Find LC resonance hidden by inductance

Step response or pulse response? (with an FFT maybe?) Can you do a difference measurement, with and without the unknown inductance?

George H.

I do not have access to the inductance

Cheers

Klaus

I see the resonance dip-and-spike. How much resolution do you really need?

Subtract the modulation on the red curve from the blue curve?

Seriously, I'm with George. Pulse it as hard and fast as you can and FFT.

Mark L. Fergerson

I've got an SRS770 spectrum analyzer, with a chirp* output. That I assume would be able to nail the peak.

George H.

One should be able to calculate it in the time domain as well...hang a capacitor off the black box and make a voltage divider and connect a high impedance voltmeter to the junction, then hit it with an impulse. After some time there should be a voltage remaining on the second capacitor.

If you know the transfer function (impulse response) of the whole system, then you can use the "final value theorem" limit to see how the final voltage on the second capacitor is related to the components in the circuit. I'm working on the math but it's a little heavy :(

Yes, C can be measured various ways. It's theoretically impossible to measure one L if the two Ls are in series. It's barely possible if the Ls are not ideal.

A magnetic field sensor outside the box might help.

I would like a more significant spike, just so any spread (unforeseen or foreseen) can be overcome. So something like at least 50%

How does that increase the sensitivity?

The parallel LC is in series with the inductance. So if I understand you, the pulse will create a current in the LC, and letting the pulse freewheel will create 2 frequencies. One from the LC and another from the inductance/parasitic capacitance ringing?

Cheers

Klaus

How do that work exactly, I don't quite follow the technique?

Cheers

Klaus

Hi Klaus, maybe it would be better to call it a network analyzer. It has both a source for stimulation as well as a spectrum analyzer. So forget the chirp function for now... it just makes getting a spectrum faster. Imagine stimulating your system with white noise. Then you take the FFT of the output and you've measured the system response... It's like your technique, except rather than step through the frequencies one at a time, you do them all at once.

The box can also average a bunch of spectra so you can get a nice clean signal. (The box is not cheap! $7.5k or so.)

Yes, but how does that give a clean value of the LC resonance? With the big L in series, it will just measure that inductance, right?

Cheers

Klaus

It's like this one:

Graph:

Cheers

Klaus

I was guessing that inside the black box there's a series L followed by a parallel LC, both in series, if that makes any sense. Could be wrong

You are correct, like this:

Only L1, L2 and C1 (the other stuff just for Spice testing)

Ahh it's not any different than your technique, you'll see some little resonance feature at ~350 Hz.

George H.

In that case, the net impedance goes infinite at resonance, which should be hard to miss. The problem in real life is that 100 uH and

Easiest, is to drive it with a FM-swept sinewave through a resistor.

Phase-lock the sinewave to a '4046 or similar Phase-lock the black box to another '4046 Look at the difference signal of the voltages driving those PLLs (you're just looking for a transient glitch-like signal). An X-Y o-scope with frequency sweep control voltage versus the difference signal will give a clear indication.

Or do a real phase-detect on signals picked off the two ends of that series resistor. There'll be phase shifts, + and -, faster than just the L contribution, at resonance.

Hi Klaus,

Maybe do some averaging to improve the SNR? The data may look more appealing on a log scale.

Here's a utility I use sometimes on the Picoscope (that you helped convince me to purchase :) A strong point of the scope is the easy export of data for averaging multiple tests.

Chester

The MLSSA audio system analyser uses maximum length sequences to measure all sort of things

That is basically measuring with white noise except it is known noise so you can do cross correlation instead of FFT, (though it might be implemented with an fft)

-Lasse