Does "one drop" pf PCM ruin DSD recordings?

[Jud]

This is a mistake - both the Geek Out and the iFi Nano iDSD are excellent DSD-capable DACs. I have both (+ the iFi Nano's big brother, the Micro) and any would let you hear DSD in all its glory

 

Thanks John, I could only speak to the Geek Out, and was about to say the same thing about it. I prefer my Semi-Customized DAC to the Geek Out, but that is not at all surprising.

[wisnon]

So let's say this is allowed to lead its own life. And why not, because it's legitimate, no matter springing from my typo;

 

So Jud, you are familiar with iZotope small chanGes, right ? Of course I am too with changing filters and such. Difference : I look through the analyzer before I am satisfied, and what I'm satisfied with is what I see (knowing what I want to achieve).

Visual changes can be very very minor, which is also how I tune my own (new) filtering these days.

The audible difference ? 2 second work, no matter what track (so to speak).

 

And then we might think that it doesn't matter all that much when turning the whole track (yes) upside down and imply changes (visually huge now) and even hope for the better when all sprung from some original, no matter that was "mangled" to begin with ?

 

Still, in my view, when PCM is converted to DSD the right way, DSD has a chanCe, because all what really happens is that we upsample AND filter further than PCM can do it (like my 16x in practice). But I said "the right" way and this is not how I see it done. But never mind that for now, because it's about the possibility for the better.

If this "right way" is not applied, we end up with piles of noise out of the audio band, and next hope that or our amps won't notice and if they do our speakers won't notice. And might we prevent that with a nice analogue filter we hope that "we" don't notice.

 

There's so much difference in the both means if you look at the resulting signal that we're better off if we don't hear the difference anyway (could be Fritz). But if we do, it beats me how it can be for the better. All IOW : no single "drop" allowed if I was asked the question. But I said that in my first post right away.

 

Please notice : this all is totally unrelated to when all can be kept native (either DSD or PCM) and is played through a DAC which doesn't mangle with it.

This is basically what I told him from the beginning.

 

I dont know where he got $10K from (I certainly never mentioned it, and I own a iFi Micro). If he must know, he can go get an iDSDNano for $200 and use HQP on trial and see what DSD sounds like on a capable player, both native and upsampled. If however, as you say, he is satisfied with converting all to PCM, then no need to bother with anything.

 

However, throwing up the $10k red herring makes you wonder why he started the topic in the first place, as he seems unwilling to experiment and experience for himself what the differences are!

[jabbr]

But I don't know what a "reasonable DAC that can do DSD" is. The UFO that Blue Coast sells? DirecStream DAC from PSAudio? I've put the $200 ones you cited from Amazon in the category of JRiver and maybe HQPlayer as being not good enough to represent the format.

 

I've just started using the iFi iDSD micro combined with HQPlayer and its great. Moreover it allows you to compare either DSD or PCM source with either DSD or PCM output and upsampling. One thing I've found is that I like my CD Redbook (16/44) source better upsampled and converted to DSD ... on this particular DAC.

 

So from my standpoint having some PCM somewhere in the chain hasn't ruined any recordings for me, rather on this particular reasonably priced DAC, conversion to DSD seems to help. I haven't studied alot of direct DSD recordings but for material transferred from master tapes in both formats (I have both the 24/192 and SACD/DSD versions of "Kind of Blue" for example), when played through HQPlayer to the iDSD, I don't hear a tremendous difference (both great).

[Jabs1542]

Funny but it appears that just about everyone here is offering the same advice - a very rare occurrence . Give DSD a good listen and see if you "get it", and if you do do you like what you get - otherwise stick with PCM and be done.

[fritzg]

So from my standpoint having some PCM somewhere in the chain hasn't ruined any recordings for me, rather on this particular reasonably priced DAC, conversion to DSD seems to help. I haven't studied alot of direct DSD recordings but for material transferred from master tapes in both formats (I have both the 24/192 and SACD/DSD versions of "Kind of Blue" for example), when played through HQPlayer to the iDSD, I don't hear a tremendous difference (both great).

 

Thanks. I am always willing to experiment, and this is helpful.

[fritzg]

This is a mistake - both the Geek Out and the iFi Nano iDSD are excellent DSD-capable DACs. I have both (+ the iFi Nano's big brother, the Micro) and any would let you hear DSD in all its glory

 

Very helpful, thank you.

[fritzg]

This is basically what I told him from the beginning.

 

I dont know where he got $10K from (I certainly never mentioned it, and I own a iFi Micro). If he must know, he can go get an iDSDNano for $200 and use HQP on trial and see what DSD sounds like on a capable player, both native and upsampled. If however, as you say, he is satisfied with converting all to PCM, then no need to bother with anything.

