|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
50.183.218.111
In Reply to: RE: "goal is bit-perfect, high-res, analogue, output from the PC" - whose goal is that? posted by Storris on September 12, 2014 at 13:07:14
"Just want to highlight that 24/666 is equivalent to 16Mbps exactly.
And you want 2-channels for stereo, right? Make that 32Mbps.
CD quality is 16/44.1 which is 705.6Kbps, x2 = 1.4Mbps
24/192 'hi-res'has a bit-rate of 4.608Kbps, x2 = 9.2Mbps
Given the state of electronics today do you think you will hear the difference between 24/192 vs 24/666? And do you want more than a couple albums at that resolution to fit on your smartphone? That's about 1GB per 4-minute tune. ;-)
Follow Ups:
Just as the difference between 1080p and 360p video is obvious, Hi-Res audio should be just as perceptible (roughly a factor of 3, the same as 24/192-24/666). I obviously can't speak definitively, but human physiology says it's a certainty. I guess we'll just have to wait to find out.
I'm not sure I want or need that level of resolution on my smartphone, the primary concern is the PC where I do the majority of my listening... but I wouldn't turn it down.
And fortunately, SanDisk has just released a 512GB SD Card with 95MB/s transfer rate. That's plenty of room for several days typical mobile listening. As I said at the top storage capacity is not an issue, on any platform.
http://www.sandisk.com/products/memory-cards/sd/extremepro-sdxc-sdhc-uhs-3/?capacity=512GB
"Just as the difference between 1080p and 360p video is obvious, Hi-Res audio should be just as perceptible (roughly a factor of 3, the same as 24/192-24/666)."
In this case the base level format had output that was nowhere near the limitations of the human eye. 1080p was much closer and so looked better. When you get to the so called "retina display" levels of resolution, you're pretty much up against the limitations of the human eye. Tripling the resolution of a retina display is not going to give you a three times better picture.
You'll find the core dispute in this forum is between those who argue that human hearing has some limitations to it and therefore at some point further refinement to the audio signal is inaudible while others argue that the limits of human hearing are still unknown and that every change to a system, however subtle, may be audible. The first group argues that for changes to be audible they should also be measurable, the second group argues that the ear can hear changes that are not measurable and therefore hearing trumps measurement. The first group thinks that "Redbook CD" is pretty darn close to the limitations of human hearing and so is "transparent," while the second group thinks Redbook is not close enough to the unknown limits of human hearing to be considered transparent, therefore they want even more elaborate formats and greater signal density, so called "Hi-Rez" music. The first group thinks this is a golden age of audio because even modestly priced equipment can be "transparent," while the second group also thinks this is a golden age of audio because however pricey and fancy the piece of equipment under consideration it can always be further improved by additional tweaking.
JE
"The first group thinks that "Redbook CD" is pretty darn close to the limitations of human hearing and so is "transparent"
The better recording engineers and subjective audiophiles have known for a long time that high resolution digital audio is needed to transparently reproduce what humans can hear. These people have simply used their God given senses to note the obvious. For a long time a different gang of pseudo scientific hobbyists, calling themselves "objectivists", have conducted experiments that they mistakenly claimed showed that the "subjectivists" were delusional and that nothing better than 44/16 PCM was necessary for transparent reproduction of music.
Recently, the ground on which the objectivists stood has been washed down the river to the sea. The linked thread (and some threads linked off this thread) discuss how this was done.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Hi Tony!
Golly! Do I have to read that entire 1,000+ post thread?
"Recently, the ground on which the objectivists stood has been washed down the river to the sea."
I'm not sure I'd be willing to declare that just yet. I've only read a few pages of your link so far. First of all I'd like someone to check the provenance of the "smoking gun" files to make sure there wasn't an issue in preparing them. Secondly, are we testing the transparency of 16bit, 44k files or the audibility of the conversion of files to that format? Finally, if the test is legit and the results correct, then that would be insanely cool! Could these results be expanded to more general musical files than some homemade files of jingling keys?
All the best and looking forward to reading more of that thread!
JE
You have to read all three threads. More like a total of 4000+ posts! You will see all of your questions, and more, discussed and debated ad nauseum. There are musical files too, from AIX records.Plus you have to download the audio selections and listen to them. To understand some of the discussions it is also useful to run the ABX software and see if you can hear differences or artifacts from the test software itself.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Edits: 09/13/14
Something tells me it's going to be neither.
Hi Tony!
"You have to read all three threads. More like a total of 4000+ posts!"
Yeow! You're going to have to give me some time for that.
JE
For full appreciation it will be also necessary to download test software and the test files and listen to them. You can have an extension. :-)
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
There seems to be passionate support for both 'sides' of this discussion. However, I'm not prepared to accept the word of one unknown individual over my own experiences on any subject, not even when his findings agree with my own. And given that we have the tools to do our own digging, and that the results of such digging will be unique for each of us, I don't see that we should even contemplate it.
