Spindle encoder resolution?

[Mcgyver]

If you were to put a spindle encoder on a machine so output drives machine motion somewhere else e.g. hobbing or an electronic leadscrew, what are your views on the minimum required resolution and why? Different ops might require different resolutions of course, just trying to get a sense whats required/used in these sorts of applications

Thanks

[Bented]

A resolution that will produce the motion that you require, no more or no less.

[MaxHeadRoom]

Generally an encoder on a spindle is for precise RPM control, where generally lower resolution is required, if positioning is required, then generally a much higher resolution is used, at least 1k pulses/rev.

[ikdor]

As much as your controller can comfortably handle but beyond 4000 seems to me of no added benefit. Of course your wallet should also handle it comfortably.

[MaxHeadRoom]

Also much depends on the capability of the controller, where the loop is closed back to etc.

[Baz]

Try working through the numbers for an actual example to see if it makes sense relative to your normal machining tolerances. Eg if you are doing screwcutting, ELS, and your number of pulses is such that each pulse equates to moving the saddle half a tenth and what you are making is a bolt to repair your lawnmower is that kind of out of proportion. There are quite a few ELS designers eulogising their gigahertz processor which is just silly.
However there are some times when a high number makes it easier. Eg, somewhere in the old posts on here is John Stevenson's original gear hobber that used, I think, a 4000 pulse unit so that a simple hardware divider chip (shock horror no processor in sight) could work with a 40:1 dividing head.

[legendboy]

Go 14bit and be done with it

[macona]

Not all controllers can handle high res encoders so you need to look at the max encoder pulse freq the control will handle and divide that by the max spindle speed to find the max encoder count.

[dian]

if you want to work to x degrees of precision, how many pulses per x do you need?

[Sparky_NY]

I can thow some real world numbers at you to give you a practical sense. The older bridgeport boss cnc mills had 256 line encoders on the axis, later machines went to 1000 lines. The newer Fanuc servo motors used widely have extremely high encoder counts, 50,000 rings a bell but don't quote me.

On my linuxcnc retrofitted bridgeport boss mill I used 1024 line encoders and result in sub tenth thousanth of a inch resolution.

Resolution on a machine is not determined only by the encoder, the ballscrew pitch and any belt drive reductions come into the calculation as well.

As others have pointed out, the bit rate of a encoder can exceed what the controllers input can handle if you do not pay attention to the calculations and specs. Encoders have a max rpm spec as well.

Lastly, most encoders are quadrature so the resolution is not directly equal to the line count.

[Mcgyver]

Thanks guys. There's a few applications so I was looking at what others have thought a decent result requires, i.e. good quality gears, no broken taps etc.....understanding that different applications may require different approaches. Some quick calcs on rigid tapping suggest 1024 should be good and hobbing, the same, both of which are slower rpm operations.

Baz thanks for the 4000 number. I wonder if its just what John had kicking about? With 1024 ppm and a 16DP hob, it should give about 2ths accuracy (0.1965 circular pitch). Maybe it should be more?

On the controller thing, both ops will be lower speed. The mill has a mesa 7i96 (worst name ever - they right it as 7I96, is that a "1", a small "L", capital "I"?) and linux cnc. Gear hobber likely home made, micro controller or arduino...but spindle speeds will be slow