OK, so in the interests of covering the ground for those that are interested.
First thing we need to do is define noise. In this area noise has a particular meaning that has some subtleties. Noise can be considered as simply that which is not part of the desired signal. Sounds trivial, but it can be further subdivided. Crucially all distortion components are noise. So to continue.
The noise people think of is thermal or shot noise in electronics. This is typically characterised as AIWN (additive independent white noise.) This terminology is very neat, as it conveys a crucial property. Additive independence means that any two noise sources may be summed, and you will still simply have AIWN. Crucially you can't subtract one AIWN source from another and lower the noise level - the noise floor rises same as always. This is because the noise is no internal correlated components. It is impossible to ever predict the value of a noise source based upon the pervious values.
The converse is noise that is auto correlated. Noise whereby after some analysis (a Fourier analysis for instance) we can derive a deterministic mechanism to predict the future values of the noise.
A real noise source may of course consist of a combination of the two. A useful thing to do is to define the AWIN noise as the residual after we have done all the analysis we can to predict the future values.
Next we need to consider whether the noise is correlated to the desired signal. AWIN that is not correlated to the signal is what is commonly thought of as noise. However we can expand on this. One can have AIWN that is correlated to the input signal - for instance shot noise in a junction will depend upon current in the junction. Considering auto correlated noise, we can have noise that is not correlated with the signal - a typical example is mains hum. Finally, we have auto correlated noise that is correlated to the signal. This is what most people think of as distortion products.
Now back to sigma delta (i.e. one bit ADC and DACs).
All sigma delta converters use noise shaping. The noise we are shaping is not AIWN. The noise is the signal correlated noise, in particular the distortion products (quantisation noise) of the converter.
The common misunderstanding of a one bit ADC is this. The notion is that the converter consists of a single comparator that every sample delivers a single bit, indicating whether the signal is greater than or less than the threshold. It is reasonably well understood that such a converter cannot work unless the input is dithered with AIWN of average level one half the input signal range. (This dither is used in all multi-bit ADCs where the level of dither must be one half the least significant bit's level.) This extension of the understanding of how multi-bit ADCs may work to the one bit ADC is sadly wrong.
A sigma delta converter is a feedback device. In its simplest form (first order) they consist of the following.
First thing we need to do is define noise. In this area noise has a particular meaning that has some subtleties. Noise can be considered as simply that which is not part of the desired signal. Sounds trivial, but it can be further subdivided. Crucially all distortion components are noise. So to continue.
The noise people think of is thermal or shot noise in electronics. This is typically characterised as AIWN (additive independent white noise.) This terminology is very neat, as it conveys a crucial property. Additive independence means that any two noise sources may be summed, and you will still simply have AIWN. Crucially you can't subtract one AIWN source from another and lower the noise level - the noise floor rises same as always. This is because the noise is no internal correlated components. It is impossible to ever predict the value of a noise source based upon the pervious values.
The converse is noise that is auto correlated. Noise whereby after some analysis (a Fourier analysis for instance) we can derive a deterministic mechanism to predict the future values of the noise.
A real noise source may of course consist of a combination of the two. A useful thing to do is to define the AWIN noise as the residual after we have done all the analysis we can to predict the future values.
Next we need to consider whether the noise is correlated to the desired signal. AWIN that is not correlated to the signal is what is commonly thought of as noise. However we can expand on this. One can have AIWN that is correlated to the input signal - for instance shot noise in a junction will depend upon current in the junction. Considering auto correlated noise, we can have noise that is not correlated with the signal - a typical example is mains hum. Finally, we have auto correlated noise that is correlated to the signal. This is what most people think of as distortion products.
Now back to sigma delta (i.e. one bit ADC and DACs).
All sigma delta converters use noise shaping. The noise we are shaping is not AIWN. The noise is the signal correlated noise, in particular the distortion products (quantisation noise) of the converter.
The common misunderstanding of a one bit ADC is this. The notion is that the converter consists of a single comparator that every sample delivers a single bit, indicating whether the signal is greater than or less than the threshold. It is reasonably well understood that such a converter cannot work unless the input is dithered with AIWN of average level one half the input signal range. (This dither is used in all multi-bit ADCs where the level of dither must be one half the least significant bit's level.) This extension of the understanding of how multi-bit ADCs may work to the one bit ADC is sadly wrong.
A sigma delta converter is a feedback device. In its simplest form (first order) they consist of the following.
- A sample and hold. This captures the signal for the time of one sample period.
[/list=1]- A subtractor. The signal from the sample hold is fed to this, and the output from an internal 1 bit DAC is subtracted.
[/list=2]- A discrete time analog filter. Typically simply an integrator. The output of the subtractor feeds the integrator.
[/list=3]- A quantiser. A comparator that outputs a one bit value based upon the input value. Indeed this is a one bit ADC. This is fed from the filter above.
[/list=4]- A one bit DAC. This is fed by the quantiser above, its output goes to the subtractor above.
[/list=5]
The output from the quantiser is also the output from the ADC. The whole system runs at the oversampled rate - for SACD at 2.8 MHz.
The loop formed by the quantiser and the DAC contain a fast moving approximation to the much slower moving input signal. The integrator will force this approximation to always move in the correct direction so that its own input toward a long term average of zero. Consequently the average error between the digital stream and the analog input will approach zero.
In a conventional sigma delta ADC the single bit stream is then fed to a decimator that will produce a lower sample rate with a correspondingly wider sample width. The SACD process omits this stage and records the output of the modulator directly.
A delta sigma converter does not use dither in the manner that other ADC systems use it to decorrelete the LSB. The manner in which the internal DAC is subtracted from the sampled input and passed to the integrator essentially has the same effect. There is a need for a form of dither, and that is to prevent the occurrence of limit tones. However this dither does not need to be AIWN, and indeed can be an auto correlated signal that can be crafted in such a way as to be eliminated mathematically in the decimator. I do not know what is used in the SACD process.
It is possible to create multiple order sigma delta converters. This basically consists of placing an additional subtractor and integrator stage ahead of the existing ADC. The same DAC output is fed to all the subtractors.
