General discussion LC-filter class-D

[egberttheone]

I'm a bit confused about the output filter, after reading a lot of posts here and some googling did not find my answers.
I want to use the classic common mode filter design.
After having put the recommended filter (STA516B) in LTSpice (22u coils) it showed quite a phase-shift about 35 degrees @ 20khz.
To lower the phase shift you need to lower inductance but doing so will also change the corner frequency.
Lowering the coil tot 4.7u and capacitors to 10n gives only 9 degrees phase-shift @ 20khz.
But the attenuation for the carrier (384khz) frequency is not high enough.
The original filter gives about -6db @ 384khz and the 4.7u coils gives about 9db @ 384khz
This is all simulated with a 100v source and a resistive 8ohms load.

I'm a bit confused, which is the way to go?
Is it worth the effort to lower the inductance and get less phase-shift or is this not hear-able?

In the picture you see a snubber network on the left and @ the right a zobler network.

[Pafi]

or is this not hear-able?

This is true.

You can decrease phase shift (and load-dependance, and distortion) with post-filter feedback, but it can cause some other problems, so i don't recommend it unconditionally.

[egberttheone]

Originally Posted by Pafi

This is true.

You can decrease phase shift (and load-dependance, and distortion) with post-filter feedback, but it can cause some other problems, so i don't recommend it unconditionally.

Thank you for your reply, so i don't have to worry about the phase shift.
What about the zobel network (please ignore the values in the picture i have to calculate those)?

It is not included in the example of the STA516B data sheet, but is it a wise thing to add? Can it cause harmonic distortion adding this network?

[Eva]

You are a victim of hi-end misunderstandings about audio.

What you have to analyze is group delay, the amount of delay involved (that will not be substantial actually) and it's uniformity across the audio band. An uniform delay is not a problem either.

Think about the phase shifts and group delay discontinuities from speakers...

One 20Khz wavelength measures 17mm while propagating through air. Just having one speaker 8mm further from you than the other will result in 180º phase shift from one channel to another in that frequency range. Fortunately our ears don't care about phase at such high frequencies