I have often seen discussion about using negative feedback tapped from after the output filter to keep it from ringing when the load impedance rises, causing the characteristic rise in frequency response, and it seems like a good way to make sure the output filter does exactly what you want it to do. But it also involves added complexity and cost.
Would it not be potentially simpler to use an analog lowpass filter prior to the amp's input at just above 20kHz to prevent the amplifier from driving the output filter at ringing frequencies? Perhaps augmented by an additional notch filter if a deeper cut is needed in the frequency range near output filter ringing.
Would it not be potentially simpler to use an analog lowpass filter prior to the amp's input at just above 20kHz to prevent the amplifier from driving the output filter at ringing frequencies? Perhaps augmented by an additional notch filter if a deeper cut is needed in the frequency range near output filter ringing.
The easiest way to configure the output filter is with a load. The main problem is that the load resistance is unknown in advance, but if the load resistance is determined, then it is best to configure the output filter with the load.
Sounds like you're talking about Post Filter Feedback (PFFB) Rory. Here's a couple information articles. More manufacturers are coming out with PFFB amps, mostly for TPA3255 based amps. Amir did a review of one by 3e Audio: https://www.audiosciencereview.com/...dio-tpa3255-260-2-29a-amplifier-review.50208/
I am aware of PFFB, but I am talking about an alternative to PFFB where the characteristics of the filter at some impedances are anticipated in a prior gain stage of the amp. If the only issue is with frequency response shape and not some other time-variant nonlinear phenomenon, it should be solvable before the input to the Class-D modulatior. If we observe that the frequency response peak formed at high impedances tends to start causing trouble at, say, 30kHz with an 8-ohm load, but lower impedances provide acceptable response shape, then we can filter for the 8-ohm peak at 30khz with a notch filter at the peak, then a lowpass filter that continues to address the peak's movement if the impedance is higher than 8 ohms. This is what I am proposing.
This will work, but from a practical standpoint it is best to match the amplifier's output filter to the load.
This will also work if you have a digital signal processor in front of the amplifier, which can very easily generate the required output frequency response. Adding frequency response compensation in front of the amplifier using passive components is not the optimal way.
Implementing a lowpass filter with a cutoff frequency just above the audible range (around 20kHz) can help prevent the amplifier from amplifying frequencies that could cause the output filter to ring. This approach is simpler and can be less costly compared to the negative feedback method.
