Cheap way to measure leakage inductance accurately?

[treez]

Hello,

We need to be able to measure the leakage inductance of our SMPS flyback transformers..however, we do not have an LCR meter accurate enough to do this. (all we have is an LCR40 meter by peakelec.co.uk)

Therefore, we wish to measure leakage inductance by the “resonance method” (as in the below schematic and LTspice simulation)

Method of measurement:
We will simply short out the secondary in the usual manner when measuring leakage inductance, and then add a capacitor in parallel with the primary, then we have an LC tank circuit, and we will switch some current into it, and then switch the switch off, and it will then ring at the LC resonant frequency, …from this frequency, we will calculate the leakage inductance……do you think this is accurate enough?….we will observe the ringing on our TDS1002 scope.

We will trust the LCR40 to be able to measure the 100nF capacitance very accurately, since cheap LCR meters always measure capacitance far more accurately than inductance, do you agree?

Is our measurement method accurate?

TDS1002 scope
https://courses.washington.edu/phys334/datasheets/Tek2kUserManual.pdf

LCR40 LCR meter:
https://www.farnell.com/datasheets/1692006.pdf

(also-does leakage inductance measure higher or lower when the measurement frequency is higher?)

 

[BradtheRad]

Shorting the secondary is putting a load on it. This will inhibit ringing in the primary loop.

My simulations show that you need to leave the secondary disconnected (or use a very high resistance). Then you should get ringing in the primary loop, as you plan.

Caution is needed as to the value of capacitor you connect. If you use a very small capacitor value, you may get very large voltage swings.

 

[treez]

the secondary must be shorted for leakage inductance testing.
In the schem of the top post, we are not worried about overvoltages.

 

[chuckey]

Another way might be to put your special cap in series with the primary and feed the combination with an oscillator via a high resistance (say 10k). monitor the voltage across the cap/primary with a CRO. When the signal generator is on the correct frequency, the leakage inductance/capacitor will go into series resonance and short out the voltage. Measure the frequency and calculate the L.
Frank

 

[Anna Conda]

You are very close as to the most accurate method to measure leakage inductance.
Yes short the sec, yes put a cap in parallel with the primary, say 100nF - accurate - low loss!
Then put a 4k7 in series and drive from a signal generator (the primary), observe the o/p from the sig gen on one channel of your scope, using a 100x probe (or 10x if you must) observe the wave from across the pri (it will be low) make sure all grounds are connected together on the other side to the 4k7 driving resistor.
When the two signals are in phase you are at resonance, and can calc the Llk from the freq and 100nF, by varying the freq up & down so that the observed waveform goes down to 70.7% for over & under freq, you can measure the Q at the centre freq (Q = fo /(fupper-flower)).
This is far and away the most accurate way of measuring Llk. You get the leakage referred to the primary.

 

[treez]

thanks but I was just going to do it in the simple way as in the schematic pdf in the top post...ie, just give it an impulse via a switch and a limiting resistor, and let it ring...measure the ringing frequency and get Llk from that.?

 

[andre_luis]

I was just going to do it in the simple way
just give it an impulse via a switch and a limiting resistor, and let it ring

For a more accurate result I would prefer this method rather than measuring the maximum output through a series LC network, due in theory the oscillation would occur locally, not being subjected to any interaction with parasitic elements becoming from the source of the injected signal; but you should deal with a spectral analyzer which should capture the measurement into a short period. Honestly, I don´t even know whether such an approach could be considered feasible due to the need for a trigger synchronized with the switch, due to a quick shot for the acquisition window and also due to the fast decaying rate of output.

 

[Anna Conda]

If you short the sec to measure the Llk you won't get much of a ring...!

 

[treez]

the llk doesn't "see" the short

 

[Anna Conda]

No, but you still won't get much of a ring....!

 

[FvM]

Any measurement setup that achieves sufficient resonant Q shows a ringing step response, either parallel C with high impedance drive or series C with low impedance. The condition is almost identical to getting an acceptable L loss angle in the LCR meter inductance measurement, because the capacitor can be assumed effectively lossless.

 

[treez]

we believe there will not be much ring, but we'll just keep switching the switch and capture the ring on the scope....we don't need much ring for that.

 

[chuckey]

The ringing test used to be a standard test for a suspect LOPT. You connected the primary to the "cal" output of your CRO, and looked at the EHT connector with the CRO. A good LOPT would give the standard exponentially decaying sinewave, where you could measure the frequency. A bad LOPT would give a pulse with a few ripples and you really could not measure its frequency accurately at all, due to the three pulses or so with amplitude variations of 50 :1.
Frank

 

[BradtheRad]

A simulation should be of some interest.

The switch was closed, then opened, at 50% duty cycle.

The capacitor value was edited on the fly, as shown.



A load of a few ohms on the secondary, has a damping effect on the primary ringing. By shorting the secondary, it will dampen ringing more.

 

[FvM]

Unless the transformer model includes finite coupling (another word for leakage inductance), that's not the measurement setup intended by treez. You can use a simple inductor in place of the primary winding to simulate it.

 

[schmitt trigger]

My first though would have been the series resonant idea, but the ringing method I also find intriguing.

I guess that my advice would be to build a couple of ideas and actually test them; and see which one provides the best results.

 

[BradtheRad]

For comparison, here is a similar simulation. The secondary load is high ohms.

Ringing continues for several cycles. It would continue a very long time if I did not insert some resistance in the primary winding.

 

[chuckey]

That's a nice wave form. Why not try to put 5 micro henries in series with each winding, move your voltage monitoring point to across the cap and change the 100K to 1 ohm. This will be in same ball park as Treez's problem. Its nice to see your first simulation, it represents what I have found in practice.
Frank