diyTransformer - Frequency response changes with input level - why?

[andia]

I am experimenting with winding transformers lately and while measuring one of these I observed something strange to me, maybe someone can explain.
I wound a 1:1 transformer trying out a different layering (1/2S - P - 1/2S) and measured the transformer. I also tried to find a zobel combination to make the response as flat as possible.
this worked out well at line level. then I tried to feed lower levels into the transformer and saw that the lower the inputlevel, the lower the lowend output has gotten. Why is that?
the first picture is the transformer loaded with the zobel, the second picture is unloaded (200k from analyzer).

 

[Philbo King]

If you can, do an impedance vs freq sweep of the transformer (input to primary, output from secondary) with the normal load on the secondary. REW can be set up to do that.

 

[gyraf]

..perhaps relatively poor magnetic coupling coefficient, helped by inter-winding capacitance towards higher freqs?

 

[andia]

thanks, for your answers.I will look if I can make the impedance frequency sweep. I also tried to calculate the magnetic coupling coefficient and came to a result of K=0,599. I just don't know how to interpret that number right? What is the normal range for where a useful transformers K shall sit?

 

[andia]

addon: I used pretty thin wire (0,07mm) and the bobbin (EI30) has a lot of unused space. could this lead to something like that?
Winding arrangement is: 500S - 1000P - 500S. Would a thicker wire be better?

 

[ct.waveform]

I suspect this will be to do with the magnetising inductance and probably how you are using the potentiometer (?) to change level - assuming that's what you are using. In a model of a transformer the magnetising inductance would be seen in parallel with the primary and normally its effect would be constant with level. But if you are changing the impedance of the setup then the rolloff frequency will move. If I'm way off then let me know and let us know what your measurement setup is.

 

[andia]

I suspect this will be to do with the magnetising inductance and probably how you are using the potentiometer (?) to change level - assuming that's what you are using. In a model of a transformer the magnetising inductance would be seen in parallel with the primary and normally its effect would be constant with level. But if you are changing the impedance of the setup then the rolloff frequency will move. If I'm way off then let me know and let us know what your measurement setup is.

I measure through a rohde&schwarz upp audio analyzer. there I use the generator output go through the transformer and back to the analyzer input. the levels are set in the software of the rohde&schwarz. (+4dBu, 0dBu,-5,-10,-15,-20,-25,-30,-35,-40,-45)
the generator is set to 25ohms outputimpedance, in the first picture I put a 120ohm resistor in series with the + of the output. the second picture is with no additional resistor, just the 25ohms.
the input-impedance in the first picture is a 10k resistor-240nf capacitor parallel to the secondary. in the second picture it is just the 200kohms inputresistance of the analyzer input.

 

[ct.waveform]

This is a real puzzle! Maybe it IS the coupling. I'm not a transformer designer but I don't think I have ever heard of non-linearities of this sort in transformers or inductors. And it's the kind of thing that would crop up in Audio! It's as though the iron isn't getting excited at low levels. Incidentally, that coupling coeff looks awfully low. I think it's usually 0.9xx. I would have thought that evenly winding the wire along the core would be better. There must be a fair bit of resistance in there too, given how little the -3dB points of the bottom curve moved for a 5:1 change in output resistance - but then I don't know what you are using it for.

I wonder if it isn't an excitation thing after all; with the inductance moving towards its air-cored inductance? The LF rolloff has moved a helluva way - from not even a dB down at your scale limit of 8Hz, to -3dB at ~62Hz. Even the puniest of transformers surely doesn't cut into the audio band that much but I can't account for the effective drop in inductance any other way. (I'm going to have to look up to see whether this is a phenomenon that's known about.) A change in resistance would do it equally but, noting my mental difficulty picturing what inductors in parallel do, I think if the resistance went up that would push up the frequency.

Hang on! I've had a genuine thought here. Maybe this does happen but at super low levels, which is what we've got here, I think. 50dB down into a 200k load. Perhaps there just isn't enough current and the reason it hasn't popped up in Audio is because it is too far down to be worth considering.

 

[cyrano]

I know next to nothing about transformers. If I encounter unexpected results, I'd make a second one to see if it behaves the same.

I was thinking about bad insulation in one of the windings, but that's just a wild guess, as I can't easily visualise transformer behaviour.

 

[abbey road d enfer]

I am experimenting with winding transformers lately and while measuring one of these I observed something strange to me, maybe someone can explain.
I wound a 1:1 transformer trying out a different layering (1/2S - P - 1/2S) and measured the transformer. I also tried to find a zobel combination to make the response as flat as possible.
this worked out well at line level. then I tried to feed lower levels into the transformer and saw that the lower the inputlevel, the lower the lowend output has gotten. Why is that?

