Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Why SACDs? (Read 39354 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Why SACDs?

Reply #50
...but you guys aren't abiding by his arbitrary rules, when it comes to the Stones and the Beatles, you have to settle for second-hand vinyl (never mind that the $30 may only cover the shipping charges).

Ok, show me where I can get the entire collection of the Dave Matthews Band or Dream Theater for only $30.

Why SACDs?

Reply #51
...but you guys aren't abiding by his arbitrary rules, when it comes to the Stones and the Beatles, you have to settle for second-hand vinyl (never mind that the $30 may only cover the shipping charges).

Ok, show me where I can get the entire collection of the Dave Matthews Band or Dream Theater for only $30.


From Amazon:

* Under the Table and Dreaming (1994) 2.25
* Crash (1996) 1.94
* Before These Crowded Streets (1998) 1.94
* Everyday (2001) 2.71
* Busted Stuff (2002) 4.54
* Stand Up (2005) 1.99
subtotal = 17.92
shipping = 17.88
total = 35.80

Not quite $30 but Busted Stuff has held it's value for some reason.

As for the vinyl examples, you guys would have to come with me to the swap meet every week and you have to be a lot more patient. But trust me, I've gotten most of the classics for only $1. The ones I like I'll buy on CD but it saves me a TON of money to buy them on vinyl first and give them a few spins.


Why SACDs?

Reply #53
Maybe SACD vs DVD-A can be interesting, after all 
My two cents..
Quote
This is also a data rate one-half that of DSD and
double that of CD, with a sampling rate of 4 × 44,100 =
176,400 Hz. It can achieve a noise floor 120 dB below full
scale up to 20 kHz, using 96 dB of noise shaping, and a total
noise power of –19 dBFS.

DSD is "1bit PCM with noise shaping", isn't it? So if you say "PCM with noise shaping is more efficient than DSD" you say "there is a more efficient noise shaping algorithm than DSD's one". So the challenge is between noise shaping algorithm here (and I must understand them better to say anything else here)

But DSD noise shaping is "analog": if the "better" algorithm you can use for PCM is digital, you require an higher resolution digital source to process it and downsample, don't you? How much high?
And how can I get it? 

Maybe using a DSD A/D converter, with its inherent noise shaping, with actual technology can be more convenient than using a high res PCM ADC, process the digital signal with a DSP, and downsample.
This is from a production perspective (DSD vs PCM). For consumers all this debate (SACD vs DVD-A) is worthless, since mp3 is the way to go

If you have any reference to technical documentation about DSD and noise shaping please post some links, I'm really curious to read it when I will have some time.


Why SACDs?

Reply #55
So 2-channel SACD has a bitrate of 5.6Mbps?
Yes.
Are you absolutely sure?  I'm having a hard time verifying this.  A follow-up question would be, how is the information for each channel split?  Everything I'm reading is extremely vague.

So if you doubt that there's no audible difference in applications of more than 2 channels and you also believe SACD is an inefficient format in terms of bw and snr, where's the trade-off?  Is SACD not capable of storing as much audio (number of channels * duration of each channel) as DVD-A?


I think SACD is probably only 2x overkill.  DVD-A is more like 100x more then is needed.  That is the trade off.  One is much more ridiculously over-speced then the other.

Why SACDs?

Reply #56
So 2-channel SACD has a bitrate of 5.6Mbps?
Yes.
Are you absolutely sure?  I'm having a hard time verifying this.  A follow-up question would be, how is the information for each channel split?  Everything I'm reading is extremely vague.

So if you doubt that there's no audible difference in applications of more than 2 channels and you also believe SACD is an inefficient format in terms of bw and snr, where's the trade-off?  Is SACD not capable of storing as much audio (number of channels * duration of each channel) as DVD-A?

I think SACD is probably only 2x overkill.  DVD-A is more like 100x more then is needed.  That is the trade off.  One is much more ridiculously over-speced then the other.
Thanks Mike (and thanks to eevan also) for clarifying.

Why SACDs?

Reply #57
"Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications":

http://sjeng.org/ftp/SACD.pdf

very interesting article, in short - SACD is bad and DVD-Audio (PCM output) is the right way to go for advanced resolution audio

Why SACDs?

Reply #58
And the test referenced in this thread indicates that advanced resolution audio really doesn't provide any tangible benefit to the end user.

