It's cheap, it's N, it's dirty, it's.... The CIRCLOMOS!!!

[Elvee]

Not especially intended for audio applications, but could easily be adapted.
Crude, but effective (34Vpp out @ +/-20V supplies).

[Elvee]

....The CIRCLOFLOP!!!

However, it does have some unusual features: it is capable of directly driving unwieldy loads, such as high Q, undamped parallel or series resonant circuits, as well as pure reactive loads, at high power and high frequencies.

Look Mum, no Zobel...!....!.:

[Elvee]

....just on life support.

Here are some more pics:

First the simulated response to a 100KHz square wave, on a pure capacitive 100nF load.

Second, the same, but in the real world (capacitor is Siemens polypropylene B32650, 0.1µ/400V)

[Rafael L]

Their simulations are without the capacitor coupling (C1).
What the function of R11? not missing this one capacitor!

[Elvee]

C1 is in circuit, and it is the main compensation capacitor (not coupling).
R11 is the output load; its value varies (from 1ohm to 10 megohm) according to the type of test performed.

[godfrey]

OK, I'll bite - what's it for then, if "not intended for audio"?

Is it running in class A or B? - I haven't simmed it but the circuit looks like it could do either. Presumably something somewhere is adjustable to set the bias?

Nice to see something different / original

[Elvee]

Originally Posted by godfrey

OK, I'll bite - what's it for then, if "not intended for audio"?

It is a "lab workhorse", an all-purpose power amplifier, designed to drive all sorts of transducers (including loudspeakers), transformer, coils, etc, at various frequencies, from DC to ultrasonic. It also does stress-testing of components or subassemblies, to simulate load-dumping conditions f.e., or inject interferferences into power supply lines.
This is why it has to be extremely stable, robust and tolerant, without any network in the output which might degrade the accuracy.
And it has to cheap, in case things go really wrong.

Is it running in class A or B? - I haven't simmed it but the circuit looks like it could do either. Presumably something somewhere is adjustable to set the bias?

It operates in AB, under near-square law conditions (insofar as the transfer function of the FETs permits). The Iq is about 120mA, with a strong negative coefficient (in the sim at least: with real components, heatsinks and non-ideal thermal couplings, things are more even).
The bias could be made adjustable by making part of R12 variable, but I used no trimmers, and selected D2 instead.
D5 compensates for Q3 Q5 Q6, and D6 for the output transistors.

Nice to see something different / original

Thanks for the appreciation

[Elvee]

First, more stress-tests.


On a (quasi) pure inductive load of 10µH:



The load dumping condition: the load forces an 80Vpp, 100KHz square wave (magenta) into the output of the amplifier (green). That's twice the supply voltage of the amplifier:




Now, a low impedance test, with a 1ohm load (for obvious reasons, the voltage is limited to 9Vpp):




Next, the load is a high-Q parallel resonant circuit:




Here is a real-life application of the above simulation: inductive heating of a screw. First, the test set-up:




And then the thing in action:

(Sorry for the pic quality, the coil was beginning to catch fire)



A close-up of the prototype, showing the heading of the thread is properly chosen:




Finally, a more serious note: a look at the distortion:

[lumanauw]

inductive heating of a screw

This is the first time I see an audio amplifier doing that. At what frequency/amplitude is it?

[Elvee]

In the case of the screw, the frequency was ~45KHz, and maximum amplitude (35Vpp).
The series resonant circuit exciting the fluorescent tube was driven at around 120KHz, also 35Vpp.