There's an "I owe an apology to BIS" thread in the HiRez Asylum, after a few public trashings of a new piece of equipment the OP had his hearing tested and discovered that his ears were at fault. For many people there is going to be no benefit gained from Higher-res audio, either due to physical impairment of unsuitable equipment.
I have an image of speakers topped with various colours of lights that represent the frequencies a listener can't hear.
The necessary files and tools are easy to download and run. Several people have reported positive results, e.g. ability to pass blind tests 20 out of 20 times repeatedly, except for an occasional time when a nearby dog barked.
Regardless, the claim that no one had ever passed such a blind test no longer stands. Of course the deniers remain unconvinced. But this just proves that these "objectivists" were not and are not objective. Reading all the threads all the way through says a lot about the personalities.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Sorry, I'm just not sold on the validity of self-administered blind tests....besides which, I don't think blind tests of any sort are necessary.
The tone generator method I described earlier will give a much more accurate result of what resolution the listener or his equipment can achieve, while fidelity can be measured via spectrograph.
This might all spoil what is supposed to be an enjoyable past-time though, so I'm not recommending this level of analysis for anyone who enjoys listening more than fiddling.
Edits: 09/13/14 09/13/14
There's an important difference between ppi and resolution when it comes to video/images. High PPI retina-type displays simply allow higher definition on smaller screens, rather than being a measure of High Definition per se.
Whether or not one can hear the difference between DSD64/24@192 and DSD384/24@666 should just require the software to produce the required frequency for the required time.
That software would allow each of us to determine the Highest-Res distinguishable on an individual basis with an objectively definitive result, provided the use of equipment capable of translating the analogue signal of course.
And I'm almost certain that I can find that software on the interwebs. I'll be back with results after MotoGP.
Hi Storris!
I hate to post this link because of the outrage it often engenders, but you may find it interesting.
JE
JE,Would it be fair to say that recording in DSD, or at the highest possible sample rate, would produce an orders of magnitude more accurate digital master than the alternative, and that (given the correct equipment) D/A conversion of such recordings will produce a much more accurate analogue output?
Presuming you concur...
Any analogue signal is continuous, but the digital signal from which it was made is not, leaving gaps between samples which should be observable on a spectrograph with sufficient resolution. At any sample rate which is below the speed of analogue, these gaps will be transferred to the analogue as silences (or noise/whatever).
Because the analogue signal is travelling at something approaching the speed of light and because of the listener's auditory capability/limitations of his equipment, none of these gaps will be heard by the listener, unless transferred as noise and left unfiltered (if transferred as noise, one would presume that they would be of a constant nature and therefore identifiable and susceptible to filtering, leaving the gaps as silence).
What this means is that the speaker is playing a broken signal. Where there is a silence, the speaker merely waits an imperceivable amount of time (0.000021secs for 48kHz)[¹] for the next bit of playable signal to arrive.
However, where there is a more accurate analog signal, the gap will be much shorter and will be followed by a signal that would not have been reproduced with a lower sample rate.
Similarly, a RedBook CD pressed from a High-Res master will contain different information to a RedBook Cd made from a Lower-Res master of the same track.
There will be audible and measurable differences between higher and lower-res music, even if you aren't actually listening/playing back at a higher resolution.
Note 1:
Whether this is actually an imperceptible silence should easily be tested with the correct software, introducing the same gap into a constant tone. My calculations a few posts up suggest that 0.00005 is the shortest perceivable time for a 20kHz signal, and 0.001 for a 1kHz signal, the inverse should also hold with the equation 1/xHz providing the minimum audible time for a given frequency. i.e. 0.000021secs of silence is less than half of what is needed to produce an audible change at 20kHz, and is thousands of times shorter than can be perceived at 1kHz.
Edits: 09/14/14 09/14/14 09/14/14 09/14/14 09/14/14 09/14/14
Hey Storris!
"Would it be fair to say that recording in DSD, or at the highest possible sample rate, would produce an orders of magnitude more accurate digital master than the alternative, and that (given the correct equipment) D/A conversion of such recordings will produce a much more accurate analogue output?"
Actually, I don't concur with that. Firstly, I know nothing about how DSD works, so I can't offer any opinion about that. I know next to nothing about how PCM works, but what little I do know makes me think that increasing the frequency of the sampling rate does not "move the samples closer together." Instead it adds higher frequency data to the signal. You already picked up the information between 20Hz to 20kHz when you recorded at 44kHz. Recording at 192kHz is only going to add in high frequency harmonics to the original 20Hz to 20kHz signal. Similarly, increasing the bit depth does not make the samples smoother, instead it lowers the noise floor.
