![]() ![]() |
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
204.245.209.24
In Reply to: The more you crank it up, the more dynamic range you get, but also the more amplifier power you need posted by Christine Tham on September 13, 2005 at 15:47:45:
I tend to agree with that statement. I certainly can't hear it. However, I also tend to believe that non-integral sample rate conversion cannot be audibly distinguished from integral sample rate conversion. And there is certainly no conlusive evidence to the contrary for that either.There is a psychological effect for me however. Just knowing that there will be temporal errors introduced in non-integral SRC, regardless of how infinitesimally small (inaudible) they may be, just rubs me the wrong way. So I prefer integral SRC.
For the same reason I prefer high bit/sample rate PCM to DSD. Why bother with a system that generates measurable noise-shaping artifacts, whether audible or not, when you can have a measuably cleaner PCM reproduction.
But that's just the engineer in me I guess. I always prefer the most elegant solution to a problem.
![]()
Follow Ups:
ALL sample rate conversion generate quantization noise and artefacts - one day we will have DSPs with sufficient power (in terms of filter taps and numerical precision) to make this a non-issue, but that day hasn't quite arrived yet.*** Why bother with a system that generates measurable noise-shaping artifacts, whether audible or not, when you can have a measuably cleaner PCM reproduction. ***
The "system" that you are not bothering is the same "system" that's embedded in just about every PCM ADC and DAC so the artefacts are not confined to DSD. Look at spec sheets.
The "measurably cleaner" PCM reproduction is because the modulation typically happens at 128x rather than 64x but the ultrasonic artefacts are still there. The only way to avoid it is to use a ladder type DAC, but that introduces other problems which is why they are becoming a rarity these days.
DSD is an interesting idea, comparable to the idea of capturing the "raw" image from a digital camera instead of converting to TIFF (PCM) or JPEG (compressed audio). There are benefits of capturing data "raw" (in this case, the undecimated delta sigma bitstream) and defer conversion to PCM, but the acquired signal is not easily editable without conversion.
However, 64fs may be good enough at one stage, but these days 128fs is the "norm" for both ADCs and DACs, and "hybrid" designs (multi-bit plus sigma delta) is common, so "raw" really should be a 5 bit 128fs bitstream rather than a 1 bit 64fs bitstream. It's kind of like forcing digital cameras to capture 1 megapixel "raw" images when the CCD is actually a 10 megapixel wafer, so the original "benefit" of DSD (not throwing away data) becomes a moot point. To me, that's a far more serious objection to DSD than ultrasonic artefacts.
![]()
"We make all our recordings in PCM. And we try to fit them into a very short total time because the problem is if you make more than 65 minutes on a hybrid SACD you have a problem with the dynamics. And that is the reason why we try to reduce the total time. Then it is possible to make a good sound in SACD.You must know that the DSD signal is very bad. They say you have a frequency response up to 100kHz. But the problem is when you come over 40kHz you have a lot of distortion, and it is not possible to cut this distortion. You must take it onto the SACD disc itself. And when you have a lot of dynamics, and a lot of total playing time, then you don’t have enough room on the SACD. So you must reduce the dynamics. And this is the reason why you find a lot of SACDs that are poor in dynamics. And that is very bad. And therefore the SACD is a big big technical problem. But when you make a PCM recording and when the playing time is not too long, you can realise a good result.
The reason why we have DSD is that they have the license for the CD. But this ability to earn money stopped two years ago, because the 25 years was over. And this is the reason why they think, ‘well, what can we do to get more money?’ And the reason is DSD, but DSD is a very problematic and a difficult format.
And when you read the instructions for recording they say that DSD is only a ‘consumer’ format, i.e. you should make PCM recordings, and when you bring it on a disc, you can use DSD. But most of the record companies don’t speak about it. It is not a good format, but the problem is that the marketing for the SACD was very good. They spent more than 50million euros for the labels to make recordings on SACD. They started to buy recordings. They went to all the independents saying: "Well, we spend the money for three or five recordings and you make it, we pay it." After three years they stopped it — at the end of last year."
![]()
Tsk, tsk... Martin, are you quoting KEKL? This sounds SO like him. If so, very funny! However, if you intended to submit the quote as a serious critcism of DSD... uhhh the word RUBBISH comes to mind. English being a second language is a fine excuse for shaky grammar, but what's the excuse for having no F***ING CLUE about what he/she's talking about?"You must know that the DSD signal is very bad"
Just another paranoid Sony hater with NOTHING to back up what he/she says. And not even willing to take credit for the ill-formed opinion that's being spewed. Try again using some ACTUAL information from a REAL person.
![]()
He was a very real person, who actually makes classical hybrid SACDs. And somebody who was actually offered "incentives" by the DSD consortium.I phoned him 'out of the blue'. He was assumming I was a typically non-technical member of the public. Sometimes that's the best way to find out how people in the industry really feel.
![]()
nt
![]()
"You must know that the DSD signal is very bad."This guy sure sounds like he knows he's talking about something.
"It is not a good format, but the problem is that the marketing for the SACD was very good."
Oh yes I've never seen such fantastically successful marketing before.
![]()
English is not his first language.
![]()
Although I have no basis for refuting the remarks.
![]()
. . . so I will respect the wishes of the source not to have their identity repeated here. I do understand that some people here will want a "byline" so to speak. But I'm afraid for the reason I stated, I will not name my interviewee.