 

However, throwing up the $10k red herring makes you wonder why he started the topic in the first place, as he seems unwilling to experiment and experience for himself what the differences are!

 

Untrue and not helpful.

[sdolezalek]

http://old.hfm-detmold.de/eti/projekte/diplomarbeiten/dsdvspcm/aes_paper_6086.pdf

 

Nice attempt to determine audibility here. Blind test, lots of participants. Excellent equipment, all tracks recorded concurrently in DSD and 176/24 with converters from the same manufacturer. Was an AES Paper in 2004.

 

Dennis: Nice reference and interesting paper. I was disappointed that the authors seemed to ignore (or at least not pursue further) two aspects of the only data that didn't support the conclusion. The first is that the only place a difference was heard was in stereo (on headphones) and never in surround mode. The second is that the recordings in which a difference was heard involved a single instrument and specifically two of those four were the human voice (speech and song).

 

Because of how familiar we are with voices, it would not be surprising that if you were going to hear a differences voices are where you would hear it. But the stereo versus surround difference might point to a different direction -- that the differences we do/can hear between formats are tied to minute timing differences. I say that because a surround recording should maintain the full frequency of the format, but having that many more speakers increases the likelihood of timing differences that should obscure the sound even as they otherwise might give more accurate directional clues. A pair of headphones makes it far more likely that what you hear, from a timing standpoint, is exactly what was recorded.

 

That, in turn might point to impulse response and ringing as a significant culprit in sound degradation. We certainly know from a lot of work that Magico has done in their speaker design that cabinet resonances can make a huge difference in speaker clarity. Which takes us full circle to the discussion we were having in the discussion about cable break-in -- namely, is the difference between 16/44 and hi-res and between PCM and DSD almost all a function of filtering and the effects of filtering on impulse response and timing, not frequency?

[esldude]

Not the old impulse and timing thing again. Jumping from the particulars of that test to that conclusion seems a leap of some considerable size. We only know the stereo recording used two microphones. We don't know the spacing of the mics or whether it would encode much timing. But it may have. Headphones also block out some of the room sound, might have made the slight glitch in switching more obvious (and often one gets positive ABX results even if consciously unaware of a slight difference).

 

Yes, the particulars of the 4 who correctly picked at least 15 of 20 are interesting. On the other hand, you would expect 3 of 145 partipants to score 75% or better and 3 score 25% or less. Notice 5 participants scored 25%. Just as a simple exercise I filled 2900 cells (145 x 20) of a spreadsheet that would randomly generate a number then rounded to show a correct vs incorrect choice randomly. Just grouped them 20 at a time. In ten such trials, the number with 15 or the number with only 5 was between 2 and 7 in all ten trials.

 

It is interesting that all 4 were tonemeister students. Yet none of the professional tonmeisters. There were 32 such students among the 145 participants.

 

So this is not good evidence that DSD and high sample rate PCM are discernably different. It would have been nice to retest the 4 successful participants. They weren't able to do so. That all were young, and tonmeister students certainly points to trained listening skills and young undamaged hearing possibly being a factor in their success. On the other hand, those 4 plus 5 at the other end of the scale are right near what you would expect purely at random from that number of participants.

[Hiro]

^It's probably high time somebody trotted out the Meyer & Moran study showing there's no difference between CD and hi-res.

[sdolezalek]

Not the old impulse and timing thing again. Jumping from the particulars of that test to that conclusion seems a leap of some considerable size. We only know the stereo recording used two microphones. We don't know the spacing of the mics or whether it would encode much timing. But it may have. Headphones also block out some of the room sound, might have made the slight glitch in switching more obvious (and often one gets positive ABX results even if consciously unaware of a slight difference).

 

Yes, the particulars of the 4 who correctly picked at least 15 of 20 are interesting. On the other hand, you would expect 3 of 145 partipants to score 75% or better and 3 score 25% or less. Notice 5 participants scored 25%. Just as a simple exercise I filled 2900 cells (145 x 20) of a spreadsheet that would randomly generate a number then rounded to show a correct vs incorrect choice randomly. Just grouped them 20 at a time. In ten such trials, the number with 15 or the number with only 5 was between 2 and 7 in all ten trials.

 

It is interesting that all 4 were tonemeister students. Yet none of the professional tonmeisters. There were 32 such students among the 145 participants.

 

So this is not good evidence that DSD and high sample rate PCM are discernibly different.

 

Dennis: I agree. If anything the study validly suggests they are not. It just struck me that IF there might actually be any difference, these would be the testers and the tests most likely to hear it. I then made the leap to asking whether if they actually heard anything, it was more likely a timing difference than a frequency difference.

 

I guess what I'm really asking you, because I think you make valid cases for their being no greater than 20kHz benefits from hi-res, whether you equally strongly believe that there can be no timing differences or other distortions, particularly to the extent those are introduced by the D/A and A/D filtering methods.

