The Ear - DSD under Fire

[mansr]

I got noisy but undistorted sound out of 96/8 PCM by using subtractive dither. There was no recognizable distortion on the music, but there was an obvious amount of benign white noise, about at the level of a typical 4 track prerecorded tape from the 1960's.

 

Dithering converts distortion to uncorrelated inband noise. It didn't seem important to go into that detail in the discussion leading up to 1-bit DSD and sigma-delta modulation. Perhaps I should have.

[tailspn]

Agreed. As with Redbook CD, Sony made the mistake of releasing a technology prematurely and as a result with too low a sampling rate.

 

Hi Tony,

 

Would you have preferred Sony have waited for Blu-ray to be commercially available? What was that, five years ago?

 

Both Sony and Philips had hardware running at higher than 64fs in the late 90's, Sony at 128fs, and Philips at 256fs. The decision to release 64fs was entirely predicated on the only available delivery technology; The DVD with its limit of 4.7GB. Of course they could have limited the distribution to stereo only, but that was seen as a so-so new technology introduction with the sound quality advantages of 5.0/5.1 multi-channel.

 

It's my belief, and theirs, that had they waited another 10 years for a supportive delivery media (Blu-ray), the much higher resolution and sound quality of DSD compared to the competing 196 KHz PCM DVD-A, never would have happened.

[mansr]

Hi Tony,

 

Would you have preferred Sony have waited for Blu-ray to be commercially available? What was that, five years ago?

 

Blu-ray came on the market in 2006.

 

Both Sony and Philips had hardware running at higher than 64fs in the late 90's, Sony at 128fs, and Philips at 256fs. The decision to release 64fs was entirely predicated on the only available delivery technology; The DVD with its limit of 4.7GB.

 

They could have used dual-layer DVD.

 

Of course they could have limited the distribution to stereo only, but that was seen as a so-so new technology introduction with the sound quality advantages of 5.0/5.1 multi-channel.

 

Of course, they didn't know then that surround music would never catch on.

 

It's my belief, and theirs, that had they waited another 10 years for a supportive delivery media (Blu-ray), the much higher resolution and sound quality of DSD compared to the competing 196 KHz PCM DVD-A, never would have happened.

 

Are you saying DSD64 has "much higher resolution" than 196/24 PCM? That's a statement I don't quite agree with.

[Hiro]

Both Sony and Philips had hardware running at higher than 64fs in the late 90's, Sony at 128fs, and Philips at 256fs.

 

I've heard that Gus Skinas at the SA Center is currently using the 128fs Sony unit. I wonder if anyone out there is (or could get) in possession of the fabled 256fs ADC built by Philips...

[Daudio]

My function vs. mechanism comment ...

Function: Draining the swamp.

Mechanism: keeping one's ass away from alligators.

 

I would reframe that as:

 

Goal: keeping one's ass away from alligators.

Function: Draining the swamp.

Mechanism: A big pump with a short input pipe and a very long output pipe.

[Hiro]

Are you saying DSD64 has "much higher resolution" than 196/24 PCM? That's a statement I don't quite agree with.

 

DVD-Audio could only store 24/96 pcm in multi channel.

[Stereolab42]

In the event that we need to perform multiple mathematical operations on the signal, we must choose an intermediate storage format with sufficient headroom that the added noise can be ignored. A common choice is to use a few more bits per sample than the final product will have, e.g. using 24-bit intermediates when producing a 16-bit CD. If we insist on using 1-bit intermediates, this too is possible by using a higher sampling rate than intended for the final product. Done right, the end result can be as accurate as we wish whichever route is chosen. The only real difference is that 1-bit intermediates require substantially more computational effort to produce an equivalent result.

 

Thanks for the explanations Tony and mansr. I did have romantic notions that DSD was "closer to the analog" but I realize that's not the case in any philosophical sense. (I don't think I'm the only one hoping that the "magic" of analog/vinyl could somehow be captured by digital computers.) It can only be considered closer in the sense that there may be fewer (possibly none) conversion steps between the output of a particular ADC and storage as DSD. But I'm no longer spooked by conversion from DSD to other formats as long as you indeed have enough headroom in the target format to avoid any audible noise/artifacts.

[tailspn]

Are you saying DSD64 has "much higher resolution" than 196/24 PCM? That's a statement I don't quite agree with.

 

Much higher. It's even noticeably higher sound quality resolution than 352.8KHz 24 bit PCM. I work with with DSD and DXD every day, switching between the two, as do many producers who also record in, and favor native DSD. It's apparent there, and more so at half the PCM rate of DXD. You in the recording biz mansr? Any experience recording in DSD, or in any position to hear content in the various stages of production?

[tailspn]

I did have romantic notions that DSD was "closer to the analog" but I realize that's not the case in any philosophical sense.

 

God that's a relief, I was worried! You're too easy Stereolab42. Get some experience of hearing a well set up for music multichannel system, and compare the free files we offer on nativedsd.com in the various formats. Then perhaps you'll understand the relationship to real analog signals, and not the bastardized use of the term to refer to vinyl and tape.

[Jud]

My function vs. mechanism comment was based on my experience as a computer network architect with a computer science/mathematics background. I note that similar distinctions are in use in biology and ecology, among other fields. My experience as an expert/consultant in computer networking patent disputes came later. The basic problem is that most people do not handle abstractions willingly, and this even extends to many engineers.

 

Function: Draining the swamp.

Mechanism: keeping one's ass away from alligators.

 

Different world views, I guess - I would think of draining the swamp as the mechanism that performs the function of keeping one's ass away from alligators. Oh well, so much for communication between an engineer and a lawyer.

[徐中銳]

The latest Cat Stevens DSD download offered by Analogue Productions...

 

Greetings Hiro,

I'll voice this thought to our polyphony, I believe the idea of recording in PDM should be introduced to live recording enterprises such as, to name one that came immediately to mind, NPR's Tiny Desk. Have you seen their Yusuf/Cat Stevens edition ?

 

And is NPR's own JW Player (link here) better sounding on your system than that of YouTube's re-encoding ?

 

Of course, I'm thinking that their outlay in turning to PDM will be recouped with popular demands by monetising DSD downloads of their Pop...

 

But if PCM proponents convince their management...

 

Only time will tell if purist DSD will flourish...

[Hiro]

Greetings Hiro' date='[/font']

I'll voice this thought to our polyphony, I believe the idea of recording in PDM should be introduced to live recording enterprises such as, to name one that came immediately to mind, NPR's Tiny Desk. Have you seen their Yusuf/Cat Stevens edition ?

 

And is NPR's own JW Player (link here) better sounding on your system than that of YouTube's re-encoding ?

 

Of course, I'm thinking that their outlay in turning to PDM will be recouped with popular demands by monetising DSD downloads of their Pop...

 

But if PCM proponents convince their management...

 

Only time will tell if purist DSD will flourish...

 

Hi 徐中銳,

 

I haven't seen/heard this NPR concert, but heard many others in the series that I greatly enjoyed, by Yasmine Hamdan, Susanne Vega, and others. I agree with you that having those performances in DSD would be very nice indeed, as DSD is an ideal choice for this type of acoustic recordings, when you don't need any special effects/gimmicks, and just want to hear the real music from analog mic feeds.

 

Will give it a listen!

[mansr]

Are you saying DSD64 has "much higher resolution" than 196/24 PCM? That's a statement I don't quite agree with.

 

That "196" was obviously a typo, though it wasn't obvious whether it should have been 96 or 192, and somehow I propagated it without realising. Not that it matters much; DSD64 can at best be considered equal in fidelity to 96/24 PCM.

[mansr]

Much higher. It's even noticeably higher sound quality resolution than 352.8KHz 24 bit PCM. I work with with DSD and DXD every day, switching between the two, as do many producers who also record in, and favor native DSD. It's apparent there, and more so at half the PCM rate of DXD. You in the recording biz mansr? Any experience recording in DSD, or in any position to hear content in the various stages of production?