Do we have a comprehensive article that addresses analog reconstruction (frequency and phase response) between PCM and SACD as it's applied in practice?

Why SACDs?

Reply #59
Do we have a comprehensive article that addresses analog reconstruction (frequency and phase response) between PCM and SACD as it's applied in practice?


I think that this is from November 2000 Stereophile by David Rich:
Quote
Because DSD uses a high-order delta-sigma modulator (7th order–stated earlier in the text), the noise above 20kHz rises very quickly. (The higher the order, the faster the rise.) To prevent high levels of high-frequency energy from getting out of the player, an analog low-pass filter at 100kHz is required during SACD mastering. Even so, a large amount of out-of-band noise might be passed on to the power amplifier, perhaps as high as a tenth of the full power output of the system. If not rolled off by the amplifier, this may be just below the energy level to harm a tweeter. This is why SACD players are required to use a further 50kHz low-pass filter on their outputs (though this can be defeated). The need for these low-pass filters works against claims that DSD has wide bandwidth and low phase shift.


I have added the italic text
If age or weaknes doe prohibyte bloudletting you must use boxing

Why SACDs?

Reply #60
I was just curious to search some data to verify my previous post.

http://www.analog.com/en/subCat/0,2879,760...F0%255F,00.html
http://focus.ti.com/paramsearch/docs/param...R_2000084|EQ|24

I made this (very little) research: it seems that all high resolution analog/digital converters from Analog Devices and Texas Instruments use sigma-delta modulation to perform the conversion, and then they internally translate from sigma-delta to PCM.

I must investigate further, but I think we can't compare PCM vs DSD without accounting that modern AD converters use "DSD" natively.

Why SACDs?

Reply #61
DSD is not bad because of its noise shaping algorithm, but because it is a 1-bit system that has an inherent SNR of 6 dB (lots of quantization noise). In order to make something useful of this, it has to be heavily noise shaped so that this noise is pushed up to ultrasonic frequencies and a decent SNR is achieved at audible frequencies (20 Hz - 20 KHz).

20 or 24 bit PCM does not need any noise shaping to get better SNR than DSD, I mean, it has better SNR than DSD without resorting to using noise shaping. But it can be noiseshaped in the same fashion and achieven even higher SNR, but this would be absurd, because without noise shaping it is already around or over the limits of real-world electronics (~120 dB SNR).

Another problem with 1-bit systems is that they can't be properly dithered, which means that they can't be made totally distortion-free from a digital point of view, whether this is audible or not. Higher than 1-bit systems can be properly dithered so that the are totally distortion-free from a digital point of view. Only distortion at the output of these digital systems would be the caused from the analog electronics of the ADC and DAC, but not from the digital nature of the system.

Edit: AFAIk modern converters are not internally 1-bit anymore, but a few bits instead.

Edit: DSD is a pain to process. PCM is much easier and in practice everything is recorded and processed in PCM and then converted to DSD for SACD. Noise shaping of DSD is not analog, it is digital.

Why SACDs?

Reply #62
I must investigate further, but I think we can't compare PCM vs DSD without accounting that modern AD converters use "DSD" natively.


Those modulators are not 1-bit as DSD is. That is a huge difference.
If age or weaknes doe prohibyte bloudletting you must use boxing

Why SACDs?

Reply #63
In order to make something useful of this, it has to be heavily noise shaped so that this noise is pushed up to ultrasonic frequencies and a decent SNR is achieved at audible frequencies (20 Hz - 20 KHz).
...and it would seem as though the product has been brought to market and a second generation seems to be in the works.

Are you suggesting that it's actually broken or incapable of delivering on the published specifications?

Higher than 1-bit systems can be properly dithered so that the are totally distortion-free from a digital point of view. Only distortion at the output of these digital systems would be the caused from the analog electronics of the ADC and DAC, but not from the digital nature of the system.
So it's not like PCM formats don't have their own difficulties and shortcomings when it comes to sampling and reconstruction.

Why SACDs?

Reply #64
...and it would seem as though the product has been brought to market and a second generation seems to be in the works.


I have this 2003 article:
Quote
SACD II will be announced in September. That is the news reaching us here at
High Fidelity Review from a number of different sources.

High Fidelity Review has learnt that SACD II will introduce at least two
significant changes; enhanced video content and improved copy-protection
measures.


But I don't know what's going on now with SACD II.
If age or weaknes doe prohibyte bloudletting you must use boxing

Why SACDs?

