Help, I'm ready to buy into either SACD or DVD-A (1 Viewer)

[KeithH]

Philip, good point. Both formats offer backwards compatibility, but I find SACD's way of handling it more compelling. You can buy DVD-Audio players for $200 now, so one need not really rely on the lower-fidelity DVD-Video mixes. With DVD-Audio, you often get higher-resolution stereo and surround-sound tracks. The DVD-Video portion of these discs only give you lower-resolution surround-sound mixes, which of course, you can't generally enjoy outside your home. By contrast, you can play hybrid SACDs anywhere. Like DVD-Audio, SACD offers cheap players for playback of the high-resolution tracks at home. Now if Sony would only offer hybrid discs.

 

[Lee Scoggins]

Wow,
No response from Kotches? Must be a busy weekend...

 

[Lee Scoggins]

John,
One more thing. Remember there was an AES paper recently that showed how dither in DSD domain can reduce noise floor to something like -160db.
If you can hear that, please let me know.

 

[John Kotches]

Lee,
Contrary to popular belief, I do have a social life from time to time I'll even be gone for a long weekend for the 4th of July...
You said this about my understanding of DSD:

This is almost meaningless definition of DSD in the context of sampling rates. If you sample fast enough, you simply capture more in time of the sampled event - live music.

A single sample in DSD can only have one of two voltage values, either +V or -V. Unless the only voltage values in the system is equal to one of these two values, one can either say it's meaningless by itself (which I've said) or that it's wrong. Is there some mystery value in between +v and -v that makes DSD not be a 1-bit format? Lest we forget, a bit is an atomic unit, ie it is not divisible.
It is the integral of the previous samples + the current value which derives meaningful information out of the 1-bit datastream. By the time you get enough samples out there to derive an aggregate voltage resembling the actual voltage over a period of time, you've had enough time to grab a 24bit/192K sample, and maybe even a 24bit/96K sample.
DSD is a brute force approach to the solution, nothing more, nothing less. Some would refer to this as the "McDonald's approach".... what the samples lack in quality (they are not highly accurate) are made up for in quantity (they sample more).
Each PCM sample is a point in time value of the waveform, with precision governed by sampling depth. In other words, each sample contains meaningful information about the wave.
I'll reply to some of the other items later in the evening.
Regards,

 

[John Kotches]

Lee,

They will do so by trading off top end bandwidth.

This paper was just released at the May AES, I suspect that V&L will talk about this in one of the next two conventions.
It's a constant back and forth between the two groups of individuals.

Each Time the DSD camp has a presentation, the PCM folks have an answer at the next convention or two.

Regards,

 

[Lee Scoggins]

A single sample in DSD can only have one of two voltage values, either +V or -V. Unless the only voltage values in the system is equal to one of these two values, one can either say it's meaningless by itself (which I've said) or that it's wrong. Is there some mystery value in between +v and -v that makes DSD not be a 1-bit format? Lest we forget, a bit is an atomic unit, ie it is not divisible.

John,

I'm not sure what your point is. As Ed Meitner points out repeatedly, the high sampling rate capture transients very well.

As for the one bit definition, the answer is very complicated and depends what part of the chain your are in, but the important point is that DSD capture the music wave by looking at changes in amplitude at each point in time, not the absolute value of the amplitude that PCM does. In any event, the music wave is very accurately captured. That is why the dynamics and tonality of Super Audio is so damn good and close to real live mic feeds and that is why so many of us in the pro community have fallen in love with the sound. High-rez PCM is an improvement but the highs don't seem right to us and the tonality seems off. Maybe this will change in time, but Super Audio may also improve as dithering techniques and mastering stations improve - listening to some new Meitner station recordings yesterday like the new YoYo Ma Silk Road was a real eye opener for me.

As Michael Bishop has pointed out on the Audio Asylum recently, the sampling rate is more important than word length. You get more information which is good. Tony Faulkner talks about his belief that 16/176 is better sounds better than 24/192. And no one will argue that Tony is not a great engineer. My Naxos classical CD collection will back me up on this!

And you don't get the instabilities that are caused by PCM when the zeros are reset to absolute zero-the so-called zero-crossing distortion. You can't have it both ways John, saying zero crossing is a non-issue but ultrasonic issues in Super Audio are so important. No one format is perfect but Super Audio, to my ears, is as close to the master tape as I have heard minus $20,000 turntable systems.

And actually, my reading of the AES paper is that audible bandwidth is not traded off.

 

[Lee Scoggins]

Sony and Philips designed DSD to capture the complete information of today’s best analog systems. The best 30ips half-inch analog recorders can capture frequencies past 50 kHz. DSD can represent this with a frequency response from DC to 100 kHz. To cover the dynamic range of a good analog mixing console, the residual noise power was held at -120 dB through the audio band. This combination of frequency response and dynamic range is unmatched by any other recording system, digital or analog.

