[Joe Bloggs]

First, some live history of someone waking up to the theoretical limitations of SACD.

Post 1
Stage = Marketroid parrot (even though I'd hardly read any SACD marketing literature)

--------------------------------------------
I think he's referring to the DSD encoding method of SACD:

Old CD uses PCM encoding: the absolute position of the waveform is specified in time.

SACD uses DSD encoding: instead of encoding the absolute position, for each time-step the position is specified relative to the previous time-step by a '+1' or '-1'.

So I suppose you could say that a square wave can't be accurately represented by DSD.

But then again, it can't be accurately represented by CD either--the time resolution is much poorer.

For all practical purposes SACD should have even better transients than CD

Oh heck, I don't seem to be making much sense here, let me try again:

Quote:

http://www.delosmus.com/sacd_explained.html DSD is the new digital recording technology employed by the SACD. A high-density format dependent on the high speed microprocessors and large capacity storage devices being created by the modern high tech boom, DSD takes a new approach to the basic problems of digital music recording. By using a completely different system of encoding, DSD avoids many of the limitations of physics, electronics, and manufacturing technology inherent to the current method of PCM recording used for CDs and other current digital audio applications. The encoding process uses a single bit (1 or 0) to represent how the music signal (acoustic or electronic waveform) is changing over time, that is, taking its trajectory rather than its absolute value. This process is repeated very rapidly--2.8 million times per second--to get an accurate picture of how the signal is shaped.

So you can see that it seems impossible for DSD to encode instant massive changes in the waveform (like transients or a square wave). But in reality since the sampling rate of SACD is so high (2.8MHz) a string of +1s should encode the rising edge of a transient just as effectively, if not more so, than CD PCM. Although PCM can jump to any value 'instantly', the low sampling rate of 44.1kHz cancels the advantage completely.

 

[Joe Bloggs]

Post 2:
Stage = someone realizing what an idiot he has been

--------------------------------------------
Belay that! I just did some reading and calculations and...

44.1kHz * 64 = 2.822 MHz

So the sampling rate of DSD is only 64x that of CD?

So, say you have a transient that goes from -1 to +1 on adjacent samples on a CD...

The SACD only has 64 samples to go from the negative extreme to the positive extreme?

So you get 64 +1s...

If you want to preserve this same transient ability of CD, you would have to be able to reach the positive extreme from the negative extreme in 64 steps...

The bit depth would be effectively [size=medium]6[/size] bits???

This would seem to make SACD completely inferior to CD!!

Am I calculating something wrong here or what?

 

[Wodgy]

The idea behind SACD is that it's much, much cheaper to make a great DAC for SACD than for PCM. Even the cheapest SACD DACs are better than 85% of the Redbook DACs out there, even some relatively expensive ones made until very recently, so yes, for most people, SACD sounds better than CD.

Unfortunately, the quality of the underlying data on a SACD is poorer than the quality of the underlying data on a regular redbook CD, for reasons in your last post. This is kind of sad, because it means we won't see the same kind of improvement in DACs for SACD that we've seen for redbook. The underlying data just isn't there.

Bottom line, in ten years we'll have wonderful redbook DACs, and the same old SACD DACs.

(There have been many academic papers written on this, but the real proof isn't in the theory, it's in practice. Recording engineers with DSD equipment generally don't work with DSD until the very end -- they'd lose information if they converted to DSD any earlier. The one exception is occasionally for data capture -- just as it's easier to build a good DSD DAC right now, it's also easier to build a good DSD ADC.)

This is also, BTW, why a lot of people have reported getting "fatigued" after listening to SACDs for a long period of time. The lows and midrange are awesome, but the highs are peculiar and unnatural.

Skeptics believe this is just a way for Sony to cut the price of their electronics while at the same time preserving their royalty stream (their patent for CDs expired recently, and they no longer receive royalties for each CD sold). Also, of course, SACD has copy-protection junk.

 

[Joe Bloggs]

Post 3
Stage = I'm not gonna buy an SACD player

-----------------------------------------------
So what the ******'s up with SACD?