Reply #65
I must investigate further, but I think we can't compare PCM vs DSD without accounting that modern AD converters use "DSD" natively.


Theres not much difference between a DSD and an oversampling PCM DAC.  TI's site is a great reference.  On some of their PCM and DSD DACs, the difference between the two modes is a single pair of muxes that switches the oversampling stage on or off.

I wouldn't say they use either natively.  The design of each is so similar its not very meaningful to make such distinctions.

In order to make something useful of this, it has to be heavily noise shaped so that this noise is pushed up to ultrasonic frequencies and a decent SNR is achieved at audible frequencies (20 Hz - 20 KHz).
...and it would seem as though the product has been brought to market and a second generation seems to be in the works.

Are you suggesting that it's actually broken or incapable of delivering on the published specifications?


He explained how they met those specifications, and you replied asking him if they met them.  Your reply doesn't really make sense.

Higher than 1-bit systems can be properly dithered so that the are totally distortion-free from a digital point of view. Only distortion at the output of these digital systems would be the caused from the analog electronics of the ADC and DAC, but not from the digital nature of the system.
So it's not like PCM formats don't have their own difficulties and shortcomings when it comes to sampling and reconstruction.


The short comings of PCM are inherent in DSD as well.  The problems with DSD do not apply to PCM.  This is the distinction.  DSD is a failed format in the sense that it attempted to improve PCM, but actually kept all of the bad things about it while making entirely new problems (dithing, etc).

Why SACDs?

Reply #66
Those modulators are not 1-bit as DSD is. That is a huge difference.

They are 1-bit sigma delta.

If you look at Analog Device site, only the AD1974 uses a strange "multi bit sigma delta" that is an Analog Devices patented technology (and I don't know if it is better and why it should be)
Also I picked one data sheet from Texas Instruments:
Quote
the
            device can be used to achieve 24-bit analog-to-digital
            (A/D) conversion with no missing codes. Effective
            resolution of 20 bits
can be maintained with a digital
            filter bandwidth of 1kHz at a modulator rate of 320kHz.

This probably is one of the best TI ADC. As you can see, "effective resolution" decrease with frequency, as expected (another I read was 22bit "effective" at 1 khz).

Quote
Noise shaping of DSD is not analog, it is digital.

I must check this, but I remember that the "noise shaping" of sigma-delta modulators is "performed" by the analog part of the modulator (the loop that contains the integrator). Btw "analog" or "digital" is meaningless, I would only say that sigma delta could be a more efficient (=cheap) way to do it.

However I agree that all audio is mixed and processed using PCM, so SACD need another PCM->DSD conversion. So my thoughts 

- sigma delta is convenient to make high res ADC
- SACD are worthless, because you will need another PCM->DSD conversion, and as you say, this seems to be problematic. 
- DVD-A are worthless too, because they aren't only over-speced for the user, but also for our actual recording equipment, since the sigma delta->PCM conversion in the ADC will retain all the sigma-delta SNR.    (maybe you can record on DVD-A a pretty fully synthetic sound )

Well, things could change if class D amps will become popular (so you will need PCM->DSD conversion anyway), if we start to process audio using sigma-delta DSP, or using better ADC that can really fill a DVD-A.

Anyway all this is stupid: mp3 (or flac) rulez! 

Why SACDs?

Reply #67
You're right, I haven't look at the links. That's something I vaguely recall from ADC lecture when I was at the university. I can't find the notes, but I'm pretty sure the professor talked about multibit sigma delta modulators.
If age or weaknes doe prohibyte bloudletting you must use boxing

Why SACDs?

Reply #68
- DVD-A are worthless too, because they aren't only over-speced for the user, but also for our actual recording equipment, since the sigma delta->PCM conversion in the ADC will retain all the sigma-delta SNR.    (maybe you can record on DVD-A a pretty fully synthetic sound )

The noise level at the output of a SACD rises quickly above 20 KHz. If I remember well, at 20 KHz the noise is already higher that in good old CD. In a good 24 bit 96 KHz or 192 KHz converter noise does not rise with frequency, it keeps a good SNR over the whole bandwidth, as should be.

Sigma-delta PCM converters use a very high sample rate and internally process a vast amount of data, but only the useful part of that data goes to the output. DSD keeps all that data, including the useless high frequency noise, and as sigma-delta systems they are very inferior to internal delta-sigma converters of PCM systems, I mean, they use a much lower sampling rate and nº of bits. That's why PCM sigma-delta converters have much better performance than SACD and work as they should.

