davewantsmoore Posted July 7, 2016 Posted July 7, 2016 that's the bit that gets bandied about a lot, seemingly. Becuase many many audiophiles are trying to post-rationalise why they find higher sampling rates sound better .... without actually understanding how it works (sampling and signals are quite complex) ..... and his conclusion seems to answer the question .... and in some ways it does. It gives you (one) reason why low-quality re-sampling of a signal will reduce the quality (as the signal may no longer be in the same place in time, and hence sound different due our ability to differentiate small signal delays) .... but it doesn't tell us we need a high rate in the first place. We cannot look at a low-sample-rate audio, and assume that it has insufficient time resolution.
LHC Posted July 7, 2016 Posted July 7, 2016 Becuase many many audiophiles are trying to post-rationalise why they find higher sampling rates sound better .... without actually understanding how it works (sampling and signals are quite complex) ..... and his conclusion seems to answer the question Dave, what you have said is completely unfair. By that I mean if there are issues with sampling and filtering that escape most people's attention, yet are responsibility for higher sampling rate recordings sounding better than redbook, then why don't you just cite some reliable peer-reviewed reports that point out these issues. And if such papers exist, then why are they not on Reiss survey list? It would really help the discussion if you actually name the source you are referring to. Furthermore, as I have acknowledged before, filtering design can have an audible impact. But it is not mutually exclusive to other possible reasons. Just because you may have found one cause of audibility, it doesn't automatically eliminate all other possible causes. Again, Reiss survey has been looking at audible effects not associated with choice of filtering; that was outside his scope.
Guest rmpfyf Posted July 7, 2016 Posted July 7, 2016 Dave, what you have said is completely unfair. By that I mean if there are issues with sampling and filtering that escape most people's attention, yet are responsibility for higher sampling rate recordings sounding better than redbook, then why don't you just cite some reliable peer-reviewed reports that point out these issues. And if such papers exist, then why are they not on Reiss survey list? (I don't mean this in a snooty way, though it's going to sound poor) they're not on the Reiss survey list because the majority of what's discussed concerns common digital signal processing - it doesn't need a list of peer-reviewed journals on the subject, just a good textbook. Please don't get me wrong, I have some nice 2L purchases I enjoy a lot. I won't pretend hires is a panacea, from a DSP perspective it just makes a few things (esp. filtering) easier to get good sound, and IMHO this is not a bad conclusion. Not a sexy conclusion that has me rushing to HDTracks to repurchase the library... but a reasonable conclusion. If the world (well... Sony/Phillips) weren't struggling to get 74 mins on a CD way back when we'd probably have more bits to play with, not the minimum known to meet the performance envelope defining known psychoacoustic theory. IMHO what's only left to discuss is whether a waveform with characteristics more abrupt than is possible with Redbook is relevant music, and whether it makes a real difference. What's an example? Dunno. Maybe scrunching paper on a dead acoustic background...
Volunteer sir sanders zingmore Posted July 7, 2016 Volunteer Posted July 7, 2016 Dave, what you have said is completely unfair. By that I mean if there are issues with sampling and filtering that escape most people's attention, yet are responsibility for higher sampling rate recordings sounding better than redbook, then why don't you just cite some reliable peer-reviewed reports that point out these issues. And if such papers exist, then why are they not on Reiss survey list? It would really help the discussion if you actually name the source you are referring to. Furthermore, as I have acknowledged before, filtering design can have an audible impact. But it is not mutually exclusive to other possible reasons. Just because you may have found one cause of audibility, it doesn't automatically eliminate all other possible causes. Again, Reiss survey has been looking at audible effects not associated with choice of filtering; that was outside his scope. But loads of the other "possible reasons" are in fact not possible
LHC Posted July 7, 2016 Posted July 7, 2016 (I don't mean this in a snooty way, though it's going to sound poor) they're not on the Reiss survey list because the majority of what's discussed concerns common digital signal processing - it doesn't need a list of peer-reviewed journals on the subject, just a good textbook. But loads of the other "possible reasons" are in fact not possible That doesn't address the question. If the sole reason for the audible gain in hi res audio is already known, and readily found in a 'good textbook', and other possibilities are in fact not possible. Then why would the Audio Engineering Society commission an extensive literature survey to baseline what is the state of the knowledge regarding high resolution audio? Shouldn't those professional engineers just open up their textbook to find the answers? None of this adds up, so someone has to be wrong.
