a.dent Posted January 28, 2016 Posted January 28, 2016 This ^ Exactly. So they are therefore "better" to listen to. And are almost certainly worth paying more for.
Volunteer sir sanders zingmore Posted January 28, 2016 Volunteer Posted January 28, 2016 (edited) Exactly. So they are therefore "better" to listen to. And are almost certainly worth paying more for. yep agree as long as we don't think it's 'better' because it's SACD (or vinyl or DSD or pick your favourite medium) Edited January 28, 2016 by Sir Sanders Zingmore 1
a.dent Posted January 28, 2016 Posted January 28, 2016 yep agree as long as we don't think it's 'better' because it's SACD (or vinyl or DSD or pick your favourite medium) Of course then we could get into the discussion about whether PCM converted to DSD sounds better than the original PCM file.
Guest Eggcup The Daft Posted January 29, 2016 Posted January 29, 2016 Read more at http://www.stereophi...JUudpQLx4PtJ.99 So I did. Especially the comments. It appears from the comments (which include comments from experimenters involved) that Meyer and Moran's data has not only been brought into question, but no longer exists. It is interesting, given the amount of criticism they got, that the data wasn't kept. The interesting stuff is right down at the end of the first page of comments, by the way. I've now read that article again, and some others that I could get at easily, reporting similar experiments. It's hard to compare results from the many differently designed experiments. Reiss is going to be in for an interesting time defending his analysis as well, I guess. Perhaps, rather than arguing about this for ever, we should just adopt the "purist" approach and say that what we really want is the option to purchase a digital copy of the master file, to have what the mastering engineer made; no matter what the format use was. What we then do to reproduce that, should be up to us (and, I guess, the designers of the playback equipment and software we choose).
JSmith Posted January 29, 2016 Posted January 29, 2016 Of course then we could get into the discussion about whether PCM converted to DSD sounds better than the original PCM file. "Despite the marketing hype, there are almost no pure DSD recordings available to consumers. This is because there is currently no way to edit, mix, and master DSD files. Only the rare DSD recordings made from a mixed and mastered analog recording, or those recorded direct to DSD without any postproduction, are pure DSD. Most so-called pure DSD recordings are, in fact, edited in PCM." http://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/ JSmith 1
a.dent Posted January 29, 2016 Posted January 29, 2016 "Despite the marketing hype, there are almost no pure DSD recordings available to consumers. This is because there is currently no way to edit, mix, and master DSD files. Only the rare DSD recordings made from a mixed and mastered analog recording, or those recorded direct to DSD without any postproduction, are pure DSD. Most so-called pure DSD recordings are, in fact, edited in PCM." http://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/ JSmith That's why I made that comment. People often swear that SACD sounds better than CD but the source is often the same. Why not buy the CD or 44.1 download and convert it yourself at home? Saves $$$ and atleast you know the provenance to some extent.
DoggieHowser Posted January 29, 2016 Author Posted January 29, 2016 You can mix in analog - and then do an analog to DSD conversion w/o PCM involved. The Tubes Only series of SACDs are superb - they even use analog microphones and analogue microphone amps. There are also numerous examples of remastering of old tapes done in the analog domain and then converted to DSD direct. Of course, the funny thing is that most DACs today - including Berkeley and Bel Canto - who swear by PCM - use sigma delta DACs in their designs so in effect you are hearing a DSD-ish reproduction of your PCM signals.
firedog Posted January 29, 2016 Posted January 29, 2016 "Despite the marketing hype, there are almost no pure DSD recordings available to consumers. This is because there is currently no way to edit, mix, and master DSD files. Only the rare DSD recordings made from a mixed and mastered analog recording, or those recorded direct to DSD without any postproduction, are pure DSD. Most so-called pure DSD recordings are, in fact, edited in PCM." http://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/ JSmith The quote is so "yesterday". There are hundreds - possibly over a thousand - titles available recorded in DSD. But they exist. NativeDSD.com even has over 100 DSD albums that have only been sold as downloads - and were never on disc. Yes, mostly classical. They can also be edited - there is tech which makes a fraction of a second PCM conversion only at the desired edit point and then converts back. So even after editing the track can still be 99.99% unaltered DSD.
