Finally figured out and did encoder shaft

Ignoramus6780 fired this volley in news:zIOdnYxF279JUZfQnZ2dnUVZ snipped-for-privacy@giganews.com:

Am I missing something, Ig?

How else would you determine the angular position of the spindle without the encoder? I thought that's what it _did_.(?)

(You do have a "zero index" on that encoder disk, right?) LLoyd

Well, the basic equation is that there is a full revolution of the encoder between the index pulses. If EMC2 sees a different number of pulses than 4096 (4096 pulses make a full revolution of the encoder), between two index pulses, thos would mean that pulses are missed. This is what I want to check somehow.

i

Why not use the same method your other rotary axis does? You mentioned once you had a fourth axis installed.

Mount a mirror on the shaft, shine a solidly-fixed laser pointer at it and mark the spot on a far wall. When you rotate the shaft so the spot comes back to the mark you have turned it exactly 1 revolution to an accuracy of about 1 milliradian or better.

I needed this on my lathe spindle. Turns out there was no way to do it in Galil/Camsoft (Let me know if EMC has a way)

Don Foreman came to the rescue. A small transistor amplified the index pulse to high speed opto 22 input. Allowed dead nuts accurate threading at 4000 RPM. He said duck soup easy, but total fricken mystery to me. Us MEs never quite trust those sparkEs.

Karl

Consider the possible and likely failure mechanisms, and balance the consequences against the added expense of monitoring.

Initially if the encoder sensor output looks clean on a scope you probably aren't dropping counts.

it ages and collects dirt or shifts out of alignment you could lose either signal. Can EMC2 compare the commanded and measured speeds and trip an error if the difference exceeds limits?

If you want a simple hardware monitor you could divide the encoder counts by 4096 with one of these:

clock a bidirectional counter down from the divider output and up from the index. the count output with LEDs and occasionally check to see if it has drifted off 0000 (all 4 off), or really that its normal pattern is changing.

Also you could trigger one-shots with the encoder pulses and light green LEDs with the outputs. This would detect a complete failure of either encoder.

These could be connected to the controller to give error inputs that it can poll occasionally instead of wasting its time looking for pulse edge transitions.

jsw

I was thinking something similar, but mechanical, but this optical approach is better.

But I would add one thing. The worry is mechanical slippage between hollow spindle and the little adapter plug iggy made, and slippage is greatest during rapid acceleration.

So, I would command the spindle into a long series starts, stops, reversals, and so on, that should leave the spindle in the same angular position as before the sequence. If no slippage, there will be little or no difference between where the spindle ended up (as indicated with the light beam) and the integral of the encoder output.

Joe Gwinn

My bigger worry is missing pulses due to shaft misalignment.

What I did yesterday was to buy a digital "laser" tachometer that has a TOTAL function.

When it arrives, I will run the spindle for 3 minutes and look at spindle position number in EMC, and will compare it with the TOTAL number from the tach. They should differ by less than 1.0.

i

Perhaps you could put the laser in the horizontal hole of a boring head.

jsw

Perhaps you could put the laser in the horizontal hole of a boring head.

jsw

--------------------------------------------------------------

That is probably more convenient, but a mirror makes it twice as accurate (1 deg of rotation = 2 deg. of beam deviation).

It's not a servo spindle, so there is no way to perform such an exercise and return to the start position. What he could do is scribe a line across his adapter and the spindle, run a series of starts/stops/reverses and then visually inspect to see if the scribed line is still in alignment.

If he's worried that his adapter is slipping in the spindle bore, which it really shouldn't since it and the attached encoder wheel have little mass, he should just install the adapter with some Locktite removable threadlocker.

As far as I know, EMC2 requires a real encoder with index on the spindle, as it needs to accurately handle the reversal when doing rigid tapping. Yes, this really doesn't apply to a lathe threading cycle, but the EMC way is usually to go for mathematical purity and completeness.

Since most lathes have an extension at the rear of the spindle, it is usually possible to rig an encoder there somehow, often using a timing belt. I figured out how to do it on a Bridgeport 1J, where it is practically impossible to add a sprocket to the spindle. I think the same scheme could be used to put 3 sensors on a timing belt pulley or gear as would almost always be found on the threading gear train of any lathe that has a threading screw.

See

Jon

Jon, the reason why I could do it, relatively easily if you do not count two months of thinking and false starts, is that my mill does not use a drawbar and I could insert an adaptor into the top part of the spindle. With a 1J head, you do not have that luxury.

i

I'd been wondering if an electronic counter would be a good solution, since there are numerous models with varying features made for all sorts of machine applications (counting/tachometer industrial applications). I thought someone more familiar with integrating them into computer applications might mention them as a good (but maybe not optimum) solution.

I have a couple of DPMs digital panel meters which are capable of reading several types of input devices, including optical TTL quadrature encoders. Other inputs are commonly used, from a fairly wide variety of signal sources.

Of course a DPM by itself often requires an enclosure (or approx. 2"x5" panel space) for mounting, but as far as a portable piece of test equipment, they can be very versatile for monitoring a wide variety of machine functions. The Totalize feature is common on most industrial DPM tach/counters, and other features are often fairly sophisticated specific programmable functions.

Quality counters are made to easily integrate into machine electrical systems, generally with output signal options that would likely be easily implemented as inputs for a PC CNC application (maybe with a simple isolation circuit).

You can't replace just the disc. The pickup has a mask in front of the sensors that matches the pitch of the lines on the disc. The sensor does not sense the passing of the individual lines on the disc -- it picks up light and dark transitions that result from the interference between the disc and mask.

Thanks, I did not know this. In any case, I am not really in a big need of tapping at high speed above 500 RPM, for many reasons but one reason is that reversing of the spindle becomes harder.

i

Ignoramus7337 fired this volley in news:bKadnfZsD9m04JDQnZ2dnUVZ snipped-for-privacy@giganews.com:

If your grating is misaligned with the two pickups, you'd likely get a skipped count on every revolution, at any speed.

Did you make the pickup head adjustable? Is the grating centered well on the shaft?

If the grating is centered and the head adjustable, you can do the "standard" limit-to-limit kind of adjustment on it, then split the difference, and you should be fine.

If not, spin the thing up to max (like you said you'd do), run it for a while, then stop it, index it, and tally up the counts.

My bet is, if it's working reliably at 500 rpm, it'll probably be fine at

4K.

LLoyd

"Lloyd E. Sponenburgh" fired this volley in news:Xns9E53610D4A612lloydspmindspringcom@216.168.3.70:

I should have added that back in my "robotics" days, I hand-built quite a number of quadrature encoders, with a few sporting 8K counts/rev, and never had any problems making them work. At the time, all I had was a fairly worn out Atlas 6x18 and a drill press.

I relied on a local photo-engraver to do reductions of my full-up

30"x30" drawings of the grating wheels. Inking the damned drawing on a 2K-spoke wheel was more work than any other part!

LLoyd

Lloyd, I will try to attach something to the spindle that can let me very accurately make exactly one revolution (by hand), and see if the number that EMC2 has, increases by exactly 1.000.

i