Transformer coupled preamp

[tca]

I forgot this BJT example by Escobedo: ultra class A superdrive power amp. It works with a 2N7000!



Cheers.

P.S. (edit)

It can be used as a preamp, I've added a buffer to increase the input impedance and a  volume control. Ads a very pleasant  compression to the guitar signal.

[R.G.]

Curiosity piqued. Do you have cites/hints handy?

I'll go try to dig them out. I think I have the paper textbooks back in the boxes. I gathered this stuff over about 25 years trying to understand how to design that silly coupling/phase inverter transformer.

[tca]

Curiosity piqued. Do you have cites/hints handy?

I'll go try to dig them out. I think I have the paper textbooks back in the boxes. I gathered this stuff over about 25 years trying to understand how to design that silly coupling/phase inverter transformer.

Am I missing something or the Mullard Reference Manual of Transistor Circuits (1961) has "all" the details about this type of amps?

Note: PDF download at the left menu on that page.

[PRR]

> "all" the details

It may indeed be the best readily-available overview.

But it tries to do WAY too much for 313 pages and is mostly throwing crumbs or fully-baked treats.

The question of driver transformer design.... a simple formula for turns ratio, with the vital V on another page. But what inductance? How much iron? Does winding layout matter?

[R.G.]

The question of driver transformer design.... a simple formula for turns ratio, with the vital V on another page. But what inductance? How much iron? Does winding layout matter?

You've worked with signal transformers, I see. 

I have done everything except sacrifice a working driver transformer to the cause for unwinding. Someday I'll do that but I think I've re-designed it well enough.

Maybe.    The trick - I think - is to wind the primary and secondaries in multi-way-filar. I think it's a ratio of five to two primary to secondary, plus a gob of additional primary winding of about a third of one fifth. The plan is to make up a cable of magnet wire with nine strands in parallel, then wind the core, putting the gob on for the additional turns on the primary. Once it's wound with half the number of secondary turns, one series connects two strands to make one secondary, another two strands for the second secondary, and then five in series with the gob for the primary.

This gets you the turns, and vanishing low leakage; at least as low as it can be reasonably done. I don't think the original was done this way. It was probably a 3-2 interleave as a matter of economy, but I've found that the stability of the whole amp is marginal on testing with the original. A multifilar winding might kick the leakage inductance pole a lot higher and get better stability.

[teemuk]

On topic of transformers in preamps.

http://oi46.tinypic.com/1e19g5.jpg

[PRR]

> sacrifice a working driver transformer

You seem to be well along the road. For lurkers--

If a working part is available for study, obviously you can measure electric parameters.

You can also "see" how big the iron is. If there's a little space around the winding, you can add a few-turn winding. Then you can determine, not just turns-ratios, but turn-counts. With true counts for each winding, and an eyeball on winding-build size, some head-banging against a wire-table will tell you the wire gauges. Cross-check: the turn count and the approx mean length of turn should work out to about the observed DCR. You also need the air-gap, and this may be difficult.

> multi-way-filar

You can measure leakage inductance but back-figuring to interleave/filar is really hard work.

I do note that even crude plans often specify the two base windings shall be bifilar, so at least both bases get the same drive (opposite sense of course). The same is often true of tube class B amps, even lo-fi (deep B as in battery radio, not so much AB as in wall-power gear).

> plus a gob of additional primary

That's still leakage. If gob is 50% of total, it still dominates pri-sec leakage.

> a gob ...of about a third of one fifth

Lost here.... sounds like a stiff drink??

Secondaries should be distributed ALL through primary. However this increases stray capacitance. However I suspect your primary is well below 1K, so stray C is not a big deal. (For old Germanium Power devices... if the goal is modern "2N3055" then the bandwidth is a real issue.)

> at least as low as it can be reasonably done

Probably very true. But hmmmm....

I don't know the design. I will guess 4:1+1. (Or is it 5.33:1+1?) Impedance 400:25 (only one secondary fully loaded at a time). DCR should scale with impedance. (Unless DCR is used as part of bias, but this is bad/cheap practice and thermally wrong.) Such low impedances suggest fat wire. When wire gets too fat to go around the bends, it may be expedient to use strands in parallel.

Ah. This will give negigible leakage inducance though a %$#@! to wind. Figure your primary gauge and length. Make a 12-filar. (If the length is short, go around two nails until you have 12 strands; longer, get 12 spools and filar it through a polished hole while winding the bobbin.)

Wind until full. You have 12 Starts and 12 Finishes.

Grab any 4 Starts and any 4 Finishes, solder in parallel. Do it again. Here are two secondaries with the same turn-count as the 12-filar, but DCR 1/4 of one 12-filar strand.

Solder 3 Finishes to 3 Starts. The last Finish and Start give 4 times the number of turns of the 12-filar, with 4 times the DCR 1/4 of one 12-filar strand.

There will be many hundreds of pFd of capacitance. However at this low impedance that gets you far above the audio band, and not too shabby next to modern power transistors.

The scheme scales with integer turns-ratios. And there's no vast difference between 3.6:1+1 and 4:1+1. 3:1+1 is only 9 strands, 2:1+1 is only 4. If OTOH the design needs say 1.6:1+1, rounding to 2 or 1 may be too far off. The other way, 4:1+1 (12 strands) is already a tangle-fest, 5:1+1 or 25 strands needs a trained spider not human hands.

Of course the hidden assumption is that you "must" preserve the clipping of this type amplifier. If you merely want a robust working organ, chisel the heatsink bare and drop in a big chip.

> worked with signal transformers

Class B base drivers, the "signal" is almost an afterthought. It is a lot like the driver for a (CRT) TV set H-sweep power transistor (or a large modern switcher forced to use BJTs), with whopping great chunks of current and rapid rise times and small margin for loss.

