Schaublin 125-CNC retrofit.

oh, there are answers in beetween....

Step/dir or analog or bus has nothing to do with tuning!

Tuning you position control is always the same. on a digital (bus system) it is easier because you define an controll window then send a command 1000 and the drives does 1000. At analog control you additionally have to do with Voltage calibration or dc offset.

If there people outside who tells that they need no tuning for step/dir control. then they don'T know what there talking about... Because with no need for tuning the drive does the position, and at an interpolation path the path following would be horrible.

linuxcnc-ethercat is fully stable working. Sascha (the developer/programmer) , myself and many others i know using this in industrial projects.....

It is not offically part of linuxcnc, but behaves as every other external hardware driver...

back then, when this was brand new (approx 10 years ago.. time for anniversary party), there have been discussions on the mailing lists about integration and sometime saschi lost lust to beg again...

That’s good to know for future projects. I was considering it but lost track and got confused. I saw TOT use ethercat for his Maho conversion and thought this would be handy but was not sure If I could get it running with LinuxCNC.
In the End I sticked to Analog Servo drives cause this is what I have already done.

anfänger wrote:

Most prefer Step/Dir to avoid tuning.....and you can tune without the servos actual attached to the machine.

You should still tune your servos with the machines installed, whether you go step/dir or analogue 0-10V. There seems to be a myth that step/dir somehow works optimally regardless of what you attach to it. The default settings may be generally stable for most applications but that simply tells you that you have a suboptimal setup unless you take the time to tune the response.

Thanks a lot. So with Ethercat drives, do you again have the option of having the drive close just the velocity loop, or both velocity and position loops?

I get the tuning. You need to tune the system to the axis inertia. Actually I did a course on motion control, and we worked through the sizing of drives based on inertia matching. I need dig into that again, and work out what motors would be needed.
Mark

that depends on the drive, all that i know have all options (torque, veloctiy, position and all combinations of them...

but the advantage of the underlaying transport system (ethercat) is, that you are able to sync the execution times from the drives. (distributed clocks, the exact timestamp when the drive should fetch and put his information from/to the bus.)

So you can also do cyclic position mode/control. for Examble an Beckhoff AX5000 does his position control at an cycle time of about 62.5µs equal to usually linuxcnc servo period of 1ms... if such an controller is well tuned (right inertia, resonace filters etc. ) that are mega precisious and fast Axis.

Wasn't there a servo drive called stimbl ( may be spelled wrong ) ? Think Andy P an Skunkworks were trying those drives out with good results. Thought it used ethercat .

I would never start tuning as system with everything attached. I make too much errors and had the Z axis oscillating in the beginning and it was not funny.

I never used a step dir cause I didn’t have system that supported this. But I seem I also believed this myth. As far as I understood with analog control you tune the drives to match a certain speed to a voltage and then tune LinuxCNC to stay on the planned path.

With my Chiron I was Lucky cause the drives were tuned and
I did my tuning without the belts and it was quite good after I realised that I had an error in calculating the max acceleration. BUT I had the acceleration read out from the old parameters and just needed to convert it the the units used by LinuxCNC.

Using velocity mode, tuning for load inertia is done in the drive so using the existing
(tuned) drives means only position mode tuning (chiefly FF1 and P) is needed

as PCW told
if you have an good tuned stiff velocity loop and no mechanical play then for position controll there is only a bit of P neccesary. (and FF1 for fine tuning of analog setpoint, in an digital drive system this is normaly 1.0)

But tuning an Velocity / Torque loop without load is nearly impossibel. If you get an todays velocity loop with acceleration feedforward, reconance filters, highpass, inertia calculated velocity observer etc. proper tuned, then the following error with only velocity loop should be +- zero.