One aspect of DSD which has not been discussed here so far, is the increased time domain accuracy. With 2.8MHz, you do not get a better reproduction of the high frequencies, and two points per cycle are enough to accurately represent frequency-domain information for a sine wave.
However, DSD still gets you increased resolution with regards to the time at which the event occurs, as well as changes in amplitude over the course of the waveform. Remember that music is not exactly a series of static sine waves; the amplitude of the wave varies against time, and not only as a function of progression through the shape of the wave (sin function). Thus, higher sampling rates will more closely follow these changes in amplitude, improving dynamic performance.
Our ear has about 30.000 phase-dependant resonators with varying center frequencies and Q factors. The reason for this is not accurate frequency-perception, but rather accurate phase perception. As a side note, the nerve cells trigger up to frequencies well in excess of 200kHz, according to recent research.
Increased sampling rate is an universal improvement for dynamic signals, I think, but the problem with DSD is the limited slew rate. As the 1/f distribution of energy means that as one axis of the vector lengthens (increased frequency), the other shortens (reduced energy), and thus the signal delta does not rise above a certain point. However, DSD does not have a sufficiently high samplerate to cover the slew rate of natural music at full dynamic range. I seem to recall that the dynamic capacity rolls off above 2kHz or so. Many, dCS among them, have suggested that 1.4MHz 2bit or 700kHz 4bit would have been a significant improvement.
It would be interesting for someone to do experiments with 1MHz 4bit or so..
However, DSD still gets you increased resolution with regards to the time at which the event occurs, as well as changes in amplitude over the course of the waveform. Remember that music is not exactly a series of static sine waves; the amplitude of the wave varies against time, and not only as a function of progression through the shape of the wave (sin function). Thus, higher sampling rates will more closely follow these changes in amplitude, improving dynamic performance.
Our ear has about 30.000 phase-dependant resonators with varying center frequencies and Q factors. The reason for this is not accurate frequency-perception, but rather accurate phase perception. As a side note, the nerve cells trigger up to frequencies well in excess of 200kHz, according to recent research.
Increased sampling rate is an universal improvement for dynamic signals, I think, but the problem with DSD is the limited slew rate. As the 1/f distribution of energy means that as one axis of the vector lengthens (increased frequency), the other shortens (reduced energy), and thus the signal delta does not rise above a certain point. However, DSD does not have a sufficiently high samplerate to cover the slew rate of natural music at full dynamic range. I seem to recall that the dynamic capacity rolls off above 2kHz or so. Many, dCS among them, have suggested that 1.4MHz 2bit or 700kHz 4bit would have been a significant improvement.
It would be interesting for someone to do experiments with 1MHz 4bit or so..
Angel, definitely valid points. I don't say that 24/192 PCM is perfect, only that the theory behing it is more convincing than 2.8MHz DSD.
Purplepeople, I think much of our capacity for directional hearing comes from experience. We remember how the voice sounded when last time we knew the person was in that other bedroom. I think it would look different if you didn't know the person or the house. But it's not that alone. Interestingly, men generally have a better directional hearing than women. This comes from ancient times - men needed it more than women. However, both genders have identical ears, it's the processing in the brain that differs.
Purplepeople, I think much of our capacity for directional hearing comes from experience. We remember how the voice sounded when last time we knew the person was in that other bedroom. I think it would look different if you didn't know the person or the house. But it's not that alone. Interestingly, men generally have a better directional hearing than women. This comes from ancient times - men needed it more than women. However, both genders have identical ears, it's the processing in the brain that differs.
The format wars...
HI!
Some interesting reading about the subject:
http://www.iar-80.com/page17.html
Regards
HI!
Some interesting reading about the subject:
http://www.iar-80.com/page17.html
Regards
For DVD-A, there is a full resolution up until the 96kHz cut-off. So I "see" a series of 144dB rectangles that are 0.01ms apart. Like steps.
For DSD, there is a an 6 dB step every 0.00035ms.
If I equalize the timing, I get 28 DSD steps of 6dB or the ability to render 168dB at 96kHz. That's 24dB of headroom over the DVD-A.
The extra headroom could make up for that problem at low-levels, which is could be why conversion to 24/96 is suitable during playback. A properly recorded and mastered SACD would be able to play -144dB down with some extra "foot" room. Of course, that's all ridiculous since quiet might be a total of -80dB.
So the question then becomes: how to use that headroom? How about taking out converting 4-bits less to match the 144dB resolution of DVD-A. Your sample rate would be 116kHz for a better top end. Let's say your analog stages are capable of 120dB so you match again. Your top end would be even better at 140kHz. That would clearly put you in the range of getting some of the harmonics measured by James Boyk.
ensen.
