Audio trafo winding help

[mitsos]

I am trying (planning to) to wind a 10K:10K audio transformer for use as input on my PRR176.

In keeping with the PRR varimu mentality of using what you have lying around, I have EI 625 lams in M4 grain oriented steel (a bit thinner than the usual 29 gauge M6 used in the US), as well as about 3/4" of 49% nickel I got as a sample.  I planned to make a 5/8 stack because anything larger won't hit (no low-profile bobbins to be had locally). It has to fit inside a 1U case.  I started calculating things yesterday but they don't make sense, see below, and my comparison to Edcor at the end to understand my dilemma!

From Ian's inductance thread, I got the rule of thumb Primary Inductance = 10K / 100 = 100 H.  From the reactance formula (XL = 2piFL), if I want to go down to 20Hz, I found I need at least 80H, so anywhere between those two is perfect for me.

I have no idea what the permability of the lams is, so I wound 100 turns on a bobbin and measured the following (interleaved 1x1):
M4 = 21mH
49% = 69mH
M4 mixed with 49% = 41mH

From what I've read online and elsewhere, L varies with the square of turns, like the AL number for inductor cores.  So, the 41mH of the mixed core is related to  turns squared which is 100x100 = 10000.  Using this number, I can set up a proportion to find out the number of turns I need. At least I think that's how it works.

So if I set up the relationship as: (0.041H / 80H) = (10000 / X)

X being the turns squared of the coil that will give me 80H.  Rearranging and solving for X gives me 19,512,195.1

Again, this is "turns squared", so, SQRT(19,512,195.1) = 4,417.21 turns.   

If this is correct so far (that's a lot of turns), going to the tables in the back of Wolpert's Audio Transformer Design Manual, tells me that to get 4417 x 2 (two windings) = 8834 turns, I need at least 40 AWG wire (max turns for square stack using 39 AWG = 8700, so it would just barely not fit, 40 AWG goes to 10,670 max turns)

OK, so even if I can fit all these turns, that leaves me with a DCR of 3377 (for 10670 turns).  Wolpert gives a mean length turn of 3.62", we can use this to estimate and subtract the DCR of the extra wire since I'll only be using 8834 turns:
10670-8834 = 1836 turns (times MLT) x 3.62" => 6646 inches = 553.86 feet. 

40AWG has 1079 Ohms nominal DCR.  I think this is for 1000 feet(?) so [(553.85/1000) * 1079] gives us about 597 Ohms for the extra 553.86 feet.

OK, so 3377 Ohms (for 10670 turns) - 597 Ohms = 2780 Ohms (for both windings) => 1390 Ohms per winding. hmm, that doesn't seem as bad BUT....


that said, I'm confused. I figured that, by mixing in a bit of nickel, I'd be able to use thicker wire or less turns than Edcor, who use M6 only, but then I don't know how they have only a 262 Ohms DCR, and fairly empty looking bobbins (my turns calculated above would pretty much fill the bobbin I'm sure).    Well, looking at their PDF again, they list Pri L as 8H, which I had assumed was a typo (missing a zero), because they list the frequency response as +/-1dB,  20-20KHz.  How do they get this frequency response out of a 10K trafo with 8H (8H would give 10K reactance down to 199Hz only)?

Am I completely off with these calculations?  The whole winding experience is mostly educational, but I would like to get a decent transformer out of it as well, not just waste my friend's time (he's the one with the simple winder), so if you guys can go over my numbers and maybe suggest changes, I'd appreciate it.

Thank you!

 

[zebra50]

Hi Mitsos,

I've been on a similar mission myself of late, although looking at mic transformers, and my first thought was "you're going to need a bigger core". But...

...one weird thing that happens with laminated cores is that the inductance goes up as the frequency goes down,  so you may get a different value for AL at 100 Hz than you do at 1kHz. Do you know at what frequency your meter is measuring?

I wondered if this might account for the odd specs on the Edcor, although I don't think a ten-fold increase is likely.

With my tests, I have 80% Ni and also 50% Ni lams. The 80% Ni have about 3 times the inductance, but cost nearly 3 times as much!

