UniGaBuf, a follower cut out to be a leader

[Elvee]

This unity-gain follower is an audiological UFO:

It is adjustment-free, requires no thermal compensation, has no feeedback, no servo, can even work with fake or reject transistors, and yet offers low DC offset and sub-ppm linearity:

[Jon Lord]

A "Current Dumping White Follower" - CDWF, not bad !

Without any further in-detail inspection this device seems only to be very linear as long as its bias/idle current into the load is not exceeded, right ?

[kenpeter]

What's Q4 doing?

[Elvee]

Originally Posted by Jon Lord

A "Current Dumping White Follower" - CDWF, not bad !

Without any further in-detail inspection this device seems only to be very linear as long as its bias/idle current into the load is not exceeded, right ?

No. This circuit uses a trick to overcome that limtation (hence the <1ppm dist. figure).
In this case, the idle current is less than 50mA, and in the sim, it operates at up to 2.5A^.

Q4 is part of the current mirror supplying Q7 a scaled version of Q5's collector current.

[Jon Lord]

Im 100% sure, what I said.
Q3 is essentially held at constant current, up to the point, where more than -Ibias have to be supplied to the load, then it rises.

This means a rise of output impedance outside this crossover point and thus lots of distortion.
For this reason the amplifier works nicely at bias currents above 4A ...

I think my guessing is more accurate than your simulation !

[Elvee]

Originally Posted by Jon Lord

Im 100% sure, what I said.
Q3 is essentially held at constant current, up to the point, where more than -Ibias have to be supplied to the load, then it rises.

This means a rise of output impedance outside this crossover point and thus lots of distortion.
For this reason the amplifier works nicely at bias currents above 4A ...

I think my guessing is more accurate than your simulation !

Sims may have shortcomings, but they do have a decisive advantage: they have no preconceptions about how a circuit is going to behave.

Anyway, I'm in the process of perfecting the last details in the ultimate simulation: the one that uses solder, bolts, heatsinks, and similar components.

I am not too anxious about the result: I have already physically tested the most critical and contentious part of the circuit elsewhere, and it works as expected.
Here, it is combined with a Taylor follower, but I don't think the end result will differ much from the sim (which, BTW is what I expected).

[Jon Lord]

Most time switching on your brains helps before doing a simulation, especially if you use a simulator, which is intended for SMPS simulations and you try to judge your amplifiers by THD1k at 10V.

And also: what is immediately obvious for some people, might be a preoccupation in the view of other people, but it nevertheless could be true ...

Did you try a DC sweep (-10..+10V) on V1 and look at the difference between input and output ?

Did you have a look at the difference between input and output with a 10Vp 20kHz sine ?
This should be a very ugly waveform with some neat spikes of switching distortion in it ...

This circuit is certainly not better than 0.01% THD20k in real life, perhaps 0.005%, if you use better semiconductors and a very very careful layout.

But you can prove the opposite, if you have a prototype running soon ...

[kenpeter]

One advantage of Taylor, CFPs, and current dumping in general, may not be immediately obvious:
Emitter that must be the "follower" can be held at constant current.
And thats quite different than simply throwing a CCS under an emitter.

[padamiecki]

[QUOTE=Elvee;2511348]This unity-gain follower is an audiological UFO/QUOTE]

did you built it?
and know the sound?

[padamiecki]

some questions after the simulations:

1. how to set the offset?
2. why larger current through Q4,5 instead Q1,3?
3. do Q4,5 need the heatsink (~45mA idle through them)
4. at higher supply voltage +-35VDC the oscillations at 20Vpp, how fix it?