The LM1875 for best

[Sangram]

Mooly, the problem with this approach is that the drive of the inverting section is taken from the "dirty" end of the non-inverting section. I usually reserve this configuration for bass duty (using chips at both ends, FWIW). A transformer would probably be a cheaper way to achieve similar results with 3x 1875s at either end, and it might even sound a bit better.

I'm not worried about unequal halves because you'll get a little less harmonic cancellation, which is not always a terrible thing.

[Mooly]

Its kind of deliberate is that, honest... because the lateral front ends sound so good, they have a unique sort of character and so using that output to feed the other half goes a way to preserving that without adding 'pure chip amp' sound to the mix.

Its all experimental and doodling at the moment

Maybe it would be something to play with though and I have a very good phase splitter from a Diamond Buffer bridge amp.

[Sangram]

That's probably a better solution. The 1875 in singles is probably not the best bet for a bridged amplifier anyway.

Your lateral amp looks good

[Mooly]

Thanks I'll keep doodling

[kokoriantz]

The problems need to resolve I see, are the following.
The crossover distortion.
The lack of power, only 16 good watts.
Bellow 2W, lack of dissipation.
Distortion 0.02%, too high for nowadays.
The pins are not enough thick.
Start/sleep noise suppression.

The crossover distortion exhibits at 0.5w for 8 ohms load.



As one can see from 4 ohms load, 0.7 times the max of 8 ohms, the problem is the bias current.

Lack of power.
This circuit is optimized for 16w only. The casing mounted directly on heatsink has 3°C/W T. resistance. It has a drop out of 4V on each rail. Paralleling two units is must for 4 ohms load to have a decent 30w, or bridged paralleled for 8ohm 60w.

lack of dissipation bellow 2w.

To keep constant die temperature, the dissipation falls bellow 2w at +/-20v. This can cause low frequency distortion adding further to the crossover. The graph stops at 14w. This curve is for 8 ohms but bassreflex speakers fall to 6.2 ohms minimum, running it at 21v it will result higher dissipation.

Distortion 0.02%.
This is for a gain of 20 (26db). It can be 20 times lower if it is running with unity gain, if further 20db extra NFB applied by the composite, 0.0002% is at reach.

The pins.
The leads are equivalent of 0.68mm thick wire. The American Wire Gauge table shows max 1A power transmission. This is RMS 1A. If two units are paralleled, 2A can be used. This results a maximum 32w for 8 ohm.

Start/sleep noise.
A quality amplifier cannot start or sleep by farthing. This issue depends upon the power supply type single or split, bridged or not bridged.

[kokoriantz]

Originally Posted by Sangram

My experience with the chip is that it is quite a bit better sounding than the 3886 (in spite of "worse" paper spec), but implementation is key. It is not as stable/reliable/indestructible as the 3886, conversely its simplicity makes it easier to get good sound.

However one must stay clear of the temptation to minimise the component count. It needs control over the turn-off thump, lacking any form of low Vs-muting. It also needs substantial number of components in the input side to ensure stability and good distortion performance. You've already covered the Zobel, this is pretty much mandatory for these chips.

I like to use it as a tweeter/mid amplifier in active systems. The current limit is a little on the low side for proper full range performance into medium sensitivity speakers. I am also keen to see how it behaves with current feedback, for more demanding loads. I have a few of these built out that run on 21V rails, and they sound superb. There is an output relay that drops out before the chip, so turn-off thump is avoided.

Keen to see where you go with this.

Thanks for the start/stop issues. If people arrived at 21v rail, no doubt it is adjusted by ears, adopted.
This music Happy Together - The Turtles (1967) - YouTube cannot be reproduced by LM3886 with 26db gain, the chorus refrain goes dirty. Can you please try out comparatively with LM1875. If you love too much the 3886 you can consolate yourself blaming the MP3, as normally people do.

[kokoriantz]

Originally Posted by adason

How about nested?

People call this composite. This circuit doesn't work, FauxFrench tried it out.

[adason]

Have you tried it?
I know its called composite.
So what if fauxfrench tried it, what does it mean, all nested designs do not work?
Or he did something which prevented it to work?
So you saying modulus-86 circuit does not work?

[kokoriantz]

"My_Fef"2
ANDREA a Democratic hybrid amp for luxury sound
These are my best composites power amps. I posted to others several other examples.

[kokoriantz]

The crossover distortion can be lowered if I bias at higher current. This can be done if the paralleled amp is of current nature, so a dc offset current can make each amp function in class A at low power. How much offset current ? The offset current will make the two amps dissipate more, as it is required to increase the low level dissipation, the dissipation vs output power curve will decide.
The amp may look like this.

The current amp is an open loop feeding the load with half the current that it needs, the voltage amp measures the error and provides the other halve up to a certain frequency determined by the output LR inductor. Above this frequency the amplifier becomes a current amplifier increasing its voltage by increasing the load impedance. Example, if the load is a full range with 100uH VC inductance L2, you see the bellow graph the amplitude rises. This principle have great chance to yield smooth very high definition tweeter sound .
Propositions, critics ant comments, are welcomed.
Hayk