I looked at your calcs and I couldn't see any problems (although I am by no means the theory expert!) Sooner or later you have to start winding something close to your target and tweak from there.

On an empirical level, make sure your stack is nice and tight if you want to maximise inductance - get as many in your bobbin as you can.

With my tests, I have 80% Ni and also 50% Ni lams. The 80% Ni have about 3 times the inductance, but cost nearly 3 times as much. It does mean that you can get a higher value from the same bobbin, or use thicker wire (and lower DCR) to achieve the same inductance.

Also I found that if I wax pot the inductor/transformer, then the inductance can go up by 10 to 15%.

Finally, it is certainly possible to make good sounding transformers yourself. Good luck!

Stewart

 

[emrr]

We were just talking about the Sowter RCA 86A interstage in another thread, which is advertised as 20K:10K+10K, and it has windings approximately 350R each.  I just got a set of their Altec 436 bridging inputs (15K:80K) and recall nothing higher than 660R when I checked them. 

Old rule of thumb says you should have 1K or less for a 10K winding.  Exceptions being some Western Electric, Langevin, and RCA iron that has DCR approaching 50% of nominal Z. 

Lassoharp has an L meter, and a few of my spare Edcor 15K:600; let's ask what he measures. 

 

[bezen4uk]

Hi Mitsos,
You have nice materials lying around 
Your calculations and measurements are correct.
I'd recommend to change targeting from L to DCR.
I found it's desirable to keep total DCR (pri. DCR + reflected sec. DCR)  5-8% of nominal load. You can find this rule applies for many commercial transformers. I tried to wind trafos with higher DCR/load ratio - they sound bad.
You can start with 2x2000 turns AWG36. If the bobbin is too empty, try again with higher turns or thicker wire. There is bobbin filling ratio (overall copper cros-section to core window cross-section ratio) - it's empirical ratio and depends of your winding skill. I'd recommend to calculate and remember this ratio - then you can more precise predict bobbin filling for the next trafo design.

 

[PRR]

> How do they get this frequency response out of a 10K trafo with 8H

I'm just guessing:

Drive it from a 600 ohm signal generator.

Edcor's background is 1970s-onward commercial sound, where everything is low-Z output.

Also... 100H is a LOT of inductance. As you found, a small core needs way too many turns of way too thin wire and thus a DCR which offsets much of what you had hoped to gain from a high inductance.

Without doing math, I'm guessing that 100H with reasonable DCR needs a fancy (Permalloy) core or a core bigger than a 1u rack.

AND that if you do the big core to 100H, the leakage inductance versus all that capacitance will put the upper resonance down in the audio range.

> use as input

So does it really need to be 10K? In my Comp5 plan the source would typically be able to drive 600 ohms. The "PRR176" has input buffers which I have not studied but can surely drive (or be made to drive) 600 ohms. 10K input is convenient when running a 600 ohm bus line to multiple recorders and monitors, but that's uncommon.

And do we really care about DCR? You are driving naked grids. "Infinite" resistance with some capacitance. Any winding bigger than your thumb has more capacitance than the grid. And winding resistance is a (two-edged) tool in taming the upper resonance.

Get a 230CT:230CT power tranny, try it. Inductance may be fine. If not, re-wind with double the turns. Leakage inductance is high, re-wind quadfilar.

 

[emrr]

> How do they get this frequency response out of a 10K trafo with 8H

I'm just guessing:

Drive it from a 600 ohm signal generator.

Edcor's background is 1970s-onward commercial sound, where everything is low-Z output.

I went there in my post, then I took it back out, because I once replaced a dead A-24 type (15K:600) that was plate driven by a 6C5 triode with an Edcor 15K:600 and it worked fine, measured fine. 

 

[lassoharp]

Lassoharp has an L meter, and a few of my spare Edcor 15K:600; let's ask what he measures.

Coulda shoulda woulda - unfortunately didn't document when loose.  Both are in circuit now.  With two 6cg7 plates on primary I'm reading 4.7H @ 120Hz but that doesn't sound correct.  I did briefly hook the meter up when loose but didn't write it up.  Seems like maybe in 40s? across entire primary.

I once replaced a dead A-24 type (15K:600) that was plate driven by a 6C5 triode with an Edcor 15K:600 and it worked fine, measured fine.

