f-error

More
30 Jun 2016 18:26 - 30 Jun 2016 18:30 #76863 by pmcstoneinc
f-error was created by pmcstoneinc
Is it normal to have more of a following error during a positive direction move to a negative direction move? Or should the following error be the same for each directional move? This is a servo position-velocity loop.
Last edit: 30 Jun 2016 18:30 by pmcstoneinc.

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

More
30 Jun 2016 19:10 #76867 by PCW
Replied by PCW on topic f-error
It depends on the mechanics
A non-counterbalanced vertical Z might have a higher error in one direction due to gravity

Otherwise I would not expect any difference unless there ware a motor/drive issue

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

More
01 Jul 2016 00:30 #76875 by tommylight
Replied by tommylight on topic f-error
Worm gear reduction tends to have a slight difference in forces from one direction to the other. High quality ones, it is barely noticeable, on cheap ones it is quite noticeable.
If you get that with ballscrews, misalignment or end bearings are due for a check.
Regards

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

More
01 Jul 2016 02:44 #76879 by pmcstoneinc
Replied by pmcstoneinc on topic f-error
Also I wanted to know the old system had a parameter for "in position servo loop gain" and a parameter for VFF velocity feed forward given as a percentage? Where would these values correspond in LCNC? Thanks!

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

More
01 Jul 2016 23:22 #76906 by Todd Zuercher
Replied by Todd Zuercher on topic f-error
I think VFF would be most analogous to FF1 in Linuxcnc. But there is not an equivalent to the "in position servo loop gain".

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

More
01 Jul 2016 23:36 - 01 Jul 2016 23:44 #76907 by PCW
Replied by PCW on topic f-error
Like most things with LinuxCNC, features like "in position servo loop gain".
can be built in HAL. All of the PID comps tuning variables are HAL pins which
means they can be connected to hal signals with net commands and changed
dynamically.

so as a first approximation:

Xcmd and Xfb .--> near comp --> mux comp --> PID P term

edit:
(commanded velocity near 0 might be a better "in position" flag as it would not change during a move)
In fact if you wanted to get fancy you could feed the commanded velocity into the lincurve comp to
set the P (or D or I) value based on velocity

Often these sorts of PID gain scheduling are done to get good high speed tracking without excessive buzzyness when static
Last edit: 01 Jul 2016 23:44 by PCW.
The following user(s) said Thank You: pmcstoneinc

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

More
01 Jul 2016 23:42 #76908 by Todd Zuercher
Replied by Todd Zuercher on topic f-error
Most anything is possible if your smart and clever enough to figure it out yourself. But there still isn't a ready made equivalent suitable for the unwashed masses to use.

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

More
02 Jul 2016 00:25 #76910 by pmcstoneinc
Replied by pmcstoneinc on topic f-error


Often these sorts of PID gain scheduling are done to get good high speed tracking without excessive buzzyness when static


This is the issue I have currently, I have a lot of hunting/buzzing going on while static or in position/stand still. Peter always a great help, thanks as always!

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

More
02 Jul 2016 00:44 #76911 by PCW
Replied by PCW on topic f-error

Most anything is possible if your smart and clever enough to figure it out yourself. But there still isn't a ready made equivalent suitable for the unwashed masses to use.


Yeah, there's always a tension between providing specific ready made features or providing
the tools to build almost any feature

LinuxCNC and HAL definitely tilt in the toolkit direction

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

More
04 Jul 2016 12:59 - 04 Jul 2016 13:03 #76961 by andypugh
Replied by andypugh on topic f-error

This is the issue I have currently, I have a lot of hunting/buzzing going on while static or in position/stand still. Peter always a great help, thanks as always!


"lincurve" was written mainly to allow you to vary PID parameters based on velocity or error. (this is why it has the -io output pins, as the gain pins in PID are (for no good reason) bidirectional pins. (Fixed in versions 2.6 onwards)

linuxcnc.org/docs/2.7/html/man/man9/lincurve.9.html
Last edit: 04 Jul 2016 13:03 by andypugh.
The following user(s) said Thank You: pmcstoneinc

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

Time to create page: 0.123 seconds
Powered by Kunena Forum