Absolute linear scales

I have this position line in NC program

G54
G00 Y-35

It results in DRO reading -34.976mm and DTG shows 0.000 If I modify NC program:

G54
G00 Y-35.024

It results in DRO reading -35.000mm and DTG shows 0.000

I believe position shown in DRO is correct so problem is with commanded position.

Tero

The GUIs can be set to show absolute or commanded position.

Also, have a look in Halshow (or with Halmeter) and see what the PID command, feedback, error and output are. (If I am right to think that this is a servo system)

Do these help? It seems that my error is in pid.y.error... Approaching direction does not have an affect. Do I understand correctly that control sees this -0.0239 postion error and tries to move Y by outputting -0.00952 volts to Servo Amp but it does not move with that voltage? Should I check if I have some setting in servo amp so that voltages below 0.01 are ignored? Or can this be solved by tuning PID?

Tero

Is .4 the highest P term you can use? (raising P would decrease the error in inverse proportion to the P value)

You could zero the 7I77 output/drive input offset by setting the PID bias to -0.00952V but down at this level, the 7I77s output and drives input drift will make this setting somewhat unstable

adding some I term will help also

0.5 started keeping noise. But I can try ie 0.45. I will also try adding a bit I to see what it does.


Tero

I think the I value has to be fairly large before it makes much difference
(a small value will take a long time to respond)

terkaa wrote: [Do these help? It seems that my error is in pid.y.error... Approaching direction does not have an affect. Do I understand correctly that control sees this -0.0239 postion error and tries to move Y by outputting -0.00952 volts to Servo Amp but it does not move with that voltage?

Yes, exactly. With only a P-Gain term you can never actually get to the set point, it is a real-life example of Zeno's paradox Or maybe not exactly.
For some small error the output of a P-only controller will be too small to move the system. So you will always settle at that small error from the setpoint. (though it would normally be symmetrical for a horizontal axis)

That is why you need some I-term to integrate the small error over time until it becomes a big enough voltage to move that final little bit.

Add some Igain and you should see the error disappear Start with about P/10 and increase from there. Too much IGain will eventually give slow oscillations.

Ok, I tuning it is then. Btw this is HMC so Y in this case is vertical axis with hydraulic counterbalance.

Tero

Ok I was able to increase P to 0.48 and also adding 0.06 to I And now positioning accuracy is within +-0.0002 mm and that is good enough for me.


Tero

Today we noticed control cutting corners a lot with G00 moves. We have following G-code:

(current pos X-725. Y-100.)
G00 Y 65.
G00 X 725.
G00 Y -100.

In the first end about 40mm before reaching Y65. X starts to move towards 725. In the other end about 40mm before reaching X725. Y starts to move towards -100. Any ideas why? Attached is drawing of path. Black is what it should be and red is actual:

Tero