That's more good confirmation.  Someone I spoke with at Edcor was kind of hesitant to say "ok" to their use as PP tube OTs and qualified them passable for small tubes, low current.  He may have been thinking I was asking if it was ok to run at the max 2W rating.  I've got them in two units using small triode drivers with balance pots and they sound very good.  Guess I will find out later on if there's any long term consequences to that application.

 

[mitsos]

wow! I wasn't expecting so many replies so quick! Thanks!  Saturdays off for all of us!

OK, I'm glad to hear my numbers aren't way off, cuz I called my friend after I calculated the turns, and we decided to do 22 layers  x 194 turns of 40 AWG for each winding, and interleave Pri-Sec-Pri-Sec-Pri.  For the primary, since 22 can't be divided by 3 evenly, we did 7-8-7 layers.  I'm pretty sure the DCR is around 1K5, but I won't be able to test it until tomorrow.

I've been on a similar mission myself of late, although looking at mic transformers, and my first thought was "you're going to need a bigger core".

That's what I thought too, but I went with the old rule of thumb "look at a existing similar trafo and go from there"
which is why I started with the Edcor (I think it 5/8 x 1/2" height, but I went with 5/8" height because that's the bobbin I had).  Plus, what about all those little 10K:10K input transformers, the size of an API input? They can't possibly handle THAT much level, can they?

...one weird thing that happens with laminated cores is that the inductance goes up as the frequency goes down,  so you may get a different value for AL at 100 Hz than you do at 1kHz. Do you know at what frequency your meter is measuring?

I wondered if this might account for the odd specs on the Edcor, although I don't think a ten-fold increase is likely.

pretty sure it's 1KHz, it's an old BK 875A. FWIW, I measured the 2.5W Edcor 10K:600 from my poorman 660, and while they are speced at 40H pri L, I got 22, so maybe my meter's battery is dying, but it still measures the same on my handwound toroid inductor though (which when in circuit, with the properly calculated capacitors, results in spot-on frequency boosts, so I'm not sure it's way off, but it's possible).

Old rule of thumb says you should have 1K or less for a 10K winding.  Exceptions being some Western Electric, Langevin, and RCA iron that has DCR approaching 50% of nominal Z. 

Good to know. I also saw an old post by PRR telling NYD something like 5-20% of Z. I may have to up the core to 3/4" height and try to get the wire size to 38 or so, if I keep this winding method.

Hi Mitsos,
You have nice materials lying around 

no so much, just a little bit I've collected over time, but I need to order more nickel, just need to find a place with low enough minimums!

I'd recommend to change targeting from L to DCR.
I found it's desirable to keep total DCR (pri. DCR + reflected sec. DCR)  5-8% of nominal load. You can find this rule applies for many commercial transformers. I tried to wind trafos with higher DCR/load ratio - they sound bad.
You can start with 2x2000 turns AWG36. If the bobbin is too empty, try again with higher turns or thicker wire. There is bobbin filling ratio (overall copper cros-section to core window cross-section ratio) - it's empirical ratio and depends of your winding skill. I'd recommend to calculate and remember this ratio - then you can more precise predict bobbin filling for the next trafo design.

I like that! Seems a more practical way of dealing with this. At least according to walcott, we can get 4200 turns of 36AWG on a 5/8 x 5/8" core. But I don't think we'll be able to. The bobbin my friend wound (I have yet to test it cuz he simply wound it but did not finish the leads, and I have no soldering iron here (at in law's for the weekend) to finish the leads and hook up to the LCR!).  But I don't think we'll be able to get that kind of fill, judging by walcotts numbers for 40AWG (should be max 10670 turns) and since the bobbin is pretty much full with 8800 turns.  But we could try 37 or 38AWG.  But I'll see how this one works out.

Also... 100H is a LOT of inductance. As you found, a small core needs way too many turns of way too thin wire and thus a DCR which offsets much of what you had hoped to gain from a high inductance.

Without doing math, I'm guessing that 100H with reasonable DCR needs a fancy (Permalloy) core or a core bigger than a 1u rack.

