SE class A regulator-chip-amp madness

[Elvee]

Here is the next in the series:
The Voltage-Regulator-Chip Amp family welcomes a new member:

It is a (relatively) efficient class A single ended stereo amplifier utilizing the correlation between L and R channels to improve efficiency.

This one is a really top-class amplifier, boasting less than one ppm THD!

Have fun!!!

[AndrewT]

I don't believe it.
It being the predicted performance as shown by the simulator.

[Elvee]

Here are some news:

I have built a prototype, and it works brilliantly.
I just had to add a stabilizing network, R10/C1.

I'll be back soon with pics and measurement results.

[Elvee]

Here are pics of the prototype.

I haven't made any measurements yet.

But I used it as headphone amplifier, to check that the SE efficiency scheme had no negative impact on real audio signals.
I don't know whether I still have any bits of brain left inside my skull: the thing puts out 2.5W/ch. Sheer madness.

Anyway, the results look good: even when there is a strong stereo differentiation, there are no discernible negative effects, and the separation remains perfect.

[AndrewT]

It looks superbly simple.
I think I will try it on HP as well.

[1543]

Quite interesting and low parts.
What is the gain of the circuit?
Is it possible for you to make some measurements and post the results here?

THX.

1543

[AndrewT]

The gain is set by the outer (Global) NFB loop.
R4/[R6+C3] + 1, ~11times in the passband.

[Elvee]

Here are some test results. Unfortunately, my poor old ST1700B didn't let me go very far or deep into the measurements.

At 1KHz, the THD indication was the same as the noise floor of the distortion-meter itself: 0.0012%.
The residue on the oscilloscope did show some second harmonic, buried in the noise of THD-meter, but it was insufficient to influence the reading on the meter.
I can safely say that the distortion (at full power, just before clipping) is <0.001%, and probably significantly lower (but probably not as low as in the sim).

The power bandwidth extends to ~30KHz. Beyond, the waveform breaks down, probably due to insufficient SR.
The small signal bandwidth (@2Vpp output) is 140KHz.

The opamp was a LF353.

All the measurements were made with both channels in service and loaded with 8R, and receiving an identical signal.

[1543]

Impressive results from a LM317/337.
What about using this circuit for a bridge mode amplifier?
This would avoid possible performance decrease due to strong differentiation of stereo Signal.
To still achieve the efficiency advantage by correlation is all to do to replace the lm317 with lm337 for correlation?

1543

[Elvee]

Originally Posted by 1543

Impressive results from a LM317/337.
What about using this circuit for a bridge mode amplifier?
This would avoid possible performance decrease due to strong differentiation of stereo Signal.

Why not, but you have to carefully evaluate what is to be gained by doing that.
In SE class A, the current source determines the peak output current.
With the 317's, that's what limits the output power on 8R (bias=800mA, ~half the max current of the 317).
By bridging, you could increase the output power, but on a load larger than 8R.
Not very convenient.

There is another path to use the correlation: instead of the middle 317 used as a current source with a relatively low voltage across thanks to correlation, you could adopt a complementary scheme:
The middle regulator would be modulated by the L+R signal, and one of the channels and outputs would have its polarity reversed.

I will probably try this scheme too, but it's a bit more complicated to implement.