[Jud]

Funny but it appears that just about everyone here is offering the same advice - a very rare occurrence . Give DSD a good listen and see if you "get it", and if you do do you like what you get - otherwise stick with PCM and be done.

 

Knew it was too good to be true, didn't we?

[christopher3393]

^It's probably high time somebody trotted out the Meyer & Moran study showing there's no difference between CD and hi-res.

 

Here ya go: Don't Buy What Neil Young Is Selling

 

Excerpt: "Though Young and Pono have failed to produce double-blind studies on the benefits of high-rate audio or their music player, inquiring minds have taken the time to do it. In a 2007 paper published in the Journal of the Audio Engineering Society, Brad Meyer and David Moran outline the results of a study in which they presented a large sample of "serious" listeners with a double blind test comparing 44.1 kHz audio from "the best high resolution discs we could find." The goal was not to show which was better, but simply to find out if people could even tell the difference.

 

"None of these variables have shown any correlation with the results, or any difference between the answers and coin-flip results," they write in their conclusion. Later they note, "Further claims that careful 16/44.1 encoding audibly degrades high-resolution signals must be supported by properly controlled double-blind tests."

 

This is how you do science. It's incredible to me the lengths that educated and intelligent people will go to say that they're somehow endowed with impossible hearing powers that necessitate a level of audio encoding that's demonstrably unnecessary."

 

And here's Michael Lavorgna's response: Gizmodo's Garbage Dump on Pono | AudioStream

 

...and the beat goes on...

[kumakuma]

Funny but it appears that just about everyone here is offering the same advice - a very rare occurrence . Give DSD a good listen and see if you "get it", and if you do do you like what you get - otherwise stick with PCM and be done.

 

I disagree. I think he should decide on the technical merits of the respective formats.

[tailspn]

I disagree. I think he should decide on the technical merits of the respective formats.

 

And other Expert advice

[wisnon]

And other Expert advice

 

Hehehehehehehe.

[esldude]

Dennis: I agree. If anything the study validly suggests they are not. It just struck me that IF there might actually be any difference, these would be the testers and the tests most likely to hear it. I then made the leap to asking whether if they actually heard anything, it was more likely a timing difference than a frequency difference.

 

I guess what I'm really asking you, because I think you make valid cases for their being no greater than 20kHz benefits from hi-res, whether you equally strongly believe that there can be no timing differences or other distortions, particularly to the extent those are introduced by the D/A and A/D filtering methods.

 

I find it hard to see where a timing problem would be. Even the 3 year old phone in my pocket can time to better than the nearest nanosecond. It seems very unlikely the timing of any decent AD/DA conversion is an issue.

 

I would grant some higher probability to filtering being a sometimes issue. I nevertheless have strong doubts about that as well. For one, if filtering is an audible issue, simply going to 96 khz would seem to be more than sufficient to fix it. No need for some of the other wild stuff that has been tried. With people who had a firm opinion digital has problems digital recording an LP and letting them hear it on a high quality system has given them pause to wonder if there is anything wrong. (note I don't think LP is a reference myself it has too many failings)

[firedog]

I wasn't sure what you point is. But if you are trying to say that Meyer and Moran is good science, it is far from it. Holes all over it and through it.

 

Here ya go: Don't Buy What Neil Young Is Selling

 

Excerpt: "Though Young and Pono have failed to produce double-blind studies on the benefits of high-rate audio or their music player, inquiring minds have taken the time to do it. In a 2007 paper published in the Journal of the Audio Engineering Society, Brad Meyer and David Moran outline the results of a study in which they presented a large sample of "serious" listeners with a double blind test comparing 44.1 kHz audio from "the best high resolution discs we could find." The goal was not to show which was better, but simply to find out if people could even tell the difference.

 

"None of these variables have shown any correlation with the results, or any difference between the answers and coin-flip results," they write in their conclusion. Later they note, "Further claims that careful 16/44.1 encoding audibly degrades high-resolution signals must be supported by properly controlled double-blind tests."

 

This is how you do science. It's incredible to me the lengths that educated and intelligent people will go to say that they're somehow endowed with impossible hearing powers that necessitate a level of audio encoding that's demonstrably unnecessary."

 

And here's Michael Lavorgna's response: Gizmodo's Garbage Dump on Pono | AudioStream

 

...and the beat goes on...

[Miska]

I find it hard to see where a timing problem would be. Even the 3 year old phone in my pocket can time to better than the nearest nanosecond. It seems very unlikely the timing of any decent AD/DA conversion is an issue.

 

Interesting detail regarding timing is that there certainly are timing consistency differences between different formats when reconstructed by a DAC.

 

Arbitrarily placed transient has varying level of waveform consistency depending on used format.