measuring and treating audio transformers

[5v333]

hi!

i am trying to find a good termination and grounding for a transfomer that sees an unbalanced source of about 500R. ratio is 1:2 and the sec acts as a centertapped phase splitter, feeding two grids in my balanced tubecircuit.

after alot of experimenting i can know see alot of different things that could be of interest and i dont know if i should try to fullfill them all.


-terminate/zobel the secondary so that a squarewave looks like the sharpest squarewave ever.

-do a XY scan and see that the phase curve runs somewhat smoothly and reach 180º with out to much fuzz.

-same as above, but see that the phase curve looks similar on both ends of the secondary and that 180º is not found an octave or three from each other.

-do a sweep looking at both pri and sec and see that the loop gain is more or less steady before it falls.


in my tries of getting a nice result, its hard to get all of theese stuff lined up. feels like i have to sacrifice something...


am i overthinking stuff.
what should i focus on?
is a sharp squarewave on the sec trivial?

 

[PRR]

> i have to sacrifice something...

If you needed response 1Hz-1MHz you would not be using a step-up transformer. So you already decided to "sacrifice".

You need response to 15KHz.

Anything over 30KHz is for the dogs. And they don't buy your music.

Get the frequency response flat past 15KHz if possible. Any rise is likely to be ringy, but brutally damping ringing tends to hurt midband transfer so don't go crazy.

In a Linear Phase system, the phase response can be completely plotted from the frequency response. If gain is smooth, phase will be smooth, and that's honestly good-enough for music. (In the few cases where zero phase is wanted, do not use transformers.)

You are not building a house. The square wave is the least of your concern. You don't hear "square", you hear frequency response below 20KHz. Gently knocking off all the higher harmonics of a Square will look roundy on the 'scope but little effect on the ear.

 

[5v333]

appreciate your input!

Another thing im not sure about is how the bandwidth/response is measured..

Just looking at the sec or the differance between pri and sec (loopgain)?

 

[abbey road d enfer]

appreciate your input!

Another thing im not sure about is how the bandwidth/response is measured..

Just looking at the sec

Yes, but make sure the source signal is constant in the measurement bandwidth and that teh actual impedance presented to the primary is that that will be used in practice. Most measurement oscillators have a low-ish output impedance; connecting directly to the primary will result in larger bandwidth than in practical use. However, some oscillators have a 600 ohm output; in that case, the measured BW will be narrower than in most practical cases, where the typical output impeadnce is around 100 ohms.

or the differance between pri and sec

This method results in better performance than in actual life, because the interaction between the xfmr and the source impedance is cancelled.

(loopgain)?

Loop gain is a term generally applicable to systems under Negative FeedBack, which is not the case here. I would think the correct term is transmission factor.
Bill Whitlock, in his paper, that could be considered the "modern Bible" of audio transformers, use the term "transmission characteristics"
http://jensen-transformers.com/wp-content/uploads/2014/09/Audio-Transformers-Chapter.pdf

 

[5v333]

My generator outputs 50r z. I always put a series resistor to match the actual circuit.

In this case the signal on the pri does fall aproximatly together with the sec.
But at a point the signal on the pri makes a significant dip.
Voltage gain rises so im thinking this could be a resonance, even though a square looks rounded without wiggle.

 

[abbey road d enfer]

My generator outputs 50r z. I always put a series resistor to match the actual circuit.

In this case the signal on the pri does fall aproximatly together with the sec.
But at a point the signal on the pri makes a significant dip.
Voltage gain rises so im thinking this could be a resonance, even though a square looks rounded without wiggle.

That is a sign that the resonant frequency is not too far from the fundamental frequency of the square-wave, so the harmonics are filtered out. Maybe counter-intuitive, but "wiggles" are a good sign; they're a sign of extended frequency response. See Gibbs phenomenon.

 

[5v333]

is it a good sign or a bad sign?

it seems that i am able to balance the response of the two sides with severe resistive loading and then a little RC across the sec for rounding of the wiggle seen with a square wave. also a small snubber on the input seems nice for smoothing out the remaining stuff of a squarewave seen on both input and output.

-3db difference between in and out is at 400khz now. this cant be the bandwidth im looking for could it?

Lundahl LL1684 is the one in question.

 

[PRR]

> -3db difference between in and out is at 400khz now

What is the sound of a 3dB unbalance at 400KHz?

