Help with Lm3886 paraller configuration

[Cyberartem]

Hello pros !

I am building a stereo amplifier with 4 Lm3886 , each channel has two lm3886 in parallel mode. Now i have a problem , when i am connecting the outputs of the two Lm3886 at the end after the 0.1R resistor they both drain a hell of a lot current and overheat in seconds i also have a major distortion at the output.
But when only one at a time is connected all works fine. I have them both in non-inverting configuration , both identical connected. They share the same input after a low pass filter. My supply is +-32V , connected to a 4ohm speaker for testing. Looks like they are going into saturation but why ?!

I have attached the schematic i used and my PCB.
All looks fine to me

What could be the problem ?
The funny thing that i have built already one amplifier similar to this using the XY Lm3886 PCB boards from ebay and all works cool.

Thanks.

[AndrewT]

You have ignored all the warnings posted on this Forum about the problems that are likely to surface as soon as one tries to couple multiple chipamps together.

[Mooly]

If you want to parallel IC's then the TDA7293 is perhaps the one to use as it it specifically designed for that (look at the data sheet).

I know Andrew follows the chip amp forums far more than me and so if he says there are issues doing this then I think that has to be accepted. Too parallel two completed and identical (and that is part of the problem... they never can be identical) amps is asking for trouble.

My thoughts on seeing this for the first time (paralleling two amps) is that the smallest of DC offsets will force current from one amp to the other... it has to by definition as if one amp has 0.00 volt offset and the other has 0.1 volts then the 0.1 volts is effectively working into a "short" or at least whatever the output impedance of the amp is.

Same for slight gain/phase differences, the amps are pumping current one to another.

You could try AC coupling the outputs !

[Cyberartem]

Thanks for the feedback guys.
I had no idea that paralleling could be such a problem , i looked at the AN-1192 and decided to build one. Build one already and it worked fine, that is why i didn't understand why this time it is not working.
I used the same components that i use always , poly caps and metal film resistors. I guess the problem is in the Lm3886's or in one of them, perhaps one is a fake. Anyway , for 4 ohm speakers it will be enough just one Lm3886 per channel. So for now i will just stick to a single chip design.

Thanks again.

[abraxalito]

Oddly, you're showing 1% tolerance on the 0.5R output resistors (where tolerance doesn't matter much) and no tolerance on the 22k and 1k gain setting resistors where it matters hugely. Also where did the idea for the 47pF shunting the 22k come from? That's going to make more likelihood of oscillation methinks.

[nigelwright7557]

You need to match components as closely as possible between the two amps.
Something like a capacitor with 5% tolerance could put a real spanner in the works.
Personally I prefer bridged amps to parallel amps.

Or it could be there is simply a mistake in your layout.

[gootee]

Your PCB layout might be a major part of the problem. If you made two separate amps, with identical layouts, there might be a better chance of success. And sharing one set of 1000 uF decoupling caps between two amps might also be creating some hidden feedback paths. And I can't see where you have any high-frequency decoupling caps near the chips' negative power supply pins.

[Cyberartem]

Hmm... didn't thought about that. Really, i am not a PCB expert or not even close, what you say may be the cause. Because the previous amp i made was built with two small identical PCBs from ebay, and they had a separate capacitors.
And on my PCB the ICs each have a different power and signal path.
I have a 0.1uf decoupling caps near the power input , is it matters if i place them close to the ICs or not ?

[djoffe]

have you put a scope on the output when it misbehaves? It would be interesting to know if this is an AC oscillation issue or a DC problem. Your schematic doesn't show the damped inductors which can be quite important to assuring the stability of the amp with speaker wire and a speaker connected...or maybe especially when looking into another amp!

[Cyberartem]

I looked with a scope and found that this is a high freq. AC oscillation bout 100khz and at about 10v p-p amplitude. Now it's clear ,one of them is going into oscillation. I will make two identical boards and connect them together then we will see what happens.