Analysis of the performance of sigma delta converters is a bit fraught. It is possible to treat the error signals in much the same way as AIWN, and if one does the improvement in resolution is roughly 1.5 bits per doubling of sample rate for first order, and 2.5 bits per octave for second. Contrast this with one half a bit per octave without noise shaping.
However there is a cost. Intrinsically the process does add more noise to the entire system. Thus it does not make sense to sue noise shaping unless the oversampling rate is high enough to amortise this basic loss.
Also there are intrinsic issues in physical implementation that prevent actual realisations from attaining the full benefit. As a rough approximation perhaps 3 bits worth of resolution.
As has been discussed the nature of noise shaping is to push the noise into much higher frequencies - where typically it can be filtered out. The shape of the noise shaping profile is reasonably distinctive. Higher order designs have a higher order function shape - thus may have lower noise in the pass band, but much faster rise in content out of the passband.
In a typical system the output of a sigma delta ADC is decimated and yields a conventional multi-bit sampled stream. Many 44.1/16 or indeed 192/24 streams are so created. When played back the opposite occurs. Indeed you can take the entire chain and reverse it. The SACD process takes the attitude that there is no point in decimating the output of the modulator and then simply interpolating it back again in the player. So the intermediate form is the modulator output.
I am interested to hear that critics compare the same recording - one on SACD, the other on conventional CD. There is a crucial issue. The CD will have been created by decimating the SACD stream. Sony call their mechanism SBMD. Check your CD, if it is SBMD you are listening to a SACD recording that has been down converted. If you prefer the CD, you can't blame the sigma delta ADC, or its noise shaping.
The other interesting device to consider is the dCS Verdi-Purcell-Elgar Plus combination. The Purcell will up-convert. It will take 16/44.1 and create DSD. The Stereophile review actually preferred this over direct 16/44.1 Curious.
- A one bit DAC. This is fed by the quantiser above, its output goes to the subtractor above.
- A quantiser. A comparator that outputs a one bit value based upon the input value. Indeed this is a one bit ADC. This is fed from the filter above.
- A discrete time analog filter. Typically simply an integrator. The output of the subtractor feeds the integrator.
- A subtractor. The signal from the sample hold is fed to this, and the output from an internal 1 bit DAC is subtracted.
Konnichiwa,
I cannot comment specifically on that.
I can comment on the following:
CD vs SACD where the CD source is "state of the art" and the CD is a seperate CD, not the CD layer on a hybrid SACD. It has been repeatedly shown (Stereophile, HiFi-News) that the CD Layer on many SACD's is crippled to well below CD's capabilities.
In those cases I prefer a "state of the art" CD Source using an "upper Mid-Fi/lower High End" SACD Player. The recordings where mostly from "Audiophile" labels. CD on average was easily as good or better, subjectively, in my system and to my taste. That is not to say taht there was no difference, but no distinct preference.
I also had the chance to compare recordings to direct feed in a studio, using dCS and other converters, 44.1/16, 48/18, 96/24 and DSD. DSD was marginally better than CD and about level with 48/18 but by no means a match for 96/24, which sounded indistinguishable from the direct feed, which was not the case for DSD or any of the lower rate PCM.
Note also that the same reviewer is an enthusiastic advocate of ASRC "upsampling", which I personally feel makes the msuic sound artificial, unnatural and failtly unpleasant, compared to no ASRC use (that indcludes the dCS item). Again, there is a clearly identifiable difference, but not neccesarily one that is to the better.
I note that in a recent article Keith Howard noted that he subjectively preferred simple truncation of > 16Bit wordlength to 16Bit releases over adding dither as well as prefering otherwise undithered signals IIRC.
Sayonara
Francis_Vaughan said:
I cannot comment specifically on that.
I can comment on the following:
CD vs SACD where the CD source is "state of the art" and the CD is a seperate CD, not the CD layer on a hybrid SACD. It has been repeatedly shown (Stereophile, HiFi-News) that the CD Layer on many SACD's is crippled to well below CD's capabilities.
In those cases I prefer a "state of the art" CD Source using an "upper Mid-Fi/lower High End" SACD Player. The recordings where mostly from "Audiophile" labels. CD on average was easily as good or better, subjectively, in my system and to my taste. That is not to say taht there was no difference, but no distinct preference.
I also had the chance to compare recordings to direct feed in a studio, using dCS and other converters, 44.1/16, 48/18, 96/24 and DSD. DSD was marginally better than CD and about level with 48/18 but by no means a match for 96/24, which sounded indistinguishable from the direct feed, which was not the case for DSD or any of the lower rate PCM.
Francis_Vaughan said:
Note also that the same reviewer is an enthusiastic advocate of ASRC "upsampling", which I personally feel makes the msuic sound artificial, unnatural and failtly unpleasant, compared to no ASRC use (that indcludes the dCS item). Again, there is a clearly identifiable difference, but not neccesarily one that is to the better.
I note that in a recent article Keith Howard noted that he subjectively preferred simple truncation of > 16Bit wordlength to 16Bit releases over adding dither as well as prefering otherwise undithered signals IIRC.
Sayonara
Francis_Vaughan said:
I preferred direct 16/44.1.
Curious...
Btw I have two new SACD discs, both for Telarc:
- McCoy Tyner "Illuminations".
Quote, on the back cover: "produced directly from DSD masters made during the recording sessions".
- Al di Meola "Flesh on Flesh".
Quote: "original recording format: 24bit PCM".
How do they sound?
Quite different.
McCoy Tyner sounds smoooooth, "dead", gives me sleep.
Funny treble, thin and not extended, muted (I've found this on many SACDs).
Al di Meola sound brilliant.
Live, direct, dynamic, transparent, impressive, a twilight zone experience.
Seems like the real thing, a live, unamplified concert.
I have no words... breathtaking.
Very quickly I took this as a reference recording.
Coincidence?
They start looking as too many coincidences for me.
DSD masters?
SACD can sound very good, but something's very wrong on the recording chain.
I have a Linn hybrid SACD that sounds much better on the CD layer.
Take the two examples above, you have both on this hybrid disc: the SACD layer sounds "dead".
I really think PCM is better, but the recording industry really likes to play games with the consumer.
