Encoders? Karl?

The servos on my mill have sinusoidal encoders which are not compatible with most CNC control stuff.

I am looking at various options.

One is to make a converter to convert sinusoidal to quadrature. It is a pain in the butt to do due to lack of documentation on what wire does what.

Another one is to reuse a converter board from the existing Heidenhain controller. Same issue as above. (and I do have documentation, it just does not say what is what).

I have opened up one of my servos to see the shaft. It is a 10mm shaft.

For that size, US Digital has suspiciously cheap encoders E7P:

formatting link
They seem like they will fit, however, I am slightly surprised at the price.

I Wanted to know if anyone has any suggestions, as I do not want to go a wrong way.

i
Reply to
Ignoramus21167
Loading thread data ...

Those are pretty low resolution, only up to 720 CPR. The encoders on the CNC machines I used to work on were all around 2,000 CPR. You have to do the math with your ballscrew pitch to figure out what the final movement resolution is, but I expect it will be far too low. Also since this is a servo setup, the minimum servo error with such a low resolution encoder could equate to an unacceptable position error at the cutter.

Reply to
Pete C.

Well, on the back of the envlope calculation is: ball screw pitch 1/8" (0.125"). Pulley ratio 1:2. So, 720 CPR results in

0.125"/2/720 = .0000868 inch per cycle OR 0.002mm

That would be acceptable to me.

i
Reply to
Ignoramus21167

.0000868" per count * servo error tolerance of +/- 128 counts (256 count window) = 0.0222208" error which is very significant and would not be acceptable to me.

Reply to
Pete C.

that sounds bad!

i
Reply to
Ignoramus21167

Bingo! That would be why the "big boys" use 2,000 CPR or better encoders.

What is the CPR on your current encoders?

Reply to
Pete C.

I've had good luck with these:

formatting link
I'd get a higher count per inch encoder. get one that gives you a very even count per inch movement number, like 20,000 or 50,000 or even 100,000. remember you get four counts per encoder pulse.

get the differential option. use their wire and connectors.

Don't be a cheap skate and use the PC 5 volt supply. get one just for the encoders. I usually get a multiple voltage out DC power supply. Seems like I always need 24, 10, 5 I don't know if EMC does analog (10V) or opto isolated digital I/O (24V)

Karl

Reply to
Karl Townsend

I am confused.

Your accuracy number for my example was 0.02".

If you increase the count by 3 times (as in your mention of 2,000 CPR) then the accuracy only improves to 0.006", also an unacceptable number.

Something is not right.

Reply to
Ignoramus21167

For a home machine, I would want accuracy at least to 0.001", so the normal servo error window would need to be ~0.0005" or so to be acceptable. The big commercial machines hold much tighter accuracies, to the extent of using liquid cooled temperature stabilized ballscrews, double servos with glass scales on the machine, etc.

Reply to
Pete C.

I think the issue is encoder "lines" vs. "counts". The encoders on the machines I worked on were 2,000 line encoders, which as Karl noted could provide 8,000 counts per revolution. The US digital site uses counts, so presumably the 720 CPR encoder is 180 line?

Reply to
Pete C.

The terminology on this is terrible. Multiply the USdigital number by four to get counts that your control will see.

karl

Reply to
Karl Townsend

OK, great. Just what I need. I printed out the datasheet and will measure mounting dimensions to order them.

OK

I will check what Jon's PPMC does, but I think that he takes 5v differential signals.

Makes perfect sense to me. Thanks Karl.

i
Reply to
Ignoramus21167

Lines = Number of physical lines on the encoder disk

Counts = Lines * 4 (leading and trailing edges of the two quadrature channel signals)

Reply to
Pete C.

FANUC's latest "off the shelf" nanometer resolution rotary encoder technonogy runs at 16 million CPR. The cheap stuff, what you'd see in a 0i is a million and it really is inexpensive.

Reply to
John R. Carroll

We've got some linear enocoders that are 100nm resolution, over fairly long scales. That's four millionths of an inch.

Reply to
Spehro Pefhany

Did you notice in the reading to put a resistor and capacitor on the lines? be sure to do this. Unless you like weird noise issues.

Karl

Reply to
Karl Townsend

Yes, I did see that, and I will do so. Should be easy with Jon's breakout.

I hope that my servo motors' mounting holes match some available options options for that encoder, that way it will be all neat and easy.

i
Reply to
Ignoramus21167

Na, neat and easy, and refitting old machines are mutually exclusive. Any algebra expert should know this.

Karl

Reply to
Karl Townsend

I will take a caliper to it and I will know, hopefully tomorrow or so.

i
Reply to
Ignoramus21167

The normal rule of thumb in servo design is that the encoder error should be no more than 1/10th the desired total system error.

Your 128 count rule would mean that your 1/8" pitch ball screw would give an error of 128 * 0.125" / 2 / 2000 = .004". Is that really what you meant? Four mils error?

'course, by my "ten times" rule, Iggy's .087 mils suddenly become .87 of designed-for error, which is a lot more than _I'd_ be willing to tolerate -- 2000 counts would give a designed-for error of 0.3 mil, which isn't lovely, but is probably getting limited more by the machine than the encoders (which is what you aim for).

Those sinusoidal encoders will give you essentially infinite resolution, so you'll never have to wonder if you're losing accuracy to quantization noise. You can still lose accuracy in lots and lots of other ways, but not to quantization...

Reply to
Tim Wescott

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.