So add on here the dynamic range that one hears in live music. -120db, gee that's pretty damn low, no wonder I don't hear the noise you speak of John

 

[John Kotches]

Lee,

All of the modern A/D and D/A convertors are using delta-sigma internally, with internal rates up to about 16fs (multi-bit), so you tell me, is that still bringing Zero crossing distortion into play? Searches for for "zero crossing distortion" and PCM come up with precious little information, other than Eds interview, and some very dated papers on CD technology. In the digital world, 8 years is a long, long time.

If this is really still a problem why isn't Ed bringing this up with a presentation at the AES? V&L and Philips/Sony are not shy about putting their ideas out through the AES for review by their peers or anyone else willing to pay the price to get the preprint.

There has to be a technical paper on it.


Regards,

 

[John Kotches]

Lee,
Should I quote your quote, or just summarize?
I think I'll summarize....
PCM filtering vs DSD filtering, redithering vs. noise shaping:
PCM standards have evolved to 96k and some 192k sampling. As such, the filtering is not nearly as drastic. It is not the brickwall that was the practice for 44.1K and 48K sampling rates. Sony has conveniently not adequately addressed 96K and 192K filtering.
Second, DSD requires the same requantization under a different term, noise shaping. Everytime you touch the DSD signal you have to run noise shaping against it. Why is this not mentioned?
The negative feedback loop paragraph:
That's an integral. Remember them from calculus?
The "look" of DSD:
Plot the corresponding voltages for PCM samples, and they too resemble the wave. That's pretty nifty, ain't it?
DSD dynamic range from Sony glossies:
Dynamic range below 5k is certainly in the 120dB range, but as delivered to the consumer no SACD player has delivered that capability across the commonly defined audible bandwidth. The best that's been measured that I've seen was the Classe` Omega, with -100dB to 10k dropping to -90dB in the 10 to 20k octave. So, that would be 120dB only from DC to 5 kHz. Then a rise up to -90dB @ 20kHz. -90dB? That's less than CD dynamic range.
Then again, DVD-A in a much less expensive player, at roughly 1/7th the cost or (thereabouts) managed to maintain -120dB all the way out to 80kHz.
So, which is it? 120dB, or 90dB? Depends on where in the spectrum you look. Here I'm just talking about the standard defined audible bandwidth out to 20kHz.
Look 'em up at Stereophile please. The players were the Classe` Omega and the Toshiba SD-9200.
Regards,

 

[Lee Scoggins]

The best that's been measured that I've seen was the Classe` Omega, with -100dB to 10k dropping to -90dB in the 10 to 20k octave. So, that would be 120dB only from DC to 5 kHz. Then a rise up to -90dB @ 20kHz. -90dB? That's less than CD dynamic range.

If this ultrasonic noise is so audible why can't John Atkinson, Mark Levinson, or Michael Bishop hear it? Why can't I hear it - I spend a lot of time in the studio with state-of-the-art equipment...

I frankly this this is just some minor technical issue that the DVDA camp has seized on to trump up to be some Achilles Heel of Super Audio.

Again, as we say often on the cable forum, measurements don't capture everything, it is the sonics that matter and that is much better, more accurately done with one's ears.

To my ears, Super Audio seems smoother, more natural, more capable of reproducing the attack and decay of notes. One can hear, in certain circumstances, better playback on vinyl but for a convenient format involving compact silver discs, I think SACD is superb.

 

[John Kotches]

Lee,

You are reading into my post something that is neither stated nor implied.....

Here's what I said verbatim:

If this ultrasonic noise is so audible why can't John Atkinson, Mark Levinson, or Michael Bishop hear it? Why can't I hear it - I spend a lot of time in the studio with state-of-the-art equipment...

Nowhere did I say "ultrasonic noise" because I wasn't talking about it. You posted information which claims an available dynamic range of 120dB across the audible band. I posted conflicting data from a US$10K+ player which falls short of that 120dB dynamic range above 5K. Want more conflicting data? Check out the other SACD players that have DSD digital black measured. I actually did find one that maintained 100dB range across the audible band, that was the Pioneer DV-AX10. This player uses a Motorola DSP engine to convert the DSD data to PCM for conversion by a Burr-Brown DAC.

You pointed out requantization, I pointed out noise shaping as the equivalent in DSD. Touch the data, and you pay the price, whether we're talking PCM or DSD. Properly implemented they are non-issues.

What about signal processing? Mixing, editing, etc. There's no such thing as a "Pure DSD" solution around. Sony? They use 8-bit 64fs. Philips uses 32bit, 8fs. Is the entire signal chain irrelevant? Sony's black box solutions will hide this, but in the end it will be processed as PCM data and converted back to DSD for delivery to the consumer.


