[Zig]

Mike,

The " box of trix " is german made to the best of my understanding. I shall try to locate the URL.

As for the linear motor having non linear thrust .. that is standard fare for any iron cored motor unles special consideration is paid to the design to minimise " coggy " nature.

BLDC and DC motors usuallyy have skewed slots within the pole pitch to try to mitigate the this problem.

The motors I have here are old ( 15+ years ) iron cored thrusters Normag units. Back way back before compnies like Balldor took over smaller players.
Linear motors with ironless thrusters are great but this is all I have.

In fact I had the motrs runing with equally old motion science servo amps where i had to provide external frequency to voltage converter to provide velocity feedback.

These worked OK but were a problem to tune compared to the AMC servos. Hence this new round of activity.

Just to add.. simulation ( LT Spice) suggests x64 pulse rate with reasonable pulse distribution followed by /4 to get quadrature signal is quite possible.

Presently I am looking into
a) disabling output in case of bad startup phasing on the output versus input
and
b) logic to steer output so it follows input phasing minus all the spurious states

[Zig]

Ok..

Here it is..

motrona interface: Frequency and Impulse Dividers, Frequency and Impulse Multipliers

[mike_Kilroy]

Thanks for correcting my comment Al, I did get carried away trying to describe my idea. Yes, quadrature means 90 degrees phase relationship so I should not have written it the way I did. I was just trying to show there are only up to 4 meaningful transitions per "count" so only x4 is available for a meaningful set of pulses.

Thanks for the Motrona link and info Zig. I never saw anyone sell such a unit. It could be very useful for use with a DRO that cannot scale the input signal! If one needs to change from 1024PPR to 1000PPR for instance since the DRO will not care about the batched unlinear pulses coming in, just counts em.

It reminds me of way back when Hiedenhain first came out with multipliers and had what ended up being termed "batched" output: they multiplied the sine wave in the head but held the multiplication for a count or two then sent the higher number out as a higher freq batch - was absolutely useless for servo control. Many years later we helped Newall solve their similar issue with sending pulses in batches on one of their first designs - the issue again was the servo they were used with went crazy and non one had a clue why until I put a scope on the encoder output and showed them. there was a very quick redesign done.

Zig, ALL motors with magnetic material in the magnetic field have cogging. It is called cogging torque and better motor companies spec this or will give it to you if asked. Depending on the motor design it can be minimized, one of many ways is skewing the magnets as you say. As motor engineers we don't typically call cogging torque unlinear Kt, torque, or thrust, just a separate entity; hence my confusion over your name. Thanks for clarifying it for me.

Bottom line is still that any multiplication above x4 to a quad TTL encoder count will produce bad servo response data. there are two reasons this will only make your velocity loop response worse:

1) You WILL have batches of pulses if you run anything other than 1 constant speed where you adjust it for; you see this right? If not let me know and I will give examples to help make sense of it and why it screws velocity loop up even worse than lower res original signal.

2) You will be adding meaningless data that cannot help the velocity loop control better. The problem you have now with too low a count is there is dead space (no data) between the pulses since they are too far apart. So you try to turn the gain up on in the loop and it can only go so high before it outpaces the next pulse of feedback, so you end up with overshoot. So you find the max gain limit, since it is lower than could be due to the dead space between meaningful pulses, the available bandwidth is lower than might be otherwise. So multipy by 10; now you DO have more pulses but they are still meaningless since they tell nothing of actual feedback; only every 10th one has meaningful data, so even tho you have 10x more pulses, he loop gain still cannot go higher than what it was with the 10x less pulses since by the time the loop says to do something in its output, the input fake pulses carry no data so don't tell it that it is overshooting. Result will be the same gain (or lower) than you started with. And you found out that with linear motors you need the velocity loop gain like 10-100x HIGHER tan for the equivalent rotary servo thru mechanical gearing and or ballscrew to achieve the same machine stiffness.

[Zig]

Mike,
Thanks for Your comments.

I had noticed a kind of bunching of pulses which wass caused by frequency multiplication based on edge detection in the initial approach I had considered.

I then switched to a phase based if I can put it that way approach which then produced much beter distribution of outptu pulses per encoder line cycle.

My expectation is that the approach will yield acceptable results and the simulation suggests that the phase difference between the output inphase and quadrature pulse trains is not 90 degrees but is sufficiently large to provide unambiguous lead lag information so that the frequency to voltage converter in the AMC circuit will make sense of the contents of the the two pulse trains.

I shall be laying out the PCB artwork today ( sunday 19th ) hope to have more info over the next couple of days.