Or they are a bunch of morons, which I really think they are.
The two only DVD-A discs I have sound horrible.
In conclusion, right now there's a very serious risk of either DVD-A and SACD diying very quickly as a format.
I've heard rumours that SACD is already dead, and Sony Music is being obliged by the hardware guys (Sony) to release SACD recordings, which they don't want to.
You do need to be pretty careful here. At least with Telarc there is some chance that the same general recording techniques have been used, and indeed it seems Telarc will tend to use the same microphones as much as makes sense. But you now have an interesting question to answer. Just what is it that you think is causing the trouble? The encoding on both SACDs is DSD. However one is a direct DSD feed, the other converted from a different source. The conversion will essentially mimic the DSD encoding process, except it will do so entirely in the digital domain. But such aspects as noise shaping will occur in the same manner. So, perhaps what you dislike is an artefact of the SACD encoder's physical realisation. Because it seems you have no trouble with an audio stream that is encoded as DSD.
One would need to be very careful about drawing conclusions here. Two comments. In general the artefacts generated with simple truncation are in a word - nasty. You don't even get harmonic distortion products, but something far worse. You get sum and difference products that are not harmonically related to the original signal. To the ear these generally sound so unmusical to be objectionable at even very small levels. However, there is an implicit assumption in the anecdote. And that is that the signal had useful information at the -110db level. It almost certainly didn't. It is very likely that the noise floor was at the mid-90's. Indeed many recording engineers have found that they actually didn't need explicit dither for a good recording - there was enough thermal noise in their microphone amplifiers to do the job for them. If the noise floor of the recording was at this level then truncation would cause no damage to the data. Indeed compared to conversion with added dither it may indeed sound better - certainly it could sound different. The different spectral density of noise between that used for dithering the sampling versus that in the recording naturally may make for a noticeable difference in sound. In this case we are comparing the effect of shaped noise (as opposed to noise shaping) on the conversion process - and potential changes in the spectral distribution of the quantisation noise.
About the only thing that can be said about upsampling is that the only differences in sound must come about because of vagaries in the DAC implementation used at each sample rate. This should really be of no surprise. However I would hesitate to draw firm conclusions about the effects in general.
Konnichiwa.
One always must be very careful about drawing conclusions. Usually there is insufficient reliable information upon which to base such a conclusion, yet draw it we must....
I draw the following conclusions based on practical experience:
1) All else being equal Multibit DAC's appear to sound better than DS DAC's, subjectively.
2) All else being equal non-DS ADC's (in other words very fast, sucessive approximation types which have all but disappeared) appear to sound better, subjectively.
3) All else being sufficiently equal DSD encoding appears to harm the "signal" more, subjectively speaking, than 96/24.
4) All else being sufficiently equal SACD currently generally fails to deliver material advances on the best available from CD.
Sayonara
Francis_Vaughan said:
One always must be very careful about drawing conclusions. Usually there is insufficient reliable information upon which to base such a conclusion, yet draw it we must....
I draw the following conclusions based on practical experience:
1) All else being equal Multibit DAC's appear to sound better than DS DAC's, subjectively.
2) All else being equal non-DS ADC's (in other words very fast, sucessive approximation types which have all but disappeared) appear to sound better, subjectively.
3) All else being sufficiently equal DSD encoding appears to harm the "signal" more, subjectively speaking, than 96/24.
4) All else being sufficiently equal SACD currently generally fails to deliver material advances on the best available from CD.
Sayonara
Seems its just the way of life that every single time discussion touches 1-bit vs. multibit the debate degrades down to SACD vs. CD/DVDA war.
Please people, have courage to look at things without prejudice and making premature conclusions. SACD is only one implementation of 1-bit.
I want to add these points.
There exists NO 24-bit parallel ADC that can come even close to 24-bit resolution. The use of delta-sigma with later decimation during ADC step is a _forced_ reality. Thus, there is no point in bashing delta-sigma vs. PCM as a means of encoding audio. DS is simply always there, including in your best sounding PCM stuff.
What remains is design decision of ADC sampling rate and means to deliver audio data. Delta-sigma resolution depends very heavily on the sampling rate, and if its picked too low, all hell breaks loose that Nth order shaping tricks can't fix.
What can be argued, is that SACD has *too low* sampling rate to deliver transparent reproduction, and all the inconveniences in processing it. That can be agreed on. What you can't conclude from all your experience with SACD is that DSD as such is worse than PCM.
Coming back to 1-bit DACs, it makes a huge difference whether you use 64xOS or 512x OS in your 1-bit DAC. Higher sampling rates require heavier digital processing to convert PCM to DSD and more stringent requirements to DAC implementations. That establishes directly the price point vs result quality. That should be obvious. And that should make it obvious that as there are different quality levels in 24-bit DACs, digital filters, there are different quality levels in 1-bit DACs.
There is quite some irony in fact that most highest grade commercial DACs raved for their 24/96 PCM performance are actually low-bit delta-sigma DACs. Eg. that same dCS Ring DAC.
Bitrate decisions for SACD were made based on some criteria. The noise level at high octave was never a problem because it goes below human audibility curve. Higher bitrate would mean less audio data on the CD, therefore it was picked high enough but as low as possible. Now it seems there is evidence that this was not a wise enough decision, but thats a problem of Sony, not that of delta-sigma.
Problem of high noise screwing audio equipment beyond DAC was probably one of the issues SACD designers failed to take seriously. I could think that originally SACD was meant to deliver 20kHz upper limit and filtering out all DSD noise with analog filters. Then as the race for ultrasonics began, the filtering frequencies were simply raised without raising DSD sampling rate. So now SACD devices show bad noise levels. You still could filter out beyond 20kHz if you wanted to make comparison to CD.
That does not apply to high OS rate 1-bit DACs, as they are free to pick DAC bitrate so high that there remains no noise in the audio band. Thats also exactly what delta-sigma ADC converters do. Delta-sigma DACs prefer multi-bit (5-6 bits) delta-sigma to reduce jitter sensitivity, use less steep noise shapers and ease up digital processing, not because delta-sigma works much differently then.
