Input & Feedback filters

[lhwidget]

OK, now I'm wondering about the two high-pass filters in the NS reference designs for the LM3886 (well, actually i was wondering all along, I just wanted to ask about the Zobel resistor first)

I really hope I'm not missing something basic here. The input filter blocks DC at the input of the basic non-inverting schematic. This filter is typically tuned to 2-7 Hz using Cin & Rin and the equation:

F3 = 1 / (2•pi•Ri•Ci) [Hz]

There is typically a second high pass filter created by inserting Ci between Ri and sig ground. Since Ri is part of the gain structure of the amp circuit, the values of Ci get high to maintain a low F3 & we sprout an electrolytic in the circuit.

Why is the second high-pass filter needed? What am I missing here? If DC is blocked at the signal entry point by the first filter, what protection is gained by the second filter on the feedback loop?

Thanks,
Jay

[pacificblue]

Chipamps have a DC offset. By inserting Ci the gain becomes 1 at DC, so that DC offset is not amplified with the AC gain. If you look at the datasheet the DC offset is given as 1 mV typical and 10 mV maximum. If you assume the worst case of 10 mV for the allowed gain range of 10 to 50 you would get DC between 100 mV and 500 mV at the output without Ci.

[lhwidget]

Thanks Pacificblue, I felt like I was missing something.

I built my first amp with Ci and both channels had less than 4 mV offset. Now I know why

[AndrewT]

Hi,
the input filters to the power amp should define the bandwidth of the amplifier.
This requires that the NFB DC blocking cap should be set at least half an octave lower than the Input DC blocking filter/cap.

The electrolytic used for the NFB filter should have zero or near zero AC voltage across it to minimise distortion that this cap creates in the amplifier. This is a second reason for increasing the value of the DC blocking capacitor in the NFB loop.

[lhwidget]

Thanks Andrew, I was wondering about that.

The F3 point of the combined filters, if they are close to the same value, begins to droop a bit.

My amp was done with the input filter tuned to 4.8 Hz (Rg = 10 k ohm & Cin = 3.3 µF) and the NFB filter tuned to 7.2 Hz (Ri=1 k ohm & Ci=22 µF), just the opposite.

This didn't have any great effect on the sound, but the overall response is down almost 1 dB @ 20 Hz, which could have been avoided.

[lhwidget]

OK, I just realized I don't actually understand what's happening in the NFB loop with Ci.

A cap in series blocks DC, right? It seems like Ci would insure that any DC voltage coming through Rf would have to appear at the - input of the 3886.

I understand the filter created by Ri and Ci, but how does the cap insure unity gain at DC?

[pacificblue]

The gain in a non-inverting ampifier is (Xf+Xi)/Xi. With practical components you get (Rf+Ri+1/(Ci*2*PI*f))/(Ri+1/(Ci*2*PI*f)). At DC the impedance of Ci will be next to infinity, which makes Rf and Ri negligible and leads to a gain of ~1.

Any DC coming through Rf would come from the amplifier output, thus had to be present at the input in the first place. DC blocking caps will not let that happen. So it could only be the DC offset at the input coming through and that is amplified with a gain of 1 only.

[AndrewT]

Originally posted by lhwidget
the input filter tuned to 4.8 Hz ........and the NFB filter tuned to 7.2 Hz................
This didn't have any great effect on the sound, but the overall response is down almost 1 dB @ 20 Hz,

Hi,
many builders and listeners claim that since the speaker response is already down many dB @ 20Hz, that another dB or so does not make any difference to the sound coming from the speaker.

I disagree.
Change the NFB from 7.2Hz to 1Hz and listen again.
Then change the input from 4Hz to 1.5Hz and listen to that.
Come back and tell us if you have changed your mind about what you can hear.

This didn't have any great effect on the sound,

[lhwidget]

Thanks Pacific, I don't see where that equation is coming from, but I apreciate it. I still have a hard time seeing diferent aspects of circuits.

Andrew, I don't doubt what you're saying, I just can't get to the parts to easily change them (point to point construction). Are you referring to the slow roll-off below 100 Hz (flat above 100 Hz, down about 0.8 dB @ 20 Hz)? I can see this subtly affecting the "warmth" of the bass response.

[pacificblue]

Originally posted by lhwidget
Thanks Pacific, I don't see where that equation is coming from, but I apreciate it.

One of many sources about op amp basics is Op Amps For Everyone