[R.G.]

> plus a gob of additional primary
That's still leakage. If gob is 50% of total, it still dominates pri-sec leakage.
> a gob ...of about a third of one fifth
Lost here.... sounds like a stiff drink??

A third of a fifth is about right for starting a transformer design in general.  
However, the fifth I meant was one of the five parallel strands in the primary multifilar. The gob is about 6% of the turns on the primary, much less than another parallel wire in the bundle, and much less than another interleave if interleaved.

Secondaries should be distributed ALL through primary. However this increases stray capacitance. However I suspect your primary is well below 1K, so stray C is not a big deal.

Good suspicions, they match mine. I figured that leakage inductance was the real problem in this particular setup, capacitance much less so. Multifilar is the ultimate reduction in leakage if you can wind it, both from primary to secondaries and secondary to secondary. I strongly doubt the original was multifilar.

... Such low impedances suggest fat wire. When wire gets too fat to go around the bends, it may be expedient to use strands in parallel.

Ah. This will give negigible leakage inducance though a %$#@! to wind.

Interesting issues. I would have thought so too, but the peak currents as measured into the output bases are about 250ma, and the glimpses of wire I can get over the edges of the interlayer insulation (it's a low-budget winding) look not huge nor paralleled. So I figure they cheaped out on secondary wire, used the same size for all of them, and just used the wire size that filled the window to 85%. This is supported by the measurement of the secondary resistance at 6-8 ohms for a wire size I calculate will fit, which matches estimated wire resistance, and an ampacity over 250ma for the size. So the second order stuff kinda fits.

Figure your primary gauge and length. Make a 12-filar. (If the length is short, go around two nails until you have 12 strands; longer, get 12 spools and filar it through a polished hole while winding the bobbin.)

Wind until full. You have 12 Starts and 12 Finishes.
...
4:1+1 (12 strands) is already a tangle-fest, 5:1+1 or 25 strands needs a trained spider not human hands.

Good points. I'll take another pass looking at whether this can fit the numbers.

Of course the hidden assumption is that you "must" preserve the clipping of this type amplifier. If you merely want a robust working organ, chisel the heatsink bare and drop in a big chip.

There's that. Except that Thomas Organ already took that out of the equation in most of the USA Vox amps. They inserted a "limiter" ahead of the power amp. This is a variable threshold clipper, which the adjustment instructions set by forcing the limiter to limit before the power amp clips. I believe they did this trying to protect the output devices in the bigger amps, which had no other kind of protection and were running on the hairy edge of both Vceo death and SOA death. But it did sound nicer.

I've tried this on chip amps with mixed results. I suspect the output damping is lower on the totem poles as opposed to the Godzilla opamp chips.

[teemuk]

The interstage transformer is limiting the amount of NFB one can safely apply so I bet it does effect damping.

...But it falls down to question if the effect to damping factor is significant or not.

In my experience the damping only drops to levels where variations in speaker impedance cause only about 1 dB differences in overall response. Nothing to write home about.

[WaveshapeIllusions]

Thanks for all the interesting examples and discussion.

tca, thanks for the Escobedo design example. He has a lot of amazing circuits out there. The Mullard reference guide is awesome. I have been looking for design textbooks. I had another one on transistors but this one seems rather handy.

The transformer design discussion is pretty hlpful too. I considered winding one, but figured it would be much simpler to get a premade one. I'd probably have messed something(s) up. Was going to grab an inductor and add an extra winding. It would have been less than optimal, to say in the least.

[PRR]

> limiter to limit before the power amp clips.

Pee-poor "protection".

Yeah, poor understanding of SOA forced some bad ideas.

> suspect the output damping is lower on the totem poles

Sure "is". Not so hard to measure. (For lurkers: if you know the amp is stable no-load, bring it up to a couple volts at maybe 100Hz with 100 ohm load, add 8 ohm (whatever it is rated) load, observe voltage drop, do math.)

As teemuk says, "damping only drops to levels where variations in speaker impedance cause only about 1 dB differences". Or if very fussy:

hi-Z (DF~~0.1, naked pentodes/transistors)
'matched' (DF~~1, some classic Fenders
low-Z (DF~~3, naked triodes, some guitar amps)
hi-damp (DF>10, all solid-NFB hi-fi boxes)

> Godzilla opamp

These are often so much more efficient than classic totem poles (note the VOX has 0.5 ohm resistors with 4 ohm load, over 10% gone already) that a DF nearer 4 than infinity can be got at the same power with simple series resistor. If the load is 4 ohms, two 1 ohm 10W resistors and clever switching gives DF of 8 4 or 2 at player's option. And there's tons of active variable-damping plans, though some restrict speaker grounding.

[R.G.]

Pee-poor "protection".
Yeah, poor understanding of SOA forced some bad ideas.
> suspect the output damping is lower on the totem poles
Sure "is". Not so hard to measure. (For lurkers: if you know the amp is stable no-load, bring it up to a couple volts at maybe 100Hz with 100 ohm load, add 8 ohm (whatever it is rated) load, observe voltage drop, do math.)
As teemuk says, "damping only drops to levels where variations in speaker impedance cause only about 1 dB differences". Or if very fussy:
> Godzilla opamp
These are often so much more efficient than classic totem poles (note the VOX has 0.5 ohm resistors with 4 ohm load, over 10% gone already) that a DF nearer 4 than infinity can be got at the same power with simple series resistor. If the load is 4 ohms, two 1 ohm 10W resistors and clever switching gives DF of 8 4 or 2 at player's option. And there's tons of active variable-damping plans, though some restrict speaker grounding.

Yep, been to all those places.