acg Posted January 29, 2016 Posted January 29, 2016 The quote is so "yesterday". There are hundreds - possibly over a thousand - titles available recorded in DSD. But they exist. NativeDSD.com even has over 100 DSD albums that have only been sold as downloads - and were never on disc. Yes, mostly classical. They can also be edited - there is tech which makes a fraction of a second PCM conversion only at the desired edit point and then converts back. So even after editing the track can still be 99.99% unaltered DSD. There are probably a thousand new Redbook albums a day/week across the planet... 1
LHC Posted January 29, 2016 Posted January 29, 2016 You guys are just being 'tribal' But there are a few interesting paragraph in that Mojo Audio blog that is worth discussing. "Another common marketing myth about DSD vs. PCM is that when blind listening tests were done comparing DSD to PCM, there was a consensus that PCM had a fatiguing quality and DSD had a more analog-like quality. This was proved to be total marketing BS. One way that marketing lie was perpetuated was with hybrid SACDs that have DSD64 and 16-bit 44.1KHz PCM on the same disk. The DSD64 tracks have roughly 33 times the resolution of the 16-bit 44.1KHz tracks so that they could make DSD sound better than PCM in comparisons. The truth is that in recent blind studies they've proved that high-resolution PCM and DSD are statistically indistinguishable from one another. Considering that nearly all DSD recordings were edited, mixed, and mastered in PCM, it is no wonder." "The DSD64 tracks on a hybrid SACDs have roughly 33 times the resolution of the 16-bit 44.1KHz PCM tracks. This was done purposely so that they could sell more SACD players by fooling potential customers into believing that they were making a fair comparison when they played music from the same disk." "DSD has significantly higher quantization noise than PCM, and the noise is much closer to audible frequencies, requiring significantly more sophisticated digital filters, as well as noise-shaping and upsampling algorithms, that can result in distortion of the analog signal." "High-resolution PCM and DSD formats are statistically indistinguishable from one another in blind listening tests. When a PCM file is played on a DSD or Bit Stream converter, the DAC chip has to convert the PCM to DSD in real time. This is one of the major reasons people claim DSD sounds better than PCM, when in fact, it is just that the chip in most modern single-bit DACs do a poor job of decoding PCM." Firstly he didn't include any legitimate references to back up his claims so it is near impossible to understand where he is coming from Secondly he seems to flip flop between whether DSD or PCM is better. On the one hand he said DSD has distortion in the analog signal, but then he also said single-bit DACs does a poor job of decoding PCM. So in the end result which one sounds better? He did say DSD and high resolution PCM sounds the same under blind testing He also said the DSD track on a hybrid SACD will sound better than the 44.1kHz PCM track on that same disc. Is he implying that PCM track is poorer quality than what one may find in the equivalent redbook CD? As he has argued there are many PCM conversions involved in editing and mastering, so a 44.1kHz track could easily be produced from one of those stages and be of the same quality as the DSD track. So I don't understand what he is saying here. 1
LHC Posted January 29, 2016 Posted January 29, 2016 Perhaps, rather than arguing about this for ever, we should just adopt the "purist" approach and say that what we really want is the option to purchase a digital copy of the master file, to have what the mastering engineer made; no matter what the format use was. What we then do to reproduce that, should be up to us (and, I guess, the designers of the playback equipment and software we choose). Totally agree with this. Although MQA is one way we could get close to accessing the master.
Guest Eggcup The Daft Posted January 30, 2016 Posted January 30, 2016 Totally agree with this. Although MQA is one way we could get close to accessing the master. An MQA file is NOT the original recording, and is likely not to be the true master either. I seriously mistrust a system where a "master quality" light comes on for the playing of a lossy compressed file - but stays off if you play a true copy of the master. As I said, a lot of this is "in your head" and that's what is in mine. If there has to continue to be an intermediate step between the master and what we play, then surely that intermediate step has to be of proven benefit. Until that is done MQA remains marketing hype dressing a way to give us something less.
