Audio trafo winding help

[CJ]

6SJ7 close to EF86 also,

glass is rare but better,

watch pin 1 on the metal jobs, it is connected to the steel can,

if it gets grounded and you mis wire the socket, could be trouble,

 

[emrr]

I haven't heard a glass 6SJ7 I liked as much as a metal one so far, but I've said that before. 

So, back on topic, mitsos had some questions about winding DCR, and we've thrown out various examples for study.  I guess we are waiting on his wind results?

 

[mitsos]

which wind results did you want? I posted what I could get from my LCR in a previous post.. I'm going to hook it up to a 2520 this week and let you know if anything blows up  ;D 

I did make one mistake though: no shields between the pri-sec sections  :-[ .  I meant to include them, but just plain forgot.

I have to say, this exercise is teaching me a lot about trafo design, and why the good ones cost so much money. I really want a CNC winder now, but unfortunately have no reason to get one.

Now that it's done though, I'm actually thinking of going in a related but slightly different direction with this thing, but I think it'll be better to make a diagram for it instead of trying to explain it, the way I tend to speak/write, it's probably best!

 

[mitsos]

OK, finally got the thing fully put together (the trafo not the test circuit) and have some more complete readings.

All readings on primary unless otherwise stated.

Pri L = getting between 120H and 123H 
25mH (leakage L?) with secondary shorted.
DCR 1,392 Ohms

capacitance measured across primary (someone please explain these?)
37.7 pF  (secondary open)
139 pF  (secondary shorted)

Capacitance measured between start of Pri and start of Sec (please explain if this means anything?)
925pF (other leads open)
580 pF (finish of each winding shorted together)

Not sure but these capacitances seem very high, but I just measured capacitance between two starts on a quadfilar API output and got 6n4, so quite a bit higher. But this is wound quadfilar so it makes sense that there be a lot more capacitance between windings, or am I confused? Wouldn't be the first time.


Maybe I overdid it on turns. I'll try again but first I'll listen to this one.  She's a pretty little thing. Looking from the side looks like all nickel cuz the steel lams are M4 (about 10 mils) vs 14 mils for the nickel.  Nice and shiny, but you can't really tell from the pic.

Thanks for the tips/comments guys. Here's a pic:

 

[lassoharp]

Looks good!

 

[CJ]

:-*

 

[lassoharp]

I'll bet you my Firestone, Aristocrat, and genuine AC Delco metal SJ7s cost less than those monkeys.

And my RCA Double "Sub-Standard" SJ7 is the A-bomb.  I won't build without one.  :-X


The JJ stuff doesn't rust as well as the old ones. 

 

[CJ]

you can also cook hotcakes on a good pair of 6l6 metals

 

[mitsos]

I'd take some hotcakes, not cooked on a tube though, probably full of lead.

here's some charts. Done with fuzzmeasure using an mbox 2 "pro" firewire guy.  First like is the interface on it's own, output to input, then the transformer by itself (1-1.5dB spike from 10K on up, probably keeps going and going, but no resolution up there.  last line is the new trafo hooked up to the input of the 312ish solid state pultec amp.  Goes 10K:10K (through a 1uF cap, forgot about that guy) to the 2520 opamp, out to an "hs56" prototype.

comments?

 

[CJ]

that looks pretty good,

drive some headphones and see how it sounds,

 

[mitsos]

haha, talk about timing!  I was just coming back here to say I ran some audio through it!  It sounds very nice indeed, but I need to listen to it some more (and on better phones or monitors). But there was little if any difference between audio through the amp and not through the amp (I already know the rest of the amp is OK,  the little wrinkle below 20Hz is from the 2520 and output trafo, goes away with a 47R in || with the output cap, just like in 312s with a 2503 output.  The little 1dB rise above 10KHz is the input trafo, not sure where that's coming from, self resonance I guess.  Need to cut the capacitance down, or is it the R? Or both?

hmmm, calculating from the 37.7pf and 120H, the resonating F would be 2367Hz...  that kinda sucks.. 

checked the math, lowering either will raise the Frequency... inductance will go to 80 but that won't make much of a difference, so dropping the C is the way to go? Pie wind? not sure how that'll come out.. Is there any way to predict where C will be with the lower turns? (Say 3500 or so, the current one is 4268).

thanks!

 

[PRR]

> the 37.7pf and 120H, the resonating F would be 2367Hz

The 37pF does not resonate with the 120H. (Re-peek the equivalent circuit.)

You want Leakage Inductance. If all your windings are one lump, this will be under 1H. That puts resonance out at 24KHz.

Capacitance changes little with number of turns. It is more about the total surface area of the winding.

 

[CJ]

computing leakage C is so complex that even the engineers say that it is at best, a hit and miss calculation,
in other words, they just sweep the xfmr and forget the math,

digging up a few more dissections, some with pics still there,

Fischer 6BQ5 Output  killer for guitar amp

http://www.groupdiy.com/index.php?topic=35998.msg446313#msg446313

Ouncer O-9

http://www.groupdiy.com/index.php?topic=15050.msg174987#msg174987

McCurdy Input MRI 2001

http://www.groupdiy.com/index.php?topic=19156.msg224082#msg224082

Newcomb TR-91 Input

http://www.groupdiy.com/index.php?topic=36473.msg448188#msg448188

API 2621 Input

http://www.groupdiy.com/index.php?topic=13460.msg156169#msg156169

API  2622

http://www.groupdiy.com/index.php?topic=818.40

Sescom  MI 46

http://www.groupdiy.com/index.php?topic=36482.msg448289#msg448289

CTC  TA-10

http://www.groupdiy.com/index.php?topic=36407.msg447326#msg447326

UTC HA-100X  LA2a Input

http://www.groupdiy.com/index.php?topic=35743.0

Telefunken V-72  Output

http://www.groupdiy.com/index.php?topic=36632.msg450015#msg450015

Kenyon T-6 Input

http://www.groupdiy.com/index.php?topic=36383.msg447020#msg447020

 

[mitsos]

> the 37.7pf and 120H, the resonating F would be 2367Hz

The 37pF does not resonate with the 120H. (Re-peek the equivalent circuit.)

