Transformer Distortion Project

[D_Ex_Patria]

I'm new to effect building, having only made a few 386-based amps and effects over the past few months, so please be patient.

While reading about power amp emulation, I got a little hung up on transformer saturation. I fished a 1:1 modem transformer out of a digital TV box someone was throwing out and banged it across the output of one of my amps, and it the sounds that came out of it were pretty much what I expected: some dB loss, and some filtering outside of the transistor's frequency range (not altogether unpleasant).

More reading turned up a similar idea in a pedal:
https://www.talkbass.com/threads/an-inexpensive-audio-transformer-with-diy-projects.1113802/page-17

I looked at their numbers (throwing 12Vp-p through their transformer, a TY-250P), and mine (10Vp-p if you run a 386 at 12V, through my Midcom 671-8001), and thought it might be possible to do a minimal version of theirs.

Here's my thinking so far:



I left out all the bits and bobs for stabilizing the 386 in this diagram, so assume all those components are in place. Since I'm losing some dBs through the transformer, I thought I'd just toss another power amp on after it, with some attenuation to keep its output unclipped. I have a breadboarded bridged LM380 that's just waiting to be hooked up for that purpose, but I thought I'd share and see what thoughts people had about the idea while I got to tinkering.

[Rob Strand]

You need to put a cap in series with the transformer otherwise a large DC current will flow.    (Even if you have a dual rail supply it is a good idea to use a cap when connecting to a transformer.)

There was an commercial preamp around in the 80's which was very popular for bass players - the Tapco IVP preamp.  Back in the day it sounded good to me too.  It used a two transistor differential amp driving the transformer differentially (one extra transistor for the current source for the diff-amp.)

Here's the schematic!
https://projectivp.wordpress.com/

[D_Ex_Patria]

You need to put a cap in series with the transformer otherwise a large DC current will flow.    (Even if you have a dual rail supply it is a good idea to use a cap when connecting to a transformer.)

There was an commercial preamp around in the 80's which was very popular for bass players - the Tapco IVP preamp.  Back in the day it sounded good to me too.  It used a two transistor differential amp driving the transformer differentially (one extra transistor for the current source for the diff-amp.)

Here's the schematic!
https://projectivp.wordpress.com/

Thanks! I've got a 470uF cap sitting after the 386's output as per the design here: http://mylkstuff.com/page35.htm

It's sitting before the transformer. Would that suffice?

[Rob Strand]

It's sitting before the transformer. Would that suffice?

Should be fine.  (Could try tweaking the value once you get it working.)

[anotherjim]

It looks as though you are on the right track.
I for one am interesting in seeing/hearing how it works out.

I've found it takes a lot more drive using these kinds of "line transformer" to get it noticeably saturating. This with 9v power. Too high coil impedance probably? Bridge drive could do better?

I've got some small audio xfmr's, meant I think for either speaker matching or power-amp to transmission line, which might be easier to saturate driving their low impedance winding. One day....

[D_Ex_Patria]

It looks as though you are on the right track.
I for one am interesting in seeing/hearing how it works out.

I've found it takes a lot more drive using these kinds of "line transformer" to get it noticeably saturating. This with 9v power. Too high coil impedance probably? Bridge drive could do better?

I've got some small audio xfmr's, meant I think for either speaker matching or power-amp to transmission line, which might be easier to saturate driving their low impedance winding. One day....

I have so many questions.

I dug up the datasheet for my transformer:

http://www.datasheetarchive.com/MIDCOM%20671-8001-datasheet.html

If I've got my math right, when it's connected, I'm sending 10Vp-p over 600Ohms of impedance for the primary winding, which should give me 16mA. The best number I can fish out regarding saturation from the datasheet is 100mADC max  for D.C. in PRI. Does this mean it will handle up to 100mADC without saturation?

If so, you're right, I need another transformer, probably the TY-250, otherwise I will get no saturation, only attenuation.