LHC Posted January 30, 2016 Posted January 30, 2016 This appears to be the AES workshop in question. http://www.aes.org/events/139/workshops/?ID=4687
firedog Posted January 31, 2016 Posted January 31, 2016 There are probably a thousand new Redbook albums a day/week across the planet... Irrelevant. The claim in the article , there are almost no pure DSD recordings available to consumers. This is because there is currently no way to edit, mix, and master DSD files. Only the rare DSD recordings made from a mixed and mastered analog recording, or those recorded direct to DSD without any postproduction, are pure DSD. Most so-called pure DSD recordings are, in fact, edited in PCM is clearly refutable. That's the point. No one is forcing anyone to like or listen to DSD. I personally have found some fantastic DSD recordings of music I like that ISN"T recorded in PCM. I also have found many of the tape>DSD remasters of classic Rock and Jazz to be better sounding than any other version of the same album I know of-PCM or vinyl. That doesn't mean DSD is "better"; but it does mean there might be a good reason to buy a release in DSD. Why does anyone have a problem with that? 1
almikel Posted February 21, 2016 Posted February 21, 2016 apologies for re-covering old ground, but I'm still confused regarding the temporal resolution capability of digital sampling at the recording end of the process. Assuming 48kHz sampling in the ADC, so a sample every 20uS, and band filtering on the input as per normal. If you had 2 drum hits 10uS apart, and the first drum hit aligned with a sample point (and of course continued for many many sample points), the 2nd drum hit will start in between samples and won't be picked up till the next sample (and of course continue for many many samples).... Is there not a 10uS error for the start of the 2nd drum hit? And when played back (let's assume played back at 48kHz FLAC so there's no sample rate differences), the 10uS error will be faithfully reproduced? I can't see how that error would represent frequencies higher than 24kHz (sample rate 48kHz), so I'm confused. I'm not saying it's audible, just trying to get my head around the theory, and very smart people like Lipshitz et al saying the time resolution for band limited signals is almost infinite. cheers Mike
Guest rmpfyf Posted February 21, 2016 Posted February 21, 2016 (edited) @@almikel (Editing as my post was originally a bit vague...) 48kHz is not a studio recording frequency (unless it's a home studio ~15 years ago, my Data Minidisc-based kit was 48kHz). It's intended for playback. 48kHz or lower with sufficient bit depth can place a waveform with far more resolute accuracy than 1/48k seconds. Studio recording is typically higher frequency, lower frequency formats are typically used for playback. Could you actually hear two sounds 10us apart? Probably not, which is why the need to play them so isn't so relevant. Of greater relevance is the shape of a waveform, and with hires (simply) more is possible, irrespective of 'how far apart' the next waveform is. You don't need 10us gaps, you need placement and reconstructive accuracy. But this is a psychoacoustic perspective, psychoacoustics simply tell us the minimum 'gap distance' (and thus maximum frequency) we can reasonably hear. It could be argued that the gap between what's audible and what's actual content shouldn't be decided by a playback medium and should be left to our heads (ears included). There is no way a cymbal crash in 48kHz (or Redbook) is the full representation of what's actually happened (or of what's probably recorded in the studio master), it's an attempt at recreating what some theories tells we can hear. In essence, therefore, the choice of output format (in your example 48kHz) does act as a filter of sorts. Edited February 21, 2016 by rmpfyf
almikel Posted February 21, 2016 Posted February 21, 2016 (edited) hi @@rmpfyf , that doesn't help answer my question, and obviously I'm behind the times......and as I said in my post I'm not saying it's audible...I'm just trying to understand the theory when Lipshitz and Vanderkooy say the time resolution is nearly infinite.... Let's say 96kHz sampling, and the 2 drum hits are still 7uS apart, with the first hit on the sample point - the start of the 2nd drum hit is 3uS before the next sample (quick maths only - sample every 10uS)...is there an error???.. or the sampling is 192kHz and the drum hits are 2uS apart etc etc I'm struggling with Lipshitz and Vanderkooy saying the time resolution is nearly infinite...help me out... I accept that it's not audible, but I'd like to understand the theory - if the error gets buried in noise I'm happy with that also...that I would understand... You asking whether I can hear the difference between two sounds 10uS apart (I can't) doesn't help my understanding, nor does your point that 48kHz hasn't been used in a studio for 15 years... ...I'm trying to understand the theory...your input throughout this thread has been great - it would be appreciated if you (or others) could address my question. cheers Mike Edited February 21, 2016 by almikel
almikel Posted February 21, 2016 Posted February 21, 2016 hi @rmpfyf , I'm still trying to absorb your edit - but thanks for the extra info... I'm still confused based on Lipshitz and Vanderkooy on how the time resolution is nearly infinite for signals that start (or stop) in between samples...clearly this is different to quantisation error (voltage error for a specific sample), and I suspect it just adds to noise - but I may be completely wrong??? Mike
hochopeper Posted February 21, 2016 Posted February 21, 2016 hi @rmpfyf , I'm still trying to absorb your edit - but thanks for the extra info... I'm still confused based on Lipshitz and Vanderkooy on how the time resolution is nearly infinite for signals that start (or stop) in between samples...clearly this is different to quantisation error (voltage error for a specific sample), and I suspect it just adds to noise - but I may be completely wrong??? Mike Hi Mike, I'm a bit limited for time so I won't make a full interactive graphic which is what I'd make ideally. I have been meaning to for a while, maybe one day. I think you'll get where I'm going with the following: Consider 2 sine waves of the same frequency. One has a slight phase shift from the other (representing time delay less than 1/sample rate) The sampling ADC has no idea in advance of the phase of either sine wave. Both sine waves are still captured completely (providing Nyquist Theorem is satisfied) Why should the minimum phase difference that can be captured be limited by the sample rate? It isn't. It's limited by the resolution of each sample to differentiate the slight change in magnitude that results from the phase shift at each sample time. Does that help or confuse more? Chris 1
Guest rmpfyf Posted February 21, 2016 Posted February 21, 2016 @@almikel Mike, you raise it fairly. I'll try again. You're right, whilst time resolution for a single sample is very, very fine, time resolution for successive samples in the same channel is limited effectively by the sampling rate. Ultimate res in Redbook for 'placing' a signal is 44.1k/(2^16), so very fine. The 'adjacent samples in one channel' argument pushes you straight into psychoacoustics and yeah, error. What's in playback is clearly not the ultimate spectral content said drums intended. Certainly there's quant error. Armchair psychoacoustics tells you that doesn't matter because it's content you can't hear regardless when assessing this content in the spectral domain. I'm not so sure - we don't listen to spectral densities, we listen to time histories. Much of the alleviation of said error is going to come down to quality of filtering higher-frequency content. IMHO early digital suffered here somewhat. But spread across two channels temporal resolution is more than sufficient to have a slight delay in one channel enough to recreate a live recording sounding really live, etc.