For DSD, there is a an 6 dB step every 0.00035ms.
If I equalize the timing, I get 28 DSD steps of 6dB or the ability to render 168dB at 96kHz. That's 24dB of headroom over the DVD-A.
The extra headroom could make up for that problem at low-levels, which is could be why conversion to 24/96 is suitable during playback. A properly recorded and mastered SACD would be able to play -144dB down with some extra "foot" room. Of course, that's all ridiculous since quiet might be a total of -80dB.
So the question then becomes: how to use that headroom? How about taking out converting 4-bits less to match the 144dB resolution of DVD-A. Your sample rate would be 116kHz for a better top end. Let's say your analog stages are capable of 120dB so you match again. Your top end would be even better at 140kHz. That would clearly put you in the range of getting some of the harmonics measured by James Boyk.
ensen.
That could be the effect of attenuating the natural harmonics of the highs. That is, the highs were less smooth to begin with and SACD smoothed them out.
------------------------------------------------------------------------------------
It seems that the digital interface and cables still hold the keys in terms of fine differences in sonics.
My dCS system with twin pll relocks all over the place and good clocks sounds significantly different in 2 cable upsampled PCM mode with different cables. The 3 cable BNC SDIF (separate clock signal) interface for upsampled DSD is also cable dependent. There is also a very signifoicant difference when the digital processor is set to internal clock mode versus relocked signal clock.
Until all these factors are under control, it will be difficult to isolate the differences due to format alone.
------------------------------------------------------------------------------------
It seems that the digital interface and cables still hold the keys in terms of fine differences in sonics.
My dCS system with twin pll relocks all over the place and good clocks sounds significantly different in 2 cable upsampled PCM mode with different cables. The 3 cable BNC SDIF (separate clock signal) interface for upsampled DSD is also cable dependent. There is also a very signifoicant difference when the digital processor is set to internal clock mode versus relocked signal clock.
Until all these factors are under control, it will be difficult to isolate the differences due to format alone.
Re: The format wars...
When it comes to signal theory (or mechanics, for the matter) that site is a notorious collection of utter rubbish, only demonstrating Moncrieff's ignorance/arrogance.
No. You are still reasoning with delta-modulation, which is fine if you want a quick and dirty grasp of the basics of the coding scheme, but DSD is not delta-modulation. It is sigma-delta, with noise-shaped feedback. As a result, you do not have 6dB per timestep, and as a result, you are limited in dynamic range. (Just as delta-modulation would be seriously slew-limited).
Tube_Dude said:
When it comes to signal theory (or mechanics, for the matter) that site is a notorious collection of utter rubbish, only demonstrating Moncrieff's ignorance/arrogance.
purplepeople said:
No. You are still reasoning with delta-modulation, which is fine if you want a quick and dirty grasp of the basics of the coding scheme, but DSD is not delta-modulation. It is sigma-delta, with noise-shaped feedback. As a result, you do not have 6dB per timestep, and as a result, you are limited in dynamic range. (Just as delta-modulation would be seriously slew-limited).
jwb said:
That would be true if the jitter spectrum is pure white noise. Then jitter on DSD is averaged out over more samples than with PCM. Unfortunately that jitter is not pure white noise but has also player specific low frequency components that give the player its “sonic signature” much in the same way as PCM.
Cheers
Re: The format wars...
Wow, what a way to fill 5 pages with text and not say anything constructive and objective. This is a joke. They are telling you how to cup your lips while saying "SSSS" to reproduce what they heard, calling it a "fatal flaw". Yet they do not present any real technical information to support their claims. They are throwing some numbers that came from old textbooks talking about oversampling. But don't say how it all affects the real waves we are trying to reproduce . This is all written to sound like hey know what they are talking to people who have no technical background. You need to show some measurements, graphs and real numbers to convince a half educated consumer these days. This was like reading a review of a power cord in Sterophile.
I'm not saying that SACD is the greatest end-all format. But that article is not explaining why it's not. It just offers biased one man's opinion on what he heard three times in a row and why he thinks that what he heard is indeed a flaw and not a strenght of the system.
I find SACD's high frequency reproduction to be much better than that of CD. Is this a proof that DSD is superior? To me it is. To anybody reading this post it's not. If I continued for another 1500 words and included some semi-relevant technical terminology maybe I could convince a few that DSD is superior. But I don't have the time or desire to do it.
Tube_Dude said:
Wow, what a way to fill 5 pages with text and not say anything constructive and objective. This is a joke. They are telling you how to cup your lips while saying "SSSS" to reproduce what they heard, calling it a "fatal flaw". Yet they do not present any real technical information to support their claims. They are throwing some numbers that came from old textbooks talking about oversampling. But don't say how it all affects the real waves we are trying to reproduce . This is all written to sound like hey know what they are talking to people who have no technical background. You need to show some measurements, graphs and real numbers to convince a half educated consumer these days. This was like reading a review of a power cord in Sterophile.
I'm not saying that SACD is the greatest end-all format. But that article is not explaining why it's not. It just offers biased one man's opinion on what he heard three times in a row and why he thinks that what he heard is indeed a flaw and not a strenght of the system.
I find SACD's high frequency reproduction to be much better than that of CD. Is this a proof that DSD is superior? To me it is. To anybody reading this post it's not. If I continued for another 1500 words and included some semi-relevant technical terminology maybe I could convince a few that DSD is superior. But I don't have the time or desire to do it.
Re: The format wars...
I have read this article a half year ago, and at first sight I was convinced of their statement that the serious flaws of DSD were responsible for the lacklustre soundquality of SACD.
I had heard a couple of SACD-consumermarket players, and I seriously disliked them. They were no match for high-end cd-players like Meridian 589 etc. Bass-reproducton of SACD was notably better, but especially the treble was not clear and showed no real dynamics. The overall impression was muddy.
Being a recording-engineer of classical music myself, I was seriously asking myself whether I should change from hirez PCM to DSD. I contacted some of my collegues who had been working with DSD for a longer time, among them were Tony Faulkner and Jared Sacks. After their straight comments on the advantages and disadvantages of DSD, I gave it another change; Now, I have found out that the majority of the SACD consumermarket-players are real scrap!
However, if you have got the chance to listen to a well manufactured, or well-tweaked SACD-player, or (even better) a Meittner or dcs DSD-DAC, you will hear something very special.
The treble is not quite as good as hirez PCM (but much better than CD), the midrange is much better than (current) hirez PCM, and the bass is quite spectacular. If one must describe the difference, it is particularly the ambience-information in the midrange that is so appealing of DSD.
The very problem of DSD however is its use in the recording-process. SInce it can't talk to existing gear, one is wholy dependant on the features built into the Pyramix system. This is a burden for the recording-engineer, but totally irrelevant to the consumer. At current state, I prefer staying with 24 bit 176 kHz PCM for my recordings, which can be quite well converted to DSD at the end of the process.
At this moment their is an effort to produce a socalled hyper-DSD (with 4 bits, instead of 1, and many times the current samplerate). This should serve as a no-compromise recording medium, that can be processed without any losses, and at the end of the process be downscaled to SACD's DSD format.
The problem with developing this new studio-format is the financing of it. Noone really feels responsible... Let's see how long it takes.
So my advice to anyone interested in SACD is: Be sure to find a good sounding player. At this time it is still difficult, but I am sure that more will come, especially when more audiophile manufacturers will start producing them. A well constructed SACD player can sound much more real than any CD-player ever could.
With kindest regards,
Lucas.
Tube_Dude said:
I have read this article a half year ago, and at first sight I was convinced of their statement that the serious flaws of DSD were responsible for the lacklustre soundquality of SACD.
I had heard a couple of SACD-consumermarket players, and I seriously disliked them. They were no match for high-end cd-players like Meridian 589 etc. Bass-reproducton of SACD was notably better, but especially the treble was not clear and showed no real dynamics. The overall impression was muddy.
Being a recording-engineer of classical music myself, I was seriously asking myself whether I should change from hirez PCM to DSD. I contacted some of my collegues who had been working with DSD for a longer time, among them were Tony Faulkner and Jared Sacks. After their straight comments on the advantages and disadvantages of DSD, I gave it another change; Now, I have found out that the majority of the SACD consumermarket-players are real scrap!
However, if you have got the chance to listen to a well manufactured, or well-tweaked SACD-player, or (even better) a Meittner or dcs DSD-DAC, you will hear something very special.
The treble is not quite as good as hirez PCM (but much better than CD), the midrange is much better than (current) hirez PCM, and the bass is quite spectacular. If one must describe the difference, it is particularly the ambience-information in the midrange that is so appealing of DSD.
The very problem of DSD however is its use in the recording-process. SInce it can't talk to existing gear, one is wholy dependant on the features built into the Pyramix system. This is a burden for the recording-engineer, but totally irrelevant to the consumer. At current state, I prefer staying with 24 bit 176 kHz PCM for my recordings, which can be quite well converted to DSD at the end of the process.
