Strange Offsett in Y and Z axis after touchoff

More
12 May 2015 14:13 #58586 by REEEN
Hello Community,

I have converted a CNC Mill with glass scales and DC Servo Motors and did the tuning of the axis on Saturday. The machine runs with gmoccapy.
While doing the tuning I recognized that there is a small offset in the X and Z axis.
When moving from position to position the Y axis is always +0.004 mm offsetted to the position it should normally stay and the Z axis by 0.008 mm.
When I started the tuning the error was bigger, while increasing P the error got smaller. I could not remove the whole offset by increasing P so I used a bit of I to get rid of the whole offset.
So for now everything was fine.

Now I used X = 0 command in G54 coordinate system to do the touchoff.
Next I did the same thing with Z and Y:

Z = 0 => and then it happened, the displayed Z position is now 0.008 and not 0.000
Y = 0 => same here the displayed position is now 0.004 and not 0.000
It worked with X!

When I move the machine to e.g. Y = 200 display shows Y = 200.004 ... and same for Z with + 0.008
Iam not quite sure whether this comes actually from tuning Iam not sure whether the offset before and now is the same.


The curious thing is that this only happens when doing the touch off...

I hope someone has a hint or so..
I need you help :)

Greets Rene

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

More
12 May 2015 18:12 #58593 by BigJohnT
That is a very small amount... 0.00015". I assume your looking at actual position instead of commanded position.

When you tuned the axes what was the following error.

I have some servo tuning information here .

JT

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

More
12 May 2015 20:10 #58599 by REEEN
I got it to less than 0.006 mm in rapid movment by not looking at the big jump at the start of the movment this was up to 0.012 but also very good I assume.

What I think ia really curious is the fact that the error isbnot present before touchoff and then it looks like the machine actually thinks the position with the offset is the right position because it holds it perfect and I can go back an forward reachin the same positions perfectly but always with this strange offset.

I don't think this is a problem of tuning it is more likely a software Problem.

I have another machine that does not have this problem.

I was looking at the actual position you are right.

And yes I used your tutorial :)

Greets Rene

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

More
12 May 2015 21:55 #58602 by andypugh

What I think ia really curious is the fact that the error isbnot present before touchoff and then it looks like the machine actually thinks the position with the offset is the right position because it holds it perfect and I can go back an forward reachin the same positions perfectly but always with this strange offset.


What happens if, instead of touching-off to zero you touch-off to 0.005?

I think this might be rounding-error in the scales/encoders. What is the actual precision of your position feedback devices?

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

More
14 May 2015 01:30 #58653 by REEEN
Hello Andy, I will try it out but I'm very sure if I touch off to 0.005 I will be at 0.005 plus the offset as always.

I can't imagine this:

Directly after homing, coordinates are G53 so I go to Y=0 and gui shows Y=0.000 and then I press Y=0 in G54 and now Iam at Y = 0.004 ...

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

More
14 May 2015 01:54 #58654 by andypugh

Directly after homing, coordinates are G53


I don't think you are ever _in_ G53.

It might be interesting to look at the position in machine coordinates and in relative coordinates.
(and also look to see what the G54 offsets are)

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

More
14 May 2015 06:26 - 14 May 2015 06:28 #58658 by BigJohnT

Hello Andy, I will try it out but I'm very sure if I touch off to 0.005 I will be at 0.005 plus the offset as always.

I can't imagine this:

Directly after homing, coordinates are G53 so I go to Y=0 and gui shows Y=0.000 and then I press Y=0 in G54 and now Iam at Y = 0.004 ...


The only way to move in the "machine coordinate system is to use G53 on that line.

For Example:
; move to machine 0
G53 G0 X0 Y0 Z0

A good paragraph to read when you can't get what you want on the DRO
linuxcnc.org/docs/html/common/User_Conce...hen_you_8217_re_lost

JT
Last edit: 14 May 2015 06:28 by BigJohnT.

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

More
18 May 2015 15:04 #58813 by REEEN
I did some tests on the weekend,

 

I found out the error comes from the axes tuning for sure.

I looked at the PID-error of the Y and Z axes and the error is equal to the offset.

But what’s really curious is that the output from the PID of the Y and Z axes shows a constant value so the PID wants to get rid of the error, but why doesn’t the error disappear ?

I’m wondering that the output of the PID does not increase if the error won’t get away ?

For the X axis everything is fine when standing still the PID shows no error and the output of the PID is 0.

I found out that if I leave the machine running the error will slightly decrease but very very slow and I don’t know if it ever would reach a zero.

 

The tuning of the axes is very dynamic.

 

So now the most curious thing if I set the I of the PID to 0 and then turn it up again the error will disappear until I switch the machine on and of then the error will return… And yes of course I saved everything to the INI file.

 

Does anybody know what I did wrong ?

 

Thanks everybody !!

 

Greets Rene

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

More
18 May 2015 22:31 #58825 by andypugh

I found out that if I leave the machine running the error will slightly decrease but very very slow and I don’t know if it ever would reach a zero.


For a very small error the P-output of a PID is also very small. Often too small to actually move the motor.
So, normally you need some I-gain to get the last bit of error to go away.

If the error is going away very slowly, it should go away more quickly with more I-gain. But there is a risk of too much I-gain causing slow oscillations around the setpoint.

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

Time to create page: 0.308 seconds
Powered by Kunena Forum