[joeavaerage]

Hi Samco,

that being said.. if the drive doesn't follow the trajectory of the controller - will you know? (then probably does it matter or do you care - but that is a bigger question)

Can you set the following error in the drive so it trips an estop when the following error is over a certain amount?
Look at something like a gecko servo drive(g320x). Its minimum following error is +/- 256 encoder counts. This is how far out the drive position needs to be before it will error.. That seems like a lot..

Yes, the controller can and will know if the drive faults 'Following Error'. Gecko 320's are pretty old school and very basic. That they have a fixed and fairly wide following error window should be no surprise.
Moden AC servo drives, and that includes the cheap as chips Chinese made units flooding the market all have a programmable following error window.

It is common practice to set the window wide at the factory so that a user may tune the servo in the machine without pesky faults. Once the tuning is complete however then you tighten up on the following error
window to suit the tolerance you need to achieve. It is if you like the final part of the tuning process. Essentially if you have a wide window then the expectation is that the servo will easily follow its commanded path
within that wide tolerance and you would not see any following errors except in exteremis. Alternately you may tighten up on the window but then you might expect the servo to occasionally lag to the extent that it
would fault. You might want to reduce the max acceleration in the CNCsoftware a little for instance and give the servo just that little extra time to accurately follow the commanded path.

The ability of a servo to follow a given trajectory is a fundamental property of the physics of the servo and its drive. For instance its torque vs its rotational inertia and whether the drive has the current capacity to achieve
the acceleration the trajectory commands or the voltage withstand to decelerate as the trajectory demands.

Does it matter whether you use a closed loop controller like Galil or LinuxCNC OR rely on the feedback mechanism built into a modern AC servo drive.......the absolute best a servo can do ai related to it
torque, inertia, current and voltage?. My argument is that they are all equivalent assuming they can all maintain the same closed loop bandwidth.

Just as an example I have an older style, still AC servo, by Allen Bradley circa 2005. Amongst the things that you can program into the drive there is a piecewise linear approximation of the current to magnetic saturation curve
of the servo. This allows the FOC algorithm to apply a current correction as a feedforward term, and thereby reduce the feedback required with a consequent increase in overall bandwidth. Have you ever heard of
a closed loop controller that has a nonlinear saturation curve programmed into it? I certainly have not. The point being that because Allen Bradley made both the drive and the servo they are able to match
the two to perfection, better even than Galill or LinuxCNC.

Were I to program LinuxCNC or a Galil I could not hope to exceed the 'smarts' built into the Allen Bradley servo drive....unless I too were to program in the non-linear magnetic saturation curve. Note that
Allen Bradley supply the magnetic saturation data for their servo, it comes with the setup and tuning software. You declare to the drive, in fact you don't even have to do that, the setup software detects the model
servo to which it is attached and automatically supplies the saturation data.

This drive was designed in the late 90's and early 2000's, making it 25 years old. In the intervening time AC servos have only ever gotten better and better.

I studied Electrical Engineering at University in the early 80's. Control Engineering was, and still is a very major part of a degree. At that time servos were still old school, resolvers,
synchro's, transconductance amplifiers. We spent hour upon hour, day upon day experimenting with these devices while predicting and analyzing them with the mathematical theory that
underpins them. I will confess to be rather snobbish that any mere feedback drive could hope to attain the 'excellence' that I had so intimately observed while an undergraduate.
Then to rather prove the point I bought an AC servo (400w Delta) just to test it out and prove that my snobbish position was correct. I was not!. Quite frankly a modern AC servo will outperform
your expectations.

It's not that the old school servos were/are bad, in fact if they are tuned to excellence they are just as good as a modern servo, but a modern servo achieves that state of excellence so much easier.

I'd recommend to you that you try them out too. I do not recommend the cheap as chips Chinese servos, not because they don't work, but because the documentation is fair at best and bad at worst and
more often than no setup software. Delta (Taiwanese brand made in China) or DMM (Canadian brand made in China) on the other hand are good quality, documentation, support and most importantly
setup and tuning software at fair prices. I promise you'll spend hours fiddling wiuth the oscilloscope function...it's fascinating. What we would spend many tedious hours (years ago) can now be done in tens of minutes.

This is no difference from a good stepper setup. it will follow the command trajectory just fine until it stalls. A decently configured open loop stepper system works just fine too...

100% correct, a well designed open loop stepper system is every bit as good PROVIDED the stepper is used within it's capability. Feedback servos may allow you to push that somewhat,
but they too have a limit beyond which they are overmastered. A full closed loop control cannot make that range of operation any better, full closed loop control is good, but it's not magic.

Craig

[suntravel]

Full closed loop with Acorn6.

I am happy with the results.

https://youtu.be/UvlV7wJggpY

Uwe