2 proxs on gear to increase resolution

[drbitboy]

catching both rising and falling edges to increase resolution works best if the prox duty cycle is 50%.

 

[I_Automation]

...or is it an infinitely variable pitch drive
they had them like 30 years ago everybody changed them to a VFD a long time ago ....

Except around here, it's still 1989.



There is a powdercoating facility that has overhead chain conveyors that are all controlled with variable speed pulleys. The nice ones have the chain pull adjustment replaced with a reversable gearmotor to adjust the speed by a reversing drum switch.


They set their speed precisely as needed by timing one chain links travel and comparing it to a chart taped on the wall, then readjust until the time is what the highly critical process requires. [yeah, that's sarcasm]

 

[JesperMP]

catching both rising and falling edges to increase resolution works best if the prox duty cycle is 50%.

As long as you can reliably sense the rising and falling edges, the duty cycle doesnt have to be symmetrical.
Count the rising edges over time to calculate the rate.
Count the falling edges over time to calculate the rate.
Sum the two values and divide by two and you have doubled the resolution.

 

[drbitboy]

As long as you can reliably sense the rising and falling edges, the duty cycle doesnt have to be symmetrical.
Count the rising edges over time to calculate the rate.
Count the falling edges over time to calculate the rate.
Sum the two values and divide by two and you have doubled the resolution.

I don't think so, because arithmetically that [sum and divide by two] - i.e. averaging - over a fixed time is exactly the same as counting both types of edges into a single counter over a fixed time.

It's a small effect, but it's there.

 

[Peter Nachtwey]

I wonder if the proximity sensors are fast enough for 500 rpm
that would mean 40*500=20000 teeth per minute or 333..333 per second or 3ms. I doubt the sensor is that fast.
At 200 rpm there is 7.5 ms between teeth. Detecting rising and falling edges will be challenge.

JesperMP's solution is good only of the sensors are fast enough. Rise and fall times are not symmetrical. Still, this only doubles the resolution and that is not enough. Still JesperMP's solution is a cheap and easy improvement even if it isn't perfect.

drbitboy was right a long time ago when he said to use a low pass filter.
Use two low pass filters! Better yet use a two pole Butterworth filter. The Butterworth filters have a sharper corner where the response decrease.
https://deltamotion.com/peter/Mathcad/Mathcad - Butterworth NG.pdf

I would use an alpha-beta-gamma filter. We have two pole and 4 pole Butterworth filters on all our RMCs and alpha-beta-gamma filters on our RMC200. An alpha-beta-gamma filter is a very simplified Kalman filter that provides 90% of the benefits with only 10% of the hassle.

 

[L33er]

Typical prox response time is .3ms. Need to see the actual wiring.
DrbItbOy is right, really?
How does one implement alpha-beta-gamma filter with Mlgx 1400?

 

[Peter Nachtwey]

Typical prox response time is .3ms. Need to see the actual wiring.

Then JesperMP's idea will work. Adding more proximity detectors that count the rising and falling edges will help.

DrbItbOy is right, really?

He is about the filtering. The question is which filter to use.

How does one implement alpha-beta-gamma filter with Mlgx 1400?

Does a ML1400 have floating point? Is it implemented in software or does the microcontroller have floating point instructions?


Translate this.
https://deltamotion.com/peter/wxMaxima/ABG.html
I had a project were I was monitoring wave action which I modeled by a sine wave with noise. The problem was that the IMU ( inertial motion unit ) only provided data every 10 ms and I needed data every millisecond for the motion controller. I simulated a wave with a 10 second period with a 2 meter peak to trough with noise and sample jitter. You can see I can estimate the actual position, velocity and acceleration pretty well.


The alpha-beta-gamma filter is built into our RMC200 controller because often we must synchronize slave axes to a feed chain that is rolling sprockets. This causes lots of jitter. The alpha-beta-gamma filter can estimate position, velocity and acceleration. The estimated velocity and acceleration are necessary for computing feed forwards for the slave actuators.


Here is a link to the simpler alpha-beta-filter
https://en.wikipedia.org/wiki/Alpha_beta_filter


The advantage of the alpha-beta-gamma filter is that is doesn't have the phase lag of the simple low pass filters you see mentioned on this forum.


If the OPs RPM doesn't change much or quickly a low pass filter will do but if the RPM changes quickly then an alpha-beta-gamma filter has a huge advantage.



Do a search!


Come on guys. Put a strain on your brain. No one taught me this stuff. I had to spend a lot of time learning on my own.

https://www.youtube.com/watch?v=vG-wytu1YZw

 

[OkiePC]

This:

Am I missing something here - you have a HSC - get a cheap $100 encoder. Voila!

 

[NetNathan]

In searching for a high resolution motor tachometers.... I did notice... for slower rpm readings...they were using photo eyes and not prox switches.

 

[I_Automation]

In searching for a high resolution motor tachometers.... I did notice... for slower rpm readings...they were using photo eyes and not prox switches.

I think every encoder I have seen or checked out online was an optical encoder, even those rated for high speeds

 

[goghie]

I think every encoder I have seen or checked out online was an optical encoder, even those rated for high speeds

You are right. For example I use servo motors, with nominal speed 3000RPM, and resolution 2^22 pulses per turn.
If you open that encoder (and I highly recommend not to open), it has optical prism. If I have seen good, and calculated good, it has frequency of 33KHz at nominal speed (for first bit).
Just think about optical cables and it's speed.

 

[David Clement]

Thanks everyone for their input.
I have asked maintenance to fab an encoder mount and guard assembly for me. It's the right thing to do.
I have it averaging right now. It works decent.
They work at a snails pace here. I could wait 3 months for this to be completed. No joke.
I am an independent contractor that has been on site for over 10 years. I try not to rock the boat. Some of you will know exactly what I'm talking about.
I have plenty of other things here to work on.
Hope I haven't wasted too much of everyone's valuable time.
Thanks a million.
David