more on separate positioning and metrology

Chris replies:

>> Has anyone here built a machine with separate drive and metrology
>> structures? I was just reading about a diamond turning lathe at

Livermore

>> labs of quite large size, > 1 m swing, I think. It had separate

structures

>> for the drives and laser metrology.

>It is quite common to use a linear encoder for feedback rather than a
>rotary encoder on the screw.

Do you know of any examples? I think that for this to work the objections
you mentioned below must be overcome, what has been done to do this?

> Not seen too much on low grade machines
>due to expense, and also because it is hard to make a servo loop
>stable if the screw has any backlash (servo loops tend to hunt between
>encoder counts - if there is backlash between the motor and a linear
>encoder, it will hunt across the entire backlash distance).

The common servo motor/ballscrew type drive probably has very limited
bandwidth. Anyone care to make a guess what the 3 dB point is? 5 Hz? 50
Hz, 500 Hz? Also has anyone measured the frequency where the phase shift is
180°?

This could get a bit tricky, as the length of the ball screw between motor
and nut changes, so the resonance properties should change too.

>> This makes a lot of sense to me. If the feedback was quick enough it

could

>> act even though the drive structure was flexing, if the metrology

structure

>> maintained its accuracy.
>> So vibration, constant load, and thermal strains would be less of a

problem

>> with this sort of design.

>The problem is that making actutors which can operate at the
>frequencies necessary to combat vibration is really pushing the state
>of the art and currently more expensive than the amount of cast iron
>or polymer concrete necessary to stabilize most machines of the size
>that we tend to talk about.

I see your point.
Many precision machines use two stages of drive: a coarser long range drive
and a precision short range drive. Piezo's are frequenty used for the short
range stuff. They are mounted right at the measured point, so there is no
problem with frequency response. If the vibration is bounded to the range
they can cover, this kind of thing should work well.
I've seen a servo use screws and piezeo and the split servo concept to cover
about 500 mm at a few 10's of nm precision.
Differential cost is really quite reasonable, since if the metrology exists,
the piezo's don't cost much and the difference in electronics and software
seemed to be simple.

>Correcting for bed twist and stuff should be possible, assuming that
>vibration doesn't render the laser measuring setup inoperable. And
>also assuming that you can model the mapping of bed twist to cut
>diameter error - not trivial as it depends on the cut diameter, tool
>shape, tool height, etc.

Yes, for a complete system this would need 6 axis measurement. That's a
puzzle.
Maybe there is some sort of hexapod metrology that would work here.

Just thinking out loud here.

There may be some cases where the weight of a cast iron or even carbon fiber
machine would be excessive, and closed loop control would then be desirable.

Elliot Burke