|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
76.103.138.79
This is quite a long article but the main idea is captured in this sentence:
"Unfortunately, there is no point to distributing music in 24-bit/192kHz format. Its playback fidelity is slightly inferior to 16/44.1 or 16/48, and it takes up 6 times the space."
http://people.xiph.org/~xiphmont/demo/neil-young.html
A discussion of the article goes on here:
http://news.ycombinator.com/item?id=3668310
Follow Ups:
.
QUick skim - looks like a great audio article.
I have found that 16 bit, 44.1kHz audio can sound really good - but 24bit/88.1-96kHz is consistently better sounding. I have one or two 24bits/44.1kHz tracks that sound great, too. I have a few 176 and 192kHz tracks - and haven't found them sounding any worse than the 24/96 I have of the same track. So while this guy is throwing out a lot of good theory, it appears, I think he, like lots of people who trust math more than their ears - might be missing something in their explanation.
I suspect the guy is 100% right about the nature of perfectly reconstructed Redbook sound - and the issue is the playback and reconstruction of the analog signal - with the handling of clocks and digital transports, internal system reflections, etc. etc. etc. I have found enormous improvements in "CD sound" when paying attention to noise, jitter, clocks, etc. This kind of gain and improvement isn't nearly as large (or at all!) with the 24bit/96kHz+ files.
Perhaps handling 16 bit / 44.1 kHz doesn't allow enough error margin to allow for the inevitable small inaccuracies that are bound to happen in consumer playback gear? If so this would be one (of several) cases where something works wonderfully on paper, but it isn't possible for the real world to implement it nearly as good as it is on paper!
After all Vinyl sounds great, and generally isn't nearly the fidelity of the Redbook CD in theory, but in practice it spanks it.
"Knowledge is knowing that a tomato is a fruit. Wisdom is knowing not to put it in a fruit salad"
This has little to do with listening to music and were derived playing sine waves.
Who listens to 50Hz to 20k sine waves? They annoy more than sounding loud.
I've been listening to some 25 Hz sine waves lately and hearing them. :-) And a lot of other tones in the range up to 200 Hz as well. And crawling around on the floor groping and twiddling knobs...
I've been also been listening to some 30 kHz sine waves, but not hearing them, as expected. Not really hearing anything above 14 kHz these days, 50 years ago it was up to 21 kHz. :-(
Also, using gain settings common for large scale orchestral music I've been listening to some 440 Hz sine waves at low level (-97 dBfs) and hearing them. These were audible in both 16 and 24 bit formats, but dither was required with the 16 bit format to hear the tones. I have too much noise in the room to hear quieter tones unless I turn the gain up to unreasonably loud levels (louder than row A concert levels).
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Good for you, but sine waves are annoying to listen to, as opposed to music, which should be pleasant.
You will find, for example, that community noise criteria are not based on Fletcher Munson but on Leq.
I find sine waves pleasant and musical. I used to listen to Morse code at around 800Hz and it was beautiful!
You want annoying? Try listening to square waves in the audible range, or even just beyond the audible range where you sense it but don't really hear it. I use loud ultrasonic square waves to ward off pests and barking dogs.
Ah, so you listen with pleasure to function gnerators. I suggest an HP for your next purchase.
If you are serious, this says something about your choice of hardware and software. If you are not, then join TL on the floor to hunt for resonances.
Great article and a great read, thanks. But in the end almost all people prefer 24 bit music.
If I were to produce a new digitized audio playback medium, I'd say the ideal bit depth would be 18 bits, and the ideal sample rate would be 60 kHz.
This is a fairly well-written article, but in my opinion, is far too equivocal in it's statements.
The writer assumes a completely-understood (and extremely simplistic) model of the hearing system.
I thought this debate was done a long time ago. For digital playback (not mastering, which needs more), I saw argued that 18-bits (properly dithered) is the theoretical limit of audibility (and I'm sure less, if the volume is low). For frequency domain, all bets are off, in my opinion, because I don't think we have an adequate model of human hearing yet, nor a definitive design for the ideal digital reconstruction filter.
If the article is correct, then upsampled redbook is superior to the original waveform (because our amps and speakers would just mess up the original wave with IM distortion). Despite a lot of other good information in the article, this is over-reaching.
You don't need a model of human hearing to determine whether, or not, the presence of ultrasonics matter in the replay of music.
And the ideal reconstructor has been known for over a century. We can even make pretty decent real-life implementations of it.
Thorin sits down and starts singing about gold.
"You don't need a model of human hearing to determine whether, or not, the presence of ultrasonics matter in the replay of music."
