FET compound pair?

PeteMcK said:
Hi Guys, another newbie question:
is it possible to make a compound pair of FETs a la Darlington, Tslaki (sp?) ? I think I've seen something similar using a BJT + a FET. Any help appreciated, cheers, Pete McK

I'm not sure if this answers your question, but a FET compound pair is called a Fetlington.
One example is the 2N7000, i believe...
Darlington FETs

Wow! A fetlington? Crazy term! I love it!

Doug Self has quite a bit to say about the CFP - or complementary feedback pair, often called a Sziklai pair.

FETs can be configured with a bipolar to form a CFP. This approach works extremely well, with a NPN driving a P-type FET from a positive rail, for example.

The advantage of this arrangement is that the Vbe of the driver is closely controlled and since it's parasitics are so low, there is no capacitive gate to drive. However, the self-oscillation tendency of FETs makes it a tad unstable unless the emitter of the bipolar carries a small resistor, typically 10R, to degenerate its gain and lessen the local feedback factor slightly.

A CFP can be configured for gain, as well. This reduces feedback factor, but costs linearity slightly. CFPs are very elegant, interesting circuit blocks, and the hybrid bipolar/FET version is one of the best.


Hi Jam,

CFPs outside the feedback loop. Hmmm. Yes, I have some experience, not too much.

To begin, I once built two identical amps with a CFP output stage on one and a conventional double emitter follower (DEF) on the other.

Same CD player, same music, same room, same speakers. Even same power supply.

The CFP, not using MOSFETs I will admit, was not musical. The DEF was. You could hear the person behind the voice.

I investigated on the test bench. On the negative rail CFP there was minor transient oscillation, despite a 10R output device base stopper. This oscillation caused loss of detail and considerable intermodulation. It was suitably scotched with a 100pF cap between base and collector of the driver, and the amp returned to the listening room.

Now the sound was completely dead. Lifeless. IOW, manacling the output stage to eliminate local oscillation completely destroys the music.

Next, I built a single ended hybrid CFP with a bipolar driver and an IRC MOSFET. No feedback. I drove it with a tube, ran 3A down the mosfets, and cooled it with a gargantuan heatsink (150W dissipation!). Vcc was 50V.

It sounded wonderful. Nelson would have liked it; the principles echo his approach. I called it the Glass Harmony.

One of the aspects of PP topology I considered was the crossover disjunction. With the CFP, the turnoff and turnon of the drivers is more abrupt (or cleaner, view it as you wish) than with DEF, which is slowed somewhat since there are two junctions controlling the event, rather than one. This makes the crossover disjunction narrower, typically 1.2V, but it does not necessarily make the switching spikes more benign than the DEF. Add to this the tendency towards instability, and the measures one must take to eradicate it, and, to me at least, the advantage is lost. Pity, it's an elegant configuration.

While high order switching spikes cannot effectively be smoothed with global negative feedback and one might be tempted to take the output stage outside the feedback loop as you suggest, this is really not practical with a PP topology because the high order switching spikes remain with or without feedback and inevitably cause listener fatigue. Of course, it goes without saying that with a Single Ended CFP topology this situation no longer prevails, and indeed, the Glass Harmony sounds so good it is more a musical instrument than an amplifier. But it uses too much energy, and I frown upon inefficiency. If only I could make a single ended push pull amplifier in AB! (Yes, I believe it can be done, and am presently chewing multiple pencils as we speak!)

There are other effects when taking the output stage out of the global feedback loop. I have tried this with the DEF configuration, and the effects are interesting:

1. A very twee, 2A3ish sound, fragile, refined, lacking in drive.
2. Wonderful sound stage due to taking the sluggish output stage outside the loop and thereby removing phase anomalies.
3. Extremely musical bass, owing to the H2 induced by some loss of damping factor. Very similar to a PP tube amplifier, actually.
4. The same old listener fatigue, unchanged.

I hope this gives you something to think on. This stuff is not easy.....




Thanks for your insight into the CFP. I had been trying to get it to work for seversl years with marginal results, I finally quit.
You certainly have answered some nagging questions that I had about the CFP.

Single ended push-pull AB? Now that sounds like a challange, if it can be done it certainly will solve a lot of problems we face.Would the bias have to track the input?

dshortt9 said:
What would one call this output configuration:
My unproven opinion is that a high bias class AB (1 or 2 amps) may be the solution as the switching transients could be swamped by the current draw of the other phase. Yes/No?

Your opinion on CFP mathes norwegian Audio wich maked a DIY class A amplifier some years ago with a BJT CFP and Fet's in the inputstage.. No feedback.
But i think the class A bias was to high... Far higher than the outputstage could handle.... heatsink should be in the area of .2degress/watt per channel.

I think i still have the magazine... I can scan the article tomorow if anyone is interrested?