[nebyeh]

I am having trouble setting up my horizontal rotary table, its 24inch diameter and weighs about 20 lbs that I connected my 850 peek oz-in servo directly to. The problem is when I power on, the table keeps oscillating back and forth about 30 degrees. If I try to rotate the table using mach, it will fault the gecko 320. I have all the wiring correct, the other two motors for the other axis work great. Could the encoder be bad? Its an E5S from us digital, I'm pretty stumped on this one. Thanks

[Al_The_Man]

Is the encoder phasing correct?
If so try running the motor without connection to the table.
Al.

[nebyeh]

yeah the encoder phasing is correct, I tried switching it and then it constantly faulted, if its not connected to the table, the motor will run and doesn't seem to oscillate, its just when its connected to the table that it keeps doing it, and it doesn't seem to matter what I set the gain or damping to on the gecko. I know the motor has enough torque to move the table otherwise it wouldn't be able to oscillate, something doesn't seem to add up

[Al_The_Man]

Have you tried lowering the accel/decel rate really low, if this cures it, then possibly the inertia ratio of motor to load is too high.
Al.

[nebyeh]

I tried setting it really low, and it didn't seem to have an effect on things, tomorrow I might try setting it really low and telling it to move by a "G1" command instead of trying to jog it, but why would the table keep oscillating?

[Mariss Freimanis]

It's inertial load mismatching. Your rotary table has a moment of inertia thousands of times higher than the motor's. Servo loops become unstable when the load inertia is only 10 times greater. The only cure is reduction gearing from motor to table. Start at 10:1.

Mariss

[nebyeh]

that makes sense, thanks! so will that also kill the oscillations?

[Al_The_Man]

that makes sense, thanks! so will that also kill the oscillations?

Usually that is the symptom, in the case of inertia ratio, it is reduced by the square of the reduction, so 10:1 will reduce it by 100!.
Al.

[nebyeh]

haha good thing you said that, I was just figuring the gear ratio out based on the motor spec was 3.75 kg/cm^2 and my table was something like 5000kg/cm^2 I would of needed something around a 150 to 300 step down, so based off those numbers what would be the proper stepdown?

[Al_The_Man]

You could use the appropriate mathematical formula, or use a free Graphic tuning program like Kolmorgen (Danaher site).
Or based on your maximum motor rpm, figure out what the maximum ratio you could go to to and still give you your required maximum table rotation speed.
This may bring you down to less than the recommended max of 10:1 and nearer the prefered 1:1 ~ 5:1.
Al.

[Mariss Freimanis]

Good to see you have calculated your table inertia; that really helps.

Your miss-match ratio is 5000 / 3.75 or 1333:1. The good news: The reflected moment of inertia decreases as the square of the reduction ratio. This means SQRT 13333 or about 36:1 reduction gives a 1:1 motor/load match.

Better news: The servo will be stable at a 10:1 miss-match so;

SQRT (5,000 / 37.5) gives a reduction ratio of about 12:1 and you will get a stable loop response (no sustained oscillation).

Mariss

[nebyeh]

I guess some of the stuff my teachers talk about I use :-) thank you both for your help! I guess all I need to do now is get a 12:1 ratio, do you know of a source for gearboxes to fit a 1/2 shaft that won't cost hundreds of dollars?

[Al_The_Man]

I usually check out ebay for Bayside Planetary (low backlash), there are also some very compact German made ones, I forget the name at the moment.
Al.

[Xerxes]

What is the reason for load ratio limitation? I managed to tune that cast iron wheel stable even when mismatch ratio is very high. I haven't calculated it but must be way more than 10:1. Maybe closer to 100:1.

Update:
Motor inertia is 2.66e-4 kgm² and wheel weights about 4-5 kg and diameter is about 25 cm so inertia could be between 0.03-0.045 kgm². The ratio is over 100:1.

[Al_The_Man]

You probabally tailored your acceleration/deceleration to accomodate it.
Al.

[Mariss Freimanis]

How's your settling time?

Mariss

[Xerxes]

There were no acceleration limiting, instead I did sharp step responses. There were overshoot (I didn't put much effor in tuning) but system was stable. Setting times were long because torque limit made acceleration of that mass quite limited.

Did I understand correctly that the problem is flexing of the shaft and not actually the inertia? That would explain why high inertia motors are not more unstable than lighter ones.

[Xerxes]

If I understood correctly the Ormec article says following:
-The problem of instability is caused by low natural frequency of mechanical system
-High natural frequency is better
-Natural frequency is determined by shaft stiffness and load/motor inertias (Figure B)

BUT:

If I have system with Jm=1 and Jl=10 I get 10:1 inertia mismatch which would not supposedly be very good. If I try to improve things by choosing motor with Jm=10, I'll get ratio 1:1 but frequecy drops by factor of 2.35!

So I don't see why it is better to select higher inertia motor. By looking at equation it seems that using low inertia motor with stiff shaft would be the best solution no matter of ratio. The equation would give infinite frequency if motor would be completedly weightless. Of course practical low inertia motor must have limited stiffness in shaft but I wonder if that would really be the bottleneck since shaft inside motor is very short.

[Mariss Freimanis]

Overshoot means ringing.

Having such a high motor to load inertial mismatch severely limits servo performance and is impractical. So much so that we have made no effort to test behavior past a 10:1 ratio.

Mariss

[Larken]

Running a servo drive with variable sampling time should be able to tune to your load. My Viper drive or others that have a parameter for sampling time will match to the inerta of the system.
The slower the motors mechanical responce the slower the sampling time needed, to eliminate over correction that causes oscillation.