In relation to what some of the users perceive as "better sounding" is completely anecdotal. The only relevant comparison would be double-blind ABX comparison with direct feed and making necessary statistical analysis to eliminate any random dilusions or bias.
To get a clue of what "better sounding" means, turn your bass and tremble controls up and listen like that for few days, trying hard to get used to it. Then change them back to flat. You'd truely realise what the claim of "sounding better" is really worth.. Its worth nothing because human auditory system perceives "more natural" what it is accustomed to. Thats reality. And that includes being accustomed to nonlinear multibit DAC sound.
So all claims of "sounding better subjectively" can be safely discarded as biased. If you want real comparisons, you have to use scientific ABX comparison to direct feed.
To conclude my part in this thread, I assert that problems with SACD lies elsewhere, not in the delta-sigma principle. Perhaps in the specific Nth order noise shaping filters used, perhaps not high enough sampling rate. I don't really care, as SACD is evil in its essence anyway.
Similarily, problems with 1-bit DAC lies not in the 1-bit, but in the ways the 1-bit stream is generated.
Please people, have courage to look at things without prejudice and making premature conclusions. SACD is only one implementation of 1-bit.
I want to add these points.
There exists NO 24-bit parallel ADC that can come even close to 24-bit resolution. The use of delta-sigma with later decimation during ADC step is a _forced_ reality. Thus, there is no point in bashing delta-sigma vs. PCM as a means of encoding audio. DS is simply always there, including in your best sounding PCM stuff.
What remains is design decision of ADC sampling rate and means to deliver audio data. Delta-sigma resolution depends very heavily on the sampling rate, and if its picked too low, all hell breaks loose that Nth order shaping tricks can't fix.
What can be argued, is that SACD has *too low* sampling rate to deliver transparent reproduction, and all the inconveniences in processing it. That can be agreed on. What you can't conclude from all your experience with SACD is that DSD as such is worse than PCM.
Coming back to 1-bit DACs, it makes a huge difference whether you use 64xOS or 512x OS in your 1-bit DAC. Higher sampling rates require heavier digital processing to convert PCM to DSD and more stringent requirements to DAC implementations. That establishes directly the price point vs result quality. That should be obvious. And that should make it obvious that as there are different quality levels in 24-bit DACs, digital filters, there are different quality levels in 1-bit DACs.
There is quite some irony in fact that most highest grade commercial DACs raved for their 24/96 PCM performance are actually low-bit delta-sigma DACs. Eg. that same dCS Ring DAC.
Bitrate decisions for SACD were made based on some criteria. The noise level at high octave was never a problem because it goes below human audibility curve. Higher bitrate would mean less audio data on the CD, therefore it was picked high enough but as low as possible. Now it seems there is evidence that this was not a wise enough decision, but thats a problem of Sony, not that of delta-sigma.
Problem of high noise screwing audio equipment beyond DAC was probably one of the issues SACD designers failed to take seriously. I could think that originally SACD was meant to deliver 20kHz upper limit and filtering out all DSD noise with analog filters. Then as the race for ultrasonics began, the filtering frequencies were simply raised without raising DSD sampling rate. So now SACD devices show bad noise levels. You still could filter out beyond 20kHz if you wanted to make comparison to CD.
That does not apply to high OS rate 1-bit DACs, as they are free to pick DAC bitrate so high that there remains no noise in the audio band. Thats also exactly what delta-sigma ADC converters do. Delta-sigma DACs prefer multi-bit (5-6 bits) delta-sigma to reduce jitter sensitivity, use less steep noise shapers and ease up digital processing, not because delta-sigma works much differently then.
In relation to what some of the users perceive as "better sounding" is completely anecdotal. The only relevant comparison would be double-blind ABX comparison with direct feed and making necessary statistical analysis to eliminate any random dilusions or bias.
To get a clue of what "better sounding" means, turn your bass and tremble controls up and listen like that for few days, trying hard to get used to it. Then change them back to flat. You'd truely realise what the claim of "sounding better" is really worth.. Its worth nothing because human auditory system perceives "more natural" what it is accustomed to. Thats reality. And that includes being accustomed to nonlinear multibit DAC sound.
So all claims of "sounding better subjectively" can be safely discarded as biased. If you want real comparisons, you have to use scientific ABX comparison to direct feed.
To conclude my part in this thread, I assert that problems with SACD lies elsewhere, not in the delta-sigma principle. Perhaps in the specific Nth order noise shaping filters used, perhaps not high enough sampling rate. I don't really care, as SACD is evil in its essence anyway.
Similarily, problems with 1-bit DAC lies not in the 1-bit, but in the ways the 1-bit stream is generated.
Konnichiwa,
DSD is specified with the same sample rate as SACD. Therefore, if SACD has too low a sample rate, then so has DSD.
My experience (which includes dCS) does not bear out that the best DAC's arelow-bit delta-sigma. All the best sounding DAC's FOR CD SIGNAL and higher resolution PCM I have heard so far where Multibit. But that is a subjective issue. There are also pretty good sounding DS and hybrid DAC's which suggest we have a more complex picture to contend with.
The criteria was existing DSD which in turn was a archival format for analogue recordings intended for easy conversion to 44.1/16 CD Format.
Yes, you would need a filter down around 50db+ @ 100KHz, that is a pretty steep filter. Maybe about the same order as the noiseshaper?
They may be free to do so, but almost all commercial items use 128 or 256 at the very maximum, both ADC and DAC. Some of the better mankers of pro-grade ADC/DAC makers get around the issues by adding several (many) ADC/DAC chips in parallel.
Your reasoning is fault to dismiss anecdotal evidence completely. It is there and should not be dismissed wholesale. ABX testing usually faisl to achieve sample sizes sufficiently large to allow any statististically significant conclusions to be drawn (not a fault of the ABX methodology per se, but of the experimenters usual implementation and statistical analysis).
What are "bass and tremble controls"?
Any quality equipment omits these. If you require controls to remaster the tonal balance of recordings ou need a much more sophisticated tool.
Actually, my still "reference" Digital replay device uses a DS DAC. I generally have not yet upgraded as very little of my "serious" music is on CD, most is on LP.
Can they now?
