Audio Transformer Inductance

[marcus4audio]

Guys, really nice measurements Btw, how accurate will be this approach? i fixed the Freq, then trimmed the R to same reading voltage at primary and resistor. used formula to calculate Henries . I tried to measure some primaries and varying the frequency changes the value of inductance. For example, 20hz=559H, 50hz=370H, 100hz=274H, 200hz=185H an so on.

 

[CJ]

for audio transformers that operate at low signal levels, you should just stick one of the volt meters in series with the transformer primary and switch it to read ma and ua,
as your frequency goes up, keeping the voltage on the transformer equal will require a larger and larger resistor, like at  1000 hz you would have 6.28 x 1000 = 6280 times your pri Henries, which for a typical 600 ohm input might be 20 H, so 20 x 6280 = 125,000 ohms pri Z,

your frequency response of the volt and ammeter instruments should be good enough to go to 2000 Hz and still be linear, for kicks you can crank it up to 20,000 and remove the core to see what happens to your readings,

 

[marcus4audio]

for audio transformers that operate at low signal levels, you should just stick one of the volt meters in series with the transformer primary and switch it to read ma and ua,
as your frequency goes up, keeping the voltage on the transformer equal will require a larger and larger resistor, like at  1000 hz you would have 6.28 x 1000 = 6280 times your pri Henries, which for a typical 600 ohm input might be 20 H, so 20 x 6280 = 125,000 ohms pri Z,

your frequency response of the volt and ammeter instruments should be good enough to go to 2000 Hz and still be linear, for kicks you can crank it up to 20,000 and remove the core to see what happens to your readings,

How different is this compared to the approach i asked above? Sounds similar to me.

 

[CJ]

you would eliminate the pot,

 

[abbey road d enfer]

Guys, really nice measurements Btw, how accurate will be this approach?

This measurement makes the ASSumption that the inductor is a perfect inductor. In reality, there are several parasitic elements that are missing from this model.
The easiest is DCR, which appears in series with the inductor. You can relatively easily take this element out of the equation by doing two measurements are clearly distinct low frequencies, like 1kHz and 100Hz, or even simpler, measure it with an ohmmeter.
There is the stray capacitance that appears in parallels with the (inductance + DCR) dipole. This can be evaluated by finding the intrinsic resonance frequency of the inductor, which appears as a hump in the impedance graph; then you can determine its value by using C=1/L.sq(2.pi.F)
The last element that appears is the loss resistance, which appears in parallels with the inductor.
When the inductor is resonating, its impedance should be infinite. In fact it's not, because there are losses in the core that absorb energy. If you measure the impedance at resonance (very tedious, you need high Z instruments) it is the equivalent loss resistance.
In order to make things more complicated, ferro-magnetic materials are not linear vs. level or frequency.
For all practical purposes, I would recommend that you conduct several test in conditions as close as possible to the intended purpose.
Ther are a few other parasitic elements (inter-winding capacitance, magnetostriction,...), but they are generally not very important in their consequences at audio frequencies and levels.

 

[CJ]

you use leakage inductance to find leakage C, not the H. value of the inductor,

see that 100 pf of leakage C with 100 Henries would = 1600 hz  rez freq.

rez peaks are usually around the 100 to 275 k hz range in small audio transformers,

now 10 pf leakage C and 100 milli henries = rez freq of 160 k h hz,

so you need mH leakage to find C-parasitic, to find leakage inductance, short sec leads and measure pri inductance,

transformers can have 1 to 3 definite rez peaks depending on the turns ratio and coil structure,

core loss at resonant peak would not be a factor as at those frequencies, the core is no longer working, critical frequency at which 14 mil lams have magnetic skin effect that limits flux penetration to 1/2 the thickness of the sheet, this freq is 12,000 for 4% Si, 1300 hz for 50/50 Ni, and 200 hz for 80 Ni, so you have an air inductor at a rez freq of say, 65 k hz, the core does not really contribute to the inductance as it has no perm at 65 k hz, there might be losses due to capacitance of the lams to the coil, but core loss is not measured up there unless you are using ferrite cores or really thin lams,

 

[abbey road d enfer]

you use leakage inductance to find leakage C, not the H. value of the inductor,

In order to do that, you need to know the value of the leakage inductance Ll, which you may determine by shorting the other winding. But the value of the Ll is so small it is wiped out by the DCR. Anyway the new resonant frequency will be much higher than the natural resonance, even can be unmeasurable with standard audio equipment. Just take the case of a low-Z output xfmr à la Neve, with 0.25H inductance, coupling factor 0.9995, the short-circuit resonance will be right above 1MHz. Even an AP cannot measure that.
By using the open-circuit method, there is no uncertainty on the inductance. The effects of DCR and loss resistance do not modify the resonant frequency.
Now, it is true that in-circuit measurements will show the resonance due to leakage inductance if the drive circuitry is low-impedance.

