aligning resolver

[carlosthejackal]

is there anyone out there who can help. i am trying to align a resolver on a servo motor but i dont know the lockup angle, is there an easy way of finding this and does it apply to all resolver types. many thanks

[bmann2000]

With a 3-phase motor, you can move the rotor to a fixed position by energizing 2 phases, although it depends depends on the unknowns your dealing with. You can eye-up resolver alignment using a scope, but the best option for a good alignment is to write LabVIEW code to sample the back-EMF phase voltages and resolver signals when turning the rotor by hand, demodulate the resolver signals and plot the resolver output against the back EMF.

[carlosthejackal]

many thanks i will give it a go

[Yanikoo]

Hello!

From some weeks I try to write a code treating the resolver signals and measuring the phase between it and the motor phase (they have the same number of poles, so both signals have the same frequency). I don't have any demodulation kit on Labview so I 've done an home-made demodulation (sinwtcosA -> cosA and sinwtsinA -> sinA) that gives sines with some imperfections.

In order to measure the phase I used a sub-VI copied from phase_measurement seen on the forum and that works nicely on clean sines.

But there it doesn't work. Is my demodulation doing so bad job? Is there another mean to demodulate?

Thanks in advance if anyone got info.

[bmann2000]

I inherited some of my code from my boss, then refined it slightly.  I used a 6052 analogue IO card.  I connected phase A and B to AI0 & AI1 and phase C to ground.  Take care not to exceed the 10V max input of your card, either turn the motor slowly by hand, or drop the voltage through some resistors.

You get a smoother plot if you capture at constant speed, so I calculate speed then trigger the capture when I reach x-rads/s.  I also had an NAI resolver decoder card to compare my results which helped.  

I used a 4-pole motor so implemented  V-commutated = (Sin(60+2Theta)*AB+Sin(-60+2Theta)*BC)+(Sin(2Theta)*AC))*(2/3) to calculate the DC volts which you can divide by average rate to get Ke.

The resolver (aka FD field director) error needs to be phase corrected, this part I found most difficult, but this will be different for you as you don’t have the 2:1 ratio between back emf and resolver.  I demodulated using library VIs.  I have attached .jpgs of the demodulation and phase correction.  In all it took me about 5 full weeks to perfect.  I can’t give you the code, but I can help answer your questions. 

[Yanikoo]

OK I've realized the demodulation helped with the AM theory, I think I've done quite the same as you. The output signal is exactly at the good frequency but with a phase shift with the original enveloppe. I read it is normal in AM demodulation (and in radio listening, it is not a pb), and that I need to make a Phase Locked Loop.

I saw some vi's there in the forum (PLL.vi) and I don't know how to integrate it in my work. Can u help me please?

[bmann2000]

I can not see how you demodulate from the attached .jpg.  I doubt you want a PLL as you will be post-processing 1-second of captured data.  My demodulation code will introduce a very slight delay, but you can run the test twice, once CW and once CCW to discern lag from resolver misalignment error.  Once you have your resolver position unwrapped and plotted, take your stator angle (calculated from the back EMF Valpha = (AB*Cos60) + ((BC*-Cos60) + AC); Vbeta = (AB*Sin60)+(BC*-Sin60); Theta (Stator) = atan2 (Vbeta, Valpha)) and compare with resover angle to see your misalignment. 

The resolver alignment error will have a slight wobble due to manufacturing imperfections, convert to rms error/max error.  Then repeat the test in both directions, adjusting the resolver alignment until you get the same error in both directions.