Audio Transformer Inductance

[CJ]

here are a few curves i dug up,

triad hs 29-sowter vs original- the yellow bass peaks are from having the scope set for ac, which puts a cap in series, so that should really be a roll off, not a boost,  , utc ls 68 vs ls 140, notice how flat the 140 is and notice the bigger res peak-such is life, you trade one thing for another, triad hs66, hs 50, haufe, ouncer O-9, tele v76,we 11c, below the we 111c freq plot is the current consumption, notice how the low end sucks current because of lower XL, while the high end sucks current because of the resonant circuit shorting out the highs,

 

[ricardo]

Thanks for these treasures CJ.  This beach bum is tempted to blow his annual income and buy you a scanner.

Are these measurements with the designed terminating and feed resistances (apart from the yellow bass peaks)?

 

[CJ]

the input side is pretty standard, 50 ohm generator into a 500/600 ohm winding usually, unless it is an innerstage, then you have a 5k load being driven by the generator, so there is a slight difference there, but innserstage transformers are the least used of all of them, so just about everything from Langevin to API has a 600 ohm input, and a 600 ohm output, this means you get a good match up between the generator and the xfmr 95 percent of the time.

now on the secondary, i could put zobels and resistors across there, and i have done it, all it does most of the time is squash down the peaks and shift them apart a bit, so i run all of the secondary leads straight into the scope, which is high Z, so i get the worst case scenario due to leakage C and L, this way it is easier to see the personality of the transformer, if you terminate everything with a load, then the differences between the transformers become less pronounced,
seeing a big spike is always fun, some folks don't like the way i test transformers because they get mad that the transformer they make no longer matches the published response curve because of the way i test them, so if you see a difference between the advertised curves and my curves, it is because of the load they may use,

the only transformer that has no spike is the API quadfillar, but that is hiding the fact that it is fairly leaky with the windings being right next to each other, but that does not matter, tom wolfe knew that the discrete api opamp could supply a ton of current, so  driving that 2503 thing with whatever it takes is not a problem.

is it tom wolfe or somebody else? i think i got the electric koolaid acid test guy somehow building opamps, 

 

[ricardo]

CJ, is that ...

50R generator directly driving the primaries with secondaries direct to the oscilloscope?  No namby pamby loads or Zobels to complicate matters.  8)

Then the curves make sense  ;D

 

[CJ]

yes, that is correct, sir.

i do hook up a 100K with a cap in series across the sec for BH curves,

it is sort of a poor man's integrator.

then you put the scope into x-y mode, feed the xfmr about 20 hz,

supermalloy curves show the different losses for the same material,

the best alloy as far as loss was the Tamura transformer i tested,

then the Neve input stuff, then the Peerless stuff, then the UTC stuff,

 

[rafafredd]

I always said: Tamura is the best transformers you can get. I don't get how those japas could do such small, high inductance and low resistance windings. Were the lams that difference? Maybe quality of annealing?

I have used tamuras for the third and fourth channel on this one:

http://www.groupdiy.com/index.php?topic=48931.msg618264#msg618264

And it kicks!

 

[CJ]

yes the less core loss, the less distortion, as you draw less magnetizing current which can distort the pri signal.

now if i were doing distortion tests on transformers, then i would terminate the sec with the right load to get that signal current flowing as to get the ratio of signal to mag current a lot higher, which should lower distortion,

but i have never owned a distortion meter, anybody have a DIY distortion box?

 

[rafafredd]

rmaa?

 

[ricardo]

raf, I don't think you can make RMAA do distortion at LF.  It is an excellent software package but needs understanding and study like any piece of complex software or hardware.

To close the loop on transformers, here is the Great Guru pontificating in 1994.  It's from
http://www.leonaudio.com.au/microphone.engineering.handbook..chapter.8.pdf
and was probably his last word on LN design.  Reading this makes me regret skiving daily study of his work.  It's the best accurate work on LN design I've seen.  The only thing he misses out is very high Z (eg condensor mikes) for which you need his Wireless World 1968 article.

Here, he reverts to the RDH 4th ed. equivalent circuit.  There's an excellent treatment of HF damping and how to do Zobels too.  His Fig 8.23 et al show the effects I'm familiar with but also peaks & dips around 1MHz that Abbey sees in his simulations.  I gotta admit I never went beyond 100kHz in my transformer investigations.

There's stuff on BH curves which might be what CJ was referring to and how to design small audio transformers.  Grab this before the knowledge is lost forever.

 

[CJ]

great article!  covers a lot of stuff without horrendous math and esoteric verbage,

one thing i wish was in there is phase shift, which is covered by ol man crowehurst,

 

[jackies]

I was gonna say how I decided that when I grow up I wanna be like CJ and smoke crack crack transformers and learn things, but CJ already moved the bar up so high with his 670 project...

So, no time for cracking and learning, I decided to limit myself with measuring a few transformers I had kicking about, and purchased a Mastech MS5308 LCR meter.

Inductance measurements are somewhat wonky, and frequency (and level?) dependent. For instance, I measured a 0.56mH high quality foil inductor I had left from my speaker experiments, the readings were 100Hz:0.598mH, 1kHz:0.566mH, 10kHz:0.553mH, 100kHz:0.549mH. Not much variation, huh? I imagine it's because of the robust construction of the inductor - no core, optimal winding geometry, etc. Situation is completely different when I measure the ever-popular UTC A-24 (notice: mine is the newer type, grey with red print, and CJ measured the old one, with an aluminum badge) both windings in series: 100Hz:1560H, 1kHz:608H, 10kHz:2.9H, 100kHz:29mH (!)
;D
So here's some more results, DCR and inductance measured at 100Hz, all primaries/secondaries are in series where available.

