LM3886 driving mosfet

[pro]

Hi at all.
I am trying to make this amplifier. I channel is under test, but I have a problem that can't explain. Averything works fine, it's also playng in this moment and I have to sey that sound seems much better than I expected, but there is a big problem . With the scope I have seen the sin wave at 1 Khz, to be clipped at +- 11V.
Rail ar +/- 41 Vcc.
Could someone give me an idea of why this happens?
Thanks at all

[richie00boy]

Probably because of R1 and C1.

[pro]

Hi Richie
I have tryied with C1 off, (100% global feedback) but it is just the same.
Regards

[pacificblue]

If you look at the local NFB loop around the LM3886, how high is the DC gain? How high is your DC offset?

[pro]

I don't know what you mean. The local gain is 26 db that is the same of the global. The offset at the exit is about 30 mV.

[mjf]

hello.
perhaps you can use this basic schematic..........
greetings............

[pro]

Thanks mjf,
but your schematic is full nfb, while mine have only DC feedback from the load and it's dc coupled.

[pacificblue]

Originally posted by pro
I don't know what you mean. The local gain is 26 db that is the same of the global. The offset at the exit is about 30 mV.

The local AC gain is 21 (26,4 dB), the global AC gain is 22 (26,8 dB).

For the local feedback around the LM3886 the gain is R24+1/(C1*2*PI*f)+R17/R17. 1/(C1*2*PI*f) becomes infinity for DC, which makes the DC gain also infinity. Only thanks to R1 you don't get DC with nearly rail voltage at the output.
So the local DC gain is infinity and the global DC gain = global AC gain is 22.

What does that mean? For DC only the global feedback is relevant. For AC you force the voltage differential between the LM3886's output pin and the amplifier output to be 1/22 of the output voltage. Is that stable across the entire voltage range of the amplifier from 0 to full output swing?
E. g. for an output swing of 22 V there would be 1 V across R1 and either across R21+M2+R5 or across (R13+R3+R7||Q1||C4)+R6+M1+R4. For an output swing of 11 V there would only be 0,5 V. For an output swing of 5,5 V there would be 0,25 V. Can that preform well?

[pro]

Hi PB
For 22 V output swing, the only place where I have 1 Volt is across R17-R24. In all of the other places you mentioned except R1 I have 22V shifted of 3,8V of the bias (+26 -18).
It is like if the 3886 is working in it's internal with an offst of 21V, that is the offset that the LTSpice gives to me if I take off R1.
Regards

[pacificblue]

No, of course you won't get 1 V. That is what should be there, if the feedback loop would work. In a real circuit the IRFPs will win the battle and force a different voltage across R1, thus rendering the feedback loop useless.

Which model do you use for the LM3886? I simulated it with teemuk's and as long as I don't short C1, the output shows ~40V DC as predicted in my first post. With C1 shorted, there is about 220 mV DC offset. An additional blocking cap in series with R17 brings that down to ~60 mV.

I don't see any way to make your feedback scheme work. You should try mjf's schematic.