just closing one contact that sends 0v to all three drives on the enable pin. The two other axis seems ok. But willcheck further today.

INI
Warning: Spoiler!



HAL

Warning: Spoiler!
Last edit: 19 Feb 2021 12:18 by bevins.

Please Log in or Create an account to join the conversation.

Time to create page: 0.165 seconds
Powered by Kunena Forum