JLM 1:4 pri 300, 5H, sec 500, 138H
Newcomb TR-91 pri 113, 4.2H, sec 3.3k, 688H
UTC H-7 pri 1k, 18H, sec 45, 0.58H
(?)Bogen (mic in) pri 16, 1.2H, sec 94, 6.5H
UTC O-16 pri 38, 5.5H, sec 2K, 1584H
Triad A-65J pri 1.15k, 834H, sec 56, 38H
LL1521 pri 1.15k, 265H, sec 1.7k, 370H
Sowter 3195 pri 20, 2.57H, sec 670, 146H
LL1636 (1:10 config) pri 9.7, 2.6H, sec 752, 136H
LL1539 pri 38, 4.7H, sec 38, 5.1H
LL5402 pri 56, 8H, sec 14, 2.3H
LL1922 pri 116, 131H, sec 1.5k, 1750H
ADC (? small) pri 18, 0.34H, sec 460, 10H
UTC H-26 pri 600, 20H, sec 140, 5H (evidently a mil equivalent of the next one)
UTC A-15 pri 640, 20H, sec 140, 5H
Magnetika reissue HA-100 pri 64, 38H, sec 3.2K, 2870H
UTC HA-101 pri 28, 20H, sec 6.3k, 3575H (we have a winner!)
UTC A-10 pri 60, 18H, sec 2.9k, 1394H
UTC A-24 pri 1.5k, 1560H, sec 85, 58H
UTC A-26 pri 87, 21H, sec 2.5k, 1276H
UTC A-19 pri 3k, 66H, sec 10k, 346H
UTC A-20 pri 43, 5.9H,sec 41, 6H
UTC A-21 pri 28, 7H, sec 41 7H
UTC A-44 pri 300, 12H, sec 45, 2.3H
Collins (mic in?) pri 9.3, 0.4H, sec 2.8k, 81H
TFK (old mic in) pri 220, 78H, sec 3.3k, 1526H
UTC B-5859 (output) pri 226, 13H, sec 22, 1.5H
Collins (output?) pri 810, 25H, sec 76, 2.4H

 

[CJ]

Thanks!

that is a bit of work,

you have enuff iron to build a 670 and SSL, + 1176?

how much for the meter?

 

[jackies]

Had a typo in the meter name, doh! Ebay has Mastech MS5308 at $215, with free shipping from China!
Don't have quite enough iron for the 670 though, my next step is to attempt to wind something similar to HS-52, (like our mysterious comrade "vari-mu" did) and if I find that to be too much work, I could, of course, splurge for Sowters...
8)
I think it's important to know the inductance of the transformers - high inductance will work well with high impedance outputs, like tubes, and transformers with lower inductance will be alright for solid state projects...

Another question - is there a practical way to measure hysteresis, i.e. some hookup schematic to see the hysteresis loop on an oscilloscope?

 

[gyraf]

You mean BH curve tracing?

CJ has made a couple of detailed posts about this - though I must admit that I don't fully understand how to interpret the outcome waveforms.

Jakob E.

 

[CJ]

inject a square wave into the side of the transformer with the least inductance/turns

this is so you can saturate the transformer, the fewer turns, the easier it is to saturate, as the B max formula has turns in the denominator, so as turns go down, flux density goes up

take a 100 k resistor in series with about a 0.1 poly cap and connect that across the other side of the xfmr,

put the scope into X Y mode, then try to figue out where to put the probes,

i bet one goes on that cap,

BH curves tell you about the core,

if it was processed for a "square loop", then you will see a straighter  BH curve  that flattens out real quick, thus forming a square,

if it is "round loop" , you will see a see a slanted BH curve that reaches saturation gradually, which will round off the corners, thus, round loop.

Core Loss -  skinny BH curve means less core loss = high octane alloy,

so if you have 2 samples of 80 Ni, you can see which one has more loss,

you can also tell the alloy type, silicon steel, 50/50, or 80 Ni will have different curves.

for saturating big silicon cores, you will need a variac at 60 cps or hopefully less, i used to feed a variac 20 hz, it did not like it (growling at me) but i got the data

a slanted BH also means an air gap if looking at outputs and chokes,

tilting the BH means saturation is more gradual, due to the reduced perm from the gap.

 

[ruffrecords]

Thnks for making all those measurements. We have so many now I think it might be more readable if I put them in a spreadsheet.

Cheers

Ian

 

[jackies]

if it was processed for a "square loop",

This is exactly what bothers me - I have some lams from some soviet equipment, and I'm afraid they were processed for the "square loop" - any idea how it'd affect the sound of the transformers?

 

[CJ]

Gus told me he likes square loop for U47/67 transformers,

if you input 1 millivolt into a good tube, you can get 100 mv out,

how much juice does a large condenser capsule put out?

100 mv will be safe for square loop as the flux will be low due to the low voltage,

flux is really caused by current, but we can sub in volts since they tend to go up and down with each other in a linear fashion E=IR,


if you raise the voltage on a winding, you raise the current, and flux,

 

[jackies]

One cunning specimen was trying to avoid the test:
UTC A-25 pri 1.6k, 64H, sec 76, 2.4H
;D

 

[emrr]

> discrete notches don't happen in real life

They do; but it is usually convenient to stop the sweep (or prune the graph) so they don't show. Honest byproduct of multi-winding construction.

Got any examples?  Abbey is trying to simulate distributed winding capacitances so this should hold even with simple 2 winding trannies.

Wow, this answers a question I've raised here many times, with no previous comments.

Examples?  Almost every UTC I've ever taken a response plot of.  They all show some notch, sometimes very shallow, sometimes deeper.  Most of them show no notch at all when they have a center tap and you hook them up push-pull, balanced to ground with the center tap.