Every audio amplifier needs an input filter to keep unwanted frequencies away. Both DC and too high frequency are undesirable to feed into the amplification.
Blocking DC is trivial: Just use a large high quality capacitor for blocking DC and controlling low frequency roll-off and the problem is solved.
More challenging is the high frequency roll-off and phase issues so I focus on improvement of the low pass filter here. The high frequency suppression is a difficult trade-off: The cut-off frequency should be low in order to keep away high frequency trash that would get inter-modulated with the music signal and form non-harmonic products in the audible range. A low cut-off frequency will cause large phase shift well below the filters cut-off frequency, which is also undesirable.
Usually there are no restrictions regarding capacitance used in the low pass filter, but when using an audio signal transformer ahead of the amplifier imput, capacitive loading becomes an issue to deal with.
For the investigation I compare a first order low-pass filter with three variants of a second order low-pass filter.
I set the roll-of frequency for the first order filter (schematic #1) and the second order filter with high frequency pole first (schematic #2) to roughly 180kHz. The second order filter with low frequency pole first (schematic #3) aims to have the same attenuation at high frequency as the other second order filter and ends up with a roll-off frequency of 250kHz.
Schematic #4 shows a second order filter with lowest possible capacitance that is optimized to work well with the audio input transformer module that I designed. Capacitor values were optimized on the bench and resulting roll-off frequency is a result thereof. This filter is not really comparable to the other second order filters because the roll-off frequency is much higher.
Both the first order (schematic #1) and the second order filter with high frequency pole first (schematic #2) show 5.5° phase shift at 20kHz. The steeper slope of the second order filter leads to the conclusion that phase shift at 20kHz could be reduced by moving the roll-off frequency higher while maintaining good high frequency suppression. This is what the second order filter with the low frequency pole first (schematic #3) aims for and reaches 4° phase shift at 20kHz.
All second order low-pass filters show significantly better attenuation at higher frequency than the first order filter. This is expected and a huge advantage over the first order filter design.
I find the minor extra complexity of the second order filter worth the dramatic improvement and by tweaking the second order filter, both attenuation and phase shift can be optimized.
The challenge with using audio input transformers is that capacitive loading additional to the compensation network needs to be minimized. With given input impedance, this results in a higher roll-off frequency than I would prefer. A solution to overcome this problem could be to increase the input impedance, which allows to increase the low pass filter resistor values and therefore lowers roll-off frequency without increasing capacitive load at the transformer output. For example with 1k65 and 26k7, the roll-off frequency would be 300kHz while maintaining -1dB in-band attenuation. The higher input impedance also improves low frequency response of the audio input transformer.