AND that if you do the big core to 100H, the leakage inductance versus all that capacitance will put the upper resonance down in the audio range.

> use as input

So does it really need to be 10K? In my Comp5 plan the source would typically be able to drive 600 ohms. The "PRR176" has input buffers which I have not studied but can surely drive (or be made to drive) 600 ohms. 10K input is convenient when running a 600 ohm bus line to multiple recorders and monitors, but that's uncommon.

glad you stopped by,  since this is specifically meant for that mutant 176 compressor with your sidechain.  I do not plan on using any line receiver ICs on input, but I am considering rewiring the input discrete opamp as a balanced unity gain line receiver, driving the trafo.  So it MAY go XLR --> 2520 opamp -->attenuator/volume pot --> transformer --> 6bc8 grids.  (Gotthe idea from Kingston, to give credit where it's due).  In this case, I don't think the trafo impedance needs to be so high, but don't the grids want to see a high impedance? 

The other plan is removing all active electronics from the input, and wiring the trafo direct to the input XLRs.  In that case I figured 10K was a good starting point, and I figured a tube grid would want to see a highish impedance, so it became 10K:10K.  An other plan in this scenario could be to go 10K:600 and use the trafo as part of the attenuation circuit, but I don't know if the grid will like the 600?

And do we really care about DCR? You are driving naked grids. "Infinite" resistance with some capacitance. Any winding bigger than your thumb has more capacitance than the grid. And winding resistance is a (two-edged) tool in taming the upper resonance.

You mean an implied RC filter?  Can you elaborate a bit on this?  I know more capacitance will lead to HF loss, would the series resistance be in parallel with this capacitance and cause a HF boost? Or something else entirely? I get a bit lost with the trafo equivalent circuit models.

My main goal was to have a high-ish input impedance, in order to not have issues with anything driving it, but I don't realistically expect to have issues with any of the stuff I would ever use it with.

And maybe I don't really need 80H after all, or 10K, just going by what the math says.  The coil looks nice though!

 

[PRR]

> inductance goes up as the frequency goes down

It's semi-constant from sub-audio to about 200H, then drops, but slower than frequency.

The "200Hz" depends how thin the iron is sliced (lam thickness).

This isn't very important if your goal is <200Hz. But it does mock you when using 400hz or 1KHz excitation to judge inductance.

> speced at 40H pri L, I got 22

I found some old data showing 1:6 difference from very-small to small excitation (still >10:1 below saturation).

It's very frequency and level sensitive.

In another thread I advised "try it!" but that was rejected as "unscientific".

> 2520 opamp -->attenuator/volume pot --> transformer

Ack! Try it! That's wasteful and bad form. Transformers are at their best with ZERO (or negative!) source impedance; not a pot. Pot, opamp, transfo.

> don't the grids want to see a high impedance?

No. Lower is better. Mostly <10K is low-enuff for audio. However it is possible to drive the vari-Mu tube into grid current. Increased input headroom (often limited by other factors). Now source should be <1K. Arguably it would be nice to get <100r. However everything is a trade-off and super-low-Z grid sources have little upside and significant cost.

I dunno this hot-shot 6BC8. With 12AU7 the input levels are very much like console levels. Run -10VU for a light trim. Run +6VU for heavy smoosh.

> Exceptions being some Western Electric, Langevin, and RCA iron that has DCR approaching 50% of nominal Z.

I had Langevin iron with high DCR and the ratio did not work out. I opened it up and found two 4K7 resistors in series with the primary. Yup, the "bridging" impedance was above the rated 20K across the audio band and still 10K at DC; these would NOT load a 600r bus.

 

[CJ]

What are we building?

Bridging Transformer  10K:10K

Coil Geometry - lots of wire with a 10K winding, which means lots of winding capacitance. If you want to have balanced capacitance, you need something like

Pri Sec Sec Pri

or

Pri-Sec with a dual coil transformer.

forget the math, your stuck with the 625EI.

all you can vary is wire size.

i bet you do not have a UTC warehouse full of copper wire,

so take the smallest stuff you have and wind as much of it as you can squeeze into the window.

sweep it with the scope, slam it into the circuit and give it a listen,

use the 80 permalloy,

check the henries on the UTC LS-140 sometime.

or the Tamura stuff.

unreal for a 600 ohm job.