You haven't mentioned it but it's obvious your transformer is wound on a ferromagnetic core.
It is well known that magnetic permeability of ferromagnetic materials vary with frequency and level.
The lower the frequency, the higher the perm. Which results in better LF performance than could be thought by looking at the midrange inductance (often measured at 1kHz).
AND the higher the level, the higher the perm, until the core starts to saturate.
That explains why the LF response is better at high level than at lower level.
This is totally independant on loading.
This is relatively well-documented, but involves some moleculat physics.

 

[Cqwet Dbdfte]

Get some transformer design books.
It is an (manufacturing) art as much as a science.
Measure (open) inductance of your coils, and also with one side shorted (leakage inductance).
Insufficient inductance will attenuate low freq, and too much inter-winding capacitance will attenuate high freq. Design is a compromise between the two. Also measure the self resonant frequency of the transformer. It should be way above the useful range.
.Audio transformers may be expected to have 4 - 5 decades of frequency range, not an easy task.
Work out the design on "paper" before opening the tool box.

 

[andia]

Thank you, abbey. Yes, I use a steel that is called z11 for this transformer. I would love to do experiments with nickel laminations but they are relatively hard to come by here. I did get some samples from japan but I dont have fitting bobbins for these yet.
Would this effect be lower with nickel?

Today I had the time to do another wind of that transformer. this time I used thicker wire (0,125mm) to fill the bobbin better. turns count stayed the same. the inductance I measured was higher with the new wind (from 6.6H it went to 7.9H). there is still a drop in level at low frequencies with lower inputvoltage but not that drastic as with the first wind.

 

[Cqwet Dbdfte]

Transformer books talk about "window" fill, likely an important factor. Wire resistance another. For bigger transformers bobbins are not used, and windings layered between layers of kraft paper. Then the whole thing is dunked in varnish and vacuumed out, great for power transformers, but not so great for audio where the varnish increases capacitance.
Too much info for this forum.
Have fun !

 

[ccaudle]

it's obvious your transformer is wound on a ferromagnetic core.

That seems like a broad categorization. Would silicon steel as often used for audio transformers not be considered "ferromagnetic?" Or that behavior is common with most transformers to some degree, but less pronounced with some alloys?
I searched for Z11 and it seems slightly different but in the same general behavior category as M6, which I have heard mentioned more frequently in audio transformer use.

I would love to do experiments with nickel laminations

I do not know if specifically transformer use has more specific definitions, but in general use nickel also falls in that category. From the Wikipedia article on ferromagnetism:
"Relatively few materials are ferromagnetic and are typically pure forms, alloys, or compounds of iron, cobalt, nickel, and certain rare-earth metals."

 

[andia]

so then all corematerials would be considered ferromagnetic.

what I can say about z11 is that I can get close to double the inductance than with m6 on the same winding. I think nickel would result in even higher inductances.

 

[abbey road d enfer]

That seems like a broad categorization. Would silicon steel as often used for audio transformers not be considered "ferromagnetic?"

Of course it would. Iron is ferromagnetic. In the absence of data I used the broadest term.
"Ferro" comes from Latin "ferrum" (iron). Iron translates as "ferro" in spanish, "fer" in french. The "fer" root can be found in ferrite.

Or that behavior is common with most transformers to some degree, but less pronounced with some alloys?

This effect exists in all cored transformers, to different degrees.

 

[abbey road d enfer]

I think nickel would result in even higher inductances.

Check Ni core vs. M6 core
Particularly the first post.

 

[bepone]

addon: I used pretty thin wire (0,07mm) and the bobbin (EI30) has a lot of unused space. could this lead to something like that?
Winding arrangement is: 500S - 1000P - 500S. Would a thicker wire be better?

no it is ok, winding geometry is affecting top end.. your bass behaviuor depends of :
1. output impedance fromthe source
2. voltage AC connected to transformer

dont forget that permeability is voltage phenomenon, more voltage , more inductance, better bass, so all normal here

 

[CJ]

Monitor your input level with a scope or voltmeter to make sure it is staying constant.

 

[HuckyBridge]

You can check the magnetization curves (BH Curve) of your core material and you see that you get lower permeability at low and high level compared to mid levels (higher slew). The magnetic core loss ist the area inside the curve. At high levels you get high loss and distortion at low frequency.
In your case (low level) you get lower level at low frequency by the higher magnetization losses of the core at low level. The magentization losses are represented by a resistor in parallel with the primary coil. But this resistor is not only linear. The losses are higher at max and minimal levels.
In your design (low core saturation) you can only compensate the low frequency loss at low levels with low source (generator) and low primary winding resistance.
In the old days with magnetic tape recorders @StuderRevox we tried to overcome this type of problems (high dynamic range, low distortion, wide bandwidth) with HF-Bias (hi frequency injection). Until today this application is not seen in audio transformer designs > go for patents.
In modern days we recreate this type of problems with computer plugins to get magnetic saturation problems in musicians hands to be creative.