Minimum Phase vs Linear Phase

[scott wurcer]

Just to interject a comment the minimum vs. linear phase discussion is somewhat pointless removed from the context of the application at hand. DAC reconstruction/interpolation filters are far from crossover networks. This came up when discussing digital RIAA. LP's historically are pre-emphasized using standard R/C or R/L/C networks and must be minimum phase. The suggestion that using linear phase de-emphasis "fixes" something is simply wrong.

 

[maty]

The suggestion that using linear phase de-emphasis "fixes" something is simply wrong.

Who has suggested it? In this thread.

 

[scott wurcer]

Who has suggested it? In this thread.

Sorry, not here but the conversation started out talking about DAC's and the same issues applied to speaker crossovers should be a different conversation IMO.

 

[MRC01]

That's an interesting question.

If you're using playback EQ to correct a minimum phase anomaly, like some room modes, a minimum phase EQ seems ideal since it will correct (reverse) both amplitude and phase.

But the anti-aliasing reconstruction filter of a DAC is not correcting a minimum phase anomaly. You don't know what kind of AA filter was used when encoding the recording. And even if you did, the playback filter isn't reversing what that filter did. Both of these filters (encoding, and reconstruction), are doing the same thing: eliminating frequencies above Nyquist without affecting anything below it. I see this as a completely different application than playback EQ.

 

[maty]

@scott wurcer

The problem is that the defenders of the linear filters are for all types of use, not only for the DAC filters. From the beginning I said that the choice of one or the other would depend on many things, one of them the recording. And that they are easily differentiable (at least in my second system).

It is too late here, bye.

 

[Ron Texas]

Sorry, not here but the conversation started out talking about DAC's and the same issues applied to speaker crossovers should be a different conversation IMO.

I did not limit the discussion to DAC's. My examples covered three different kinds of digital signal processing and if I had thought about it I would have included crossovers either active or passive. My understanding is passive crossovers are minimum phase and analogue equalizers are too. However, someone with more expertise can chime in on that definitively.

 

[Guermantes]

@scott wurcer

The problem is that the defenders of the linear filters are for all types of use, not only for the DAC filters. From the beginning I said that the choice of one or the other would depend on many things, one of them the recording. And that they are easily differentiable (at least in my second system).

It is too late here, bye.

Who is saying that linear filters are ideal for all types of use? This "problem" sounds like a straw man argument to me.

@MrC and @scott wurcer are talking sense: There are situations where one or the other is ideal while other times the differences are more theoretical than critically audible.

 

[maty]

In other threads and forums. It is a recurring discussion in which until now I had decided not to participate. And so many other audio discussions, in which a few always extreme their positions, widespread phenomenon since the massification of the Internet on any subject.

 

[mansr]

I did not limit the discussion to DAC's. My examples covered three different kinds of digital signal processing and if I had thought about it I would have included crossovers either active or passive. My understanding is passive crossovers are minimum phase and analogue equalizers are too. However, someone with more expertise can chime in on that definitively.

Several applications of filters have been mentioned: reconstruction filter, equaliser, and speaker crossover. Each serves a different purpose and thus has different requirements.

A reconstruction filter, also known as interpolation filter, recreates the waveform represented by the sample values. This is achieved by filtering out all the image frequencies above Nyquist while doing minimal damage to the spectral content below this cut-off. A steep linear phase filter is clearly ideal here. So-called "pre-ringing" is a non-issue since its frequency is above the audible range.

An equaliser is different since it operates within the audible range. A steep cut here can indeed cause audible "ringing" artefacts. Using a minimum phase filter helps ensure they are masked by the rest of the signal. Even then, cuts should not be too steep or too deep, or else things can get weird.

Speaker crossovers are a different category again. While also operating in the audible range, their purpose is to split the signal into two or more bands to match the properties of the drivers. When the bands are recombined, any "ringing" from linear phase filters will be cancelled out, even if audible when one band is listened to in isolation.

 

[maty]

A reconstruction filter, also known as interpolation filter, recreates the waveform represented by the sample values. This is achieved by filtering out all the image frequencies above Nyquist while doing minimal damage to the spectral content below this cut-off. A steep linear phase filter is clearly ideal here. So-called "pre-ringing" is a non-issue since its frequency is above the audible range.

Then a DAC only needs two linear filters, slow and fast. The theory is right but the reality is other I am afraid. The sound may be cleaner and spectacular but if we lose the ability to transmit emotion, unlike what happens with the other filter, then it is a pyrrhic gain, we lose in the most important.

I have heard very well adjusted rooms with REW and UltraCurve, with recordings that sounded great right at the listening point. After the session, I play my reference themes, of all kinds of music and recording quality but usually with high/very high DR (very good rips from vinyl, CD and SACD), and then the fault lies with the music that I carry in the USB memory! Speakers with one or more subwoofers.

Probably very aggressive adjustments have been made and throughout the audible band. Music with synthesized instruments sound great but my problem is that most of the music that I carry in the USB memory is with acoustic instrumentation (some is electrified) and voices without Autotune.

And, of course, I am unable to differentiate between MP3 at maximum quality and FLAC 16/44, both files created by me carefully. I return to my house, and I have no problem differentiating them in my two audio systems.

 

[SIY]

And, of course, I am unable to differentiate between MP3 at maximum quality and FLAC 16/44, both files created by me carefully. I return to my house, and I have no problem differentiating them in my two audio systems.

No ABX logs, can we assume?

Do you ever wonder about why no-one takes you seriously? Could I suggest a way for you to fix that and gain at least a smidgen of credibility instead of being automatically dismissed as a wordy nutcase? I'd certainly rather have you posting useful and correct things and enjoying the interactions with people who could have some respect for you.

 

[MRC01]

Then a DAC only needs two linear filters, slow and fast. The theory is right but the reality is other I am afraid. The sound may be cleaner and spectacular but if we lose the ability to transmit emotion, unlike what happens with the other filter, then it is a pyrrhic gain, we lose in the most important.
...
Music with synthesized instruments sound great but my problem is that most of the music that I carry in the USB memory is with acoustic instrumentation (some is electrified) and voices without Autotune.
...
And, of course, I am unable to differentiate between MP3 at maximum quality and FLAC 16/44, both files created by me carefully. I return to my house, and I have no problem differentiating them in my two audio systems.

I find that the music's ability to transmit emotion depends more on the listener than the playback system. I keep a database of music listening session notes (both technical & artistic) and over the years have observed that my perception of the sonics varies a little, while my perception of the artistic expression of the music varies a lot. This suggests that for me, critical listening to the system is mutually exclusive with listening to the music. In other words, it's hard to appreciate the subtleties of artistic expression when I'm focusing on hearing nitnoid little differences in the sonics. A good system is more transparent and steps aside revealing more of the music, making it easier to appreciate it artistically. But that is a much wider range of "good enough" than most audiophiles admit.

If you're serious about ABX testing, that's largely a learned skill. Trained listeners consistently outperform others. The software (brain/learning) is more important than the hardware (ear/hearing apparatus), except in obvious cases of deficient or damaged hearing. Some here are good at it, and would offer advice and references. But that's a whole 'nuther subject and thread.

The choice of material, electronic vs. acoustic, is another interesting subject. Electronic music can be better for A/B testing since it can use the full frequency spectrum and also have unusual transients -- almost like using test signals to highlight differences that are hard to hear in music. But electronic music has no absolute reference of realism or what it should sound like. Acoustic music can give that reference, but it tends not to have much energy in the very top & bottom octaves which can help differentiating A from B.

 

[maty]

Emotion

I just heard the end of Tchaikovsky's 5th on Radio Clásica (RNE), recorded in Amsterdam last month (Proms festival, maybe the August, 23). Spectacular sound and interpretation, I have started to lead the orchestra. And that was from the FM radio. With my little coaxial. I have been listening to that station and Catalunya Música (FM) for many years, they are my traditional streaming equivalents. And I learn about ancient, baroque, classical music ...

ABX test

It was made with foobar2000 and spectacular skin. I did not know if it was MP3 or not. I struggled but did not succeed, as expected by the other person. After that I did not even try 16/44 FLAC vs 24/96 FLAC. On that occasion with the UltraCurve without the appropriate modification of the analog section, something that should be imperative, I say.

Acoustic instruments

I have heard a lot of live music, especially traditional Catalan, Basque and Celtic. Less classical - especially for the OMB - and opera. Over the years, interpretation quality matters more to me than the sound quality, if not is bad! I prefer the imperfection of Franco Corelli that compensates with his enormous capacity to transmit to the sublime technical perfection of Alfredo Krauss. FC is like my cheap/modded AVR Marantz SR4500 and AK is like a modern class D amp

 

[Ron Texas]

Do you ever wonder about why no-one takes you seriously?

That's quite a claim. How do you plan on proving it? Maty can be verbose at times, his English needs improvement, but he has occasional nuggets of worthwhile information. He is an avid DIY enthusiast which puts more skin in the game than just pulling out your Mastercard.

 

[Julf]

That's quite a claim. How do you plan on proving it? Maty can be verbose at times, his English needs improvement, but he has occasional nuggets of worthwhile information. He is an avid DIY enthusiast which puts more skin in the game than just pulling out your Mastercard.

He makes endless claims and never provides evidence to support his conjectures. It gets old.

 

[Thomas savage]

Can we dial back the personal attacks and member critique please.

Cheers

 

[MRC01]

... Electronic music can be better for A/B testing since it can use the full frequency spectrum and also have unusual transients -- almost like using test signals to highlight differences that are hard to hear in music. ... Acoustic music can give that reference, but it tends not to have much energy in the very top & bottom octaves which can help differentiating A from B.

Now that I said that, I realize I left out an important exception: I find there is nothing like acoustic music to assess how realistic and natural a system sounds, which for me is primarily about its midrange voicing. I've heard systems that make it easy to hear subtle differences between recordings, yet are difficult to listen to because they sound so unnatural. One example of this is the Sennheiser HD800 headphones. They're detailed and low distortion, making it easy to hear subtle differences. But they are so artificially bright and with perceptually uneven frequency response, I find them impossible to listen to for musical enjoyment. They make acoustic music sound like a caricature of the real thing.

[I don't mean to single out the HD800, it's just one of several different examples I've encountered over the years, but they're popular enough that I think people can relate to it.]

Playing good recordings of human voice, piano, and a variety of different acoustic instruments may be less effective for discerning subtle differences between systems, but it is still an essential part of my auditioning process because it determines whether I can actually live with this system, listening for musical enjoyment. Thus I use both kinds of recordings to get the full picture when assessing a system. The ideal system does well with both.

 

[SIY]

ABX test

It was made with foobar2000 and spectacular skin. I did not know if it was MP3 or not. I struggled but did not succeed, as expected by the other person. After that I did not even try 16/44 FLAC vs 24/96 FLAC.

Congrats, that's big step. And you actually did succeed- now you have one less thing to worry about and you KNOW instead of BELIEVE.

 

[MRC01]

Congrats, that's big step. And you actually did succeed- now you have one less thing to worry about and you KNOW instead of BELIEVE.

Yet knowledge is not a "place", but a "path". It moves forward. This is just the first step. Now, @maty can continue for example by encoding the original track at different MP3 quality levels. Start with something easy to differentiate like a rate of 64k or VBR level 9. Then incrementally walk up the quality chain, doing progressively more difficult comparisons. I found this an effective way to learn critical listening skills.

Also you can find online blind listening test training guides having useful tips. Some of it is about where to focus your hearing and what to listen for. Some of it is about mentally disabling the "filters" we normally use when listening to music, making yourself aware of everything your ears are telling you but your brain was filtering out or suppressing.

Oops, I just forked this thread into something I said earlier was a whole 'nuther topic.

 

[maty]

I was going to write a long off-topic. In short, everything you have been thinking about more than ten years ago that I did, including tests with different compressions in FLAC and 16/44 vs.... 24/192 FLAC. And OGG Vorbis -q9 is my choice to the phone, from 24/96 or 24/192 FLAC. BTW, OGG at 24 bits and about 350 kbps the last Sunday. Stravinsky Conducts Music For Chamber And Jazz Ensembles
https://www.audiosciencereview.com/...-are-we-listening-to-right-now.40/post-224277