 

[mitsos]

> inductance goes up as the frequency goes down

It's semi-constant from sub-audio to about 200H, then drops, but slower than frequency.

The "200Hz" depends how thin the iron is sliced (lam thickness).

This isn't very important if your goal is <200Hz. But it does mock you when using 400hz or 1KHz excitation to judge inductance.

> speced at 40H pri L, I got 22

I found some old data showing 1:6 difference from very-small to small excitation (still >10:1 below saturation).

It's very frequency and level sensitive.

So maybe they are measuring at 100Hz, and I'm at 1KHz. I don't think they specify, but I'm pretty sure mine is at 1KHz.

In another thread I advised "try it!" but that was rejected as "unscientific".

Isn't that exactly how much of this vintage stuff was designed? Or at least, finally tweaked?

> 2520 opamp -->attenuator/volume pot --> transformer

Ack! Try it! That's wasteful and bad form. Transformers are at their best with ZERO (or negative!) source impedance; not a pot. Pot, opamp, transfo.

Oops!  Is there a good way to attenuate prior to a unity gain diff amp?

> don't the grids want to see a high impedance?

No. Lower is better. Mostly <10K is low-enuff for audio. However it is possible to drive the vari-Mu tube into grid current. Increased input headroom (often limited by other factors). Now source should be <1K. Arguably it would be nice to get <100r. However everything is a trade-off and super-low-Z grid sources have little upside and significant cost.

So then a 600:600 should be perfectly fine?  and with the opamp driving it, it should interface well with anything outside,  and from the point of view of the tube, it will be even better?  That changes things.. a 600:600 would be much easier to wind.  thanks!

 

[mitsos]

What are we building?

Bridging Transformer  10K:10K

Not sure anymore (see above post).  I was thinking 10K:10K as a modern input to a 6BC8 varimu.  But since it might get a differencial amp in front of it, and the tube would want a lower impedance anyway, maybe a 600:600 would be OK. So I'm not sure anymore, but I have one 10K:10K coil wound to test.

Coil Geometry - lots of wire with a 10K winding, which means lots of winding capacitance. If you want to have balanced capacitance, you need something like

Pri Sec Sec Pri

We ended up doing Pri-sec-pri-sec-pri to lower capacitance, not sure how it worked out yet.

forget the math, your stuck with the 625EI.

all you can vary is wire size.

i bet you do not have a UTC warehouse full of copper wire,

so take the smallest stuff you have and wind as much of it as you can squeeze into the window.

sweep it with the scope, slam it into the circuit and give it a listen,

use the 80 permalloy,

check the henries on the UTC LS-140 sometime.

or the Tamura stuff.

unreal for a 600 ohm job.

I used 40 AWG wire, went by the tables in back of the wolcott manual to guesstimate turns/fill. Sorta worked out, but my buddy hand tensions, so we would never get quite the amount of turns as wolcott says are possible, but we got enough. 22 layers of 194 per coil, 4268 turns per coil.

I only have enough nickel (I think it's 49%) for 1.5 trafos, so I planned to mix it 1:1 with the steel, which I have plenty of.

But I've got some tamura trafos, mic input and channel outputs, the inputs seem to have tons of hanries, haven't checked the outputs yet.  I'm planning on using the 1:3 input trafo backwards as the interstage, once more the "use what you've got" mentality.

This clears some things up.  I'm going to see how this thing works in some circuit, or just sweep it until the parts I don't have arrive.  I'll also try winding a 600:600 and see if I can compare. Should work out better, at least be easier: less turns, less DCR.

 

[emrr]

> Exceptions being some Western Electric, Langevin, and RCA iron that has DCR approaching 50% of nominal Z.

I had Langevin iron with high DCR and the ratio did not work out. I opened it up and found two 4K7 resistors in series with the primary. Yup, the "bridging" impedance was above the rated 20K across the audio band and still 10K at DC; these would NOT load a 600r bus.

Pardon if I remember wrong, but I think we went around on this side track before.  There are bridging transformers with resistances built in, but I'm not referring to those.  Those that have the R almost always show the R in the drawing on the side of the can. 

