Output transformer query

[retailer]

I found this diagram of a Leak TL/25 output transformer winding layout on the internet and noticed that only the 16 ohm speaker connections uses all secondary windings, the 4 ohm connection uses secondary 1&3 windings leaving secondary 2 not connected, the 8 ohm speaker connection uses secondary 1&2 and only part of secondary 3. A circuit diagram of the TL/20 shows the OT with 4,8, and 16 ohms taps so I'm assuming the OT is wound the same way.
I was under the impression that it was "best practise" to utilise all windings in an OT and that there is no easy way to achieve this with a secondary winding having taps for the various impedances, to utilise all windings the secondary winding leads need to be bought out to terminal strips and jumpers set for the various impedances.

I wound a pair of UL output transformers for an acquaintance a few months back he sent me a text message the other day, he had completed his amp and was claiming 20Hz - 50Khz response with no feedback and that primary inductance was 40H, if he not mistaken then it is a surprise as his winding layout was nothing special. Given that Leak amplifiers do provide good performance and indeed these days seem to be regarded as high end - are there any real performance gains to be made by using all secondary windings for every impedance setting - selectable taps are very convenient and certainly a lot more convenient than setting jumpers.

[Radio Wrangler]

I've seen various different layups for the output transformers of Leak amplifiers, some more intricate than that. None of them had come from Leak, though. Ed and Mike will likely have info from transformers they've rewound.

You're right that the simple tap selection of speaker impedance is not ideal and doesn't fully utilise all the copper in the secondary windings, nor does it offer ideal flux distribution on both settings. It is probable that the design intent was an amplfier for 15 Ohm speakers, and that the 4 Ohm setting was a bit of a compromise. This would fit in with the speaker impedances in use in the period.

The frequency response without feedback only tells part of the story and the easier part at that.

To get their highly advertised point one percent distortion, Leak used about as much feedback as you can get away with in a valve amplifier with transformer output. The limit being set by the risk of the onset of loop instability. This means that by the time the gain around the feedback loop has dropped to unity, the additional phase shift around the loop must have built up to less than 180 degrees. How much less is called the phase margin. The greater the phase margin, the greater the damping if the amplifier is excited into a ring on a fast edge like a pulse or higher frequency squarewave. Fast pulses aren't exactly part of music, but they are a good diagnostic tool to allow you to quickly judge stability margin. They allow you to stimulate a bit of misbehaviour that wouldn't otherwise be visible.

The high frequency roll-off comes from the valves and the circuitry around them, plus the output transformer.

In an ideal world, the roll off would be dominated by one factor, giving a 6dB/octave roll-off and 90 degrees max phase shift.

In a less ideal world, multiple roll-offs start coming into play, and there isn't just one dominating... things like the leakage inductance in the transformer for example come into play and although this gives you a faster roll-off and brings the gain down to unity more quickly, it gives a faster than proportionate growth of phase lag. The result is poor stability or oscillation.

The valves alone can work to many megahertz, but the choice of anode load resistors in the stages before the output devices working with the stray and grid capacitances can bring this down appreciably, though you get more gain per stage from high load resistance.

So there are several interacting compromises going on. Leak was rather artful at steering a difficult path through them.

Another factor in the output transformers is the thickness of the laminations. Finer lams act to reduce eddy current losses in them, and this affects the phase behaviour at the high frequency end.

Back in the day, anyone could buy valves, resistors and capacitors. It was the cleverness in the output transformers which gave Radford, Leak and Quad their edge.

Also back in the day, the maths behind studying feedback stability was seen by many as a dark art. It isn't. It's something that has to be right in every single amplifier or it won't work right, but the viewpoint survives in the anti-feedback cult.

If you seek refuge in transistors, then you can do without the output transformers, but the transistor's more abrupt non-linearities demand far higher levels of feedback, and we're back into the race between gain roll-off and the build up of phase lag. There is no escape!

So the heart of the design of these amplifiers lies in the output transformer. A lot of thinking and experimenting must have cone into them, so any comments from Ed or Mike will be rather interesting.

David

[Ed_Dinning]

Hi David a good summary.
I would only add that the more sections there are in the windings the closer the coupling and the lower the leakage inductance (coupling is basically the difference between the primary inductance with the sec o/c and the sec s/c ~simplistically without resort to maths).
Then the downside comes in in that this increases the pri/ sec capacitance which will mess with the feedback. There are also other minor effects.

Ed

[snowman_al]

There is of course the overall limiting factor - cost.
If you have an amplifier with feedback you can do what Leak and many others do and use the 16Ω tap to provide the feed back regardless of the impedance selected for the speakers. That precludes the use of all of the windings for each ratio selected.
You either do what Radford, for instance, do and have 3 feed back components, one to suit each impedance selection. Or have a separate feed back winding on the transformer. Both of these are additional costs.

[retailer]

If one replicated say a TL/20 the circuit, which is easy enough to do and also wound a working transformer, with the end result being a stable working amp then given that the heart of the amp is the output transformer, would it be fair to say that the transformer near enough meets Leak original specs.

[Radio Wrangler]

Not strictly true, but then you'd have a fine working amplifier, and there wouldn't be a lot to gain from arguing the details. Near enough would be a good description. Leak hadn't built it, so you couldn't ethically make and fit a Leak badge to it. You could sell the transformers as replacements and they might be as close as many re-winds. These sorts of things would come down to market acceptance.

Originality is a mystical ingredient, and a large contributor to the price original Leaks, etc. sell for. Leak was pushing it a bit, there is a degree of uncertainty over whether every amp he shipped quite met the claimed specs.

Valve amplifiers are relatively simple things and they allow plenty of the characteristics of the valves plugged into them to pass through into the performance of the completed thing. Hitting original specs would be somewhat influenced by having or not valves which had characteristis like the originals had when they were new.

It all gets a bit nebulous...

David