Chipamp for sub, parallel lm3875 or lm3886

[pacificblue]

Originally Posted by AndrewT

It is quoted by a few of our experts on the Forum. I had previously used insufficient NFB and smoothing capacitance. So little that the NFB was the main bandwidth limit.
I tried this calculated method of progressively scaling the filter frequencies and found it worked well, every time. As a result I promote the information this Forum gave to me.

I learned two methods to cascade filters. The best way is to set the corner frequencies all to the same value to avoid strange Q-factors. The second best way is that the first filter filters the bigger part out to relieve the circuit of unnecessary work at the earliest possible stage. Maybe the reason for that factor of 0,7 is to guarantee that the NFB filter frequency is always equal to or smaller than the input filter frequency, even considering the 20 % tolerance electrolytics usually have.

Originally Posted by AndrewT

If the input filter is omitted or set to a low frequency then the NFB filter becomes the band-limiting mechanism.

One more reason not to mix AC- and DC-coupling. The NFB capacitor should not be used as the band-limiter. Its main purpose is to block DC.

Originally Posted by AndrewT

I have reservations about the HF performance of the electrolytic cap in the NFB loop. I think that adding a parallel film cap (about 1% to 2% in value) helps make the sound more palatable. This could be due to the film cap being more able to pass the HF components that allow the NFB to operate all the way to the MHz region.

Yes, a proven improvement. There are several mechanisms at work. One is the same as in power supplies. The film caps short out the parasytics ESR and ESL of the electrolytic capacitors with their own lower parasytics. Another is that across parallel capacitors current is shared according to capacitance at low frequencies and according to resistance at high frequencies, which confirms your assumption.