The famous $$$$ WE input transformer proclaimed as holy grail is something like 220R for rated 600 ohms, and there are multiple RCA 600 ohm preamp outputs that have very high DCR for rating, like 440 for 600 and 120 for 250.  CJ unwound one of those RCA for me, and it was just a lot of wire; to what end?  Here's where I wonder if there are times it's a purposeful loading mechanism built in by the designers.  Would doing so even out the impedance/frequency curve? 

 

[mitsos]

EMRR, do you have the winding specs of any of those RCAs?  I've "re-done" HS56's and they sound very nice, but those are only about 50 Ohms per 600 Ohm winding, or 8.33%.  I attached one RCA output CJ winding sheet I found, is that it?  120 Ohms DCR for 500 Ohm winding, that's 24%.

The famous $$$$ WE input transformer proclaimed as holy grail is something like 220R for rated 600 ohms, and there are multiple RCA 600 ohm preamp outputs that have very high DCR for rating, like 440 for 600 and 120 for 250.  CJ unwound one of those RCA for me, and it was just a lot of wire; to what end?  Here's where I wonder if there are times it's a purposeful loading mechanism built in by the designers.  Would doing so even out the impedance/frequency curve? 

I'm also intrigued by the possible use of DCR for setting up poles?  PRR hinted at this above as well.  I would assume that 220 Ohms means very thin wiring, which would allow lots of interleaving of windings, pie winding, interlayer insulation, all to lower capacitance? But if these old guys actually calculated specific R's and C's in the trafo equivalent model, then they could do lots of fun stuff I think.  And maybe that's the secret sauce that all these audiophiles are paying thousands for?

 

[mitsos]

RE: the 600:600 possibility (which will take a while cuz my friend lives 2 hours away now, and I won't see him for another couple of weeks), the same calculations as above, substituting 6H for the 80H, gives me SQRT  [(6H * 10000) / 0.041H] = 1,209 turns for a 600 Ohm winding.  Wolcott says I can fit 2784 turns of #34 wire, that would fill the bobbin to max, and give 219 Ohms DCR total for both windings, with about 110 extra feet of wire, or 26 Ohms extra DCR.  So ideal wire tensioning conditions would give about 190 Ohms, or 95 Ohms per winding.  I'm pretty sure we won't hit that with hand tensioning, so we'll probably be looking at 110 Ohms or so per winding, or 110/600 = 18.3%.

This is with hardly any insulation layers, so maybe go to #35, taking a further DCR hit and play with tape between layers, maybe pie winding, or other fun capacitance-lowering tricks.

 

[emrr]

Somehow that CJ chart still lives, it's wrong, he replaced it immediately way back when.  Yes, that's it, except 25k:250. 

 

[emrr]

In fact, you should take that RCA  chart out of your post so it doesn't propagate any further.

 

[lassoharp]

maybe that's the secret sauce that all these audiophiles are paying thousands for?

it was just a lot of wire

This is straying a little from the question emrr was looking to have answered but for the iron that CJ unwound there's a context to the high impedance primary that's somewhat interesting.  They are loading a triode like a voltage gain stage.  I haven't done the math or measurements yet but somewhere the lines for increased triode gain (higher pri Z) and step down loss intersect and the question is which one overcomes the other at what point.  I'm not sure how or if that applies to the numerous other (input, interstage) transformers with high DCR readings.  We may get some performance clues from the "86 interstage" thread where someone is using the Sowter repro.  Maybe someone out there has a Doc Hoyer rewind of that one or a BA6 IS who will chime in.

 

[lassoharp]

except 25k:250

So officially not 40-50K for either variant?

 

[mitsos]

Somehow that CJ chart still lives, it's wrong, he replaced it immediately way back when.  Yes, that's it, except 25k:250. 

haha, CJ posted that very file (somewhat) recently, in a thread about the UA 610 output trafo.  What is this trafo from?

I took the file down, do you mind posting the correct one? Is the only mistake the impedance numbers? Turns, core, etc is all correct?

@Lassoharp:  I'm not sure I follow what you're saying (I'm thick) so I'll await your further analysis!