T+A DAC 200

[Zaphod Beeblebrox]

6 hours ago, Miska said:

Issue with those super long filters is that once hit, they create constant ringing around the original Nyquist.

Sorry if this is OT:

 

This argument is no different than what has been used by the proponents of minimum phase and apodizing filters on the 'dreaded, terrible' pre-ringing of linear phase filters which cannot really be seen nor heard with true music signals but can be easily simulated with artificial test signals (impulse, step, 22.05kHz signal with exactly two samples/cycle)  and does not have much empirical evidence to support it.

 

Yes, in theory any abrupt change in the frequency domain will lead to ringing in the time domain and vice-versa, some of which can be mitigated with windowing. However, ringing at the Nyquist is a phenomenon that is overblown for multiple reasons.

• First the Nyquist is 22.05kHz for CD rates far beyond human audible range.
• There is rarely any significant energy near the Nyquist with real music.
• With true music signals where it has already gone through a bunch of processing and filtering, it is very hard to demonstrate/show  ringing near Nyquist after oversampling with a good quality brick wall filter.

There is also a nice blog post here (it talks about other things too but the main focus is the ringing):  Archimago's Musings: MUSINGS: Digital Interpolation Filters and Ringing (plus other Nyquist discussions and "proof" of High-Resolution Audio audibility)

 

ps: PGGB 256 does away with the concept of ultra long filters and taps, but it is still equivalent to using a near ideal sinc filter.

 

 

[Zaphod Beeblebrox]

2 hours ago, Miska said:

No so much, especially with typical processed loudness wars RedBook. A bit of hard digital clipping commonly prominent on modern recordings is enough to trigger it.

Sorry, I had not included hard digital clipping in the universe of real (band limited) music, my bad. Yes of course hard digital clipping can look very much like a square waveform!  Here is the right channel of track 4 from 2008 Magnetic Death (sorry Metallica fans) which is notorious for its place in the loudness wars. Blue is original, Red is upsampled 16x. The levels are different because of the gain reduction to avoid clipping due to inter-sample-overs.

 

2 hours ago, Miska said:

In addition, it is right where many metal dome tweeters have +20 dB resonance peak.

 

I'm not fond of having constant 22.05 kHz whine, especially at +20 dB levels.

But then again 'ringing forever' and 'constant whining' are not true either unless the whole of the music track looks like above.

Here is a snapshot from few seconds later and no 'ringing' and it quite easy to guess which of the two one would rather listen to.

 

I will stop here perhaps further discussions need to happen outside of this thread.