Ok, I guess I don't understand it well enough
It's not really a matter of "powering" the black box, it's more a matter of putting in a waveform, altering the original waveform a bit, then measuring that waveform... If you have a nice clean waveform to start with, then measuring the output is easier. I've tried square waves, and they produce enough noise at the corners to mess up the output signal. It is possible to create a sine wave by filtering a 50% square wave, thus it's not necessary to use microprocessor resources to create the input signal.
Here's another go at a resolver description, with some (hopefully) helpful links.
Think of a resolver as a rotating transformer of a sort. There are some decent pictures that will help my discriptions on page 27 of this link. (Unfortunately, I can't figure out how to steal a picture from a *.pdf)
http://facultyfiles.deanza.edu/gems/...torOutline.pdf
Look at slide 62 of the link to help with this description.
The stator is made of 3 sets of coils, arranged so that they are rotated 90 degrees from each other. Say the input coil is at 0 degrees, at the top. Rotated 90 degrees from that would be the "detection coil S" or sine coil. Rotated another 90 degrees from that would be "detection coil C" or the cosine coil. With this arrangement, there is no need for brushes to get the signal onto the rotating part.
The transformer part is really the flat laminations that are lobed or semi-circular. Their job is to reflect the magnetic signal generated by the input coil. Imagine for a moment that they are simply round. Ok, now the input coil gets a fairly high frequency (say 10kHz) signal. It's important that the input signal is fairly high frequency so that the output signal can be high resolution, as I will attempt to show later. Now, if the transformer laminations that are on the motor shaft were simply round, both of the output coils would get the same output signal. Specifically, we're talking about the AMPLITUDE of the output signals, because the frequency would also be replicated. If the transformer laminations are lobed, then the amplitude of the signal recieved by the detection coils would vary in proportion to the distance between the laminations and the coils.
Thus, there will be a sinusoidally varying amplitude of the input signal seen by the detection coils as the motor shaft rotates. Since they are 90 degrees apart, the different amplitudes will be 90 degrees apart, like a sine and cosine wave.
Now here part of the trickiness, and this is shown by slide 63 on page 27 of that link.
Notice how the input signal is a fairly high frequency sinusoidal signal. Below that pic is another pick of a signal composed of a high frequency sinusoidal signal varying in amplitude sinusoidally. Now, all we REALLY care about is the peak amplitude of each of the high frequency signal's peaks. Imagine we put a dot at each of those points. Later, we come along and draw a curve based on all those dots. (I think this is called an "envelope filter." Anyway, we need to generate two envelopes, one for the sine and one for the cosine. Then we compare the amplitudes of the envelopes to calculate the position. Also, you can see that if we used a lower frequency input signal, the dots would be spread apart more and the envelope created by the dots would loose resolution. This may not be a problem at low speeds, but when the motor is spinning at 10,000 rpm, this resolution would be important.
***whew***
I've gotten pretty close to producing the waveforms needed to use the resolver. I'm just stuck in one of my "gee - it could be REALLY simple if we didn't bother with a specific angle thoughts." In other words, when you look at the calculations for field oriented control, you don't really need all 360 degrees, only part of it.
Or we could say screw it - we've got the computing power, just calculate the *** angle!
Or we could just use one of those resolver>digital translators to keep the project moving.
- E*clipse
Quote:
Originally Posted by MPaulHolmes
OK I think I need one more go at how the resolver works. Let's take the perspective that it is a black box. I have a controller with +5v, ground, and 3 A/D inputs available. The resolver needs a periodic waveform from the controller for it to operate (instead of a +5v power supply)? If I supplied a 10KHz 50% duty square wave, would that suffice for "powering" the black box? I'm thinking it would be a good idea to first get the controller working with a PMSM motor. While the rotor flux angle is the only differentce between the 2 controllers, it's a significant step with a number of uncertainties. If I knew the code was good in the PMSM case, I could then focus exclusively on the ACIM rotor flux angle without all the uncertainty of other stuff. There really wouldn't be much to change. For testing, I could deal with 2 sine waves that were 90deg out of phase for computing the angle using the A/D. I would just need those 2 sine waves coming in.
|