Wanted to know the lead inductance of a motor. No RLC bridge, so I take my reliably-50-ohm output impedance generator, vary the frequency, measure the amplitude.
One ought to be able to fit a model circuit to that pretty easily. Except that the voltage should be proportional to frequency at the bottom end. Instead, it's proportional to f^0.75.
Grr. Off to do more testing.
Should be, "Off to do more thinking" ...Jim Thompson
There's ohms in the winding as well as henries. Things will also get weird if there's enough amplitude to make a bit of saturation.
In a just world (and on other motors) it acts like an LC highpass filter, with a slope equal to f at low frequencies and level at high frequencies. It's the damned non-integer exponent that's getting me.
I suspect it's something about the way it's wound. I just powered the damned thing up with PWM and verified that it wasn't going to make things go up in smoke -- beyond that I refuse to worry.
Well I sometimes do my best thinking during testing.
Tim, I have no idea about motors... change something and see what that does... load it down?
George H.
What range was your frequency sweep? Could it be so high the eddy current loses increase?
Mikek
Yeah. Not much point sweeping outside operating range.
Mark L. Fergerson
Around 10kHz, because that's the frequency of the PWM.
And yes, it could be that dammed Eddy Current guy.
Um. What frequencies are you planning to feed it?
Um. Gonna feed it PWM? How sharp?
Mark L. Fergerson
From my relatively limited experience with them, laminated iron cores seem to give constant-loss results. Give or take. That is, at the eddy current cutoff frequency, real permeability goes from ~constant, to dropping
2nd-order like. Meanwhile, complex permeability goes up, peaks at cutoff, and comes back down, with linear slopes on either side.That is to say: losses (parallel equivalent) increase linearly with frequency, until they become dominant, at which point impedance remains constant (permeability becomes imaginary, and drops linearly with frequency).
Physics tells us, it should be more of a f^0.5 (diffusion) slope. Depends on materials and construction. It could very well be that the exaggerated hysteresis loss of silicon steel (as compared to lower loss materials like Vitroperm or ferrite) contributes another f^0.5, more or less; or that the excitation B_pk is exploring an ever-smaller portion of the B-H curve, and thus permeability measures anomalously low (steel is particularly awful about initial permeability, that is, mu at very small B_pk).
At even higher frequencies, the capacitance between plates matters, and the permeability drops even faster (in the process, wrapping around the circle -- it goes from imaginary, to negative real: which is to say, the core looks capacitive).
Tim
In a similar (in absolutely no way) situation, we measured at a range of frequencies and extrapolated to 0Hz.
--sp
Everyone who needs a DC measurement does that. Measuring down at 0.01 Hz takes at least two measurements, a minute or so apart... so there's damn few 0Hz measurements that have ever been made (it takes infinite time to satisfy Nyquist).
That argument applies to a black box. When you know your measurement was over orders of magnitude longer than the longest time constant in the circuit you can reasonably say it /is/ a measurement of conditions at dc. No exrapolation required.
NT
Have something to add? Share your thoughts — no account required.
Ask the community — no account required