LHC Posted July 7, 2016 Posted July 7, 2016 The other thing that don't add up is this. On the one hand you guys say this is all common knowledge found in sampling textbooks. Yet, then in other posts you say it is really difficult to do sampling and filtering well, that is why redbook audio don't sound as good as hi res. Hang on, digital audio has been with us for more than 30 years. Sampling theory has been invented decades earlier. You would think the smart engineers and academics would have already prefected the implementation of sampling and filtering techniques by now, and become standard practise in the industry. But that is not the case, and instead we have an industry promoting hi res as something intrinsically better. So again the question, why don't the engineers in this industry just say - lets go back to the textbook and do sampling and filtering 'right' for redbook? I understand MQA is moving in some of that direction. But this is 2016, and CD was introduced in the 80s.
betocool Posted July 7, 2016 Posted July 7, 2016 I should stay away from this, but, @@LHC, "engineers" and "sales people" always always have different opinions. I'm quite sure that the industry does not promote what engineers can do (and have done) right in the digital realm. MQA is another can of worms, there's a whole thread dedicated to it.
Guest rmpfyf Posted July 7, 2016 Posted July 7, 2016 That doesn't address the question. If the sole reason for the audible gain in hi res audio is already known, and readily found in a 'good textbook', and other possibilities are in fact not possible. Then why would the Audio Engineering Society commission an extensive literature survey to baseline what is the state of the knowledge regarding high resolution audio? Shouldn't those professional engineers just open up their textbook to find the answers? None of this adds up, so someone has to be wrong. Because the state of knowledge of hires audio is a greater challenge that what hires provides from a DSP perspective alone. Most of the argument you've provided concern DSP perspectives. That filtering is easier, can be tighter, isn't necessarily audibly better, can be implemented with less penalty etc is all known. The stuff about 5us misalignment perception thresholds gets thrown around as though it's relevant to the discussion... it's not. The AES is seeking a broader perspective because - simply - the effects of hires from a system perspective is considerably broader than arguments broadcast from the 'sampling interval' pulpit. The answer is unlikely to be be that it's outright better or that the limits are consistently, audibly greater, but that's not really a problem and neither a loss. It still means hires can be good and in many cases provide a better listening experience than Redbook. There are some broader questions about waveform shame and neurophysical perception limits but again, this stuff is a broader and valid argument. The other thing that don't add up is this. On the one hand you guys say this is all common knowledge found in sampling textbooks. Yet, then in other posts you say it is really difficult to do sampling and filtering well, that is why redbook audio don't sound as good as hi res. That's unfairly simplistic. Difficult doesn't mean it's impossible, and the applicability of your filtering depends on the material you've got. Horses for courses. You're less sensitive at hires, period - that doesn't mean 'outright better'. Hang on, digital audio has been with us for more than 30 years. Sampling theory has been invented decades earlier. You would think the smart engineers and academics would have already prefected the implementation of sampling and filtering techniques by now, and become standard practise in the industry. But that is not the case, and instead we have an industry promoting hi res as something intrinsically better. So again the question, why don't the engineers in this industry just say - lets go back to the textbook and do sampling and filtering 'right' for redbook? The sheer notion of a discrete Fourier transform (as opposed to an infinite Fourier series) means that ultimately, there is no perfect filter. Ever. If you want perfect filters, go to a concert and listen live. If you want filter design that's less stressed, make the bins smaller, which means upping the frequency. Which means there's music that's intrinsically easier to make sound better in mastering at hires, and stuff for which it really doesn't make a difference. Because it depends on the source material. I understand MQA is moving in some of that direction. But this is 2016, and CD was introduced in the 80s. Reluctant to agree here. Adaptive filtering isn't new (maybe for audio but it's not new) and psychoacoustics hasn't changed. We haven't discovered we can hear higher frequencies or time misalignment/phase shift not explicable by Redbook. Filtering is better than it was in the 80's, and there is no reason CDs can't adequately replicate realism if done properly. It goes deeper than that. Nor do replication limits stop at the input to your DAC.