I'm sure others will be happy to leap in and correct me if I'm wrong!
Further, what comes out of a DAC is analog. Any digital "gaps," (which are there because of the nature of digital but which don't matter because in a bandwidth limited system there is only one, unique analog waveform that correctly corresponds to the digital information) have been filtered out in the conversion to analog. It is an analog signal passing through your amplifier and from there to your speakers. The signal is not "broken" into discrete pieces.
Even if it was, your speakers are not quantum devices. They do not move from one state to another with no intervening states. Even if you could build a quantum speaker, they would have to push against ordinary air which would thereafter propagate an analog wave.
By the way, another link to that guy who is roundly despised here.
JE
I'll go and read up some more on exactly what gets sampled and when. I'm basing this on the premise that everything gets sampled, and increasing sample rate also increase sample range.
I appreciate the analogue is a single continuous entity but the breaks do exist, in the digital at least. If the DAC is knitting the seams together, or filling the intervening space with algorithmically estimated data, the result is an analogue signal that could be improved, depending of course on what gets sampled when. I'm off to do some more reading.
Regards
Hey Storris!
Remember, we're dealing with bandwidth limited signals. That's one of the keys to digital: we're not dealing with infinite bandwidth, but only a limited set. The theory does sound crazy, but you don't even think about it when you hear voices coming out of your telephone or watch a TV show.
JE
The theory is that the signals are bandwidth limited. Theory is only an approximation to reality because a finite duration signal can only be approximately band limited. For a system to work well in practice there must be a significant guard band between the audio signals and the theoretical bandwidth determined by half the sampling rate.
Also, just to set things straight from another post in this thread, if one has a sine wave it is not sufficient to sample it twice per cycle. One must sample it more than twice. (To see this, consider that the two samples might just have hit the positive and negative zero crossings of the sine wave.) If there are only slightly more than two samples, then the filter has to be very sharp, e.g. "ring" for a very long time, to figure out what the original sine wave was.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Thanks for the links by the way, very informative stuff...I get what's being said, and think it sounds perfectly sane when applying it to a constant tone, but not with music. As far as I can tell, there is no physical way that 48,000 images of a thing can give you a picture as complete or accurate as 3,000,000 when that thing is changing at an infinitesimal level, constantly. Details will be missed.
This is true for video and still images in digital and analogue, even eyesight, so how not for A/D/A sound?
Anyway, I've put some questions up on his Wiki.
EDIT
Unless the Bit Depth is involved? 16Bits/MoreFrequencies = MoreFrequencies per Bit = Less detail than 16Bits/FewerFrequencies. But Bit Depth is Dynamic range...?
Also,
In response to other questions on the video's wiki page, he or someone at least, has written this;
"Does this mean that we should have better output if we increasing the sampling from 44.1KHz to 192KHz?"
"You'll have exactly the same 20 kHz sine wave at 192 kHz as 44.1 kHz... Only two sample points are needed to perfectly recreate a sine wave... --Leorex"
"It's counterintuitive, but try and think of it like this. You know the input signal (analog) is band-passed to 20kHz, so there are no frequencies higher than 20kHz to be reconstructed. Now look at the 2.2 samples per period; try and draw a continuous line through all the samples without using any frequencies above 20kHz. So in fact, there is only *1* solution for the line you draw through the sampling points. You can 100% recreate the original analog signal from the sampling points. You will not get better output by increasing the sampling rate to 192kHz, because you have already reconstructed 100% of the signal. -- Nhand42"
To my mind these answers only suffice for constant frequency/amplitude signals. When you add a constantly changing frequency/amplitude signal, the ability to "draw a continuous line through all the samples" will become much more difficult. We are not trying to recreate a Sine Wave. Try illustrating an orchestral piece with his lollipop diagram.
Edits: 09/15/14 09/15/14 09/15/14 09/15/14 09/15/14 09/15/14 09/15/14
Sorris:
"I get what's being said, and think it sounds perfectly sane"
No, if you got what was being said you would not think it sounds perfectly sane.
"I get what's being said, and think it sounds perfectly sane when applying it to a constant tone, but not with music. As far as I can tell, there is no physical way that 48,000 images of a thing can give you a picture as complete or accurate as 3,000,000 when that thing is changing at an infinitesimal level, constantly. Details will be missed."
You think this because you have no understanding of how digital audio works. Perhaps this will help.
Way back when, people wanted to record analog events around them (people speaking, or making music}. They only had analog tools so they used them. These analog tools created a one-dimensional analog wave form that people could fiddle with and send over the air, or impress onto vinyl, or use to magnetize iron paritcles glued onto tape. It didn't really matter. Those formats were all used to create analog waveforms that were then amplified by amplifiers and used to drive speakers.