 

No, but I have a pretty good understanding of the mathematics involved.

[Hiro]

That "196" was obviously a typo, though it wasn't obvious whether it should have been 96 or 192, and somehow I propagated it without realising. Not that it matters much; DSD64 can at best be considered equal in fidelity to 96/24 PCM.

 

Are you serious? PCM at 96khz (2fs) is still subject to brickwall filtering, rounding, ringing, aliasing. A 2.8MHz DSD converter, on the other hand, such as Grimm AD1 is absolutely stellar sounding with THD @ -129db and ultra-low jitter. It's really a force to be reckoned with, and anyone who's heard it will vouch for that.

[mansr]

Thanks for the explanations Tony and mansr. I did have romantic notions that DSD was "closer to the analog" but I realize that's not the case in any philosophical sense. (I don't think I'm the only one hoping that the "magic" of analog/vinyl could somehow be captured by digital computers.)

 

96/24 PCM surpasses both vinyl and studio-grade tape. Higher-resolution formats are way beyond any analogue storage medium.

 

Of course, a digital file on a computer can never capture the "magic" of pulling a 12" vinyl from its shelf, admiring the cover art for a moment, carefully taking the disc out of the sleeve, placing it on the turntable, and finally dropping the needle. If that ritual is important to you, by all means keep playing vinyl, but do not delude yourself into thinking it actually sounds better than a high-resolution digital file.

[mansr]

Different world views, I guess - I would think of draining the swamp as the mechanism that performs the function of keeping one's ass away from alligators. Oh well, so much for communication between an engineer and a lawyer.

 

That was my thought as well.

[mansr]

Are you serious? PCM at 96khz (2fs) is still subject to brickwall filtering, rounding, ringing, aliasing.

 

For playback, DSD64 is filtered at 50kHz, a close enough approximation to the top end of 48kHz for 96kHz PCM. I also do not believe all filters are inherently evil.

[Hiro]

For playback, DSD64 is filtered at 50kHz, a close enough approximation to the top end of 48kHz for 96kHz PCM.

 

But not with aggressive brickwall filter... best to listen to some mic feeds captured with Grimm AD1 and compare them with any 24/96 pcm.

[Stereolab42]

96/24 PCM surpasses both vinyl and studio-grade tape. Higher-resolution formats are way beyond any analogue storage medium.

 

Of course, a digital file on a computer can never capture the "magic" of pulling a 12" vinyl from its shelf, admiring the cover art for a moment, carefully taking the disc out of the sleeve, placing it on the turntable, and finally dropping the needle. If that ritual is important to you, by all means keep playing vinyl, but do not delude yourself into thinking it actually sounds better than a high-resolution digital file.

 

The "magic of the ritual", so to speak, is indeed the main reason I stick to tube amps. I no longer delude myself that the sound is something that can't be matched by TOTL solid-state amps (with tube-like EQ if necessary). I actually don't listen to vinyl because of the hassle involved, but I love the idea of vinyl's essential simplicity and analogness, and was clinging to the idea of DSD as a proxy for that. But I'm now satisfied that's not the case.

 

To address tailspn's point, which is that DSD tracks should just sound better than their PCM counterparts regardless of math and philosophy, I haven't made up my mind yet, even with a lot of listening. Unfortunately my interest in the answer to this question wanes by the day since I see no sign that (downloadable) DSD is progressing beyond a handful of niche genres/labels. Call me once underground black metal groups on Bandcamp start putting out albums in DSD... (yes, there are metal bands that care about SQ!)

[Tony Lauck]

96/24 PCM surpasses both vinyl and studio-grade tape. Higher-resolution formats are way beyond any analogue storage medium.

 

Of course, a digital file on a computer can never capture the "magic" of pulling a 12" vinyl from its shelf, admiring the cover art for a moment, carefully taking the disc out of the sleeve, placing it on the turntable, and finally dropping the needle. If that ritual is important to you, by all means keep playing vinyl, but do not delude yourself into thinking it actually sounds better than a high-resolution digital file.