A model and $2.50 will get you a cup of coffee. You need experimental data or you just need to use your own ears. Or you can be a skeptic and ignore the issue.
"And the ideal reconstructor has been known for over a century. We can even make pretty decent real-life implementations of it."
The ideal reconstructor does not work with the data encoded on most CDs, and that's because the recordings were not made in accordance with a correct version of the sampling theorem. Most recordings have energy immediately below and at Fs/2 and this implies that they also contain non-harmonic aliases in this region due to a poorly designed converter. To remove the aliases and preringing requires a different filter.
My experience remains that it is difficult to fully capture the music on a good cassette tape in the 44/16 format. Results are much more satisfactory in higher resolution, with the majority of benefits coming from going to 48/16 or 44/24. It is very disappointing to me when a recording that I have made at 88/24 that sounds excellent has to be fit into the 44/16 format. More times than not the results are less than fully satisfactory, despite the humble source material. The benefits of higher sample rates may be small, but they are the same size as they were in 1983. The cost of the extra bits have fallen hundreds of times in the past 30 years. The extra cost for bandwidth needed to download an album in 88/24 vs. 44/16 is less than 10 cents based on low volume Amazon cloud pricing of $0.12 / gigabyte. If one wishes to put out a quality product, it does not make sense to skimp on the manufacturing and distribution when the costs of higher quality are still much less than the cost of credit card processing fees.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
> > You don't need a model of human hearing to determine whether, or not, the presence of ultrasonics matter in the replay of music.
> > And the ideal reconstructor has been known for over a century. We can even make pretty decent real-life implementations of it.
It's nice to know that simple mathematics is all that is required for perfect audio.
The simplistic model I mention, is that the ear "only" performs a fourier transform on the signal, an cannot respond to phase, or timing of complex transients or high-frequency sounds.
If this model is true, then 44.1 kHz is far too much. We should have 40.001 kHz, with digital oversampling. Of course, now a transient that has 20 kHz information, has that information spread out over a second of time, rather than occurring at the proper time. I guess this makes no difference? After all, only the frequency graph matters ...
High-res music of the same mix sounds different than redbook on the majority of systems. I applaud real researchers who try to find out WHY. If the high-res simply allows electronics to be more forgiving of errors, or shifts jitter to sound more pleasing -- that is still a valid reason for going to high-res. I just want good music. I have heard redbook played well (the Bryston BDP-1/BDA-1 combo is nice). I also realize that mastering plays a big part.
High-end audio has too many superstitions. I theorize this is because music is not about logic, it's about emotion. People want to hear music that engages them. If high-res does that, we make up all sorts of (seemingly) logical reasons. Some of them may be wrong, as science can show. Just don't tell us that high-res is worse sound, because that's false. We don't live in the world of theory - we live in the world of listening.
"High-res music of the same mix sounds different than redbook on the majority of systems. I applaud real researchers who try to find out WHY. "
You state that as a fact. Where then is the evidence in the academic world?
Thorin sits down and starts singing about gold.
Not bad.
But there is an error in it:
"16 bit audio is commonly said to have a dynamic range of 96dB (each bit doubles the range and a doubling is about 6dB so, 6dB*16=96dB). This is incorrect.
The incorrect '96dB' figure ignores the spectral power density of a signal. 16 bit audio can go considerably deeper than 96dB"
The above is in fact not correct.
When discussing dynamic range it is the norm to factor out the noise spectral density, after having integrated ('summed') the noise over the bandwidth of interest, possibly with a weighting function or filter used.
So knowledgeable people understand that a channel with 96dB dynamic range and a uniform noise distribution will indeed have its spectral noise density line much lower than -96dB, and as such this system can resolve spectrally pure signals (like sines) to levels much below -96dB. There is no magic here.
And the reason we humans can perceive such sounds apparently burried deep below the summed noise is that only a fraction of the total noise is located in a bandwidth adjacent enough to the signal of interest so that it can mask it.
Most of the noise is located where it has no masking power over the actual signal.
"If the same transducer reproduces ultrasonics along with audible content, harmonic distortion will shift some of the ultrasonic content down into the audible range as an uncontrolled spray of intermodulation distortion products covering the entire audible spectrum. "
This is not very relevant. The spectral distribution of typical music rolls off towards the treble, so that when the ultrasonic region is reached there is not that much signal left. It won't do any harm in a half-decently engineered sound system.
The second half of the article looks like an ad for H2-audio. Boring.
Thorin sits down and starts singing about gold.
I scanned it, but will have to read it in depth.
Regards,
Geoff
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: