desirable distortion

[andreasmaaan]

Well, yes, probably in the world of pop music. But, "everyone" is something of a stretch.

The minority of us who prefer classical music don't really have the same issues forced upon us. Yes, there is still an art to classical music recording, and there has been some manipulation and "sauce" added by some labels, EQ for example. But, there is still mainly an emphasis on "naturalness" for the best of classical recording, also including dynamics and spatial imaging. Those of us who go to live classical concerts have a pretty good sense of this, and we don't appreciate obvious manipulation of the recordings.

I agree with this. I think the aim of classical music recording and engineering normally must be to achieve a natural and realistic representation of the acoustical sound. But although I've never recorded in the classical field, my understanding is that quite a share of panning, track levelling, EQ, compression and sometimes other effects normally go into achieving this aim.

 

[Fitzcaraldo215]

I agree with this. I think the aim of classical music recording and engineering normally must be to achieve a natural and realistic representation of the acoustical sound. But although I've never recorded in the classical field, my understanding is that quite a share of panning, track levelling, EQ, compression and sometimes other effects normally go into achieving this aim.

It depends on the label and recording team.

There is not much to pan in classical, other than to adjust the mix for mic positioning, lest a natural sense of the performance space and hall ambiance become too distorted. Many classical releases today are done from live concerts with an audience. It's cheaper. Also, spot mics have been deemphasized by many modern labels/teams, though they may still be there among others. But, it is nothing like multitrack pop studio mixes with one or more mics per performer, artfully mixed into a contrived sense of space, which might be nothing like the performers' actual positioning in the studio. And, of course there is no singer in a booth.

Yes, of course, there is a reference level each team determines and levels to across tracks on an album, although there is no universal standard level for music as there is for video. You cannot have tracks in an album wildly at different average levels. There may also be some compression, but considerably less so in hirez. EQ, yes, but not in all cases. Recordings natively in DSD are not amenable to that without conversion to/from DXD, which is avoided except for edits and "splices".

I should say that my knowlege here has long been influenced by releases in hirez Mch primarily from small, artisan, mainly European labels. I have thousands, as well as some engineering friends in that segment of the business. I have lost touch pretty much with CD releases by major labels, and I have no engineering contacts with those.

 

[andreasmaaan]

EQ, yes, but not in all cases. Recordings natively in DSD are not amenable to that without conversion to/from DXD, which is avoided except for edits and "splices".

Good summary of the process. Agreed that there is far less going on than on pop releases.

The sentence quoted above is the only part I didn't understand. What do you mean by "not amenable"?

 

[Fitzcaraldo215]

Good summary of the process. Agreed that there is far less going on than on pop releases.

The sentence quoted above is the only part I didn't understand. What do you mean by "not amenable"?

DSD is such a tricky representational format that there are few devices that can operate on it for common operations in recording production, like EQ. Or, at least, DSD is such a small niche that Sony and others have not seen fit to develop the tools for DSD. And, they would be much more complex than their PCM equivalents. There are DSD DAWs and such, but they are more limited in functionality than PCM equivalents.

Part of the problem is that there is no instaneous signal level represented in DSD. Signal level is really a function of the cumulative integral of the bits since the beginning of the track. Each bit in DSD is sort of like a delta to the signal magnitude, but it is actually more complicated than that. DSD is just Sony/Philips' name for pulse density modulation.

But, often for editing, engineers will take a short snippet of DSD, convert that to 352.8k PCM = DXD, manipulate in PCM, then convert back to DSD. Mostly, it goes unnoticed, but engineers don't want to overdo it.

So, why do some labels and engineers love it? Beats me. But, they claim it sounds better. Others dislike it, probably mostly for the recording production limits it imposes. However, it also generally means that engineers have kept their hands off most usual, easy recording tricks, possibly resulting in a "purer" recording.

It does sound a little different to me, typically softer in the highs than hirezPCM when no EQ is applied. But, when I play SACDs, I normally convert DSD to 176k PCM and apply Dirac room correction. IMHO, that changes the sound for the better much, much more than DSD vs. PCM.

It is also true that the absolutely, most beautifully true to life recordings I have ever heard were unedited, non-commercial DSD256 Mch recordings by an engineer friend. John Atkinson, Kal Rubinson and music critic Andy Quint all agree after having heard some of them. They do sound spooky real in DSD playback, but with limitations in my system. My normal PCM 176k/Dirac playback of them is extraordinary.

 

[andreasmaaan]

Another great explanation - many thanks

Do you have any theories as to why (or even how it could be that) DSD seems to sound different to PCM to your ears?

 

[Fitzcaraldo215]

Another great explanation - many thanks

Do you have any theories as to why (or even how it could be that) DSD seems to sound different to PCM to your ears?

I think the most likely cause is audible PCM filter artifacts in a-d and d-a, possibly including time domain pre-and post-ringing. DSD does not have that, and hirez PCM has decreasingly less of it audibly with higher sampling rates above RBCD. But, my engineer friend, a Grammy-winner who is a very credible guy with an amazing system, claims he can hear it even in 176k/192k PCM vs. DSD. I don't have the program material or the interest to verify that myself.

Since I am a Mch listener, DSD has other major drawbacks for me. You cannot readily apply DSP directly to it in playback. So, no speaker distance correction, no bass management and no Room EQ is easily possible. Although, there is a PC software package, HQPlayer, that claims it can do all that in DSD. I have not attempted to play with it.

 

[andreasmaaan]

I think the most likely cause is audible PCM filter artifacts in a-d and d-a, possibly including time domain pre-and post-ringing. DSD does not have that, and hirez PCM has decreasingly less of it audibly with higher sampling rates above RBCD. But, my engineer friend, a Grammy-winner who is a very credible guy with an amazing system, claims he can hear it even in 176k/192k PCM vs. DSD. I don't have the program material or the interest to verify that myself.

Since I am a Mch listener, DSD has other major drawbacks for me. You cannot readily apply DSP directly to it in playback. So, no speaker distance correction, no bass management and no Room EQ is easily possible. Although, there is a PC software package, HQPlayer, that claims it can do all that in DSD. I have not attempted to play with it.

Very interesting, thanks again.

It seems hard to believe that high-res PCM contains audible ringing. Does anyone know if any scientific studies have been conducted on this topic?

I also seem to recall Bruno Putzleys having some misgivings concerning DSD, but don't recall what they were now. Will try to dig up the article.

Have to admit, DSD is not something I'd even looked into before this discussion.

 

[SIY]

I think the most likely cause is audible PCM filter artifacts in a-d and d-a, possibly including time domain pre-and post-ringing.

I am not the only one to observe that he couldn't hear the difference between direct mike feed and the mike feed D-A -> A-D with a PCM system. So i wonder if it's just stuff that those observing this can't hear or whether DSD is the inaccurate medium?

 

[andreasmaaan]

I am not the only one to observe that he couldn't hear the difference between direct mike feed and the mike feed D-A -> A-D with a PCM system. So i wonder if it's just stuff that those observing this can't hear or whether DSD is the inaccurate medium?

You've definitely piqued my interest in the topic

 

[Fitzcaraldo215]

Very interesting, thanks again.

It seems hard to believe that high-res PCM contains audible ringing. Does anyone know if any scientific studies have been conducted on this topic?

I also seem to recall Bruno Putzleys having some misgivings concerning DSD, but don't recall what they were now. Will try to dig up the article.

Have to admit, DSD is not something I'd even looked into before this discussion.

I think the ringing in PCM shows up routinely on scopes. The question is audibility, especially in hirez. I think scientific listening testing would be tough, though. If there is an easy way to eliminate it, and if it could be clearly shown to be audible, we would have eliminated it by now. And, if simply altered filtering could remove it, are you hearing the different filter or are you just hearing the elimination of the ringing? Heisenberg Uncertainty Principle? As I said, though, to my ears, higher rez makes it increasingly inconsequential. And, these effects are slight and they tend toward decreasing audibility.

I do think those digital filter artifacts, even in hires, may have become part of the sonic signature of digital audio as we know it, like it and accept it. In the absence of a higher level, purer, more natural standard beyond the digital recordings that we know, we might find it to sound "right" due to our preconditioning And, if it adds perhaps a slight added "crispness" to the natural sound, and possibly other things, we are OK with that. Is DSD with less of that more accurate? Who knows for sure?

I may be foolish for bringing him up, but Bob Stuart has been a maven about this for some time, especially pre/post ringing, including as part of MQA. But, Stuart, among others, hates DSD, for better or for worse, although it might appear to solve technically some of his hot button issues. Go figure.

I lose no sleep over MQA or even DSD. I am just a very happy man listening to great music recordings with by far the very best sound of my lifetime - in or converted to 48, 88, 96, 176 or 192k 24-bit PCM Mch with Room EQ. No MQA, thanks. Debates or concerns about DSD vs. PCM just don't seem to matter to me or be that significant. I do have tons of music in DSD, however, just because that is how it is distributed, and it feeds my inner Mch junkie. And, it sounds really great even after conversion to PCM and use of DSP.

 

[DonH56]

The main difference is in the respective filter circuits. Which is better is (of course) up for debate. I generally prefer PCM and steeper filters but YMMV.

There are oodles of articles about ringing in digital filters and the audibility of them. Since the ringing occurs at the transition frequency it is essentially inaudible to most of us (too high in frequency). I could relate plenty of stories either way but no point in wasting our time.

 

[andreasmaaan]

The main difference is in the respective filter circuits. Which is better is (of course) up for debate. I generally prefer PCM and steeper filters but YMMV.

There are oodles of articles about ringing in digital filters and the audibility of them. Since the ringing occurs at the transition frequency it is essentially inaudible to most of us (too high in frequency). I could relate plenty of stories either way but no point in wasting our time.

By transition frequency, do you mean nyquist frequency?

And do you recall off-hand any of those articles? I'd like to delve into this question further...

 

[Blumlein 88]

I loved the sound! It has that smoothness and lack of bite that's so hard to find.

Yes, that is what I like about such recordings.

If you have a really good well resolving system, these kind of recordings sound very interesting to me. I wouldn't expect to dump them to a memory stick and listen on my morning commute in a car to them. Nor will their best quality be found listening over poor systems.

I like SIY's comment on one of them, "no artificial sweeteners, no GMO". I think of some monkeyed with studio music as artificially sweetened or hyped or made ready to sound like it has pizazz with low quality systems or in conditions like listening in a noisey car. I get the reason for some genre's or types of music, and it fits and that is fine. What I object to is studios become so enamored of doing this you'll rarely get them to just record a singer, and guitar and leave it alone. Or record some acoustic instruments with some singers and let it be. Especially in regards to reverb and compression.

I hear SIY's complaint about the room on the Mulvey recordings. Most guys would just slap a little reverb on it right away. Next to sort of hype it and cover up the reverb they'd compress it a bit. If done with a tasteful light hand, that might not be bad. Or might help as much as it hurts. That light hand is hard to come by sometimes however.

 

[Blumlein 88]

Isn't the main thing to be able to rest assured that you're hearing exactly the type and amount of distortion that the artist intended?

Who is the artist here?

They guys playing the musical instruments? If your goal is a sound close to live, they don't know what they sound like a few feet away. Not to even get into the fact a musician has a hard time simply listening to the music like you or I or anyone else. It is almost impossible because they are listening to how they play, how they nearly messed up etc. Some experienced guys and gals come to learn about a particular sound they are shooting for. The artist's intent is one of those things that isn't nearly the definite arbitrator toward fidelity many audiophiles believe it is.

 

[Blumlein 88]

On your recordings; What we’re really discussing here, is taste. And I can guarantee you, your taste is better for you.

Yes, and no. He was there in the room and had the best location to know what that music sounded like in the room it was performed. More so than the musician. If you would like to hear what it was like had you been there listening, SIY has the best access to that.

Now would the musician's preference be that same sound? Not always. Would most listeners for entertainment prefer that kind of sound? Not always (maybe even usually not).

 

[SIY]

Well, this gets back to "high fidelity" versus "sound I like." My aim was the former, others prefer the latter. But then, I like orange juice better than Tang.

And admit it, you only liked the recordings because I used crossed ribbons.