Guest rmpfyf Posted July 7, 2016 Posted July 7, 2016 (edited) And @@LHC - you're 100% right to push this discussion. IMHO the truth of any advantages to hires are a much harder sell than what it's often sold to do. Edited July 7, 2016 by rmpfyf
davewantsmoore Posted July 7, 2016 Posted July 7, 2016 There is an article in this months Audio Voice.... being sad about people on the AES foums asking "is it worth it to record any higher than 44.1khz" They author of the article represents what is broken in the world IMHO. There is nothing wrong with asking this question, and expecting a more reasoned answer than just "high res sounds great", or "no nyquist said so" ...... WHY does is sound great? What IS a factor? What ISN'T a factor? WHY is or isn't it worth while to raise the rate? He goes on to recount stories about how he's used every format in the world (and the newer ones sounded better) .... but provided no discussion about why, or what is at play. He simply perpetuates the the blind march towards "high res" consumer formats, by saying "it sounds great trust me". This is a professional magazine >_< The real kicker. The title of the article is "It's always good to know". ​This is what it sounds like many people: Dear Readers. You just KNOW that 24bit PCM with high sampling rates is good, so stop thinking. He mentioned this: As Dr. Reiss says: "One motivation for this research was that people in the audio community endlessly discuss whether the use of high-resolution formats and equipment really makes a difference." Yes, indeed. He might go on to wonder WHY there is so much conjecture.... it is simply that when you look at the details of what is happening inside the electronic boxes..... there is no "Yes or No" answer to this. "Performance" is a very complex measure..... just because one might reasonably want to feed high rate audio into their converter, doesn't mean audiophiles should re-buy their music at a premium.... without a much stronger value proposition. 1
davewantsmoore Posted July 7, 2016 Posted July 7, 2016 the truth of any advantages to hires are a much harder sell than what it's often sold to do. I unliked this post, so I could like it again. MQA are making a play for "selling" a value proposition for audio quality ..... although I do fear what "walled-garden" it could lead to. Perhaps we need "OpenMQA"
davewantsmoore Posted July 7, 2016 Posted July 7, 2016 Dave, what you have said is completely unfair. By that I mean if there are issues with sampling and filtering that escape most people's attention, yet are responsibility for higher sampling rate recordings sounding better than redbook, then why don't you just cite some reliable peer-reviewed reports that point out these issues. And if such papers exist, then why are they not on Reiss survey list? Because he simply did an analysis of results presented of listening tests. He does not attempt to uncover why any of the high rate material sounded the way it does. It would really help the discussion if you actually name the source you are referring to. There is no "the source", aside from basic mathematical proofs..... All the textbooks in the library which explains how signals and sampling works.... there is not a single result anywhere which contradicts what I am saying, as it is all just relatively basic math (representing signals, sampling them, etc.) The "proof" is in the fact, that you can right now, test the things that I am saying and you will see they are correct. More interesting results perhaps, are things like: Audibility of high frequency signal components Audibility of filters involved in digital sampling (ie. how filters are implemented in real devices which have finite computing power) Where you would be aware that the very few come close to demonstrating the first .... but the later is fairly well audible. You need to understand that if there is a claim of "44.1khz is not enough" .... then the burden to demonstrate exactly how it is "not enough", lies on the person making that claim. The reality is that there are plenty of things which can and do go wrong when using that rate to represent <20khz audio ..... but none of them have to be issues. The aren't theoretical problems, only practical problems which can be addressed to one degree or another. This thread is about "why 192khz matters" ... and I think we have done a fairly good job of unpacking that.
