alt Hacker News> Annoyingly the input pins had a constant DC offset of over 2 volts, which went away when shorted to ground but reappeared moments after I removed the short.
Mic inputs on audio codecs always have DC bias, because they are DC biasing the microphone inputs so that signals can pass. By default, you would expect ~2-2.7V of DC bias on each microphone channel. This is typically connected through a high impedance (~1kOhm) source impedance. The actual impedance of the input should be extremely high resistance, though, typically in excess of 50kOhm.
The proper way to connect this would have been to a line-in jack on your motherboard. There should be no DC bias on those jacks. Or use an audio input that is 'retaskable' to select the line-in function instead of the mic-in function.
> This large of capacitance is not easily available as C0G. The usual X7R MLCC capacitors are piezoelectric and will change voltage when bumped, which can add interference to audio. Electrolytic caps are an option, but will not fit well on pads meant for 0402 ceramics.
C0Gs are not needed for passing something like this. X5R/X7Rs are commonly what is used for DC blocking capacitors in ADC/DAC solutions and you can still keep extremely good audio performance even with the theoretical microphonic problems for Type 2 dielectric capacitors.
> If we wanted to improve audio filtering, we could design a replacement PCB (four-layer for impedance matching and signal integrity?) with footprints and traces to install a second-order Sallen-Key filter. This would be about the same difficulty as the HDMI2SCART, but likely reusing the existing case and screws unlike the HDMI2SCART's 3D-printed case.
Why not just add your filtering circuit externally via the 3.5mm jack? No limits to what you could do. Cascade as much as you want at that point. Might be a bit ugly, but it wouldn't require re-designing the board.
Both of your points were the very first things I thought reading this.
A Sallen-Key filter is trivially easy to design if you work within certain constraints - a 2-pole Butterworth filter has the feedback capacitor exactly twice the value of the "second" capacitor to ground, if both resistors are the same value. If you pick 10kΩ for both resistors, 1nF for the feedback cap, and 470pF for the cap to ground, then you'll get pretty damn near a Butterworth response (Q of 0.707, maximally flat in the passband and then as fast a transition as possible to the stopband) at around 23kHz.
This is perfect.
And guess what? If you want to scale the cutoff, just scale the component values! If you use 15kΩ resistors you get 15kHz, if you use 22kΩ you get 10kHz, and so on. The minor error in Q will not be audible.