Funny, we did exactly that at a friends studio (agreed, the "PCM" ADC's used in testing where in fact DS based, the DAC's MB notabene).
HOWEVER:
All results of small scale ABX Tests can be safely discarded as fundamentally statistically flawed.
The DS Principle IN THEORY appears to offer excellent performance potential. Most currently deployed DS Audio Systems do not appear to realise this in any usable form. My suspicion is that the sample rates are by a significant factor to loow, forcing the use of digital manipulation that impacts negatively on resulting sonics, primarily because the human hearing does not operate anything like the measurement instruments applied to audio gear.
Great, we agree on all counts then. Your two lines above carry 100% agreement from me.
Sayonara
wimms said:
DSD is specified with the same sample rate as SACD. Therefore, if SACD has too low a sample rate, then so has DSD.
wimms said:
My experience (which includes dCS) does not bear out that the best DAC's arelow-bit delta-sigma. All the best sounding DAC's FOR CD SIGNAL and higher resolution PCM I have heard so far where Multibit. But that is a subjective issue. There are also pretty good sounding DS and hybrid DAC's which suggest we have a more complex picture to contend with.
wimms said:
The criteria was existing DSD which in turn was a archival format for analogue recordings intended for easy conversion to 44.1/16 CD Format.
wimms said:
Yes, you would need a filter down around 50db+ @ 100KHz, that is a pretty steep filter. Maybe about the same order as the noiseshaper?
wimms said:
They may be free to do so, but almost all commercial items use 128 or 256 at the very maximum, both ADC and DAC. Some of the better mankers of pro-grade ADC/DAC makers get around the issues by adding several (many) ADC/DAC chips in parallel.
wimms said:
Your reasoning is fault to dismiss anecdotal evidence completely. It is there and should not be dismissed wholesale. ABX testing usually faisl to achieve sample sizes sufficiently large to allow any statististically significant conclusions to be drawn (not a fault of the ABX methodology per se, but of the experimenters usual implementation and statistical analysis).
wimms said:
What are "bass and tremble controls"?
Any quality equipment omits these. If you require controls to remaster the tonal balance of recordings ou need a much more sophisticated tool.
wimms said:
Actually, my still "reference" Digital replay device uses a DS DAC. I generally have not yet upgraded as very little of my "serious" music is on CD, most is on LP.
wimms said:
Can they now?
wimms said:
Funny, we did exactly that at a friends studio (agreed, the "PCM" ADC's used in testing where in fact DS based, the DAC's MB notabene).
HOWEVER:
All results of small scale ABX Tests can be safely discarded as fundamentally statistically flawed.
wimms said:
The DS Principle IN THEORY appears to offer excellent performance potential. Most currently deployed DS Audio Systems do not appear to realise this in any usable form. My suspicion is that the sample rates are by a significant factor to loow, forcing the use of digital manipulation that impacts negatively on resulting sonics, primarily because the human hearing does not operate anything like the measurement instruments applied to audio gear.
wimms said:
Great, we agree on all counts then. Your two lines above carry 100% agreement from me.
Sayonara
Konnichiwa,
Actually, please note that "bass" & "treble" controls are completely useless to correct tonal flaws in either room acoustics, speakers or recordings.
You need much more precisely usable tools, such as a Parametric equaliser AND the skill to use it. I have covered that issue to quite some detail in my review of the Behringer DEQ8024 at www.enjoythemusic.com plus some more notes available in my yahoo group.
For re-qualising music only the Cello Palette EQ forms a good solution, to equalise speaker and room problems parametrics or parametric & graphic EQ's are nesseray.
Sayonara
phn said:
Actually, please note that "bass" & "treble" controls are completely useless to correct tonal flaws in either room acoustics, speakers or recordings.
You need much more precisely usable tools, such as a Parametric equaliser AND the skill to use it. I have covered that issue to quite some detail in my review of the Behringer DEQ8024 at www.enjoythemusic.com plus some more notes available in my yahoo group.
For re-qualising music only the Cello Palette EQ forms a good solution, to equalise speaker and room problems parametrics or parametric & graphic EQ's are nesseray.
Sayonara
I'm not saying it's not manipulation. But that's a moot point. All audio is about sound manipulation. And that's one reason I don't understand what audiophile's goal is. I simply don't know what they are looking for. If I did, I too MIGHT have been an audiophile. But whatever it is they look for, it isn't music.
wimms said:
Indeed.
To put it in a cruel way as reality is, Philips and Sony had to make another format, because the CD-DA patent was expiring.
It has expired.
Now maby the "new" formats will die and we end up with a format that doesn't pay royalties anymore: CD.
Who wins?
The big labels, not the hardware industry.
I'm glad we are eventually finding some agreement.
However, 1-bit DSD is not recommended archiving format either. They invented DSD-wide for that. Which has less steep noise shaping and is 8-bit DSD. In other words, much much better.
I don't care if one DAC "sounds better" than the other. I only care if it is transparent, or it isn't. Thats the only criteria that has any meaning here. And that can only be determined by direct feed comparison testing. No amount of listening experience is going to be reliable means to assess quality of any DAC without immediate original reference.
And in that we agree that we have a more complex picture to contend with than simple "PCM is good, DS is bad".
There is irony in that. SACD with its 5th order noise shaping was supposed to please audiophiles because less steep LP filters are required. Well, um.
I'm sorry, I really meant delta-sigma, without implying specifically Sony's DSD. Like say 2-bit DS at 4Mhz with lower order shapers..Kuei Yang Wang said:
However, 1-bit DSD is not recommended archiving format either. They invented DSD-wide for that. Which has less steep noise shaping and is 8-bit DSD. In other words, much much better.
I do not disrespect your experience, I just do not trust it. There is huge number of experienced mastering engineers arguing that PCM is fundamentally flawed, and that only DSD resembles analog input. So the controversy is large. Its the subjective preference that raises my caution.
I don't care if one DAC "sounds better" than the other. I only care if it is transparent, or it isn't. Thats the only criteria that has any meaning here. And that can only be determined by direct feed comparison testing. No amount of listening experience is going to be reliable means to assess quality of any DAC without immediate original reference.