At this moment their is an effort to produce a socalled hyper-DSD (with 4 bits, instead of 1, and many times the current samplerate). This should serve as a no-compromise recording medium, that can be processed without any losses, and at the end of the process be downscaled to SACD's DSD format.
The problem with developing this new studio-format is the financing of it. Noone really feels responsible... Let's see how long it takes.
So my advice to anyone interested in SACD is: Be sure to find a good sounding player. At this time it is still difficult, but I am sure that more will come, especially when more audiophile manufacturers will start producing them. A well constructed SACD player can sound much more real than any CD-player ever could.
With kindest regards,
Lucas.
I happen to spot this... Is this simmilar to the solution Lucas_ G points at?
www.merging.com/download/dxd_Resolution_v3.5.pdf
www.merging.com/download/dxd_Resolution_v3.5.pdf
Re: Re: The format wars...
Lucas_G said:
I have read this article a half year ago, and at first sight I was convinced of their statement that the serious flaws of DSD were responsible for the lacklustre soundquality of SACD.
I had heard a couple of SACD-consumermarket players, and I seriously disliked them. They were no match for high-end cd-players like Meridian 589 etc. Bass-reproducton of SACD was notably better, but especially the treble was not clear and showed no real dynamics. The overall impression was muddy.
-----------------------------------------------------------------------
How can you draw conclusions by comapring consumer SACD units to high end CD. Do the same on dCS ar high end Sonys and Marantz and you can make deductions.
Some SACD units convert DSD to PCM!
I have atttended a demo and there the quality of SACD was definitely better than that of the CD, using the same program material. Period.
I don't think it makes much sense to compare SACD versus DVD-A since there are labels that support either one or the other so your only choice is a multiformat player anyway.
Only very few labels offer all three formats (like Chesky for instance).
This docu
http://www.iar-80.com/page17.html
sounds really like a joke to me. How can something sound muffled and airy at once ??? Maybe the SACD lacked some articfacts that let one belive that CD has better treble quality.
Using a fifth order delta-sigma modulator like the one proposed by Sony, the theoretical SNR at 20 kHz is still very high at slightly below 200 dB. Even if it is only 100 dB in practice, it is still much better than the dynamic range of our hearing (have a look at Fletcher-Munson).
I believe that, even though the levels of usual high frequency content are very subtle, we don't need the highest dynamic resolution at the upper end of our hearing range but at midrange frequencies.
One of the disadvantages of PCM is that it's necesarily steep antialiasing filters crap the temporal resolution at the upper end. I think this is still worse than a restricted top-end dynamic range.
Every recording format has it's disadvantages, both DSD and PCM.
And it is clear that at the point when the patents for the redbook- CD are going to run out, some commercial war between an open standard and one that is backed up by the ones wo look for a new source of money from licences, will take place. And it is expected that this one will not always be fought in a fair and honest manner.
Regards
Charles
I don't think it makes much sense to compare SACD versus DVD-A since there are labels that support either one or the other so your only choice is a multiformat player anyway.
Only very few labels offer all three formats (like Chesky for instance).
This docu
http://www.iar-80.com/page17.html
sounds really like a joke to me. How can something sound muffled and airy at once ??? Maybe the SACD lacked some articfacts that let one belive that CD has better treble quality.
Using a fifth order delta-sigma modulator like the one proposed by Sony, the theoretical SNR at 20 kHz is still very high at slightly below 200 dB. Even if it is only 100 dB in practice, it is still much better than the dynamic range of our hearing (have a look at Fletcher-Munson).
I believe that, even though the levels of usual high frequency content are very subtle, we don't need the highest dynamic resolution at the upper end of our hearing range but at midrange frequencies.
One of the disadvantages of PCM is that it's necesarily steep antialiasing filters crap the temporal resolution at the upper end. I think this is still worse than a restricted top-end dynamic range.
Every recording format has it's disadvantages, both DSD and PCM.
And it is clear that at the point when the patents for the redbook- CD are going to run out, some commercial war between an open standard and one that is backed up by the ones wo look for a new source of money from licences, will take place. And it is expected that this one will not always be fought in a fair and honest manner.
Regards
Charles
Of course are there possibilities like apodizing filters etc but they first have to be introduced into equipment that can be purchased in a shop - while SACD already has the good temporal resolution intrinsic to its working principle.
I by myself am still having a look-around for a decent multi-format player, since I don't want any format being the factor that is restricting what recordings I can play. Simple as that.
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.