Worst amplifier ever made ?

[Tekko]

I cobbled together this circuit with simplicity to build in mind.

This is likely the worst amplier ever made since it seem that the hotter it gets, the faster it runs away despite the compensation transistor. I guess PNP transistors are the opposite to NPN and need a reverse operation thermal comp:

http://i.imgur.com/gMIfi.jpg

It needs a fan to keep cool.

[davidsrsb]

No, your thermal compensation is correct.
Another 0.1R in series between the Q5 emitter/Q1 collector junction and OUT would help a lot

[richie00boy]

In fact that is what you should do and remove R7. As it is R7 is just burning power.

[Bigun]

This is a quasi output - I'm not familiar with this. The lower pair, Darlington, have two Vbe in series to produce thermal runaway and here the Vbe multiplier wants to be able to 'see' the temperature of the output device and having R8 a bit larger will help it to be more stable. The upper pair are in CFP configuration, the current/temperature of the output device is controlled by it's driver Q5 and so it's the emitter of Q5 that needs a resistor. You don't really need resistors on the output device.

[Tekko]

Emitter resistors are always used and should be used for current sharing reasons.

It was meant to have NPN outputs but i blew up the 4 NPN's i had trying to get this thing to work

[AndrewT]

Compare Q5 to Q9.
Q5 emitter is driving the output direct.
Q9 emitter is driving R4//Q2base//cap=Q2 Cob
You have to add a network to the Q5 emitter to mimic the load on Q9,(network = Baxandall diode and better network = JLH cap+Bax diode).

I would also duplicate R8 in the collector of Q1

R17 defines the current through Q3 (+ a little bit from Q7 base).
R5 defines the current through the LTUP (LongTailUnequalPair).
You must add a trim somewhere to make Q5current = Q3 current.

How do you measure equal currents? There is no emitter degeneration and there is no collector load on Q4.

Your RF attenuation is far too gentle.
You have no HF load for the output.

[Tekko]

R17 is a 1k trimmer, i go by dc offset on the output since minimum dc offset = most equal LTP currents.

I already pulled the plug on this thing. i just kept working on it as i figured why not blow up the remaining two PNP's i cad since they are useless without their NPN companions anyways.

When i plugged it in, i expected it to rail stick like yesterdays attempts and blow up but it dident, instead it oscillated so i stuck in the 82pF cap on VAS and oscillations stopped, then i trimmed down the dc offset to around 11mV and after that it worked, though the bias compensation is useless, above 60mA bias it goes into thermal runaway without a fan and above 30mA it gets hot enough to fry an egg without the fan.

I'm not gonna work on improving it or apply the suggested tips, i'll just run it as is into the ground until it explodes and then go back to the previous designs with proper current sources and current mirrors, or maybe even crap my entire line of amplifiers and start building the apex audio ones instead since they are tried, proven and approved.

This circuit is junk and should NEVER EVER be attemted by anyone!!! judging from heat dissipation vs bias current, i'd say this amp is oscillating at a frequency to high to detect with my 25MHz scope. I may keep this design around as a "how not to design an amplifier"

*Thread Closed*

[nigelwright7557]

If it is getting hot then your bias current might be too high.

I have got away with less than 10mA in some of my designs and not got crossover distortion.

[Tekko]

This heatsink can keep a stereo amp cool with 100mA bias per channel and only get luke warm.

But something heats this amp up like if it was pulling over 1A bias while the meter only read 50mV over each of the 0.1 ohm resistors = 50mA bias.

Below 35mA the amp sounds like clipping all way from zero volume and up.

[nigelwright7557]

The only other current path is the driver transistor network.
Could be a problem there.
Are they getting hot ?