Why SACDs?

Reply #69
- DVD-A are worthless too, because they aren't only over-speced for the user, but also for our actual recording equipment, since the sigma delta->PCM conversion in the ADC will retain all the sigma-delta SNR.    (maybe you can record on DVD-A a pretty fully synthetic sound )


This is a really, really strange conclusion given that modern audio ADCs achieve the highest SNR of basically any type of ADC in production, and more or less the best possible performance physically possible without cryogenic cooling.

Why SACDs?

Reply #70
Quote
I wouldn't say they use either natively. The design of each is so similar its not very meaningful to make such distinctions.

If I have well understood, the design is so similar because both them use sigma-delta, so both have the same limits.
Quote
The short comings of PCM are inherent in DSD as well. The problems with DSD do not apply to PCM. This is the distinction.

If you use sigma-delta modulators to make ADC, all the problems with DSD will apply also to PCM.
Quote
DSD is a failed format in the sense that it attempted to improve PCM, but actually kept all of the bad things about it while making entirely new problems (dithing, etc).

Of course I agree it is a "failed" format. You can't store the same data of a PCM with "less space", unless you accept some compromise ...or simply you can use Flac .
Also DSD can be fully represented as 1 bit PCM with noise shaping. It is the "noise shaping" the compromise. DSD is nothing but a noise shaping technique.

Quote
Sigma-delta PCM converters use a very high sample rate and internally process a vast amount of data, but only the useful part of that data goes to the output. DSD keeps all that data, including the useless high frequency noise

Even PCM from sigma-delta will keep all the high-frequency noise, unless you filter it. But if you filter it, you will loose some of high frequency signal.

Quote
This is a really, really strange conclusion given that modern audio ADCs achieve the highest SNR of basically any type of ADC in production, and more or less the best possible performance physically possible without cryogenic cooling.

Wow! (at every frequency?) Just a curiosity: can we make full use of DVD-A without using cryogenic cooling?
(edit: sorry I miss a thing: thermal noise gives us a bottom limit. But which is the upper one? We can have more than xxxdb of dynamics if I set the lower reference at 1Volt) 

I'm sorry maybe I'm making a lot of confusion in this thread because I'm speaking about DSD without accounting SACD specs, and because of my english. Anyway I hope my perspective to account actual recording technologies is giving some fun to SACD vs DVD-A fight, and giving some information even to casual reader. 

Why SACDs?

Reply #71
Quote
I wouldn't say they use either natively. The design of each is so similar its not very meaningful to make such distinctions.

If I have well understood, the design is so similar because both them use sigma-delta, so both have the same limits.


This isn't true.  Using sigma delta puts a lower bound on the problem, but not an upper bound.  I can certainly have all the problems of a sigma delta ADC, and still have plenty more unrelated ones too. 

And indeed, this is what happens with DSD.

Quote
Sigma-delta PCM converters use a very high sample rate and internally process a vast amount of data, but only the useful part of that data goes to the output. DSD keeps all that data, including the useless high frequency noise

Even PCM from sigma-delta will keep all the high-frequency noise, unless you filter it. But if you filter it, you will loose some of high frequency signal.


I already explained this:

Quote
If your metric is bandwidth*SNR/bitrate, then PCM is a indeed a much more efficient encoding. Every time you add a bit to your word length, you half your error energy. For 1 bit DSD, to half your error energy, you need a 4 fold increase in sample rate, which means you need exponentially more data to give a linear increase in SNR, compared to PCM where you need linearly more data for exponential increases in SNR.


http://www.hydrogenaudio.org/forums/index....7491&st=25#

The efficiency is much higher in the PCM case as you can trivially verify.  Why are you trying to complicate such a simple thing with irrelevant details about filtering?  Just compute the bits needed to double the SNR for both encodings at a 20kHz bandwidth and you'll see why none of this matters. 


Just a curiosity: can we make full use of DVD-A without using cryogenic cooling?
(edit: sorry I miss a thing: thermal noise gives us a bottom limit. But which is the upper one? We can have more than xxxdb of dynamics if I set the lower reference at 1Volt)


You can't go much higher then 1 V anyway, and since we're on a log scale, it doesn't really matter.  You could maybe get logic working at 5v if you spent a whole lot of money.  Thats not even a 10dB gain.  And it'd be cheaper to buy chillers for your gear . .  .