![]()
Chances are, if you are listening to a PCM recording that has undergone dither at some stage in it's lifecycle, it will contain ultrasonic noise, in many cases similar to DSD ultrasonic noise.Check out the following URL for comparisons of noise shaping curves for various dithering algorithms. What the diagrams don't show of course is that the noise shaping continues into the ultrasonics, and most algorithms work on the principle of shifting noise from the audible band up into the ultrasonic spectrum.
In this respect, DSD is no better or worse than any of the commonly used algorithms.
![]()
"What the diagrams don't show of course is that the noise shaping continues into the ultrasonics"The noise energy can only be distributed within the systems bandwidth if pcm is used.
(Upper part being ultrasonic of course)Beyond the upper bandwidth limit of a pcm system the noise will be filtered.
Also in this repect DSD has the disadvantage that the noise shaping needs to be done in the ad converter and as a consequence the noise is *always* included in the recording.
With pcm this depends on the hardware/software algorithms used.
Frank
*** "What the diagrams don't show of course is that the noise shaping continues into the ultrasonics"
The noise energy can only be distributed within the systems bandwidth if pcm is used. ***If you look at the diagrams, they all cut off at 30kHz. What I said was (and I think you are agreeing with me) that the noise shaping continutes beyond 30kHz into whatever the upper limit is.
*** With pcm this depends on the hardware/software algorithms used. ***
Well, that is a moot point, since there will be embedded noise in PCM whenever delta sigma A/D is used. The only way you can avoid ultrasonic noise is by using ladder A/D (which has linearity issues) or by generating waveforms synthetically.
There is also additional ultrasonic noise generalted by sigma delta DACs, above and beyond noise embedded in the recording. My sound card implements a filter that is down -2dB at 50kHz to counteract this noise.
![]()
..Well, that is a moot point, since there will be embedded noise in PCM whenever delta sigma A/D is used. The only way you can avoid ultrasonic noise is by using ladder A/D (which has linearity issues) or by generating waveforms synthetically...Not quite...
Modern AD converters using 1 bit conversion can have higer internal sample rates and can have a flat noise floor for the entire usable PCM bandwidth.
Any rise in noise floor due to a noise shaper beyond the nyquist frequency is filtered out.
With DSD this rising noise floor is always found within the usable bandwidth.
With PCM the noise floor can be flat across the useable bandwidth.
(The latest BB ADC goes flat up to 50kHz. pdf page 10)
Frank
![]()
'If you look at the diagrams, they all cut off at 30kHz. What I said was (and I think you are agreeing with me) that the noise shaping continutes beyond 30kHz into whatever the upper limit is.'It really cannot continue beyond the upper band limit with PCM.
(Unless a bitstream dac with a noise shaper is used. But that's a moot point in a discussion abbout about adding noise shaping during 'processing' of the audio data. )
... you know that there *are* actually options for ultrasonic noise distribution and noise shaping methods.
I know.But that isn't an issue here.
What is is that in principle noise shaping isn't part of PCM and can be avoided with hires.
With DSD the noise shaping is required in principle. The noise shaping option in DSD tooling is on top of the noise shaping in the analog to DSD converter.
.
![]()
Just as DSD noise-shaping artifacts are inaudible (I think we all agree here), so too are PCM noise-shaping artifacts, espcially considering how much more pervasive the noise-shaping is in DSD.So we again come back to my original premise: high-rate PCM is more elegant than DSD. If DSD were audibly superior to high-rate PCM then the issue of elegance would be moot. But of course no one has shown conclusively that DSD can be audibly distinguished from high-rate PCM.
Another thing I've been wondering lately...if the original CD patent had specified 44.1/20 instead of 44.1/16 would we even be having this conversation? I'm beginning to think anything above 48/24 may be overkill...
![]()
*** So we again come back to my original premise: high-rate PCM is more elegant than DSD ***I'm not sure what you mean by "elegance" but in the case of ultrasonic noise it's a completely moot point since as I've pointed out real life PCM recordings have just as much ultrasonic noise as DSD so I don't know what you mean when you say noise shaping is more pervasive in DSD.
Using a different criteria, one can easily argue that DSD is more "elegant" than PCM. In fact, I think Sony and Phillips *are* advancing such arguments. But their arguments are equally moot because many SA-CDs are mastered from PCM recordings.
I would say - forget about the elegance and just enjoy the music!
*** if the original CD patent had specified 44.1/20 instead of 44.1/16 ***Be thankful they did not use 14-bits as originally proposed. It's easy to moan and whinge in hindsight but you must remember at the time it was impossible even to achieve 14 bit linearity, much less 16 or 20. The only reason we can get 20-bit accuracy these days is due to delta sigma (which DSD is based on, so you are accusing something for not being elegant when it is in fact the technology that broke the linearity barrier!)
*** I'm beginning to think anything above 48/24 may be overkill ***
On my soundcard, 44.1/24 generates the "best" results, as you can see from the attached link
![]()
"On my soundcard, 44.1/24 generates the "best" results, as you can see from the attached link"
'so too are PCM noise-shaping artifacts'There is no such thing as 'PCM noise shaping artifacts' those are just plain noise shaping artifacts.
24 bits PCM at 48ks or higher doesn't require noise shaping at all.
Just plain old dither would be sufficient and probably the best option to use with 48ks. Higher sampling rates give the option to move the noise spectrum higher up the inaudible frequency range and use a less agressive noise shaping algorithm.Noise shaping is only a trick to enhance the perceived resolution on delivery formats with lower resolution.
Frank
This post is made possible by the generous support of people like you and our sponsors: