Spindle Encoder - Speed limitations?

[travasky]

Im back on my Kitamura project again. Its been a while.

Im in the process of selecting an encoder for the spindle. Max speed is 10,000 rpm so my options are somewhat limited not only due to speed but also space/mechanical constraints. Best bet right now seems to be a non-contact RLS brand ring encoder. Output is the typical RS422 differential line driver format so connectivity shouldnt be an issue.

Typical spindle encoders seem to be right around 4096 counts. The closest RLS offers in my needed size is a 4864 count encoder ring. Their 4864 count encoder can do the following speeds @ a given edge separation between the signals.

11837RPM @ .07us edge separation
11837RPM @ .13us edge separation
11514RPM @ .50us edge separation

My question is, can the Oak reliably read these signals at those frequencies? I would imagine the .50us would be the "safest" bet, but I still don't know how quickly the oak is sampling these inputs.

[cncsnw]

I am pretty sure this has been asked before, but I do not recall whether there was an answer:

What does RLS mean by "edge separation", if it is independent of encoder counts and RPM?

I am assuming the output is quadrature (two square waves, offset by 1/4 cycle); and I am assuming that when they say an encoder has 4864 counts, that means 1216 full cycles per revolution (with each cycle having four counted edges: A rising, B rising, A falling, B falling).

If you have 4864 counts (edges) going by in each revolution, and you are running at, for example, 11514 revolutions per minute, then that would yield 56,004,096 counts per minute, or 933,402 counts per second. The reciprocal of that would be 0.00000107 seconds per count, or 1.07us per count.

Why would they say the edge separation is less than half that amount? Are they saying that the A and B phases are not necessarily 1/4 cycle out of phase with each other, so that the time between successive edges can be more and less than 1/4 of the time for a four-edge cycle to go by?

In any event, if you have a quadrature differential spindle encoder signal, and it provides no more than 7200 counts per revolution (no more than 1800 cycles per revolution), then you can run it at 10,000 RPM and reliably read the signal with a Centroid control board such as the Oak.

1,200,000 max counts/sec * 60 sec/min / 10000 revs/min = 7200 max counts/rev.

7200 counts/rev / 4 counts/cycle = 1800 cycles/rev.

[travasky]

The full manual for the LM13 is here: https://static.rls.si/media/wysiwyg/dat ... sheet_.pdf

But here's the relevant screen cap:

Timing Edge Separation.JPG (16.49 KiB) Viewed 5544 times

[eng199]

The short answer is that there will be no problem with 4864 counts per revolution at 10000 RPM.


Normally, encoders are rated with the frequency response of A and B channels. It is uncommon to have a rating above 1MHz. This means any MPU11 based control (Oak, ALLIN1DC, etc.) will handle any normal encoder without even switching to high speed filters (parameter 323). Low speed filters will handle channel frequency up to about 1.5MHz, and high speed filters will allow up to 6.25MHz.

Minimum edge separation is 4 times the normal channel frequency rating. This is convenient because it is also the encoder count frequency. It is also confusing because most encoder manufacturers don't use this specification.

In the datasheet I see minimum edge separation in the part number. I don't know where the RPM ratings come from, and the rings are a different datasheet. I don't think the posted RPM and time (us) ratings correspond to each other. I would think the read head should be matched to the application, making option "C - 1 µs (1 MHz)" the one to use. The faster models would work, but ideally the slowest possible rise times are used to reduce noise in the system.

[travasky]

Thanks for the detailed responses gentlemen.

Eng199, the read head seems to be pretty universal for all of the rings but the rings come in a wide variety of poles an signal widths depending on the ring OD. For this application Ill select the .50us response time ring paired with the appropriate LM13 read head. Should be either option B or C, but Ill let RLS help me decide for certain.

[tblough]

Be sure and check with RLS because the 13 read head speeds are slower. They have a special chart for the 13 head you have to email them to get.

[travasky]

Good to know Tom. Thanks!

[travasky]

So Ive been on the phone with Renishaw today (they own RLS) and there is a problem with using the ring size I want at the speed I want while still having 4000+ counts per rev. It occurred to me that 4000+ counts is probably overkill on a spindle. What is the recommendation on spindle encoder resolution? Wouldn't something like 1024 work perfectly fine?

[tblough]

1024 lines = 4096 counts works perfect for a spindle encoder. You don't gain anything going to higher resolutions.

[travasky]

4096 counts is going to be tough to achieve at 10,000 rpm for these magnetic radial encoders. I need something closer to 1000 counts. So my question is would that be sufficient resolution for a spindle?