How to perform a sine sweep using torque control servo amp

[windell74712345]

Hi,

I'm trying to figure out how I can characterize the natural modes of my mechanical system. I have a servo amp in which a voltage is fed and in turn a PWM is generated. So the voltage fed is proportional to the torque I want to apply. The feedback is an encoder. However, when I do a sine sweep test I'm really looking at the phase and magnitude difference between the commanded position and actual position.

1) So how do I perform a sine sweep on my system if the commands i'm giving to the servo amp (using only as an amp, not using feedback) is a torque (voltage) and i'm reading the encoder back?
2) I was thinking that I could differentiate the torque signal (just get another sine wave) and infer the commanded position, but I don't know if this is the proper way to do this.
3) are there any texts that go over examples on how to do a sine sweep using different kinds of amplifiers to identify modes of the plant?

thanks,
windell

 

[Skogsgurra]

Look at DNA, Dynamic Network Analysis. That is usually about much higher frequenies, but the principle is the same for millihertz, kilohertz and gigahertz.

I recently did a resonance search on an 11 kV 1600 kVA transformer + harmonics filter using a 900 W audio amplifier (that was what we could get hold of in that part of the World) an arbitrary function generator that could output a swept sine and a recorder with FFT analysis. Found a surprising 195 Hz resonance where there shouldn't be any resonance at all.

We didn't need to look at phase angles, but that would certainly also have be possible.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[Skogsgurra]

The acronym DNA has been taken over by the analysis of social networks, which seems to be the dominating interpretation these days.

Google Bode Plot instead.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[windell74712345]

I understand how to construct a bode plot, however what I am sending to the servo amp is a voltage which is proportional to torque. What I am reading out from the encoder is position. So somehow I need to either convert my torque commands to a corresponding position command or convert my position values to corresponding angular accelerations. By doing this conversion, I will then be able to do the ratio properly to form the bode plot.

1) I'm trying to figure out what the best way is to do this and how this is done in industry.
2) Will the bode plot for Accl_out/Accel_in be the same as Pos_out/Pos_in?

 

[FeX32]

Bode plots for electric motors are generally done simply inputting a sinus with constant amplitude while varying the frequency from a low to a high of interest.
You must define the characters you are interested in though. This is the part I think you don't quit understand.
If you are interested in the position then you are really looking at the amplitude and phase of the fluctuating component of the position.
For example:

http://ctms.engin.umich.edu/CTMS/index.php?example=MotorSpeed&section=ControlFrequency

 

[blacksea]

I hope that you have possibility to measure gain/phase of position from encoder vs. supplied sine as well. The slope of Bode for such case must be (-40dB/decade). And any resonnace will be reflected as pair of anti-resonance and resonance frequencies.