davewantsmoore Posted July 10, 2016 Posted July 10, 2016 A good way to anecdotally see, this is to look at what MQA is doing. Start with the least molested version of the audio (so it hasn't been harmed by someone else) Convert this audio to redbook in a way which doesn't harm it (or improves it if possible) Distribute it Convert this audio to a playback rate. (A DAC normally does this poorly) .... so use the MQA decoder to do this instead of the DAC where possible. The temporal resolution can survive "redbook". Again, this isn't my "opinion". It's just a fact. I can show you that it does.
LHC Posted July 10, 2016 Posted July 10, 2016 Hi @@davewantsmoore I will repeat what I wrote earlier. I do have to say I truly and personally admire your drive, ingenuity and initiative. I wish we have more people like you in our country (in a general sense). Now it is painfully obvious that you and I just interpret written information differently. Nothing wrong with that of course. But it means there would be impasse where we can't agree. I am fine with that. I also treat what you write with respect regardless. So I think I will just leave it at that.
Guest rmpfyf Posted July 10, 2016 Posted July 10, 2016 I have posted this video before. Here is Bob Stuart talking about how MQA works. Sure it involves new filtering, but as I wrote before it is the objective that is important. The objective has to be about achieving a temporal resolution not achievable by redbook. Filtering is a means to and end, it is the end that is important. I have posted earlier a blog post with Bob explaining what he means by 'temporal resolution', and it was consistent with Kunchur's. That pres is 50% an attempt to make things simple and around 20% BS. Any statement that talks about the temporal resolution of two adjacent sounds needing hires is full of it and treading on 50+ years of psychoacoustics and mountains of well-proven data. Two sounds that need hires to discern between them (not absolute positioning or channel delay) are a higher frequency than you can hear. Period. (Kunchur's conclusions were stuffed here). The brickwall filter itself does not smear anything. It simply rejects anything above the pass frequency. You can do this in sampling all day long. It's simple - you sample, you convert to the frequency domain (where the challenges start), you cut off every component above a higher frequency. Now: This works 100% when what you're cutting out discrete high frequencies ('compound of sine waves' argument). This is challenging when you're cutting out spectral components that shape the waveform of lower frequency stuff you can hear. This is challenging when the filters applied to creating the original spectra are poorly focused - because what the brickwall takes out in spectral content mightn't have been a best spectral repreentation of what you had. You may have what's called spectral leakage outside the target window (what the image in your link suggests). In a frequency domain context, some of the spectral energy has 'leaked' to other frequencies. When you flip spectral content back to the time domain and there's leakage, the recreated waveform is made of more content from spurious frequencies than it should have. The digi-**** term for this is 'smearing', though nothing is actually 'smeared'. it's just reconstructed within limits. Those limits depend on the filters employed, which are resolute to the sample frequency. Leakage around any filter is inherently smaller when the filter is higher and the resolution finer. This has nothing to do with hearing two noises at a given temporal resolution. Relevant filtering is characterized by both amplitude and phase distortion ('how loud stuff is' and 'how late stuff is' relevant to the original source). Short version - the main advantage of hires isn't temporal resolution, it's the spectral resolution of the filters employed. Anything that attempts to use adaptive filtering (e.g. assess the next bit of the source material, picks a best filter, uses it accordingly) is a directionally-correct thing, though this is fraught with some conjecture as it's a rapid departure from a consistent thing. That's another argument altogether. So's the 'what's music?' argument. I can recreate a waveform (not now because I'm working on a Sunday) at sub-20kHz that Redbook won't be able to recreate faithfully. This doesn't address: Is that practical to music? At least what you listen to? Can you recreate that difference? If you can recreate it, can you hear it? Would suggest, as @@Newman writes, that much of this is down in the flys**t regards relevance. My experiences are as per yours - there is very little a good Redbook system can't do that a good hires system can, and good Redbook systems take work. Much of the difference is in ease of mastering and DAC filters. Where there are limits - extremely busy and abrupt periods of music - there are other practical challenges (ask anyone that's ever mic'd and recorded an orchestra) that make direct comparisons impossible, and present related challenges for analogue formats. It's a net result you're after. Just enjoy it. It would be unobtainium to suggests one has to read many books and rely on years of practical experience. I don't think that's poorly intended (it's at least how I learned). DSP isn't easy, even for engineers. The sheer magnitude of YouTube audio 'experts' is a testament to as much.