Guest rmpfyf Posted February 21, 2016 Posted February 21, 2016 Why should the minimum phase difference that can be captured be limited by the sample rate? It isn't. It's limited by the resolution of each sample to differentiate the slight change in magnitude that results from the phase shift at each sample time. This is a 500% better explanation than mine
davewantsmoore Posted February 21, 2016 Posted February 21, 2016 apologies for re-covering old ground, but I'm still confused regarding the temporal resolution capability of digital sampling at the recording end of the process. Assuming 48kHz sampling in the ADC, so a sample every 20uS, and band filtering on the input as per normal. If you had 2 drum hits 10uS apart, and the first drum hit aligned with a sample point (and of course continued for many many sample points), the 2nd drum hit will start in between samples and won't be picked up till the next sample (and of course continue for many many samples).... Is there not a 10uS error for the start of the 2nd drum hit? No. You are focussing too much on your own interpretation of what the "sample points" mean. Everything between the sample points is captured perfectly. As long as the frequency of the signal is below the nyquist limit (half the sample rate). If you search, there are many sites which explain this to you in pictures. If you have sample points every X .... then the signal can start/stop anywhere in between the points, as long as the frequency (slope) of the signal, is less than X/2 1
davewantsmoore Posted February 21, 2016 Posted February 21, 2016 A way to 'visually' illustrate this to your self is. Draw a "signal" (however random you like). Amplitude on vertical, time on horizontal. Now. Look at the signal you have drawn. The slope of the line, ie. how quickly it rises and falls. Represents the frequency of the signal at any given place. This determine the sample rate you need to capture it correctly.... and so in practise, you will need to either increase the rate, or filter the signal to remove the some high frequencies (ie. make the rise/fall slope of the lines gentler). Ok. Now draw sample points, equally spaced out on the horizontal axis. Start wherever you like... anywhere. Everything between those points, is captured perfectly .... as long as the rise/fall of the line is not too steep, ie. as long as the frequency of the signal is not too high. Because you can start your sample points anywhere.... the issue you were mentioning, about the location in time, of a certain signal. Is not an issue. I hope that helped, rather than confused. So much easier in person with a whiteboard, cos we can focus on (and keep going over) whichever aspect of it is unclear. 1
a.dent Posted February 22, 2016 Posted February 22, 2016 I'm as confused as Mike is here regarding this concept of (say) two drum hits 10us apart. Are you saying that the DAC can look back in time (say 5us) between two samples and reconstitute the start of the second drum hit even if it occurs between samples? Or does it look forward in time (spooky) and predict where that second drum hit starts and start making allowances for it in the reproduced sine wave before it actually happens? I believe sampling theorem is correct but I just have trouble grasping it.
hochopeper Posted February 22, 2016 Posted February 22, 2016 I'm as confused as Mike is here regarding this concept of (say) two drum hits 10us apart. Are you saying that the DAC can look back in time (say 5us) between two samples and reconstitute the start of the second drum hit even if it occurs between samples? Or does it look forward in time (spooky) and predict where that second drum hit starts and start making allowances for it in the reproduced sine wave before it actually happens? I believe sampling theorem is correct but I just have trouble grasping it. It doesn't need to sample at the 'start'. It doesn't need to look forward or back. A sound of any frequency (within bandwidth limits of Nyquist) will have a particular magnitude after a particular time. So when the sample is taken within the context of the adjacent samples the result can reconstruct the complete (bandwidth limited) audible signal. There are thousands of samples and combined they have all of the information about the signal that was received.
Recommended Posts