Regards,

 

[Lee Scoggins]

I posted conflicting data from a US$10K+ player which falls short of that 120dB dynamic range above 5K.

John, you are simply wrong here. DSD was designed to meet this 120db standard and there are ADCs that have it. As far as playback equipment I will look into it, but I seem to recall some very high numbers from Stereophile tests. Playback equipment is relatively young as well as the whole format. Implementation may be lacking in this area as it was in the early days with high end redbook CD players, but that does not mean that the format does not have the entire usable 120db dynamic range. I will look for some examples though.
But let's look at the statistics of the recording systems: you have better dynamic range (96db versus 120 for SACD), better frequency response (5-20Khz versus DC-100khz for SACD), more accurate waveform encoding, higher sampling rates (and hence better transient capture), and easier implementation in a format that is completely backward compatible to the redbook CD standard (not with DVDA). I think Sony and Philips have delivered a gem at the end of the day.
Just look at the 10khz waveform replication on the Sony site www.superaudio-cd.com and compare with the PCM output.
It looks much better with DSD!

 

[Philip Hamm]

Also, S/N ratio doesn't tell the whole story. Vinyl has a miserable S/N ratio but sounds fantastic nonetheless.

However, John has a great point about the studio signal chain. The whole industry is incredibly deeply entrenched in PCM technology. I can not see DSD being a long term success except as an "audiophile" curiosity.

 

[Mike Broadman]

SACD is better because it has a Charles Mingus album and DVD-A doesn't.

'Nuff said.

 

[Lee Scoggins]

However, John has a great point about the studio signal chain. The whole industry is incredibly deeply entrenched in PCM technology. I can not see DSD being a long term success except as an "audiophile" curiosity.

Disagree. Many leading mastering suites are using the Sony DSD equipment and many labels are using DSD just for archiving. A lot of well regarded recording engineers liek Michael Bishop, Tony Faulkner, Bob Ludwig, etc. are using DSD instead of PCM because they think it sounds better.
At one point, you could have said that PCM can never take hold because analog tape is so firmly entrenched.
Engineers will change if something sounds better! We are in a period where the equipment is new, getting better every day and dropping in price. Check the above Sony site and you will see a news article on dozens of studios that have adopted DSD as a primary recording system.

 

[Lee Scoggins]

SACD is better because it has a Charles Mingus album and DVD-A doesn't.

And it will soon have Coltrane!

 

[John Kotches]

Coltrane's "Blue Train" has been out as a DAD for a while from Classic Records.

Regards,

 

[Shane Martin]

Maybe I should mention double blind tests and lets really get this thread a brewin!
In all seriousness, I'm learning alot by this verbal jabbing. Lots of good information to help me make a decision.

 

[Adam Barratt]

Spirited discussions are encouraged here, but not when they degenerate into personal attacks or sniping. Please bear that in mind.

Adam Barratt
HTF Administrator

 

[John Kotches]

Lee,
You said requantization is an issue, it is not. Filtering is not the problem it once was, and the pre/post ringing examples demonstrate this.
I'll show you the exact pages that demonstrate SACDs inability to deliver the claimed 120dB dynamic range:
Pioneer DV-AX10 Measurements -100dB @ 20kHz *Converts DSD to PCM for D/A Conversion
Sony XA-777ES -90dB @ 20kHz
Sony SCD-C555ES -75dB @20kHz
Classe` Omega SACD player -90dB@20kHz
Philips SACD-1000p -90dB@20kHz
DSD measurements not provided for the Accuphase Transport+DAC combo, or the Sony SCD-777ES or SCD-1 player. It is possible that the test tools didn't exist at the time they were tested.
Just for kicks and grins, let's compare that with the tested DVD-A/DVD-V players:
Technics DV-A10 -110dB @ 20K (with 16bit data)
Toshiba SD-9200 -130dB @ 20K (with 16bit data) *Might be an inaccurate test, in which case the numbers would be somewhat higher.
So, every player with published data that maintains DSD through the D/A conversion has 90dB or less @ 20kHz to the noise floor. By design it's supposed to deliver 120dB, and is not.
DVD-A was designed to deliver up to 24-bit data, and the best tested player (the Toshiba SD-9200) has delivered at the 22-bit level.
The question about remembering integrals from calculus is valid. I mentioned integral, you countered from the Sony/Philips literature with the Negative Feed Back loop. The NFB loop is the integral (Summation from time 0 to present sample -1). Sometimes, I like to think of it as an attempt to estimate the area under the curve that is the waveform.
Regards,
* edited to clean up links more than once