diyTransformer - Frequency response changes with input level - why?

[Cqwet Dbdfte]

An optimized transformer design uses a core, EI, L L, C, toroid, whatever appropriate to the application, with a core area, path length, window size, air gap size, wire gages and geometry, and magnetic material. Everything is a tradeoff. It does not happen empirically, the math and the science is well understood and documented. The art is selecting tradeoffs to suit the application. Putting it together requires skill and the right equipnent.
Get some good books on the subject. There are also quite a few software packages....
https://www.downloadcloud.com/transformer-design-software.htmlMany commercial and some free ones...

 

[CJ]

How many kilo gauss are you testing at?

M6 steel saturates at 18 KG

Find voltage constant for flux,

Faradays Law

100000000/4 x 1 x 1000 x 10

4 is stacking factor
1 is core cross section in cm
1000 is turns
10 is frequency in hertz (arbitrary)

Flux is 2500 E

E is sine wave RMS

18 K Gauss (saturation)/2500 is 7 volts.

Run at half that for low distortion so 3.5 volts RMS.

Are my eyes right in seeing that the xfmr drops of when running -50 db or 0.003 volts?

If so then you are wondering why the transformer drops at
.003/7 = 0.04% saturation level?

The reason is you are very low on the hysteresis loop (BH curve)

You are on what is called the Initial Permeability part of the curve. So your inductance is dropping at -50db.

Here is a BH curve. Notice the slope of the tangent line at ui (initial perm)

In other words , the steel core needs a certain level to start working and you are barely there. Not enough excitation current generated by 0.003 volts into your DCR.

 

[bepone]

this is single ended or PP tranformer, i didnt get it? i was thinking we are speaking of PP transformer, with excitation without DC bias, if so, i would rewind it again and make some taps on primary (several) and see the behaviour if existing on taps... or test the core and see what number of turns fits best

 

[bepone]

to me still looks like lack of inductance(low permeability-low level excitation) but after tap test i will know more

 

[andia]

The reason I tested for so low levels is that I wanted to see if this could be used for a mic level splitter.

 

[bepone]

add 1000 turns more and recheck

 

[abbey road d enfer]

The reason I tested for so low levels is that I wanted to see if this could be used for a mic level splitter.

For that you'd better use multifilar winding. You can get HF response up to MHz.
What size is your core?

 

[andia]

For that you'd better use multifilar winding. You can get HF response up to MHz.
What size is your core?

EI30

 

[andia]

add 1000 turns more and recheck

wouldnt it be a generaly good thing to keep dcr low? if that has anything to with that

 

[CJ]

gotta have a nickel core i am afraid, or at least 50/50.

 

[abbey road d enfer]

wouldnt it be a generaly good thing to keep dcr low? if that has anything to with that

A good rule of thumb is to make DCR less than 1/10th of nominal impedance (150-200 ohms for mics).

 

[abbey road d enfer]

this is single ended or PP tranformer, i didnt get it?

It's not an output xfmr for a valve amp. It's a distribution/isolation xfmr. It runs without DC.
However, it does not mean there's no magnetization. The signal is what magnetizes the core.

 

[abbey road d enfer]

Get some good books on the subject. There are also quite a few software packages....

Most, if not all these books and software deal with power transformers and inductors. They optimize power at fixed frequency operation.
An audio xfmr must operate on a very wide frequency range (typically 1:1000) and with low distortion.
Power xfmrs work at fixed frequency and their distortion is very large (10-20%).
Audio xfmrs pose a very different set of constraints, like leakage inductance vs. nominal, DCR, capacitive parasitics, and so on. Not mentioning the effects of running them at initial magnetization and Barkhausen effect.
I haven't seen yet a software package that does it with confidence. Spice uses the Chan model, but I found that the results are often very off the mark, when one wants to achieve very low distortion.
Also, using the Chan model requires access to parameters that are not divulged by core manufacturers.

 

[CJ]

Prerequisite reading for all future transformer geeks>

 

[bepone]

It's not an output xfmr for a valve amp. It's a distribution/isolation xfmr. It runs without DC.
However, it does not mean there's no magnetization. The signal is what magnetizes the core.

ok, but formula for -3dB should be where internal impedance of the source (mic 100-150 ohms) is equal to Inductance of the transformer.. so if is dropping on higher freq. means inductance is not enough , thats why to put more turns

 

[CJ]

You mean lower frequencies?

The transformer does not need a core after about 10 kC, maybe even lower.

To see this put the xfmr on a scope and pull out the core.

In fact , the core is a hindrance at high freqs as it increases capacitance.

 

[abbey road d enfer]

ok, but formula for -3dB should be where internal impedance of the source (mic 100-150 ohms) is equal to Inductance of the transformer

That is correct.

.. so if is dropping on higher freq. means inductance is not enough , thats why to put more turns

The problem with increasing the number of turns also increases DCR, as well as parasitic capacitance. The latter may not be an issue, but the former is significant because it increases losses and impairs the noise performance.
The graphs published by the OP just show that the core material is not optimum for the purpose.

 

[bepone]

You mean lower frequencies?

The transformer does not need a core after about 10 kC, maybe even lower.

To see this put the xfmr on a scope and pull out the core.

In fact , the core is a hindrance at high freqs as it increases capacitance.

yes, sorry for my english, we are speaking only for the bass area ofcourse.. and -3dB on min side of the spectrum, and "higher" statement from upper quote means higher from desired, like 80 Hz is "higher" than 30.. is starting to drop on 80Hz ("higher" than 30) means missing inductance for 30 Hz or any other calculated desired f.

 

[bepone]

That is correct.

The problem with increasing the number of turns also increases DCR, as well as parasitic capacitance. The latter may not be an issue, but the former is significant because it increases losses and impairs the noise performance.
The graphs published by the OP just show that the core material is not optimum for the purpose.

yes, but you can add some paper every several layers to cut this and also is stated that bobbin is not full , so wire size can be increased..

 

[bepone]

btw..agree from above that 1:1 interstage bifilar winding has no any problem , and is going over 1MHz, on C core specially