I think I can guess... since upper frequencies are lower in amplitude the designers of DSD supposed they could get away with the format not being able to represent a full-swing at 22kHz?

But it still doesn't make any sense...
Say you want to retain 16-bit depth...
65536 steps / 64x sampling rate = 1024 CD-size time-steps to go from 0 to max.
That would mean the upper frequency limit for a full-scale signal is 44.1kHz / 1024 = ~ 43Hz!!!

 

[Joe Bloggs]

So much for the history...

Now, for me the puzzle is not why SACD is accused of having the shortcomings it has, but how it could be coaxed into sounding decent, let alone engineered and advertised until it is hailed as the next big high quality standard.

 

[Wodgy]

Joe, I answered your question in a post above (you might have missed it in between your posts).

FYI, a much better format (theoretically and practically) would be 24bit/96kHz or higher PCM. DVD-A does this, but it's hobbled by audio "watermarking" in the audible frequency range, so I'm not particularly excited about it either.

 

[Joe Bloggs]

... how 'audible' is it?

With 96kHz ability in the digital realm and ~48kHz ability after filtering, it should have been easy to watermark outside the audible range?

And I suppose SACD could be 'bad', but not THAT bad... for example take a look at this graph, taken from a stereophile article http://www.stereophile.com/showarchives.cgi?374:1

'Fig.1 DSD encoding, spectral analysis, DC-200kHz, 1kHz tone at 0dBFS (log. frequency scale, FFT bin width 10Hz). After Sony. '

The noise levels are above what could be achieved by 24/96, but in the audible band still far lower than the 16 bit level, and nowhere near a '6-bit' level!

Although I don't know what to make of this since this clip was not actually trying to encode anything except a 1kHz tone...

 

[KR...]

Quote:

Originally posted by Joe Bloggs So much for the history... Now, for me the puzzle is not why SACD is accused of having the shortcomings it has, but how it could be coaxed into sounding decent, let alone engineered and advertised until it is hailed as the next big high quality standard.

That's not true, when SACD's were 1st released, many people pointed out these flaws. I am sure you can find old articles on this subject matter on-line if you look around a bit.

 

[Wodgy]

Quote:

Originally posted by Joe Bloggs With 96kHz ability in the digital realm and ~48kHz ability after filtering, it should have been easy to watermark outside the audible range?

Absolutely! But then it would have been too easy to defeat (e.g. you could burn a CD copy of a track and all the watermarking information would disappear).

Quote:

And I suppose SACD could be 'bad', but not THAT bad... for example take a look at this graph, taken from a stereophile article ... The noise levels are above what could be achieved by 24/96, but in the audible band still far lower than the 16 bit level, and nowhere near a '6-bit' level!

There's much more to sound than noise. Many people prefer vinyl to CD/SACD, and let's face it, there's way more noise on vinyl

Seriously though, the human hearing threshold at 1kHz is between -90dB and -100dB. Whether SACD pushes noise in the audible spectrum down to -200dB is mostly marketing. Especially when SACD inevitably introduces a new kind of noise -- ultrasonic junk. Look at the graph -- performance above 20kHz is poorer than even CD.

Moreover, even not considering noise, SACD is poorer at representing actual sound.

The fidelity of SACD is an inverse function of sound frequency; bass is represented extremely well, but highs are represented less so. Ultrasonic sounds (above 20kHz) are mostly junk; a side effect of the DSD process.

 

[KR...]

Here is part of an article that I found, it is however pretty one sided :

Quote:

Direct Stream Digital (DSD) DSD was chosen as the encoding format because it avoids using PCM, which is claimed to be an unnecessary, intermediate format and DSD is the best format for archiving. DSD is claimed to offer high quality audio with lossless compression and to be more future proof than PCM. Parameters include 100kHz frequency response and 120dB dynamic range on all channels. Philips and Sony recommend that recording studios use DSD and convert to CD audio using a process called Super Bit Mapping Direct. But independent studies have concluded that DSD (also called 1-bit sigma delta) suffers from a number of problems that makes it unsuitable for archiving and, possibly, distribution. These problems include non-linearity and high frequency noise. DSD is also not easy to edit without converting to PCM. One conclusion is that DSD makes digital to analogue conversion easier and, in the past, cheaper, but PCM provides a more reliable and accurate representation of the music.