And in that we agree that we have a more complex picture to contend with than simple "PCM is good, DS is bad".
Yes. Which means that lower order noise shapers require equally lower order LP filters.
There is irony in that. SACD with its 5th order noise shaping was supposed to please audiophiles because less steep LP filters are required. Well, um.
Those ADC's are done as multibit then, 4-6 bits at such low OS. Its a matter of jitter tolerance mostly. Though other issue is that comparators' precision depends on sampling time. In other words, reality bites.
I've seen it too much. Snake oil anyone? It *really* works, btw. As long as you do not admit your perception of "good sound" is at mercy of your mind, you are their potential customer, more or less. $30k pricetag on any decent DAC would make it "sound good", no less.
Please reread that part. My point is that what is pleasing the listener isn't always "true to original". Its precisely the job of the right tools to do the pleasing side of the perception, and is not the job of the DAC. In that meaning, any sort of "sound" of a DAC is a fault.
Yes, and thats the most credible and interesting part of your experience you've conveyed here. Unfortunately, your test has not been documented well enough, and no ABX testing log can be seen, so its not possible to deduce if the test was "clean" or still the outcome "proved" expectations.
No, small scale ABX tests are not fundamentally flawed. They don't permit to make conclusions about whole population, but they brutally show whether _your_ perception is real or random imagination. Please, don't tell that you can not be deceived. Anyone can. If you haven't done properly conducted ABX session, you don't imagine how shocking a revelation it can be. In short - you can *clearly* hear difference where there isn't any.
I'd agree on SACD part, but don't think that shortcoming is by significant factor. DSD-wide easily exceeds, by far, any performance requirements. IMO, thats just IMO as I have no solid math background to assert that, imo SACD would have been fully transparent if it had been designed as 2-bit@2.8M DSD with 2nd-3rd order noise shaping.
Konnichiwa,
To put it short and succinct, a credible illusion of listening to music, or as an alternate a great ease of the suspension of disbelief.
Sadly for you this will not easily translate into engineering parameters, moreso as what what hinders the "suspension of disbelief" varies to a fairly substantial degree with the individual.
I could list what "does it for me", but that would moot.
You may be right, they do not look for music, but THE EXPERIENCE OF MUSIC. Just some more or less pleasant background noise is not sufficient. What is needed is a good illusion engine....
You may (or may not) find this little article by Lynn Olson illuminating:
Illusion Engines By Lynn T. Olson
You may argue that what is asked for is not possible. From where you stand that may be so indeed.
From where I stand, it is possible. That is the difference between a craftsmen and a magician (as in stage magic, as opposed to real magick). A craftsmen can give a facsimile of something according to a number of objective parameters and he will argue that the representation is accurate (according to his measures) and true to life (according to his measures). The magician will simply give the impression of whatever it is being real, or close enough that the target audience is fooled into accepting it for reality.
In the context of audio the best allegory may be the technomages from Baylon 5:
"We are dreamers, shapers, singers, and makers. We study the mysteries of laser and circuit, crystal and scanner, holographic demons and invocations of equations. These are the tools we employ, and we know many things.
Fourteen words to make someone fall in love with you forever, seven words to make them go without pain, how to say goodbye to a friend who is dying, how to be poor, how to be rich, how to rediscover dreams when the world has stolen them from you."
And if you don't get it, well, then you don't....
Sayonara
phn said:
To put it short and succinct, a credible illusion of listening to music, or as an alternate a great ease of the suspension of disbelief.
Sadly for you this will not easily translate into engineering parameters, moreso as what what hinders the "suspension of disbelief" varies to a fairly substantial degree with the individual.
I could list what "does it for me", but that would moot.
phn said:
You may be right, they do not look for music, but THE EXPERIENCE OF MUSIC. Just some more or less pleasant background noise is not sufficient. What is needed is a good illusion engine....
You may (or may not) find this little article by Lynn Olson illuminating:
Illusion Engines By Lynn T. Olson
You may argue that what is asked for is not possible. From where you stand that may be so indeed.
From where I stand, it is possible. That is the difference between a craftsmen and a magician (as in stage magic, as opposed to real magick). A craftsmen can give a facsimile of something according to a number of objective parameters and he will argue that the representation is accurate (according to his measures) and true to life (according to his measures). The magician will simply give the impression of whatever it is being real, or close enough that the target audience is fooled into accepting it for reality.
In the context of audio the best allegory may be the technomages from Baylon 5:
"We are dreamers, shapers, singers, and makers. We study the mysteries of laser and circuit, crystal and scanner, holographic demons and invocations of equations. These are the tools we employ, and we know many things.
Fourteen words to make someone fall in love with you forever, seven words to make them go without pain, how to say goodbye to a friend who is dying, how to be poor, how to be rich, how to rediscover dreams when the world has stolen them from you."
And if you don't get it, well, then you don't....
Sayonara
Konnichiwa,
Hmmm, that indeed may be something. 8Bit at 2.822MHz noiseshaped, or in effect 22.5MHz bitrate. In effect the tenfold increase I was thinking might make all the difference.
I do not expect you to. I do expect you to do your own empirical research, which may lead you to the opposite conclusions from my own. Which is fine. But I prefer for people to make their choices on first hand knowledge (gnosis) rather than send and third hand doctrine.
Funny, I must have missed them. One of the greatest intial high profile proponents of DSD (Tony Faulkner) is back to Hi-Rez PCM and Analogue as the holy grail.... ;-)
Given that ANY objective observation is absoltely and completely impossible (one of the few impossibilities in this universe for us humans) the only valid point that can be made is subjective preference. If I like it I like it, if not it sucks. Full Stop.
I agree. On diretc feed testing I noticed that DSD via dCS gear had a sonic footprint about as large (but not in the way) as ancient 48/18 stuff. Compared to that 96/24 was transparent (as not identifiable). On a preference scale ancient 48/18 was less offensive in the sonic footprint it presented than DSD.
Yet the less steep filter leaves so much ultrasonic noise that it causes (IMHO) major problems.
