thermal tracking in quasi-complementary

[sam9]

Here's a sort of puzzel. In an EF BJT output Self and others recommend thermal tracking by placing the Vbe multipler on an output device or at least close by on the heatsink. In a CFB output the recomendation is to place it on one of the drive transistors.

This leave me with the question: which is appropriate for a quasi-complementay output stage. You could do it either as long as you pick match the technique to either the "EF" side or the "CFB" side. Is there a conventional wisdom regarding this, or is this so obscure anyone building such an amp has to figure it out themselves?

[Nelson Pass]

I don't think it's a very big deal. Often as not it's an arbitrary
layout choice, and other times it just hits the compensating
mark better as determined through experiment.

[sam9]

Thanks, I was thinking of building one as my next project just to say I had done it. Then it ocurred to me that if there was definately a "wrong" way that was almost certain to be the one I tried first and had visions of the output transistors going up in smoke.

[sreten]

Its a very good question without a good answer as far as I know.

Possibly mount drivers and outputs on a flange to the
heatsink and mount the Vbe multiplier to the flange.

D. Self implies only the Baxendall version quasi should be
considered initially but effectively discounts it later on for
normal class aB (optimum quiescent) operation.

He then implies quasi can be used for Class A operation.

For stability reasons in the CFP side I wouldn't build a quasi.
(Along with a list of other reasons)

sreten.

[sam9]

I'm not too worried about the stability on the CFB side as I've built a couple of pure CFB amps that worked out, although it took a willingness (Pigheadedness?) on my part with alternate layouts untill I got stability. BTW, a single output pair seems to be easier to get stable than parrallel pairs.

My main interest in doing a quasi is self (no pun) education. SPICE models are nice, but I seem to lean something every time a physically build a unit that I would not have grasped with books and simulations.

[Easyamp]

I believe quasi out amps have an equal number of advantages to disadvantages to comp pairs to make them desirable especially if you have a bunch of npn output devices with no comps.

It's easier to build a comp pair amp with regard to design but a well designed quasi amp will work well if done right.

I have designed and built a triple quad quasi amplifier that I had posted previously, and though no one liked it, it sounded and worked well?

I also have a baxandall quasi out amp I built which I'm listening to right now that sounds ok but needs work.


Cheers

[sam9]

My SPICE simulations tend to agree with D.Self -- namely that on just a strictly distortion basis they are almost as good a CFB. That's one thing I'll be looking for when I build one. Another consideration will be how easy it is to get a layout that avoids parasitic oscillations. This is especially with multiple output pairs. I've found I can do it with a single pair without too much anguish, but adding a second pair often gets you back in the oscillator busness again. Despite some authorities saying there's no harm is there small parasitics, I really feel better when I don't see even a little bit of "fuzz" on the scope. If the resolution to the problem lies somewhere between EF and CFB in difficulty, I would be inclined to think quasi isn't so "old fashioned" as thought.

[sreten]

Well if your going to build one Neil McBrides excellent Naim clone
site is as good a place as any to get good designs and advice.
(Naims are quasi with allegedly very high Ft output trannies)

Personally I don't like CFP. Stability is one issue. Biasing another.
But mainly because low power distortion is much worse then the
EF output stage and the distortion advantages of a CFP are only
apparent at continuous powers near full level, which IMO has not
much similarity with real world music use.

My SPICE simulations tend to agree with D.Self -- namely that on just a strictly distortion basis they are almost as good a CFB.

Limitations of your Spice modelling. D.Self's real world measurements
of quasi output stages indicates they come third compared to the other
options, and that the Baxendall diode helps much less than expected.

sreten.