[anotherjim]

My knowledge falls short.
I would not have thought you'd want any DC in a winding, so that spec' could well be a way of describing saturation level. I honestly don't know.
A few articles on audio transformers.
http://sound.whsites.net/articles/audio-xfmrs.htm
Find AN008 at the following.
http://www.jensen-transformers.com/application-notes/

... none of which is all that interested in making a transformer distort.
And I don't see much on saturation specs, DC or otherwise.

[ashcat_lt]

I've had trunks finding anything like a transfer curve for transformer saturation.  Does anybody know where I can see that?  Like, how does it compare to other forms of distortion like diode clipping or tube saturation?

[anotherjim]

The hysteresis curve, like this...


As for drive level  -
Jenson audio transformer specs often say "used at up to 20dBu". That about 22v p-p isn't it? Presumably over that, distortion become significant?

[ashcat_lt]

The hysteresis curve, like this...

So the curve is different depending on which direction the signal is going?!?  That means it's time dependent?  How much time?  Is there text somewhere to go with that picture?

[Mark Hammer]

Actually, discussion about transformers in pedals began here a number of years ago, largely in response to the many amp emulators of the type you see at runoffgroove.  The discussion was essentially "Well, it may be possible/reasonable to try and sub FETs for tubes in the preamp of a given amp, but we are missing what the power stage contributes".

[D_Ex_Patria]

A few tidbits I picked up in my web lurking and from a friend that works mostly with power transformers (he had nothing to say about audio directly, but had some input).

• Distortion happens most easily in the low end, and disappears quickly as frequencies get higher (which could explain why bass players dig it).
• The distortion is characterized by odd harmonics.

Thanks anotherjim for the links; there's some bits in there about cap sizing and boosting bass frequencies, which could come in VERY handy once I start fine tuning that cap, as Rob suggested.

In any case, I should be wiring it up tonight, and maybe recording some audio in the sliver of time I have tomorrow morning between the  wife leaving for work and the start of my commute.

[anotherjim]

That hysteresis curve happens because every half cycle leaves residual magnetism in the core.
As I understand it...
Going positive, the core is magnetised to a polarity.
Going negative, it must first de-magnetise (overcome) the residual before it can re-magnetise to the opposite polarity.
However, although hysteresis is real, that diagram maybe exaggerates the shape for clarity.

I snatched the first good looking curve I saw, the relevant page is here...
http://electronics.stackexchange.com/questions/177863/can-anyone-explain-why-a-b-h-curve-with-hysteresis-doesnt-result-in-a-highly-no 
... which is quite interesting.

You'll also have seen at Jenson that the basic, cheap, silicon steel cores have x10 more distortion than the fancy nickel alloy in high-end xfmrs at low signal level. But the silicon THD is still only 1% and that falls off with rising frequency. Then all type cores leap into distortion over +20dBu.

[ashcat_lt]

That link actually ends up asking the same question I had.  That curve shows how the magnetic flux changes,  but what does that actually mean in terms of Voltage? 

In fact, I have this frustration with tubes and diodes too.  Try to find an actual Vin>Vout transfer curve for any of those things.  You can't.  You can find a V/I curve for a diode, and a "load line" whatever thing for a tube, but none of that really answers the only question we ever care about in real world audio work.  Ours is a voltage world. As long as things don't burn up, we really don't care what happens with current except because of how it might manifest in a voltage change.  Likewise, I really don't care about the instantaneous flux density around the transformer.  How does it affect the voltage on the other end?  That's all I want to know.

And yes, I understand that it's never that simple.  The answer to my question is always that it depends on everything else in the circuit, but can we at least narrow it down to some "typical" application?

Sorry for the rant.  I hope it's not too far off topic.

[PRR]

Transformer distortion is low if source impedance is low.

Little loudspeaker amps are very low output impedance.

You got a "modem" transformer.
http://datasheet.datasheetarchive.com/originals/scans/Scans-057/DSA2IH003337.pdf

Throw 470r or 1K in series with the primary, so it has some slack and distortion happens.