 

[CJ]

DCR can be 10 ohms on a good 600 ohm winding,

Neve original LO-1166\a has 3 peaks with no zobel network,

1- 100 k hz 2 - 200 k hz and 3 -  300 k hz,

zobel takes out bump 1 and 3, leaves the 200 k hz,

Neve 2567 has 1 peak at 275 k hz unloaded,

Carnhill has 1 major peak at 375 k hz

Sowter has 1 - 200 k hz 2 - 325 k hz and 3 - slight bump at 1 meg hz.

 

[marcus4audio]

you would eliminate the pot,

Cool. Thanks. Yes, this is primary of the small signal high permeability core (mumeta), but can I consider this measuring as accurate? Of course, it depends from meters tolerance, but for overall info I'm happy with even + / - 10%

 

[abbey road d enfer]

DCR can be 10 ohms on a good 600 ohm winding,

Neve original LO-1166\a has 3 peaks with no zobel network,

1- 100 k hz 2 - 200 k hz and 3 -  300 k hz,

zobel takes out bump 1 and 3, leaves the 200 k hz,

Neve 2567 has 1 peak at 275 k hz unloaded,

Carnhill has 1 major peak at 375 k hz

Sowter has 1 - 200 k hz 2 - 325 k hz and 3 - slight bump at 1 meg hz.

Are these measurements made with zero-ohm source impedance? Most of the times they are made with the generator output, which can be anything between 50 and 600 ohms. With 50 ohms, the resonance is damped, making it almost impossible to visualize, with 600, it becomes more visible.
Measurement with the secondary shorted also damps the circuit so much the resonance is hardly visible.

 

[marcus4audio]

[/quote] Are these measurements made with zero-ohm source impedance? Most of the times they are made with the generator output, which can be anything between 50 and 600 ohms. With 50 ohms, the resonance is damped, making it almost impossible to visualize, with 600, it becomes more visible.
Measurement with the secondary shorted also damps the circuit so much the resonance is hardly visible.
[/quote]
50 ohm signal generator with 10V p/p. Sec open. This tx have to work as tube output (mic amp) so should I add 10k is series?

 

[abbey road d enfer]

you would eliminate the pot,

Cool. Thanks. Yes, this is primary of the small signal high permeability core (mumeta), but can I consider this measuring as accurate? Of course, it depends from meters tolerance, but for overall info I'm happy with even + / - 10%

Over its normal operating range, mumetal permittivity changes as much as -40%, so the inductance at nominal operating level (0.5 to 0.7 T) will be much lower than the inductance you measure with a weak test signal.

 

[marcus4audio]

Are these measurements made with zero-ohm source impedance? Most of the times they are made with the generator output, which can be anything between 50 and 600 ohms. With 50 ohms, the resonance is damped, making it almost impossible to visualize, with 600, it becomes more visible.
Measurement with the secondary shorted also damps the circuit so much the resonance is hardly visible.

50 ohm signal generator with 10V p/p. Sec open. This tx have to work as tube output (mic amp) so should I add 10k is series?

 

[abbey road d enfer]

Are these measurements made with zero-ohm source impedance? Most of the times they are made with the generator output, which can be anything between 50 and 600 ohms. With 50 ohms, the resonance is damped, making it almost impossible to visualize, with 600, it becomes more visible.
Measurement with the secondary shorted also damps the circuit so much the resonance is hardly visible.

50 ohm signal generator with 10V p/p. Sec open. This tx have to work as tube output (mic amp) so should I add 10k is series?

That would give you a better idea of the actual in-circuit behaviour.

 

[marcus4audio]

Thanks, sounds reasonable. But how accurate is that measuring?

 

[marcus4audio]

Something like this?

 

[abbey road d enfer]

Something like this?