LHC Posted July 10, 2016 Posted July 10, 2016 @rmpfyf Thank you for putting in the effort to write a thorough response in post #453 above. Please note that it was appreciated.
davewantsmoore Posted July 10, 2016 Posted July 10, 2016 But statistical testing like Meyer and Moran are publishable. But not usually published (nor undertaken afaiui). However, if you're trotting out Meyer Moran, then I fear you've misunderstood me (and probably them). You seem to be approaching me like I am saying that there is no possible benefit to using higher samplng rates .... and that studies which show audibiliy of high rates must be flawed ..... and studies which show no audibility must be correct. That's not what I think at all. M/M result says people couldn't hear the difference. There's a zillion reasons why this could be. Their result doesn't tell us that higher rates are unnecessary, or not beneficial..... or why that might be. Similarly, studies which show that a difference could be heard. There's known reasons why this could be. Simply recording that people could hear "it".... doesn't tell us what "it" is, or how "it" works. We already know what "it" is. What sampling rates do and don't achieve how/why/and when .... is very well understood. It isn't something which is up for debate, or "opinion". You can see this in any good text book... or in the MQA patent .... however there is not somewhere which is going to spoon feed it. Certainly, how audible various difference between digital audio are .... IS debatable, as obviously "audibility" is both subjective, but also is not a binary (yes/no) answer.... THAT is the question these studies are looking at (becuase those are up for debate). How digital audio works, isn't something I get to have an "opinion" about. It just "is".
davewantsmoore Posted July 10, 2016 Posted July 10, 2016 Any statement that talks about the temporal resolution of two adjacent sounds needing hires is full of it ... and this is the important part of this discussion. "needing hires". What do we mean by "needing hires" .... we need high(er) sampling rate capabilities if we want to: Avoid sample rate conversion in the production and distribution chain Apply filters to audio which attempt to repair audio that has been previously damaged Render audio into a format which doesn't handicap the playback hardware These things are potentially desirable. So we can see that in practice, we do need higher rates (or they are at least a solution to some problems). ... but we can also see from this that nothing can be directly inferred about the quality of audio in a low rate vs a high rate. As a couple of people have said. Not many people care (they just want what sounds good, which is perfectly uderstandable) ...... However for anyone looking to understand "why", this is essential/basic knowledge. 1
davewantsmoore Posted July 10, 2016 Posted July 10, 2016 Hi rmpfyf. I think I will have to agree to disagree with you and Dave at some point to give this a closure. If you are interested in the "why" .... then I think that's a sad outcome. The semantics of what is being discussed is important (in so far as it represents an understanding of how digital audio does/doesn't work).
Guest rmpfyf Posted July 10, 2016 Posted July 10, 2016 @rmpfyf Thank you for putting in the effort to write a thorough response in post #453 above. Please note that it was appreciated. No worries. It's a robust discussion. I like @@davewantsmoore argument - put another way, what should hires sound like? DSP is an amazing thing that allows a digital interface to filtering effects we take foregranted in audio all the time. Room effects, crossover and loudspeaker design, the lot. There's a transfer function for each and all. When people come over sometimes, to make a point, I play them the same passage of audio in two takes. One sounds much clearer and time-resolute than the other. They assume it's hi-res vs Redbook. They're actually both Redbook, I'm simply running DRC on one attempt. And it's an indirectly analogous thing to suggest, as there's a phenomena going on in an uncorrected room that is much akin to spectral leakage or 'time smearing' or whatever the 'net snobs are calling it this week. Point is... what's it supposed to sound like?
davewantsmoore Posted July 11, 2016 Posted July 11, 2016 Thank you for putting in the effort to write a thorough response in post #453 above. Please note that it was appreciated. The dot points in 453 are very similar to what I've been attempting to say in layperson speak. Ric is a professional, however.\ There are really no "sources" for them. As mentioned, aside from a good text book.
JSmith Posted July 11, 2016 Posted July 11, 2016 Point is... what's it supposed to sound like? 42... JSmith
Guest rmpfyf Posted July 11, 2016 Posted July 11, 2016 The dot points in 453 are very similar to what I've been attempting to say in layperson speak. Correct.
Recommended Posts