You want Leakage Inductance. If all your windings are one lump, this will be under 1H. That puts resonance out at 24KHz.

The windings are divided 3:2, P-S-P-S-P. Granted this was done on a very simple (junkyard parts) winder, just a DC motor with foot switch-pot, and a counter made from a modified encoder, but no tensioner, so tension and left-right movement was manual. Little insulation between layers and, as I said, I forgot to include shields. 

Is Pri leakage L with the secondary shorted?  this was 25mH. So this resonates with the 37pF? If so, that puts us way above audio band, unless I made a mistake converting to Farads, 164KHz.  less worrying.

Capacitance changes little with number of turns. It is more about the total surface area of the winding.

I have a hard time visualizing the surface area of these spirally things, but I think I get it.

computing leakage C is so complex that even the engineers say that it is at best, a hit and miss calculation,
in other words, they just sweep the xfmr and forget the math,

Here is the graph I posted above but showing just the trafo by itself, and with a different vertical scale (set automatically by the program).  The low end looks nice, 0.1dB down at 20Hz.  But the 1dB rise (at 20KHz) starting at about 3K5 looks much worse like this.  Is there a way to even this out without resorting to zobels?

thanks again

 

[emrr]

That looks a lot like many many many vintage transformers, that people know and love. 

 

[mitsos]

That looks a lot like many many many vintage transformers, that people know and love.

  Well that's mighty nice of you to say sir!!  Made my day!  ;D  That's good enough in my book, but I'd still like to know what kind of tricks can be used to reduce this kind of thing.  will post back with results from further experiments.

 

[emrr]

never hurts to improve!

 

[CJ]

rewind with a couple of shields and that res curve will move up to 40 k hz,

just a prediction but give it a try,

you do not need copper foil, just use magnet wire and leave one end open,

plot the low end until the xfmr craps out, we like to see where it drops 3 db,

2 hz?  4 hz?  16 hz?

you can trade some bass for a smaller core, this would lower the leakage C.

use smaller wire,

or try sec-pri-pri-sec

or try pri-sec-pri,  use two windings in parallel for each pri, one on inside of coil, one on outside, this will improve your leakage L.

most 600/600 transformers are about 1200 turns for pri and sec, 2 shields, 10 to 30 henries, and flat to 80 k.

you need a hi-octane core for this, only want 2 or 3 layers per section,

hmmm.

.

.
.

.

.
OK, i got it,

for a 600:600,

wind 960 turns bi-fi on that core,

120 turns per layer, 8 layers,

no need for layering, just pri and sec wires getting wound side by side, at the same time of course,

so "pull two"  as they say in the xfmr biz,

you should have much lower leakage, in the micro henries, and there will be no bump, dcr should be about 65 ohms each if you use #33. maybe 18 henries with the juiced up core,

should be flat from dc to daylight,

 

[lassoharp]

. . with an Edcor 15K:600 and it worked fine, measured fine.

Guess it would probably be a good idea to clarify which one since they now offer the XPP -1 15K:600 "PP tube OT".

I had assumed the XSM in my earlier response.

 

[mitsos]

Hi CJ, been reading you post over and over, to make sure I understood what you're saying.

rewind with a couple of shields and that res curve will move up to 40 k hz,
just a prediction but give it a try,
you do not need copper foil, just use magnet wire and leave one end open,

I asked the same friend to wind me an almost identical trafo (less turns of thicker wire being the difference) with shields, and winding only L->R, cuz the last one was wound back and forth.  I also asked him to wind me a 10K:600, using the same turns as the new 10K:10K, don't have the notes in front of me, but it was a bit less than 3500 turns.  600 Ohm secondary was around 800-900.  Maybe I already posted it, if not I'll post it all when I test the trafos and sit down here with my notebook.

plot the low end until the xfmr craps out, we like to see where it drops 3 db,

will do

use smaller wire,
or try sec-pri-pri-sec

All of them are P-S-P-S-P  so far cuz they are all tons of turns.  On the first we used #40, on these new ones the 10K is #38, and the 600 is either #33 or #34. Again, I'll repost the exact early the following week (I'm picking these up next weekend).

or try pri-sec-pri,  use two windings in parallel for each pri, one on inside of coil, one on outside, this will improve your leakage L.

I'm a bit confused by the part in bold. Do you mean something like this:  P1a, P2a, S, P2b, P2a...  Put P1a in parallel with P1b, and P2a in parallel with P2b and then put the doubled coils in series to get the full winding?

OK, i got it,

for a 600:600,

wind 960 turns bi-fi on that core,

120 turns per layer, 8 layers,

no need for layering, just pri and sec wires getting wound side by side, at the same time of course,

so "pull two"  as they say in the xfmr biz,

you should have much lower leakage, in the micro henries, and there will be no bump, dcr should be about 65 ohms each if you use #33. maybe 18 henries with the juiced up core,

should be flat from dc to daylight,

will try to convince him to do this one too, though I don't know if it'll make it in this batch, not sure how his schedule it.

As I said before, I need more nickel lams.  (and a coil winder at home...). 

thanks again for all the comments everyone!