diyTransformer - Frequency response changes with input level - why?

[andia]

Damn, that is some square response. I wouldn't have expected the T38 material to manage that.. You sure about the 1MHz?

/Jakob E.

I am sure that the signal generator was set to 1MHz, I am absolutely not sure what it means .

 

[gyraf]

That's a really nice wide-band transformer.

Now it would be interesting to see what your maximum working levels are like

 

[bepone]

excellent graphs.. what is the interleaving, if any? some pics of the transformer?

 

[bepone]

i have wound something for REDD47 1:7 input and output transformers on small double c core, after i pour them in wax, i need to find them i finish the project (some day)

 

[andia]

That's a really nice wide-band transformer.

Now it would be interesting to see what your maximum working levels are like

I suppose what is interesting is level at 20Hz and when 1%THD is reached, as I saw in the Jensen datasheets.

this is interesting, 1%THD @20Hz is reached at about +9.9dBU but skyrockets after that. at +12dBu it shows 16%, at +18dBu it is over 52%. I suppose this is the material?
0dBu is 0,08% THD, +4dBu is 0,15%THD @20Hz

@1kHz it is at the bottom of what my analyzer can measure 0dBu = 0.0003%THD, 18dBu = 0.0006%THD.

 

[andia]

excellent graphs.. what is the interleaving, if any? some pics of the transformer?

no interleaving, just the 2 coils wound bifilar at the same time.
here a picture of the transformer and the core. ( https://www.mouser.at/ProductDetail/871-B64290L615X38 )

 

[gyraf]

1%THD @20Hz is reached at about +9.9dBU but skyrockets after that

Is to be expected, overload and the core magnetic field essentially breaks down. Scope'ing the signal can show you clearly what happens..

With that extended high-end, and waay low leakage inductance, I think the transformer is a good candidate for even more turns: More turns reduces flux-per-volt, which again makes more headroom at low frequencies. Trading off some of your (perhaps unneeded) high end for more (possibly wanted) low end overload margin, so to speak

/Jakob E.

 

[ccaudle]

TDK T38 material.

I have seen ferrite material used for high frequency transformers before, but I did not think it was usable down to 20Hz. Is that common for ferrites, or is the TDK material unusual in some way? The TDK datasheet lists the "optimum frequency range" as 10kHz to 100kHz.
Any idea why they are not more commonly used for audio? It seems you had good results.

is there anything that speaks against such a kind of transformer for mic level applications?

1:1 seems unusual for mic applications. With no connection to mains power there is typically not the need for very high CMRR at low frequencies that transformers can provide, and 1:1 winding does not take advantage of potential noise advantages for a preamp of having part of the first stage gain in the transformer instead of active devices, or of reducing the output impedance of the driving device on the microphone end.
What would be the goal of such a 1:1 transformer?

 

[Cqwet Dbdfte]

With that high-sh pri-sec capacitance, isolation, and lack of voltage stepping, it is like a fine capacitor, except it provides "ground" isolation. Much like a pulse transformer

TDK T38 datasheet says initial perm (ui) "10000", Micrometal who also has Sendust, and many other materials, their -38 material has ui of 85. ( same measurement unit?)

 

[andia]

I have seen ferrite material used for high frequency transformers before, but I did not think it was usable down to 20Hz. Is that common for ferrites, or is the TDK material unusual in some way? The TDK datasheet lists the "optimum frequency range" as 10kHz to 100kHz.
Any idea why they are not more commonly used for audio? It seems you had good results.



1:1 seems unusual for mic applications. With no connection to mains power there is typically not the need for very high CMRR at low frequencies that transformers can provide, and 1:1 winding does not take advantage of potential noise advantages for a preamp of having part of the first stage gain in the transformer instead of active devices, or of reducing the output impedance of the driving device on the microphone end.
What would be the goal of such a 1:1 transformer?

1:1 like in a microphone splitter application was the idea.

 

[MisterCMRR]

An interesting thread I just scanned for the first time. I don't want to be the "wet blanket" here, but I have two questions. First, I haven't seen a source impedance specified for the plots. Low-end frequency response will always look great from very low driving impedances. Tests should be done with a source Z consistent with the application. Second, with that much primary to secondary capacitance, of course the HF response will be extended because coupling becomes entirely capacitive. Which means very little isolation/CMR in applications like mic input and splitter. Jensen mic splits, for example, use double Faraday shielding because noise rejection is so important in splitter apps.

 

[andia]

An interesting thread I just scanned for the first time. I don't want to be the "wet blanket" here, but I have two questions. First, I haven't seen a source impedance specified for the plots. Low-end frequency response will always look great from very low driving impedances. Tests should be done with a source Z consistent with the application. Second, with that much primary to secondary capacitance, of course the HF response will be extended because coupling becomes entirely capacitive. Which means very little isolation/CMR in applications like mic input and splitter. Jensen mic splits, for example, use double Faraday shielding because noise rejection is so important in splitter apps.

I have redone the measurements according to test-circuit 1 in the jensen transformer documents, ~150ohms source impedance (more like 140, I don't have the correct resistors at hand), 1kohms load impedance. @second: I suppose there is no way to incorporate faraday shields into this kind of transformer?

frequencyresponse and outputlevel @1V inputlevel:

freqresponse at lower levels

phase/frequency
THD @2dBu, -10dBu, -20dBu
THD @20Hz(green), 30Hz(red), 50Hz(vio)

 

[gyraf]

Yeah, guilty of assuming some finite driving impedance in test setup, here in my setup it's always 200 Ohms and it makes me kinda not think of it when eyeball'ing plots

Your 150-Ohms plot still looks quite decent - a hf-bump just starting to develop at/above 80KHz if i read it correctly. For further optimization, I would still pour on more turns until this peak came down to some 30-40KHz at realistic (application dependent) driving impedances. At least, this is how I usually cut the cake, and I've been lucky so far

Faraday shield would kinda negate the bifilar idea: The reason for the very-good high-frequency response in bifilar-wound (like the API output transformers) is the winding capacitance transferring highs at a first-order highpass, ideally where leak inductance looses it. But this makes for poor output balance, as one end of winding usually "sees" more capacitance than the other

otoh, if you want, faraday shields are possible in toroids. Just add a layer of copper/conductive foil between pri and sec windings, making sure NOT to close a complete turn around -into core, i.e. isolating beginning and end. Solder on and bring out a single wire.

as for the previous question of T38 ferrite material being qualified for low frequency work, I can assure you it is: I use it in several places for low-frequency inductors, although in pot core form. I have test pilot reports indicating that its pattern of misbehaving and breaking up is quite agreeable to human tastes.

/Jakob E.

 

[Cqwet Dbdfte]

If you use a faraday shield pri-sec you can always restore the lost HF response with a capacitor ;-)
This transformer only provides ground isolation, it does not "transform".
If this was the objective, mission accomplished.
More turns would make the core saturate with smaller signal. How much is enough? The turns to inductance ratio is a square.
Line level job or microphone level?

 

[gyraf]

actually, more turns would make the core saturate at higher signal level even if it is counterintuitive

 

[Cqwet Dbdfte]

Of course, you are right. I had my mind into DC ampere turns.
More turns more inductance, more impedance, less turns per volt, more resistance losses etc.