davewantsmoore Posted July 8, 2016 Posted July 8, 2016 (edited) Furthermore, as I have acknowledged before, filtering design can have an audible impact. But it is not mutually exclusive to other possible reasons. That is of course correct. People have found the filters to be audible. They have found other things to not be audible. What else can be shown to be audible? I don't know, of course .... but we can only work with what has been demonstrated, and people have been working on it for decades. I don't really know what else to say, expect that you need to understand digital audio in great detail to be able to know what IS and ISN'T wrong with it .... and where those things will happen. Just because you may have found one cause of audibility Not me. The people who designs and build you DACs and CD players, or you recording studio .... mathematicians / engineers, etc. All I can claim is that I've never been able to contradict a published non-controversial result ..... which should be quite unsurprising. Reiss survey has been looking at audible effects not associated with choice of filtering; that was outside his scope. Sorry, I don't understand. Filters are (essential) in every digital audio device/system. They are a very possible cause of what people heard. Either filters in the recording equipment, or in the playback equipment. It is certainly outside the scope of things he could analyse (as he doesn't have the equipment there to look at) .... he only has the listening tests. He found that some people could "hear a difference" with higher rates. I have absolutely no problem or surprise with that (you seem to be treating me like I am attempting to argue against his study). All I am saying is that people need to be aware of WHY 192khz can and can't deliver a different result..... as it is frequent misunderstood and over-generalised .... especially by people looking to explain WHY they got a result (ie. post-rationalising the fact they hard high-rate sound better .... by assuming the [details of the] cause). Please don't get me wrong, I have some nice 2L purchases I enjoy a lot. They are excellent high rate recordings .... which should be enjoyed directly in that high rate, unless you very very carefully resample them (Aside: I wonder how carefully resampled the version on the 2L site are?) One interesting comment from 2L is they contain less than 16bits range (unsurprising) ..... which should perplex people making related claims about "24bits offering higher quality" (same, complex details exist under that rock - but same result, it's not strictly necessary) If the world (well... Sony/Phillips) weren't struggling to get 74 mins on a CD way back when we'd probably have more bits to play with, not the minimum known to meet the performance envelope defining known psychoacoustic theory. In the AES journal "CD Story" .... says that if it wasn't for the byte (8 bits) arrangement on the computer tapes which were used .... then we would probably have either 13 or 14 bit digital audio system. 8 was obviously not enough, and 13 or 14 was seen as an inefficient use (why not just have 16). That would had of been a shame, as it would probably not have been enough (or at least restrictive / challenging) .... but surely well dithered and shaped 16bit is enough?! (insane situations aside). Then why would the Audio Engineering Society commission an extensive literature survey to baseline what is the state of the knowledge regarding high resolution audio? Shouldn't those professional engineers just open up their textbook to find the answers? None of this adds up, so someone has to be wrong. It is in the blurb of the paper. " to assess the ability of test subjects to perceive a difference between high resolution and standard audio" The reason he states the "causes are unknown", is because his research did not investigate the causes, or reasons why .... and so he cannot say based on this study (there are many well understood reasons why, of course). None of this adds up, so someone has to be wrong. There is no binary answer to such a complex issue. The only simple answers are: Audiophiles should strongly consider playing audio at the rate it is provided to them in.... but they should not assume anything about the quality of the audio based on the rate. Edited July 8, 2016 by davewantsmoore
davewantsmoore Posted July 8, 2016 Posted July 8, 2016 On the one hand you guys say this is all common knowledge found in sampling textbooks. Yet, then in other posts you say it is really difficult to do sampling and filtering well, that is why redbook audio don't sound as good as hi res. That's right. The reason is that the filters called for in digital audio can be difficult to implement. Most are implemented with quality which could be improved. For example. When your DAC converts the incoming digital audio the new (internal rate) it resamples the audio. The filters it uses to do this have limited computing power available, and they must also work in real-time. This means your DAC is likely harming the quality of the audio. This isn't a problem with sampling theory, or not a high enough rate, etc. etc..... it is a problem with the DAC. However when your DAC has incoming data which is a higher rate .... it may use different filters, that are easier to implement ... and this may mean higher quality. It all depends.