As an extreme abuse, ignore Rob's very-excellent advice "...put a cap in series with the transformer" and let the DC flow. The '386 will run warm. The transformer's DC resistance (~~108 Ohms) at 6V will be less than the transformer's DC mA rating (100mADC). The transformer iron will approach saturation. Now audio may come out even more bent.

[D_Ex_Patria]

Thanks all! I'm a little overwhelmed with info. I'm still trying to bend my head around the science.

On the practical side of things, I wired it up moments ago while the wife was sleeping.



The good news is: something kind of like my bass guitar is making it to the speaker. This is promising. The bad news is, the circuit is super noisy, and goes from zero to 11 VERY quickly. Just touching a string too hard makes it bark.

A lot of this is probably wires being wires, but there are also some soldered components in the 386 portion that might be contributing. There's a zobel network in there that probably shouldn't be, and the whole thing is a rat's nest.

I'll try again tomorrow morning when I can make more noise without waking the wife.

Then it's off to buy components for breadboarding another 386 and trying all your suggestions.

[D_Ex_Patria]

Digging into the Whitlock chapter as we speak.

Now, if I'm putting that cap in between the 386 and the transformer, will it behave like a highpass in combination with the resistance of the transformer, like the output cap and the speaker on most 386 projects? If so, is it the DC resistance, the impedance, or some other variable like load that dictates the cutoff?

[anotherjim]

If you are lucky and you know the DC resistance (or measure it) and the inductance (Henries), then you have a good chance of working out frequency response.
http://sim.okawa-denshi.jp/en/RLChikeisan.htm
Actually, I can't find one that includes series R of the coil - so just add it to the source R?
Yes, it's a high pass.
Telecom bandwidth is only 400Hz to 4Khz, so a modem xfmr only has to meet that, but could well exceed it in practice - what I'm trying to say is that you may not be able to stop it being bass light.

I had all the magnetics stuff drummed into me about 35years ago. Back then, you maybe didn't just build PSU's (no wall warts), you might have had to wind the transformer from a kit, maybe some chokes too. A component catalogue was about 1/5 magnetics - cores, formers, enamelled wire etc; with handy inductance formulae filling out the white space.
Now I don't remember so much of all that -  never had to use it.

I like Paul's idea of injecting DC. Also what I read does indicate high source resistance raises distortion. Have even thought of glueing a magnet on for partial saturation.
Maybe, if DC is there, and it drops quickly on power off or rises quickly on power on, it could generate a significant voltage spike out of the secondary? I've been shocked by this before.

[D_Ex_Patria]

When I got a chance to make some noise this morning, I felt a little better about the sound I was getting. It's gainy and a has a lot more top end than I expected. There's a weird fizz-out happening where I was hoping for sustain, which I have noticed with a few of my 386 builds, which i think is exacerbated by the second power amp. I recorded a representative example below (tiny crappy speaker(tm) was used for testing purposes, so don't expect much in the tone department).

https://soundcloud.com/hogfiend/transformer-bass-note-fizzle

I'm going to start working with a breadboarded 386 and see if I can work out some issues there. I'm using the LM386N-4 right now. I'll try swapping for an NJM386D and see if that makes a difference. Plus there's all the output resistors and caps I can try out.

I also gave the Rod Elliott article a good read, and it answered a LOT of the questions I had last night, so thanks for posting that, anotherjim.

From my understanding, I really need to find out the load of the transformer, and treat it as an LC or RL circuit, instead of an RC, like I had assumed. The article had a method for finding that, which involves a signal generator, which I don't have (maybe my phone? I'll look into it).

[D_Ex_Patria]

That RLC calculator is super helpful! 

I'm played around with some of the Henries values in the article (~2-5) and some of the cap and resistor values I've been considering (2.2uf-10uf, 200-1000ohms), and they all seem to fall in the low end area (20-75Hz). Dropping to 100nf brings it up into the 300-400Hz range, which might bring it into the transformers functional range. Off to the electronics shop!