Another strange design

[kenpeter]

I told DFenriz I'd post one weirder... Here's my 1st attempt to do so.
Look again... Everything you once knew as "normal" turned inside out!

And I got another one I'll post later. Same basic output as this one,
but with an almost but not quite conventional, differential input stage
that recovers from clipping a lot cleaner...

I didn't copy DF's input stage (yet), but I should probably try that too.
His diamond compares to a fraction of the output rather than ground
(so non-inverting). And got some strange turned backward transistor
and CCS, I havn't quite figured every possible abuse it might be good
for yet...

And this amplifier class still has no official name. Shall we call it ABa?
Its basically an AB. Except there is a constant current minimum, and
nothing but Schottky diodes is ever allowed to actually "B" turn off.

And here we further abuse Schottkys to create quasi-complimentary
square law curves. Complimentary MOSFET curves are not used, as
these IRFs are well known not to be wonderful compliments anyway.

The function of Q5-6 is to hold a fixed voltage across D1-4 that is
insufficient for all four Schottky diodes to turn completely on, thus
they are forced to sum this voltage by operating on the curve.

There are unresolved issues here keeping me about 3.5 volts from
either rail, but its still a work in progress. Current limit kicks in at 7A.

[sumaudioguy]

Can you say Nakamichi? I recall something that looked a lot like this in their 1976 power amp. Of course it was not FET.

[kenpeter]

I can say it (in very poor Engrish), but can you post a schematic for comparison?
My choice of FETs was only cause the 4Volts VGS-On proved a convenience.
Since power devices do not turn off, and don't cross upon on native temperature
dependant curves, it hardly matters BJT or FET so long as you can bias them.
Throw some Darlingtons at it, I'm sure you could get something to stick.

[sumaudioguy]

Sorry, do not have the schematic. There claims were similar I believe.

[kenpeter]

You mean I'm not the only one making completely delusional claims? Whoa...

Anyways, I just discovered you can cross the collectors of Q1 and Q4 to
the emitters of Q3 and Q2, change R15 and R16 to 100 Ohms. This seems
to behave a little smoother into and out of clipping.

[sumaudioguy]

I do not believe all the claims are delusional... however this sort of high symmetry design really ask for thermal tracking and matching of transistors. Also you may wish to glue transistors together to get them to track better.

Have a look at a few of these parts and see if you can use them instead of all the discrete parts.

Intersil Home Page and search transistor array
Transistors (BJT) - Arrays | Digi-Key

[Elvee]

Originally Posted by sumaudioguy

however this sort of high symmetry design really ask for thermal tracking and matching of transistors. Also you may wish to glue transistors together to get them to track better.

I don't think so. Not more than any standard design anyway.
D1 to 4 have to track Q5-6, but I see no problem there, Q7-10 and Q8-9 require a moderate level of tracking/matching as usual, and at first sight, the input stage would require an intersex matching between P and N devices.
But in fact, if the P and N devices are not too far apart, and the N-N and P-P pairs are matched, errors will compensate.

I like very much this design, IMHO it is very close to perfection: it is nice, clean, elegant, symetrical, well controlled, and has very smooth and clean transitions between positive and negative output.
In addition, it has a quasi rail-to-rail output, without warts or anything.
And who cares about the class? Call it "Class KP", it is deserved.
There are issues of personal taste: I do not like too much the idea of MOSFETs fighting with the schottky's to reach the Vbe of Q5 Q6, but the bias can be shifted to suit anyone's taste.

[kenpeter]

As long as the full-on drop across the 4 Schottkys
remains less than 2 BJT emitters, I don't see any
need for tracking. I don't make any futile attempt
to regulate what small currents of an underbiased
Schottky crossing might be. Underbias is the Key.

Keep in mind the quiescent current of the Schottkys
can drift all the way down to Visch B. I don't think it
ever would, but if it did, it wouldn't matter. Its a drop
in a bucket compared to the 120mA quiescent of the
power devices.

[kenpeter]

Now here's my other, slightly more "normal" input section for your entertainment.
We got the sum/difference of four currents colliding to make two drive signals.

[sumaudioguy]

I think I liked the first one better.