The myth about filter ringing

[Mnyb]

On case anyone missed it .

http://archimago.blogspot.se/2018/01...gital.html?m=1

Great blog by Arch as usual .

The conclusion these ringing graphs are a red herring . They ate interpretted out of context.
They are good if you want to understanf how the flter works .

But properly recorded music do NOT ring because of this

[andy_c]

On case anyone missed it .

http://archimago.blogspot.se/2018/01...gital.html?m=1

Great blog by Arch as usual .

The conclusion these ringing graphs are a red herring . They ate interpretted out of context.
They are good if you want to understanf how the flter works .

But properly recorded music do NOT ring because of this

There's a really interesting post on this subject in, of all places, Audio Asylum. A pro DSP guy named Werner set up an experiment wherein he made a minimum-phase low-pass filter (with no pre-ringing, but plenty of post-ringing) with a cutoff frequency of about 20 kHz. He looks at the impulse response of the filter. Then he places an FIR filter after it, with a cutoff frequency just a tiny bit higher than the 20 kHz value of the minimum-phase filter. Does it add pre-ringing to the result? No. That's because the pre-ringing of FIR low-pass filters is due to the presence of spectral content at the cutoff frequency of the FIR filter. But that content, in this case, has been removed by the minimum-phase IIR filter ahead of it, so there's no pre-ringing in the combined response at all. Here's the link.

[darrenyeats]

See my link to PFM at the end of the comments section of Archimago's article. There you will find more recent comments from Werner.

[Archimago]

Yup... Thanks Mnyb for the post.

Yeah, all these years the audiophile industry and press have been suggesting that impulse responses are such a big deal yet we see all the time new products that:

A. Have no filters - NOS
B. Have weak filters - MQA
C. Have very strong filters - Chord

And yet among audiophiles, we see all kinds of testimonies for each of these products! The bottom line really at the end of the day is that filter effects are subtle. The real question is, does a person want good filtering in order to achieve proper audio output - flat response to 20kHz, no aliasing/imaging, and reduce jitter? If so, then go with a DAC that performs with a high quality filtering algorithm (not MQA - very weak and introduces its own overloading distortions).

As for the whole ringing business, there's nothing to worry about so long as the mastering was done properly! The use of the impulse response "image" with pre/post-ringing and long duration to scare people into believing that short duration, no pre-ringing are "good" traits are ridiculous examples of an industry that has nothing better to do but create fear, uncertainty, and doubt to stimulate consumerism.

[Daverz]

There's a really interesting post on this subject in, of all places, Audio Asylum. A pro DSP guy named Werner set up an experiment wherein he made a minimum-phase low-pass filter (with no pre-ringing, but plenty of post-ringing) with a cutoff frequency of about 20 kHz. He looks at the impulse response of the filter. Then he places an FIR filter after it, with a cutoff frequency just a tiny bit higher than the 20 kHz value of the minimum-phase filter. Does it add pre-ringing to the result? No. That's because the pre-ringing of FIR low-pass filters is due to the presence of spectral content at the cutoff frequency of the FIR filter. But that content, in this case, has been removed by the minimum-phase IIR filter ahead of it, so there's no pre-ringing in the combined response at all. Here's the link.

I didn't get what he was trying to get at with these plots until I read the letters section of the issue of Audio Critic (from 1991. The math goes back to 1915!) that one of Archimago's commenters links to:

"The pre- and postringing of the ideal brick-wall filter does not in any way introduce precursors or postcursors (?) which were not already present in the original bandlimited signal which is being reconstructed from its samples. To believe otherwise is a serious misunderstanding of the mathematics involved, and hence of the true outcome. This may seem counterintuitive, but it is correct. For example, if the input analog signal was bandlimited by a causal brick-wall filter approximation (e.g., a minimum-phase analog antialiasing filter), which thus had no pre- cursors in its impulse response, an ideal sin x/x reconstruction will not introduce any precursors ."

http://www.biline.ca/audio_critic/ma...ritic_16_r.pdf

Bottom of page 6, top of page 7 (PDF pages 8 & 9).