Well, I am not doing tests to publish. I am doing them to answer to me personally, to my satisfaction certain specifi questions. I do not claim universal applicability, however, I would not that most of my tests are at the very least sufficiently blind to eliminate obvious bias. I am aware of the limitations, but the problem with most ABX or blind tests (including my own) is that the sample size is by far to small to allow a statistically significant conclusion to be drawn.
Nope. They do not tell that, simply because their risk of type b/2 statistical errors is huge if you use .05 significance, sufficiently high to reliably eliminate subtle but percieved differences from detetction.
If you select a significance level that balances the likelyhood of type a/1 errors and type b/2 you find that you have no certainty about what was hear or what was not heard, if you select .05 significance you have certainty that most listeners under most circumstances will not show a significant detection of any difference but you achieve near certainty that any small differences will not be detetcted.
Only with large datasets can you balance type a/1 error likelyhood and type b/2 error likelyhood so that you can be certain that any audibile difference detected is not due to "luck" and EQUALLY that any failure to detect audible differences is NOT due to excessive levels of type b/2 errors.
I have been part of both that which you call "properly conducted ABX session" and of actual ABX sessions implemented to not confirm reliably the null hyphotesis (as seems the aim [concious or uncincious] of most of those conducting these tests) but to actually answer the question "what is potentially audible" and "what is potentially preferable". The invariably low sample sizes lead to a drastic lack of statistic significance (sadly).
As said, I do not test to publish (I disdain modern academia - what a boundless bog of sullen inertia and "NIH" bovine excrement) I ask questions to get answers. And I get them.
Significant enough, at this stage of mastering, recording and replay gear. It MAY get better (probably not as SACD together with DVD-A will soon Video 2000 and Betamax) just as CD did. I am in many ways not really concerend with the potential, but with the achievement of the format.
Very possibly.
And CD may very well have been good enough from the start if it had been implemented as 18Bit System with 48 or better 64KHz sample rate (which was easily possible in the late 70's but would have cut play time).
And at the Fox Hunting this weekend in England, had the Dog not sat down to take a <r@p he might have caught the fox in contravention of the new laws.
But CD is what it is, SACD is what it is and the Dog did sit down. So what?
Sayonara
wimms said:
Hmmm, that indeed may be something. 8Bit at 2.822MHz noiseshaped, or in effect 22.5MHz bitrate. In effect the tenfold increase I was thinking might make all the difference.
wimms said:
I do not expect you to. I do expect you to do your own empirical research, which may lead you to the opposite conclusions from my own. Which is fine. But I prefer for people to make their choices on first hand knowledge (gnosis) rather than send and third hand doctrine.
wimms said:
Funny, I must have missed them. One of the greatest intial high profile proponents of DSD (Tony Faulkner) is back to Hi-Rez PCM and Analogue as the holy grail.... ;-)
wimms said:
Given that ANY objective observation is absoltely and completely impossible (one of the few impossibilities in this universe for us humans) the only valid point that can be made is subjective preference. If I like it I like it, if not it sucks. Full Stop.
wimms said:
I agree. On diretc feed testing I noticed that DSD via dCS gear had a sonic footprint about as large (but not in the way) as ancient 48/18 stuff. Compared to that 96/24 was transparent (as not identifiable). On a preference scale ancient 48/18 was less offensive in the sonic footprint it presented than DSD.
wimms said:
Yet the less steep filter leaves so much ultrasonic noise that it causes (IMHO) major problems.
wimms said:
Well, I am not doing tests to publish. I am doing them to answer to me personally, to my satisfaction certain specifi questions. I do not claim universal applicability, however, I would not that most of my tests are at the very least sufficiently blind to eliminate obvious bias. I am aware of the limitations, but the problem with most ABX or blind tests (including my own) is that the sample size is by far to small to allow a statistically significant conclusion to be drawn.
wimms said:
Nope. They do not tell that, simply because their risk of type b/2 statistical errors is huge if you use .05 significance, sufficiently high to reliably eliminate subtle but percieved differences from detetction.
If you select a significance level that balances the likelyhood of type a/1 errors and type b/2 you find that you have no certainty about what was hear or what was not heard, if you select .05 significance you have certainty that most listeners under most circumstances will not show a significant detection of any difference but you achieve near certainty that any small differences will not be detetcted.
Only with large datasets can you balance type a/1 error likelyhood and type b/2 error likelyhood so that you can be certain that any audibile difference detected is not due to "luck" and EQUALLY that any failure to detect audible differences is NOT due to excessive levels of type b/2 errors.
wimms said:
I have been part of both that which you call "properly conducted ABX session" and of actual ABX sessions implemented to not confirm reliably the null hyphotesis (as seems the aim [concious or uncincious] of most of those conducting these tests) but to actually answer the question "what is potentially audible" and "what is potentially preferable". The invariably low sample sizes lead to a drastic lack of statistic significance (sadly).
As said, I do not test to publish (I disdain modern academia - what a boundless bog of sullen inertia and "NIH" bovine excrement) I ask questions to get answers. And I get them.
wimms said:
Significant enough, at this stage of mastering, recording and replay gear. It MAY get better (probably not as SACD together with DVD-A will soon Video 2000 and Betamax) just as CD did. I am in many ways not really concerend with the potential, but with the achievement of the format.
wimms said:
Very possibly.
And CD may very well have been good enough from the start if it had been implemented as 18Bit System with 48 or better 64KHz sample rate (which was easily possible in the late 70's but would have cut play time).
And at the Fox Hunting this weekend in England, had the Dog not sat down to take a <r@p he might have caught the fox in contravention of the new laws.
But CD is what it is, SACD is what it is and the Dog did sit down. So what?
Sayonara
To be perfectly honest, I do not understand the audiophile thing. What I meant with "But whatever it is they look for, it isn't music" is that you listen for that "other" stuff. Surely anyone can hear the difference between a good and a not so good sound system. But the music lover is so occupied enjoying the music that the "other" stuff doesn't really matter.
Lets get serious here for a while. We disagree on many things. At least I don't see being a nuthugger is the main purpose in life. But I understand you very well. Both as a vinyl junkie and as an interaction designer.