 

My rule of thumb has been that 48 kHz is roughly equal to cassette tape, 96/24 roughly equal to 7.5 IPS 2 track, and 192/24 roughly equal to 15 IPS 2 track (or multi-track on wider tape). Some historic master tapes that were transferred at 96/24 and 192/24 sounded clearly better at the higher speed. If people are hearing the difference at two speeds, then by my lights, the highest speed needs to be doubled for safety margin.

 

Comparing DSD to PCM is difficult, because they have different types of distortions which are evaluated differently according to the type of music, the style of recording and the listener. In general, however, a good rule of thumb is that more bits is better if available and affordable.

[mansr]

My rule of thumb has been that 48 kHz is roughly equal to cassette tape,

 

That's a bit harsh. Anyone with half an ear will agree that CD (44.1 kHz) is far better than cassette tape.

 

96/24 roughly equal to 7.5 IPS 2 track, and 192/24 roughly equal to 15 IPS 2 track (or multi-track on wider tape). Some historic master tapes that were transferred at 96/24 and 192/24 sounded clearly better at the higher speed.

 

I have yet to see a high-resolution tape transfer with any meaningful content above 35 kHz. Digital captures almost always go significantly higher (depending on the instruments, obviously). 24-bit resolution is enough to give noise levels well below what the analogue electronics are capable of.

 

If people are hearing the difference at two speeds, then by my lights, the highest speed needs to be doubled for safety margin.

 

People have a tendency to hear what they want to hear, and there is no shortage of people insisting they hear differences that physically can't exist. If these factors are corrected for, then I agree.

 

Comparing DSD to PCM is difficult, because they have different types of distortions which are evaluated differently according to the type of music, the style of recording and the listener.

 

Don't forget that tape also has distortions.

 

In general, however, a good rule of thumb is that more bits is better if available and affordable.

 

Within reason, yes. Beyond some point, adding resolution becomes pointless because a) we can't hear it, b) our speakers can't reproduce it, and c) there was no sound there in the first place.

[jabbr]

No, that's not the real world, it's the ratios. To perform DXD processes on 64fs DSD content in DXD (352KHz PCM) it's an 8:1 fold down. With a Nyquist frequency of 176KHz, that's a pretty steep filter from 2.8MHz to 172KHz (4 octaves) in which to drop 120dB. It's easy to hear the results of that if you listen to instrument detail and spaciousness cues of acoustic music.

 

These issues are all implementation details, not to belittle the importance of these, because implementation is what actual products are all about. That said, the discussion of theoretical differences in PCM vs. DSD are amusing, because in both cases you start with audio and end with audio, and so the fidelity is a function of sampling rate and bit depth. As I said, numbers are numbers. So perhaps DSD64 is not sufficient, then there is DSD128, 256, 512. Similarly for PCM 192, 384, 768. As some point the available information is adequately represented.

 

From a practical POV, HQPlayer can upsample DSD64 to DSD512 and so the filter is much less steep. Similarly convert PCM 16/44 to DSD etc. Not to say that 16/44 is actually adequate, rather that at some point, some rate is actually adequate.

[jabbr]

...

Comparing DSD to PCM is difficult, because they have different types of distortions which are evaluated differently according to the type of music, the style of recording and the listener. In general, however, a good rule of thumb is that more bits is better if available and affordable.

 

Yes but as the resolution increases these differences diminish and, given optimal implementation, actually vanish. The point at which the differences vanish is the point at which increasing the sampling frequency has no further benefit. I will leave the actual mathematical proof of this to the reader

[徐中銳]

In general, however, a good rule of thumb is that more bits is better if available and affordable.

 

Whether « thumb » up or down ?

Well, I've listened to a counterpointing school of thought from Marantz' Ken Ishiwata, to paraphrase him (for I've not time right now to locate and transcribe verbatim) « number of bits are not that important, what's important is the sampling frequency. » Full 56 minutes interview at iTunes Podcasts

I recall that he also advocates towards « variable bitrate »