E.S.-JLH: Please, evaluate and comment

[Elvee]

Hi everyone,

I'm trying to develop an energy-saving version of the topology widely known as "JLH". This circuit has been popularized by John L. Hood, but it already existed in tube versions, and has been dubbed "virtual complement" by Bengt Olsson.
I want to make a multiple, all-purpose lab amplifier, not specifically aimed at audio applications, but capable of driving very accurately all sorts of low-impedance, awkward loads: transducers, actuators, motor phases, etc
As I intend to put 6 or 8 of these amplifiers in a box, the idea of operating them in class A isn't very attractive, hence the E.S. tentative. I need however to retain the low, stable output impedance of the original: with low impedance loads, any variation of the output impedance in the crossover region deteriorates badly the linearity, and in some cases, I won't be able to use global feedback because of stability issues. This means that the output stage alone has to be excellent in this respect.
In my version, the "substitute" for class A is provided by a quiescent current servoing circuit that eliminates the need for adjustment and linearizes the hand-over from one half to the other.
The schematic shows a special version, adapted for simulation and comparison purposes: the normal, all-NPN version is not practical in this respect, because of its voltage gain and the input being referenced to the negative rail. Using a PNP as input transistor solves these issues, but severely degrades the performancess due to its emitter loading impedance. Here, I have added a NPN, Q7 that presents a low impedance to the emitter of Q4; the collector of Q4 is returned to the emitter of the current generator Q3 in order to improve the headroom; the current substraction takes place in there instead of at the base of Q1, but the global "mechanics" of the circuit remain the same.
The hearth of the servoing circuit is Q5: it senses the current in the output transistors and adjusts the current source accordingly.
An initial bias is provided by R8, but this bias alone is insufficient to make Q1 and Q2 conduct, thus preventing any risk of thermal runaway.
The role of the odd-valued R10 is to equalize the gains of the upper and lower halfs: the upper half has its output taken from the emitter instead of the collector, meaning the gain is slightly larger and the impedance slightly lower; R10 restores the balance in a satifactory way, but I'm not too pleased with this solution: not only is it inelegant and awkward to implement in a real circuit, but it degrades the "good" side to match the bad one; the reverse would be preferable. Any idea to improve this aspect is welcome.
In the sim, I've checked the linearity using a 2V sweep around the crossover region (there is a 2Vbe shift between in and out).
The trace shows the ouput residue after input substraction; it may look a bit ugly, but remember the scale is about 100µV p to p, and there is no feedback.
Please, in your comments do not concentrate on the components used: they are the ones available in LTspice, and if it works properly with not very sexy semiconductors, it will work better with proper ones (here, the bandwidth extend to 1MHz); similarly, the choice of a 2 ohm load is simply to emphasize the imperfections.
The all-NPN output is a big advantage for me, as I need a very fast version: transmitting transistors only come as NPN. I'm concerned however that with the NPN input there will be some asymmetry at high speeds: for negative-going outputs, there will be a lot of drive available for the lower transistor, and the upper transistor will be actively turned off; by contrast, in positive-going output excursions, the output transistors will more or less be "left to themselves", probably yielding a slower response; once again, suggestions are welcome.
Cheers, LV
PS: the trace is in the next message

[Elvee]

HEY THERE!?!

Anybody alive?

[lumanauw]

Hi, Elvee,

This amp is singleton input stage (not differential). This means it will do little of harmonics canceling. Will it suitable for your lab purpose?
If the actual amp is not stable, between emitor of Q4-Q7 you can try 100R resistor.
Q5-Q6, are they balanced or not? Q6's emitor is directly to rail, while Q5 has resistor?
What is the quiscent current flowing from Q1-Q2?

[Elvee]

This amp is singleton input stage (not differential). This means it will do little of harmonics canceling.

It is only the driver/output stage section; depending on the implementation, it could be driven via a conventional input dif+VAS or a high performance opamp.
Q6 compensates for the Vbe of Q5, therefore they should be matched (and in thermal contact). Q5 senses the current into the resistor; it's sort of common base stage if you like.
The current flowing through Q1 and Q2 is around 100mA.