davewantsmoore Posted July 8, 2016 Posted July 8, 2016 (edited) why don't the engineers in this industry just say - lets go back to the textbook and do sampling and filtering 'right' for redbook? They DO (but there are finite budgets - ie. you can't fit a computer easily inside a DAC) .... however these errors we're discussing can also be baked into recordings. You've heard of things like hqplayer, xxhighend, MQA, etc.?! Yes. .... that exactly what these people are doing. They are running filters on a computer, that are much higher quality than the filters inside your player or DAC.... and subsequently trying to either avoid the filtering inside a DAC, or compensate their own filters to when added to the DAC filters, the result is better. Edited July 8, 2016 by davewantsmoore
davewantsmoore Posted July 8, 2016 Posted July 8, 2016 (edited) Reluctant to agree here. Adaptive filtering isn't new (maybe for audio but it's not new) and psychoacoustics hasn't changed. We haven't discovered we can hear higher frequencies or time misalignment/phase shift not explicable by Redbook. Filtering is better than it was in the 80's, and there is no reason CDs can't adequately replicate realism if done properly. It goes deeper than that. I understand MQA is moving in some of that direction. But this is 2016, and CD was introduced in the 80s. MQA really aren't doing anything especially "new".... MQA are just bringing it to market (which is a BIG task) High quality filtering is part of that, but really it is deeper. Control over the filters used to re-sample audio that leaves the studio (to make it appropriate for size distribute, and format for backwards compatibility) .... allows the errors in the filter to be "undone" later (if you control the filters used later) Encrypting the content so consumers will know if it has been changed from what the studio endorsed Sending a rate and format to the DAC chip, which bypasses (or compensates for) the internal, (potentially low quality), filters in the DAC Edited July 8, 2016 by davewantsmoore
Guest rmpfyf Posted July 8, 2016 Posted July 8, 2016 In the AES journal "CD Story" .... says that if it wasn't for the byte (8 bits) arrangement on the computer tapes which were used .... then we would probably have either 13 or 14 bit digital audio system. 8 was obviously not enough, and 13 or 14 was seen as an inefficient use (why not just have 16). That would had of been a shame, as it would probably not have been enough (or at least restrictive / challenging) .... but surely well dithered and shaped 16bit is enough?! (insane situations aside). Well... thanks to Sony's epic whinge on this topic... we (thankfully) got 16 bits. But that anecdote highlights the original point - CD was designed to produce what's required, broadly insensitive to a fully-digital signal path, and if you treat it right then you're fine. In an age of bitstream DACs, hard-coded and cheap DSPs, digital volume controls and the like.... Redbook is more challenging than it was intended for. This means nada regards its ultimate ability to impart a faithful recording.
Newman Posted July 8, 2016 Posted July 8, 2016 ...why don't the engineers in this industry just say - lets go back to the textbook and do sampling and filtering 'right' for redbook? They DO (but there are finite budgets - ie. you can't fit a computer easily inside a DAC) .... however these errors we're discussing can also be baked into recordings. You've heard of things like hqplayer, xxhighend, MQA, etc.?! Yes. .... that exactly what these people are doing. They are running filters on a computer, that are much higher quality than the filters inside your player or DAC.... and subsequently trying to either avoid the filtering inside a DAC, or compensate their own filters to when added to the DAC filters, the result is better. HDCD Regarding the bit in bold, I have a small, low-end Wadia 'DAC' (they call their DACs decoding computers) explained in the manual thusly: "At Wadia we refer to our digital-to-analog converters as Decoding Computers. This is because in function our DAC’s work in much the same way as a personal computer. We write software programs for our DAC’s that run on powerful multi-purpose processors to accomplish a task (in this case decoding and perfecting digital audio waveforms). Digital audio data received from a source, is buffered in memory, and then processed via our software and circuit designs, much in much the same way as a conventional computer." Turns out this includes proprietary digital filtering using an 8256-gate FGPA that one reviewer cutely described as a digital carwash.