With this, you have two meters in parallels with the elements.
ASSuming the reactance is about 100-200k, you would need meters with at least 1Meg input Z to make their influence negligible. That means electronic voltmeters. They cannot be of the mains-powered type (too much leakage capacitance to earth).
I use a much simpler method, which consists, after having evaluated the DCR of the inductor, in measuring the frequency response with a large-ish source impedance (10k for a "600 ohm" winding, 150k for a "10k" winding). There is a range where the the inductor will behave almost like a perfect L+R circuit, which gives an accurate enough value for the inductance, then the fist peak will show the resonance between the nominal inductance and the stray capacitance.
The amplitude of the first resonance is an indicator of the loss resistor.
Then I would either use the "zero-ohm source" (for lo-Z windings) or the short-circuit method (for Hi-Z windings) for evaluation of the leakage inductance.
Then I would enter the values in LTSpice and check that the simulated graph more or less matches the measurements. If it doesn't, I would declare the measurements wrong; I will not let reality interfere with my paradigm!  ;D

 

[marcus4audio]

Something like this?

With this, you have two meters in parallels with the elements.
ASSuming the reactance is about 100-200k, you would need meters with at least 1Meg input Z to make their influence negligible. That means electronic voltmeters. They cannot be of the mains-powered type (too much leakage capacitance to earth).

My meter has 10meg, at least that's written in the manual :/

I use a much simpler method, which consists, after having evaluated the DCR of the inductor, in measuring the frequency response with a large-ish source impedance (10k for a "600 ohm" winding, 150k for a "10k" winding). There is a range where the the inductor will behave almost like a perfect L+R circuit, which gives an accurate enough value for the inductance, then the fist peak will show the resonance between the nominal inductance and the stray capacitance.
The amplitude of the first resonance is an indicator of the loss resistor.
Then I would either use the "zero-ohm source" (for lo-Z windings) or the short-circuit method (for Hi-Z windings) for evaluation of the leakage inductance.
Then I would enter the values in LTSpice and check that the simulated graph more or less matches the measurements. If it doesn't, I would declare the measurements wrong; I will not let reality interfere with my paradigm!  ;D

cool. Is there an formula for calculating the inductance? Sorry if I ask dumb question.

 

[abbey road d enfer]

Something like this?

With this, you have two meters in parallels with the elements.
ASSuming the reactance is about 100-200k, you would need meters with at least 1Meg input Z to make their influence negligible. That means electronic voltmeters. They cannot be of the mains-powered type (too much leakage capacitance to earth).

My meter has 10meg, at least that's written in the manual :/

Is it battery-powered?

I use a much simpler method, which consists, after having evaluated the DCR of the inductor, in measuring the frequency response with a large-ish source impedance (10k for a "600 ohm" winding, 150k for a "10k" winding). There is a range where the the inductor will behave almost like a perfect L+R circuit, which gives an accurate enough value for the inductance, then the fist peak will show the resonance between the nominal inductance and the stray capacitance.
The amplitude of the first resonance is an indicator of the loss resistor.
Then I would either use the "zero-ohm source" (for lo-Z windings) or the short-circuit method (for Hi-Z windings) for evaluation of the leakage inductance.
Then I would enter the values in LTSpice and check that the simulated graph more or less matches the measurements. If it doesn't, I would declare the measurements wrong; I will not let reality interfere with my paradigm!  ;D

cool. Is there an formula for calculating the inductance? Sorry if I ask dumb question.

In the "almost perfect" region, the impedance Z is equal to sqrt[Rdc²+(2pi.F.L)]. At LF, the DCR is dominant, but as frequency increases, the impedance gets much higher than DCR, then you can use L=Z/2pi.F as a very good approximation.
You should draw a graph of Z vs. frequency, with a log scales.
You can see clearly the DCR-controlled part (parallel to the X-axis), the L.omega part (6dB/octave) and the resonance peak.

 

[marcus4audio]

Is it battery-powered?

Yes, nothing fancy but it works http://gsmserver.com/shop/equipment/measuring_equipment/multimeter_mastech/digital_multimeter_mastech_my64.php

In the "almost perfect" region, the impedance Z is equal to sqrt[Rdc²+(2pi.F.L)]. At LF, the DCR is dominant, but as frequency increases, the impedance gets much higher than DCR, then you can use L=Z/2pi.F as a very good approximation.
You should draw a graph of Z vs. frequency, with a log scales.
You can see clearly the DCR-controlled part (parallel to the X-axis), the L.omega part (6dB/octave) and the resonance peak.

This part I still can't figure out. maybe I have a problem to visualize the connections and therms. I still don't understand what is Z in this situation :/ any drawing please ?