Why SACDs?

Reply #72
Quote
You could maybe get logic working at 5v

I see... the upper bound!

Quote
Every time you add a bit to your word length, you half your error energy.

I'm starting to remember something... If you increase the bandwidth of a PCM signal, you  can reduce the quantization noise in audio band, because you spread the same noise power across all frequencies.
I remember that sigma-delta is a much more efficient way to trade more bandwidth for less noise.
(I was lucky, all these formulas are on wikipedia http://en.wikipedia.org/wiki/Sigma-delta_m...heory_formulas)

Of course this means absolutely nothing, because you may want to choose a different metric to measure "efficiency". Maybe SNR(audioband)/space(bytes)? Who is so brave to make this calculus? ^^''.

Also maybe you can find a better digital noise shaping filter to achieve better performance with PCM, if you have a better PCM source to process. The paper someone linked seems to say this.

Anyway 1 bit "plain" PCM needs much more bandwidth to achieve the same SNR in the audio range than 1 bit DSD (="1 bit PCM using DSD noise shaping"... yea, it is obvious)
So this could be relevant if each "bit" you add to PCM  "costs" more than bandwidth, for example for making ADCs, where complexity grows exponentially with bit-rate. (and you don't have a better digital source to process...)

Again, this is all related to ADCs. If you have a better digital source to downsample, and a better noise shaping filter, why the hell should you go to DSD??

Quote
Why are you trying to complicate such a simple thing with irrelevant details about filtering?

Because I was focusing in my mind the advantages of sigma-delta (when bit depth is not an option)

Why SACDs?

Reply #73
Quote
Every time you add a bit to your word length, you half your error energy.

I'm starting to remember something... If you increase the bandwidth of a PCM signal, you  can reduce the quantization noise in audio band, because you spread the same noise power across all frequencies.
I remember that sigma-delta is a much more efficient way to trade more bandwidth for less noise.
(I was lucky, all these formulas are on wikipedia http://en.wikipedia.org/wiki/Sigma-delta_m...heory_formulas)

Of course this means absolutely nothing, because you may want to choose a different metric to measure "efficiency". Maybe SNR(audioband)/space(bytes)? Who is so brave to make this calculus? ^^''.


I can't parse any of this.  It would help if you didn't write in the form of a stream of consciousness narrative.  Punctuation is also recommended. 

Again, this is all related to ADCs. If you have a better digital source to downsample, and a better noise shaping filter, why the hell should you go to DSD??


I don't think this has anything to do with ADCs or noise shaping.  Its a simple matter of minimizing RMS error / bit.  You're overcomplicating a simple idea by confusing everything you read with some nonsense about ADCs.  Theres nothing mentioned here that has even the slightest bit to do with ADCs aside from your original observation that most DSD and PCM ADCs and DACs are nearly identical.   

Quote
Why are you trying to complicate such a simple thing with irrelevant details about filtering?

Because I was focusing in my mind the advantages of sigma-delta (when bit depth is not an option)


The advantage of sigma delta is that it allows for very high bit rate ADCs at reasonable cost.  However, this has nothing to do with anything you've mentioned.

Why SACDs?

Reply #74
Hi-res PCM (DVD-A) is much more efficient and better from an objective quality point of view.

From the document that krabapple referenced:
Quote
Practice so far shows that DSD is at least as sonically transparent as 192kHz/24 bit and better than 96kHz/24bit. However, one channel of DSD takes up only 2.8Mbit/s, whereas one channel of 192kHz/24bit takes up 4.6Mbit/s.


Yes, and that is one line I wanted to know the most about, but Mssrs. Putzys et al. didn't expand on that.
No published ABX data that I can find.  What 'practice' are they referring to?  IT'S THE VERY CRUX OF THE ISSUE and yet, again, frustratingly, no listening data is presented.  It isn't a technical argument.  So referencing it is an us another appeal to audio authority and anecdote, and nothing more.

"Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications":

http://sjeng.org/ftp/SACD.pdf

very interesting article, in short - SACD is bad and DVD-Audio (PCM output) is the right way to go for advanced resolution audio



Careful.  SACD technology changed soon after that 2001 paper  --DSD Wide was introduced -- and it has been argued that its objections no longer hold.