Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[RD James]

In recent years, DAC manufacturers have recognized that there are multiple ways to implement conversion, including many different reconstruction filters, including linear phase, minimum phase, apodizing, closed form etc. Those filters can also be implemented in software, for example when the software is doing delta-sigma conversions. By implementing multiple filters, a software player can allow far more choices than an individual DAC manufacturer typically provides. For example, HQPlayer's main design approach is to  do  delta-sigma modulation with a large selection of filters available.

This is just people trying to create unique selling points for their device/software by being different rather than better.
Pre-ringing is a non-issue for non-clipped audio tracks, and it is worse with minimum phase filters when playing a clipped audio track.
 
One thing which could improve audio quality would be implementing noise-shaped dither - particularly if this was combined with upsampling to shift the noise far outside of audible frequencies - but improvements are going to be minor for 24-bit audio. The switch from RPDF to TPDF is far more significant.
 

[dtc]

This is just people trying to create unique selling points for their device/software by being different rather than better.
Pre-ringing is a non-issue for non-clipped audio tracks, and it is worse with minimum phase filters when playing a clipped audio track.

Sorry, but DAC manufacturers and software players offer different filter options because they sound different and some people prefer one filter to another. For me, providing people with options that improve their listening experience is useful.  If you do not hear differences, that is fine. But why dismiss the fact that others do hear differences?

[RD James]

Sorry, but DAC manufacturers and software players offer different filter options because they sound different and some people prefer one filter to another. For me, providing people with options that improve their listening experience is useful.  If you do not hear differences, that is fine. But why dismiss the fact that others do hear differences?

I did not say there were no differences, but different does not mean better. Professional gear uses linear phase filtering for a reason.

[dtc]

I did not say there were no differences, but different does not mean better. Professional gear uses linear phase filtering for a reason.

"Better" is what sounds better to the listener.  Linear phase is often used because it provides time alignment, which is a desirable goal.  However, in practice different people hear different filters differently.  The idea that there is one right way does not square with the fact that different people prefer different filters.

For example, one listening test of linear phase versus minimum phase filtering was done a few years ago by Archimago and published on his website.  His conclusion was the more people preferred minimum phase over linear phase, but it was dependent on several factors, including the actual track.  My point in referencing this is not to try to recommend one over the other, but to point out that in actual listening tests linear phase is not always a conclusive winner.

http://archimago.blogspot.com/2015/07/the-linear-vs-minimum-phase-upsampling_10.html

My recommendation is simple - provide options and let the user decide.  Many designers do that and users pick the option that sounds best to them.  The fact the human hearing works at all is pretty incredible and it is no surprise given its complexities that different people hear differently.

I do not think this is the right place to continue this discussion. It has been hashed out over and over again on other websites and in multiple articles.

[RD James]

"Better" is what sounds better to the listener.  Linear phase is often used because it provides time alignment, which is a desirable goal.  However, in practice different people hear different filters differently.  The idea that there is one right way does not square with the fact that different people prefer different filters.

And some people claim to like how their expensive NOS DAC sounds when they aren't using any upsampling at all.
It's objectively bad, but because it sounds different and cost them a lot of money, they convince themselves that it sounds better.

For example, one listening test of linear phase versus minimum phase filtering was done a few years ago by Archimago and published on his website.  His conclusion was the more people preferred minimum phase over linear phase, but it was dependent on several factors, including the actual track.  My point in referencing this is not to try to recommend one over the other, but to point out that in actual listening tests linear phase is not always a conclusive winner.

http://archimago.blogspot.com/2015/07/the-linear-vs-minimum-phase-upsampling_10.html

I recommend that you read their 2018 update on the subject: http://archimago.blogspot.com/2018/01/audiophile-myth-260-detestable-digital.html
I don't like that they are pushing an intermediate phase filter (which is 90% linear) as it muddies the discussion, but there are a lot of problems with minimum/intermediate phase filtering which is why you won't find it on professional gear.

My recommendation is simple - provide options and let the user decide.  Many designers do that and users pick the option that sounds best to them.  The fact the human hearing works at all is pretty incredible and it is no surprise given its complexities that different people hear differently.

I'm against it because it's a waste of JRiver's finite development resources.
"Everyone's perception is different" is a terrible argument to make when you are going against objective measures and general consensus.

[dtc]

I'm against it because it's a waste of JRiver's finite development resources.
"Everyone's perception is different" is a terrible argument to make when you are going against objective measures and general consensus.

I already said in my first post on this subject that I did not expect JRiver to implement this type of option. The latest comment is a general comment, not one specific to JRiver development.

Objective versus Subjective - Yes that is exactly what the ongoing discussion is about - measurement versus actual listening. It is the same old arguments, with no resolution.

Arichmago's "update" article is a measurement article, not an update of actual listening tests. His listening test results stand. And his conclusion in the new article is based on perfectly mastered music, something that is far from true in general.

"General consensus" - only among some objectivist.

Let me provide another reference, including in the agonizing discussion on the CA thread about John Atkinson's recent article. Forget the article itself or the various arguments. The following is part of a post from Miska, a well respected participant on that forum:

"minimum phase filters are in my personal opinion better sounding. But I don't even try to force that on everyone and that's why I provide both linear- and minimum-phase versions of same filters."

https://www.computeraudiophile.com/forums/topic/49609-john-atkinson-yes-mqa-is-elegant/?page=22

I do not understand why people reject the idea that people hear differently.  Remember the recent audio that was widely circulated on which some people hear Yanny and others heard Laurel. The difference is in how each individual brain interprets the sound.   There were lots of explanations put forward about the various frequencies involved, etc.  but in the end the difference is how peoples' brains function that makes the difference. The human brain is complex and every one is different.

Enjoy the music.

[kr4]

I do not understand why people reject the idea that people hear differently.  Remember the recent audio that was widely circulated on which some people hear Yanny and others heard Laurel. The difference is in how each individual brain interprets the sound.   There were lots of explanations put forward about the various frequencies involved, etc.  but in the end the difference is how peoples' brains function that makes the difference. The human brain is complex and every one is different.

Perhaps it is semantics but this has nothing to do with "hearing differently" but with taste and preference.  If you hear reproduced music differently, then you also hear live music differently and would still need to have a linear and transparent reproduction chain in order for them to sound the same to you.  If you want something different, it isn't your hearing that underlies the choice (since it is a constant) but a subjective preference.

[dtc]

Perhaps it is semantics but this has nothing to do with "hearing differently" but with taste and preference.  If you hear reproduced music differently, then you also hear live music differently and would still need to have a linear and transparent reproduction chain in order for them to sound the same to you.  If you want something different, it isn't your hearing that underlies the choice (since it is a constant) but a subjective preference.

Yes, I was imprecise here. There are several things in play here. People do hear differently. Some hear details that others do not, and that certainly changes with age. In addition, as you point out, we have subjective preferences.  Given that people have subjective preferences, different filters are one way to satisfy those preferences.

My preference is for music from vinyls played through tube equipment. It is certainly easy to "prove" that this is "inferior" in terms of specs to digital music played through a digital system. But, vinyl and tubes still sounds better to me, as it does to many others.

[dtc]

A commonly used tool by many using MC is Dirac Live, a software package that does room correction. It helps correct for the problems introduced by the shape and contents of a room and by speaker placement. It is very helpful for both stereo and multi-channel systems. Dirac live uses a mixed phase design, which is neither linear phase or minimum phase. They believe it is the best way to accurately deliver the sound in a real world situation.

One quote from the Dirac website

"Minimum-phase and linear-phase room correction filters can’t physically optimize the acoustic impulse response in a room. At best, they can minimize problems caused by the application of a filter. Room-acoustic responses are non-minimum-phase, which is why Dirac Live uses mixed-phase correction."

Their reasoning is discussed in depth in this paper:

https://static1.squarespace.com/static/55c8a274e4b09cb562cd3ea0/t/5694b0c3d82d5e959756b637/1452585157921/on_room_correction.pdf

Take a look and decide for yourself.

[RD James]

Room correction is a completely different subject from standard audio playback, since you're doing a lot to alter the audio and it's the last DSP step.
Media Center does resampling very early in the audio chain, where a non-linear resampler would not be appropriate.

[dtc]

Let's go back to my original post, which was in response to the idea that there was little left to do with audio. I commented that some people were taking what have been traditionally DAC functions and moving them into the software player, e.g. HQPlayer.  In doing so, the software can introduce different filters that the user can select to give the the sound they like the best. That approach is fundamentally different than what MC does today. As such, I do not see JRiver adopting that approach, as I said in my original post. I simply offered it as an example of something that could be done in a software player to enhance audio functionality. HQPlayer has a large following because many people like the approach they are taking. If you think that approach is flawed, that is up to you. But many people do like the approach.

Room correction is just another example of specific capabilities that are outside what MC provides today and one that does not use linear filters.