Volunteer sir sanders zingmore Posted July 8, 2016 Volunteer Posted July 8, 2016 HDCD Regarding the bit in bold, I have a small, low-end Wadia 'DAC' (they call their DACs decoding computers) explained in the manual thusly: "At Wadia we refer to our digital-to-analog converters as Decoding Computers. This is because in function our DAC’s work in much the same way as a personal computer. We write software programs for our DAC’s that run on powerful multi-purpose processors to accomplish a task (in this case decoding and perfecting digital audio waveforms). Digital audio data received from a source, is buffered in memory, and then processed via our software and circuit designs, much in much the same way as a conventional computer." Turns out this includes proprietary digital filtering using an 8256-gate FGPA that one reviewer cutely described as a digital carwash. Do they really call their DACs DAC's ?
Newman Posted July 8, 2016 Posted July 8, 2016 (edited) Do they really call their DACs DAC's ?No, they call them Decoding Computers. But yes, I asked myself that same question when I was cutting and pasting the above straight from the manual. I believe that the apostrophe is allowable, but not recommended. Edited July 8, 2016 by Newman 1
JSmith Posted July 8, 2016 Posted July 8, 2016 I really don't see what the harm is in using very high rates (aside from data rates being too high for via the internet) Unintended inter-modulation distortion... 44.1 isn't too low... 60 is better... there really is no need for anything above 96. Once this level is passed there is too much possibility of unintended distortion. JSmith
davewantsmoore Posted July 8, 2016 Posted July 8, 2016 (edited) Unintended inter-modulation distortion... Potentially an issue.... although not really a massive reason against high rates (and can be sorted out in the electronics if an issue)... xiph really shouldn't have bothered with it IMHO, and just stuck to explaining why high rates weren't inherently necessary in general. 60 is better... there really is no need for anything above 96. See, while I know what you mean.... We are just back at generalisations. Enough for what?.... the assumption that such or other rate will generally do X, doesn't hold .... which causes all the "debates" where people are trying to argue for either "Yes", or "No". The generalisation people need to take .... is that high rate doesn't mean high quality .... it means: The audio might be less damaged (as it has been resampled less) The audio might work better on your DAC (as your DAC performs better with high rates) See that neither of these things mean you can't have high quality with low rates. Edited July 8, 2016 by davewantsmoore
davewantsmoore Posted July 8, 2016 Posted July 8, 2016 Regarding the bit in bold, I have a small, low-end Wadia 'DAC' (they call their DACs decoding computers) explained in the manual thusly: "At Wadia we refer to our digital-to-analog converters as Decoding Computers. This is because in function our DAC’s work in much the same way as a personal computer. We write software programs for our DAC’s that run on powerful multi-purpose processors to accomplish a task (in this case decoding and perfecting digital audio waveforms). Digital audio data received from a source, is buffered in memory, and then processed via our software and circuit designs, much in much the same way as a conventional computer." Turns out this includes proprietary digital filtering using an 8256-gate FGPA that one reviewer cutely described as a digital carwash. Indeed. Wadia were some of the first people to run filters like that. These days many high end DACs do it ..... some of the most extreme versions (example: xxhighend) use a really powerful computer (like a Intel Corei7) connected to a custom built filter-less DAC where they run all the resampling and filtering in the computer. (Not direct to you Newman, just in general) As mentioned before, it takes a lot of words to start going into the how and why for why this is desirable ..... and as mentioned somewhere, a good text book is the essential place to start - otherwise the reasoning is hard to follow.
Newman Posted July 8, 2016 Posted July 8, 2016 Happy not to have all those words here. BTW every Wadia product has done it that way, starting with their 1988 model 2000 CD player. Personally I think that, today, the importance of filters is overstated. Demonstrably audible with specific signals and specific filter poles? Sure, if they say so. But today?... using commercial music products in the same player?... taking the high-res product and converting down to 16/44 in the studio?... after the original high-res product has been processed and cleaned up as part of its routine production process?... then played back through the modern player's sigma-delta DAC with massive oversampling and simple filter? No, I still think we are obsessing at the wrong end of the playback chain. Still mistakenly driven by a 'garbage in garbage out' principle that puts the front end on a pedestal that made sense when front ends were dramatically error-prone, but ceases to be a useful principle when front ends are either inaudible or near-as-drat inaudible.
Recommended Posts