The crucial point to remember is that a signal consisting of a one dimensional, varying voltage, was used to capture and to simulate the sound of music or other noises.
Digital Audio is NOT a description of reality. Digital Audio is a description of the one dimensional, varying voltage that, until now, has been used to simulate sound and music.
So long as you keep trying to use "Digital Audio" to describe "reality" you are going to go insane. Only once you get the idea that "Digital Audio" is only useful to describe "recorded audio" will you get to be comfortable here.
All the Best!
JE
"varying voltage"
Two words that have fixed everything, cheers.
You just needed a new perspective. Just as there are no "quantum speakers" neither are there any "quantum microphones." When you think about it, is there any part of recorded audio, digital or analog, that can't be attributed to a voltage variation?
JE.
"I hate to post this link because of the outrage it often engenders"
No outrage here. I think the problem is at worst a matter of willful ignorance. I think the author is a fool whose bogus world view has recently been shown to be inconsistent. If he is a person of integrity he faces a rude awakening. I feel sorrow, not outrage, in the face of ignorance.
Now if you want to talk about the prices of hires downloads, that might be different. There is no justification for a huge premium on high res versions of new releases as there are insignificant extra costs involved.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Tony:
"I think the author is a fool.."
Your concept of "fool" must vary widely from mine, and must encompass a lot more people.
"Now if you want to talk about the prices of hires downloads, that might be different. There is no justification for a huge premium on high res versions of new releases as there are insignificant extra costs involved."
Again our opinions differ. I see prices as market place signals. I don't think prices have anything to do with the underlying costs of the product.
JE
I understand. As you undoubtedly know, the economic term is price discrimination. For a few years I was responsible for a line of computer peripherals that interconnected mini computers to phone lines. We took one product and through a minor configuration change turned it into two products with almost a 2x price difference. That way we could sell a cheap large version to universities and an expensive small version to Ma Bell. That way I could keep the people who were marketing to universities and the people marketing to the phone company both happy with a "reasonable price". I thought this was a great idea, after all it paid my salary. :=)Blue Coast Records has a huge price premium on their DSD files, but for the past year they have been putting all formats on a single low introductory price for a few weeks. Eclassical.com does something similar with their new releases and there is only a small premium on their "daily specials".
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Edits: 09/15/14
No, not at all.
The same way, I didn't fell "outrage", when some years ago I was passing by an institution in Philly, called something like "Institute for Mental Health and Mental Retardation", and seeing some of the clientele wandering around and sitting on the stairs. Anyone who's been to Philly knows that half of the population there could potentially use the services of that institution, but those who (forced to) actually do - they are "crème de la crème", so to speak.
Did I try not to look? Sure - not too successful, though. That's why I read that article, too.
Thanks for the link. I've been thinking about the enormous range of frequencies available through 24/xxx myself.I don't see the benefit of recording them, storing them or trying to reproduce them at the speaker.
Any harmonic effects in the studio have already had their effect on the audible range, and that was recorded. Recording the high frequencies and reproducing them through the speaker;
allows them to have a second harmonic effect (this is equivalent to 'feedback' isn't it?),
makes it more difficult to produce the audible range,
and disrupts the audible range's ability to produce its own natural harmonics in the listening room.-------------------
This being the case, for most of us sample rates above 32kHz are pointless, and as for DSD... So, a new objective...16Mbps from a 32kHz sample rate means a 500-Bit Depth!
-------------------Reconsidered...
Ok, so 16Mbps is the maximum amount of auditory processing, listening, I can do at any given time. Meaning that if 16Mbps of music (i.e. a constant 16kHz tone at highest perceivable volume) was presented to my ears, I would be incapable of responding to any other sound, or my brain might explode, maybe. 16Mbps is not desirable.
Any sample rate above 44.1kHz is wasted, in every sense.
I've only got 16bits, but they provide 65,000 points of resolution.
Ok, JE, I think I've got most of it. A couple of things I'm not clear on, such as how a given sound is digitised, say a 1 sec beat and decay of a C# bass drum. But I guess that would apply to all A/D signals (FM, TV etc...) and isn't strictly relevant.
Edits: 09/13/14 09/13/14 09/13/14 09/13/14 09/13/14 09/13/14 09/13/14 09/13/14
I tried NCH Tone Generator but it is pretty limited, only allowing to test maximum @ 1Khz due to inability to adjust tone length below 1ms, or to adjust sample rate above 192Khz. Still works good for what it does do.
This looks more promising for high-end systems - http://www.esseraudio.com/en/test-tone-generator-windows-software-generate-test-signal-sine-pink-noise-crest-factor/ttg-dl-en.html - It's heavy overkill for me, but with sample rates up to 24/384, it should be perfect for testing resolution and tuning the highest-end set ups.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: