Linear scale resolution and viability

Hello.

At what resolution do linear scales become viable autonomously?

In other words I currently have 4096 ppr rotary encoders on a 10mm pitch ball screw, meaning 1638.4 pulses per mm.

The usual combination is for low res linear to supply positional info with higher res rotaries supplying velocity information via various PID loops.

If I am trying for a resolution of .005 - .01 mm at the table, will using 1um provide enough resolution to provide sufficient data for velocity and position closure calculations?

So, would a 1um linear scale provide enough resolution to act autonomously of rotaries entirely i.e. is 1000 pulses (vs 1638 pulses) per mm sufficient for realistic velocity and position closure feedback all on its own?

As I understand it, usually the problem is the mechanical disconnect between the motor and the slide. Whether backlash or springiness between the two, it makes tuning difficult. This is why they use a velocity sensor.

I also once read somewhere that you want minimum 5x more resolution then what you to hold to. I wouldn't take that as gospel but it does makes some sense to me.

Is there a reason you can't use two sensors?

Chris

Thanks Chris.

Actually there is nothing stopping me from using both sensors via a pair of PID loops (PD on the rotary encoder and I on the linear).

The query was primarily one of simplicity, in other words, whether the linears were of sufficient resolution whereby the dual PIDs may become a mute point and one would do. Your point of retaining a feed back device on the ballscrew is well taken, as given enough lost motion there may be some excessive wind-up on the motors until motion is seen on the linears.

Globally though, I am still rather interested to know what resolution (at least in theory) may be sufficient for a linear encoder to hold .01mm at the table and still have enough resolution to give adequate velocity feedback information. As magnetic linear encoder resolution is steadily increasing coupled with increased affordability, I suspect this question will be broached more and more moving forwards.

unless you dont use real servos
there is no real ned for encoders liniar as you cand correct

pl7i92 wrote:

unless you dont use real servos
there is no real ned for encoders liniar as you cand correct

linear encoders tell you where the slide is, rotatory encoders tell you where the motor is.
These are not necessarily the same place.

linear encoders can compensate for temperature caused growth for instance.
They can also be used for homing.

Now most people probably don't need that precision but you certainly could do it.
In fact dual feedback has been done in linuxcnc - though there was no writeup about how much better it was.

Chris

Quite right Chris,

In this case 4096ppr rotaries are driving 10mm pitch C5 bscrews. The machine had the option to fit linears directly to the table back in the day when it was made back in '96. Thus to retrofit 3 linears and given there already exists 3 empty encoder positions above and beyond those already in use will be interesting.
Especially when affordability of ever increasing resolution magnetic linears are available. In this case 1um.

Of greater interest on this journey is the following...
One day those C5 bscrews will need replacing. If, with the benefit of the linears, a C7 replacement will perform adequately and close to the same accuracy, it will be a journey well worth it.

The electronic and software side will always evolve rapidly. To me, as the owner of a machine which I have working just the way I want it, the greatest question is, how to future proof the iron side by leveraging the benefits of the evolving electronic side...

Hi Andy.

Thanks for the reply. I have another thread open which is similar but more recent and topical at:

forum.linuxcnc.org/10-advanced-configura...planner-et-al#145295

I would be VERY interested in your thoughts, particularly on the feedback arrangement I am proposing which is a block diagram I have attached to one of the posts.

In this context, here is also some discussion on the best arrangement for cascading loops based on a position/velocity cascade, driving a servo motor.

Again thanks!,
Cheers,
Marty.