[22906]

Hello Head-fi; I'm back. This time to advance my views on phase linearity in the audio signal chain.
 
1. Only planar transducers (or phase-linear transducers) should be used in audio
2. Digital filters are completely unnecessary and counterproductive to hi-fi reproduction
3. The maximum number of analog filters that should be used is one, and that is directly after the microphone, and only if the microphone's frequency response extends past half of the sampling frequency; and if it used, it must be phase linear
4. Once the phase of the signal has been altered, it is non-trivial to revert the signal to the original condition; preserving the original phase of the signal should be of the greatest priority in designing audio systems
 
This is assuming all of my arguments in my other thread, Hi-fi Audio signal chain, are true

After reading this about audibility of phase shift:
http://www.audioholics.com/room-acoustics/human-hearing-phase-distortion-audibility-part-2

We need to apply basic philosophy of science to hi-fi audio. It is generally impossible to "prove" a scientific theory with an example, but it is possible to disprove one. This is in contrast to mathematical theories. Audio reproduction is a scientific field, so even if many people cannot hear phase differences, engineering decisions should not be made based on pseudo-scientific claims, but rather err in the way of greater caution, and theoretical optimality wherever possible. This means "making it sound good" takes second priority to phase linearity, DC linearity, and frequency response. I look forward to the discussion. I won't be responding at all this time.

 

[castleofargh]

  This is assuming all of my arguments in my other thread, Hi-fi Audio signal chain, are true
 

oops. we can close the topic then.

 

[22906]

  oops. we can close the topic then.

No, actually, you can't.

 

[Speedskater]

This discussion has gone on for several decades but has not gathered a great deal of support.  Many of the highest rated loudspeaker systems have large amount of phase shift.
If you enjoy listening to loudspeakers with little phase shift, I think there our other reasons for that enjoyment than just the lack of phase shift.

 

[Speedskater]

For those interested in phase shift, digital crossovers are a good thing.  A digital DSP crossover giver the loudspeaker designer now tools for adjusting phase shift.

 

[cjl]

 
This is assuming all of my arguments in my other thread, Hi-fi Audio signal chain, are true
 

You can't just abandon one thread (since you keep losing all the arguments there) and then start a new one with "assuming all my arguments from the last disproved thread are true."

 

[Don Hills]

  You can't just abandon one thread (since you keep losing all the arguments there) and then start a new one with "assuming all my arguments from the last disproved thread are true."

 
He's clearly just "here for the argument".
I can hope that no-one bothers rising to the bait, but there's always someone...

 

[22906]

Thank you to all readers; I would like to expand on points #2 and #3 for the benefit of the community. The reason I am against digital filters is that I believe once the signal has been sampled, and therefore transformed to discrete time, further manipulations can only hurt signal integrity and obscure details. Every time we alter the amplitude or timing value of a PCM sample, which is what digital filters are doing, we are changing the signal in ways that will add all kinds of non-linear effects in the reconstructed signal after the DAC, for example, high frequency aliasing, and possibly phase shifts. In many ways, digital filtering is something that, if it were to be acceptable in digital audio, would have to be done with final playback system in mind. Therefore, it is my belief that the complex signal modification required for low-pass filtering is best performed by a well-characterized analog filter.

I think it is self-explanatory why we should only have one low-pass filter in the signal chain at maximum. Nonetheless many points have been covered in my other thread, Hifi Audio signal chain. The gist of the matter is that the transducers should be designed in such a way to bandlimit the physical audio signal before it is converted to an electrical signal. If the microphone does not provide adequate bandlimiting, the second line of defense we have against high-frequency aliasing folding back into the audio band is increasing the sampling frequency. Increasing Fs gives us the double benefit of increasing the bandwidth of the signal, and also providing a frequency "buffer zone" so that higher-than-Nyquist frequencies that are sampled into discrete time, will not fold back into the audible range when converted back to continuous time in the playback system. It is only in the last resort that an analog low-pass filter should be used in the audio signal chain, due to the possibility of non-linear phase changes. Additionally, I can't think of a situation in digital audio where more than one low-pass filter would be required. As long as the electrical audio signal is bandlimited to 0.5 Fs at sampling time, any aliases produced at playback, if not filtered out by the natural bandwidth of the devices in the playback chain, will be inaudible by definition.

 

[castleofargh]

so you just opened a second topic to keep talking about the same thing. repeating the same lies based upon the same lack of understanding. many people wasted time to explain stuff to you, gave you links to many reliable information, all to no avail. you're worst than a bot.
as my last demonstration, I'll filter you out with a good old "block member" and test for myself that the section's accuracy has indeed increased thanks to proper filtering.

 

[22906]

I hope those are not illogical personal attacks

 

[Don Hills]

  I hope those are not illogical personal attacks

They seem perfectly logical to me.

 

[sandab]

The reason you want linear phase response is that the derivative, dTheta/df (phase change over frequency change) is group delay.  If Theta/f is a straight line the group delay is constant and amounts to a propagation delay.  Phase itself however is not terribly important, it's really the group delay that matters.   Hence the strong desire for linear phase response filters.
 
http://en.wikipedia.org/wiki/Phaser_(effect)