Other than this audiophile thing, I'm with you a 100 pct here. This isn't about CD vs. SACD or one-bit DAC vs. multi-bit DAC or whatever. This is about how things are. If the SACD format failed because Sony and Philips bungled, then be so. After all, everybody on this forum would have loved for the SACD being all it was made out to be, and perhaps could have been. And I would guess that your perceived animosity toward the SACD has to do with the fact that lots of people buy into this "hi-rez" thing at face value and ignore how things actually are. Which is exactly how I feel.
Lets get serious here for a while. We disagree on many things. At least I don't see being a nuthugger is the main purpose in life. But I understand you very well. Both as a vinyl junkie and as an interaction designer.
Other than this audiophile thing, I'm with you a 100 pct here. This isn't about CD vs. SACD or one-bit DAC vs. multi-bit DAC or whatever. This is about how things are. If the SACD format failed because Sony and Philips bungled, then be so. After all, everybody on this forum would have loved for the SACD being all it was made out to be, and perhaps could have been. And I would guess that your perceived animosity toward the SACD has to do with the fact that lots of people buy into this "hi-rez" thing at face value and ignore how things actually are. Which is exactly how I feel.
phn said:
Exactly.
But I don't have any kind of animosity against SACD.
I just say that a good PCM recording is superiour.
Even then, I gave an example of a very good sounding SACD disc (although from a PCM master).
Please, if you find it, buy it.
What I do thing is that the story repeats again and again.
People are pushed to new formats by the marketing guys, while sometimes it's not a step forward.
It happened with CD (against vinyl), it happens again with SACD, the story is the same.
Reading the maketing talk is all advantages, it's a "new and very superiour, high-resolution format".
Is it?
Konnichiwa,
I am not listening for "stuff". For me I have two requirements for music.
One is for "pleasant background platter". This is satisfied via a range of fairly lowish spec systems, including a simple PC sound system at work (I can't turn it up at much more than whisper to not upset co-workers).
The other is for giving me an experience of music, that is sufficiently similar to going to a concert that the difference makes me not to bother and go out and go to listen to real music instead. For that music replayed must be riveting, grab me, make me want to listen.
And it requires the experience quality (to me this is largely a natural sound that most people call "huge", meaning with a cavernous acoustic space (if that is on the recording) and the dynamic (macro and micro) range real (unamplified) music manages. I am not sure if this is audiophile or not, I simply require a certain ability of the system that goes past that what most manage to make me "bother" to listen to music intently and as single occuplation.
If a system (or recording) fails to do the above I drift off and do other stuff.
That is missing the point.
Many a "good" sound system fails to give me any involvement or inclination to listen. My bedroom system (6.5" Fullrange speakers in wallmounted sealed boxes and a small and very cheap "all in one" stereo with some modest tweaks) is better drawing me into the music than most Mega$$$ Stereos, despite the expensive stuff being observably better in the trebles, better in the Bass's, going louder more cleanly, having better soundstaging and so on with all that audiophile verbiage.
Perhaps, but you know what really bugs me?
You can release recordings on CD that have a stunning degree of realism, of all the stuff that makes me want to listen to music. Clarly CD is "good enough" for that. Maybe 1 in 50 or 100 released CD's manages that sort of recording. Many more old LP's BTW manage such recordings.
If sound engineers and producers actually for funks sake started making decent recordings for the majority, instead of this compressed to near DC, aphex over-excited and multitracked garbage they usually turn out (even on many Jazz and Classical recordings), all this argument about the quality of the medium would maybe start making sense.
Until then SACD is as much kneecapped as CD by poor recordings and we might as well use downloaded MP3's at 64kbps as music source!
Then again, maybe if we had a format of sufficient inherent quality sound engineers and producers would start hearing just how much they mangle music and would change their working practices (just as in Film large screens and high resolution forced more serious approaches to Photography and in Film Sound large fairly high quality sound systems forced care in producing the soundtrack - never mind it's artificiality).
Okay, that was my little rant.
Sayonara
phn said:
I am not listening for "stuff". For me I have two requirements for music.
One is for "pleasant background platter". This is satisfied via a range of fairly lowish spec systems, including a simple PC sound system at work (I can't turn it up at much more than whisper to not upset co-workers).
The other is for giving me an experience of music, that is sufficiently similar to going to a concert that the difference makes me not to bother and go out and go to listen to real music instead. For that music replayed must be riveting, grab me, make me want to listen.
And it requires the experience quality (to me this is largely a natural sound that most people call "huge", meaning with a cavernous acoustic space (if that is on the recording) and the dynamic (macro and micro) range real (unamplified) music manages. I am not sure if this is audiophile or not, I simply require a certain ability of the system that goes past that what most manage to make me "bother" to listen to music intently and as single occuplation.
If a system (or recording) fails to do the above I drift off and do other stuff.
phn said:
That is missing the point.
Many a "good" sound system fails to give me any involvement or inclination to listen. My bedroom system (6.5" Fullrange speakers in wallmounted sealed boxes and a small and very cheap "all in one" stereo with some modest tweaks) is better drawing me into the music than most Mega$$$ Stereos, despite the expensive stuff being observably better in the trebles, better in the Bass's, going louder more cleanly, having better soundstaging and so on with all that audiophile verbiage.
phn said:
Perhaps, but you know what really bugs me?
You can release recordings on CD that have a stunning degree of realism, of all the stuff that makes me want to listen to music. Clarly CD is "good enough" for that. Maybe 1 in 50 or 100 released CD's manages that sort of recording. Many more old LP's BTW manage such recordings.
If sound engineers and producers actually for funks sake started making decent recordings for the majority, instead of this compressed to near DC, aphex over-excited and multitracked garbage they usually turn out (even on many Jazz and Classical recordings), all this argument about the quality of the medium would maybe start making sense.
Until then SACD is as much kneecapped as CD by poor recordings and we might as well use downloaded MP3's at 64kbps as music source!
Then again, maybe if we had a format of sufficient inherent quality sound engineers and producers would start hearing just how much they mangle music and would change their working practices (just as in Film large screens and high resolution forced more serious approaches to Photography and in Film Sound large fairly high quality sound systems forced care in producing the soundtrack - never mind it's artificiality).
Okay, that was my little rant.
Sayonara