 

[5v333]

almost like reel to reel bias.

 

[5v333]

sorry, one aligator clip had jumpd out.

-3db difference between in/out is about 280khz. when looking only at the sec, -3db is about 165khz...

 

[abbey road d enfer]

sorry, one aligator clip had jumpd out.

-3db difference between in/out is about 280khz. when looking only at the sec, -3db is about 165khz...

That means you can expect signal cancellation at around 300kHz; is it a serious issue?

 

[5v333]

280khz is huge, i didnt expect it. Its fine, i would have accepted 1/4 of that. Could it be a problem further on though..?
Im more interested to know how to control flattness and other stuff we dont like to see in a transformer. And how to monitor it.

The dip on the pri was around 150khz-300khz and looked like a 10-12db valley refered to the sec.




I

 

[abbey road d enfer]

The dip on the pri was around 150khz-300khz and looked like a 10-12db valley refered to the sec.

When a resonance is measured at the secondary, that means that it delivers more power that goes to the load or to the various causes of loss. That also means that more current is taken from the source, or in other words, the impedance reflected at the primary decreases. When driven from a source that has more than zero impedance, it creates a dip. The frequency where the dip is more shallow coincides with the actual overall resonant frequency.

 

[5v333]

Im going to revisit all transformers in the system now and see what the outcome will be with this new view on how to measure them.

Thanks again guys!

 

[5v333]

something just popped into my head on my morning walk today.

thoose high bandwidth numbers ive been talking about, i thought they were just too high to be true.
and i reallized that i had been measuring over the full primary but only half secondary since its a phase splitter. thatwould suggest that i monitored only half the voltage gain/change. in other words my former -3db points were actually -6db points.

and if thats true, the initial resonances i had that looked like 10-12db dips on the primary were infact 16-18db dips.


this also explains why i get so high numbers when measuring my pushpull OT. half of what i got should be more realistic.

am i right?

 

[abbey road d enfer]

something just popped into my head on my morning walk today.

thoose high bandwidth numbers ive been talking about, i thought they were just too high to be true.

No. These figures are typical for a bifilar-wound low'Z transformer.

and i reallized that i had been measuring over the full primary but only half secondary since its a phase splitter. thatwould suggest that i monitored only half the voltage gain/change. in other words my former -3db points were actually -6db points.

No. -3dB is relative to mid-band response; whatever the transformation ratio. Do you mean you checked the point where the secondary voltage was half the primary voltage? How was the primary connected? Series or parallels?

 

[5v333]

maybe a bad way of testing but i start by dialing up midband freq around 0º phaseshift.
put both signals on scope. trim them equal in amplitude on the scope. sweep until the differance is 75% (which probobly isnt 3db difference but im thinking half of half or half of 6db.)
i adjust the gain on my generator as signals starts to fall so i can refernce with a full vertical span on my scope and read of the scales.

pri paralell
sec series CT.
1:2CT

i think the 45º shift isnt found up around 400khz aswell..!

 

[abbey road d enfer]

maybe a bad way of testing but i start by dialing up midband freq around 0º phaseshift.
put both signals on scope.

Which "both signals"? One should be the reference i.e. the generator's output, not the primary voltage, when you insert resistors for source impedance tuning. Remeber that there is nothing like an intrinsic frequency response of a xfmr; it is always the result of its terminations.

trim them equal in amplitude on the scope. sweep until the differance is 75% (which probobly isnt 3db difference but im thinking half of half or half of 6db.)

Actually 70% is close enough to -3dB. Not significant in the context.

pri paralell
sec series CT.
1:2CT

So actually you're measuring 1:1 when you probe between CT and either end.

[/quote] i think the 45º shift isnt found up around 400khz aswell..!
[/quote]  Phase should steadily increase (leading), going through 45° (when the amplitude graph slope reverts), then 90° at resonance and then decrease (lagging) until it reaches an ultimate 180° with a 12dB/octave slope.
So there are two 45° points, both each side of the resonant peak.
400kHz for the second resonant peak is consistent with ca. 300kHz resonance; the first one would be around 220k.

 

[5v333]

attached is a pic of how i measure without any termination except for the impedance set resistor. this is wrong?! you mean i should measure right off the generator output?

i forgot to check which way the phase curve went of course when i did the sweeping!!!!

 

[CJ]

check out Crowehurst: "Measuring Up a Transformer"