 

[Blumlein 88]

By transition frequency, do you mean nyquist frequency?

And do you recall off-hand any of those articles? I'd like to delve into this question further...

If talking 44.1 kHz rates, you normally would like to see flat response to 20 kHz. Then you need a filter starting there, that drops response to -96 dbFS or -120 dbFS by 22,050 Hz. The transition band is 20,000 to 22,050 hz as that is the range the filter works over. Transitioning from full response to no response. So any 'ringing' is confined to those frequencies. Double that to 88.2 khz and everything doubles. So 'ringing' is only happening between 40 kHz, and 44.1 kHz. Surely we can agree we aren't going to hear what is happening at more than 40 kHz.

Below I made a track with a series of single samples with one bit at maximum. I copied it, used a brick wall filter just below 20 khz. I inverted that file and subtracted it from the original. Didn't do anything else. You get the FFT I show below. Everything at minus 200 dbFS with some ringing artifacts above 20 kHz only. Plus if you are recording music, you can't have something like the top signal as the ADC will apply filtering and remove anything that could ring in the transition band above 20 khz. That single sample impulse contains infinite frequencies, and the higher ones get removed by the ADC because not doing so violates the Nyquist limit of maximum frequencies at half the sample rate.

 

[andreasmaaan]

Who is the artist here?

They guys playing the musical instruments? If your goal is a sound close to live, they don't know what they sound like a few feet away. Not to even get into the fact a musician has a hard time simply listening to the music like you or I or anyone else. It is almost impossible because they are listening to how they play, how they nearly messed up etc. Some experienced guys and gals come to learn about a particular sound they are shooting for. The artist's intent is one of those things that isn't nearly the definite arbitrator toward fidelity many audiophiles believe it is.

See my posts above. I don't define the artist simply as the musicians creating the performance. Rather, I define it as the creator(s) of the recording (including but not necessarily limited to any performers). To me, the recording is the artwork that the audio system reproduces, not the performance.

 

[andreasmaaan]

If talking 44.1 kHz rates, you normally would like to see flat response to 20 kHz. Then you need a filter starting there, that drops response to -96 dbFS or -120 dbFS by 22,050 Hz. The transition band is 20,000 to 22,050 hz as that is the range the filter works over. Transitioning from full response to no response. So any 'ringing' is confined to those frequencies. Double that to 88.2 khz and everything doubles. So 'ringing' is only happening between 40 kHz, and 44.1 kHz. Surely we can agree we aren't going to hear what is happening at more than 40 kHz.

Below I made a track with a series of single samples with one bit at maximum. I copied it, used a brick wall filter just below 20 khz. I inverted that file and subtracted it from the original. Didn't do anything else. You get the FFT I show below. Everything at minus 200 dbFS with some ringing artifacts above 20 kHz only. Plus if you are recording music, you can't have something like the top signal as the ADC will apply filtering and remove anything that could ring in the transition band above 20 khz. That single sample impulse contains infinite frequencies, and the higher ones get removed by the ADC because not doing so violates the Nyquist limit of maximum frequencies at half the sample rate.

Fantastic answer to my question and exactly what I was imagining, articulated much more clearly than I was able to - thanks.

 

[DonH56]

By transition frequency, do you mean nyquist frequency?

And do you recall off-hand any of those articles? I'd like to delve into this question further...

Fortunately for me @Blumlein 88 did an excellent job of explaining it. It is usually around Nyquist but does not have to be, and remember Nyquist can be much higher for oversampled converters (delta-sigma or otherwise).

And no, I don't have any articles bookmarked, sorry. I have all sorts of reference material but it is more technical than usually appropriate for here, and is mostly in the form of reference books or IEEE articles. I have done web searches now and then but, since I have worked in the data converter area for a long time, I have my own pile of references I tend to go to when needed. I have several grad school books on filters that are fairly old now but the basics have not changed, plus a pile of books and articles accumulated since then. Someplace in the black holes laughingly referred to as "my office" and our "storage shed".

Aside: It occurs to me that the use of noise decorrelation (dither) could be considered to be a form of "desirable distortion". I consider it noise rather than distortion but figured it was worth a mention as one of the non-musical things we add.