You can read the full article here : http://www.disctronics.co.uk/technol...vdaud_sacd.htm

 

[KR...]

Here is a more technical article, I will just give you the link :

Bitstream versus PCM debate for high-density compact disc

 

[Joe Bloggs]

Whatever other failings these articles mention, there is not talk of the '6-bit' worst case performance of SACD DSD. I still think we made some mistakes in calculation there...

 

[elambo]

I love specs. I could talk specs all day and never get bored. But the problem with specs is that they're a technical and scientific way to describe what is in all actuallity a non-scientific phenomenon - music. An emotional experience triggered by auditory stimuli. That's the other side of the brain. I've read all the arguments for SACD's deficiencies, and although some make complete sense some others lack merit (i.e. the relevance of noise, filtering, etc. above 20KHz. The reason higher sample rates might be considered superior is largely because phase-shifting caused by 1st, 2nd or 3rd order low-pass filtering is inaudible when pushed above 20KHz, not because the presence of audio above 20K contributes to what's perceived below 20K in the audible range, as is commonly stated), these calculations don't always accurately represent real world experience. Obviously specs are an emperical method to attempt to describe subsistent phenomena, but the factors are too great for science to describe, or are at least, in some cases, outside thresholds perceivable by human senses.

One 'for-instance' is the vinyl vs CD argument. Many specs for vinyl recording and playback are quite inferior to those of redbook, yet a large portion of audiophiles continue to prefer vinyl to CD. Why is that? I would also relate this to SACD vs CD. I understand the point of illdefined squarewave playback where SACD is concerned, but none of my music calls for complete and accurate playback of squarewaves. On the other hand, for the first time ever I have heard sound in my headphones that completely fooled my mind into believing I was hearing live sound. A trumpet from one of the tracks on the Chesky SACD sampler was more realistic than any recording I've ever heard. My mind was completely fooled (except I knew that a pair of headphones was sitting on my head).

I've read all the above shortcomings of SACD, but I've had the opportunity of listening to the best that CD has to offer and the best that SACD has to offer (so far) and my ears tell me which is superior, even if my eyes and mind attempt to tell me otherwise. I don't argue with the science (for the most part), but I do disagree with the statement that SACD=crap.

Don't get me wrong, I'm not advocating these deficiencies, I'm only trying to point out how effective SACD CAN be at giving me the chills when I hear Mozart's 40th symphony in SACD. For a few seconds there I actually thought I was sitting in front of the orchestra.

 

[Joe Bloggs]

Heh. Can't you see now I'm trying to figure out why SACD is NOT crap?

Other than that, have you listened to any DVD-As?

Any ideas about that 6-bit issue, anyone?

It occured to me that since the Nyquist frequency of CD PCM is 22.1kHz, corresponding to a length of TWO samples, the time alotted for DSD to rise should be 128 steps, not 64--so that gives us 7 bits

But still doesn't seem right at all... I wonder if the DSD technology is being misrepresented or oversimplified to laymen like us.

 

[Wodgy]

Quote:

Originally posted by Joe Bloggs Whatever other failings these articles mention, there is not talk of the '6-bit' worst case performance of SACD DSD. I still think we made some mistakes in calculation there...

No, your calculations are correct. SACD does have poorer performance in the highs than regular old CDs.

Do a google search for 6-bit and SACD in Google and you'll find many articles.

elambo, while I agree that specs are largely meaningless to listeners, the theory is not meaningless to engineers, who have to design the DACs that we listen to, and who strive for the ideal case that might one day be possible in theory. Fact is, an ideal redbook DAC will sound better than an ideal SACD DAC. The ideal redbook DAC will be much more expensive though.

Plus, frankly, there are many people who have complained about the weird highs in SACD just by listening, without understanding anything about theory or specs.