24bit vs 16bit, the myth exploded!

.

Quote:



I made the mistake of trusting them with my wallet.

I think this thread is great, wish there were more like it.

Quote:



Maybe you're the one that's fooling yourself, here. There are professional studio monitors like the Adam and the Yamaha that go up to 35-40kHz. The thing is, you don't have to be able to hear up to 40kHz in order to perceive the information there. And of course, this knowledge is proven through lab equipment that measures the frequencies.

Now practically, in the studio, these headphones and speakers that have an extended frequency response will have more space to the sound and therefore the music will be more, precise, smooth, airy and dynamic with a better details in the image. This is sound resolution and it's the same with digital. You don't need to be able to hear up tp 96 or 192 kHz, but there is information that is stored there. Of course you need to work with a chain of studio equipment that supports high resolution. Eventually when you're going to dither down to 16/44, or just a lousy mp, your mix will be much more rich and this is the benefit of working in high resolution.

Here is more info about dithering http://www.apogeedigital.com/pdf/UV22osquick.pdf

Quote:



Quote:



If there is nothing there to record, if none of the equipment is capable of recording freqs in that range and if none of the playback equipment is capable of reproducing it then, yes I can be sure.

Look, it's quite simple. No microphone can pick up anything much beyond 20kHz, yes some of their specs go to 35 or 40kHz but just look at the response roll-off. If there was anything up there, mics cannot record it. Secondly, yes, there are speakers that will in theory go up to 40kHz but what are they going to reproduce, nothing can be recorded there and you couldn't hear it if it were. Of course 40kHz can (in theory) be recorded using a sample rate of 96kFs/s. So now you want to double this sampling rate to 192kFs/s so your audio limit is now 96kHz. 96kHz is more than double what can either be recorded or that your system can reproduce.

This is not just an opinion, it's very simple fact. No instrument produces any notes beyond about 8kHz, the only thing present anywhere near 20kHz is harmonics. Each subsequent higher harmonic is lower in amplitude than the previous harmonic. This is simple basic acoustics. By the time we get to around 20kHz the harmonics are so quiet that they are starting to disappear below the noise floor. By the time we get to 30kHz the harmonics are already way below the noise floor. Even the finest mics have very little response at 30kHz (let alone able to record something below the noise floor), so there is simply no physical way to record these harmonics. If there is something stored in these ultra-sonic frequencies it can only be system noise generated by electronics in the signal chain.

If people think 192kFs/s sounds better, I'm sorry but there is no sensible alternative to the fact they are fooling themselves. The only possible alternative is that their DAC has some kind of malfunction which so negatively affects the re-construction of signals at 44.1kFs/s and 96kFs/s that 192kFs/s sounds better. It really is just another case of consumers' expectation that more data = better quality.

Just to make absolutely clear, there is nothing produced by any instrument that exists above the noise floor once we get to about 30kHz and there is certainly nothing at 96kHz. Even if there were, it is not possible to record it because it is way beyond the capability of studio mics. Even if something does exist and we could record it, your system could not reproduce it and even if it could, it's well beyond the hearing capabilities of a dog, let alone a human being! Remember also, there is no reliable proof that anyone can hear beyond 22kHz and we are talking here (with 192kFs/s) about extending the range of encodable frequencies from 48kHz to 96kHz!

Lastly, in answer to the last sentence I quoted from Acix: If you decimate the sample rate from 192kFs/s to 44.1kFs/s a brickwall filter has to be applied to completely remove all audio frequencies above 22,050Hz. If all the frequencies above this point are not removed the re-sampling fails! So whatever may or may not be above 22kHz is totally and permanently removed. If there is nothing there, it cannot therefore make the CD sound "rich".

G

PS. For those who are not aware, the sample rate (correctly notated as kFs/s) is always double the highest audio frequency (notated as Hz or kHz). This is a basic tenet of the Nyquist-Shannon Sampling Theory.

what about 'scissors' so to speak?

ie, the argument about speed of light - that nothing can go faster - yet if you combine 2 things that move at the speed of light and they move toward each other, you have a problem with a+b still not being greater than C

so, suppose you have 2 high freq waves that are at slightly diff frequencies. combine them on a scope and look at the places where they overlap - you need higher frequency components needed to be able to reproduce that, don't you?

perhaps restated: which do you think is higher fidelity: recording 2 musicians with 1 mic (combined) or recording 2 musicians each with their own mic and each having their own speaker and amp for playback? I would think that each would be able to capture the single musician easier than 2x the amount of 'wave activity' in the air.

so, it was never the single 20k tone that gives the need for higher capture rates but the fact that when you add more and more 'stuff' its harder and harder for that 20k to reproduce the entire set of sounds at record time.

is this correct or is there a flaw in this logic?

Quote:



That is not strictly true. The pattern is not linear , it tends downwards but some harmonics are higher than their predecessors.


Quote:



That depends on which noise floor you are talking about. With CD the noise floor at -96db is low enough for harmonics at 20khz to be comfortably above the noise floor and in the cymbals crash test I did when the fundamental was at -5db the high harmonics were at ~ -63db at 20khz. Now I cannot hear that and you would need a quiet hall to get them but it is far from implausible and in a quiet listening room no problem. What renders them moot is masking, if you have a signal at -5db and one at -60db the one at -60db is not going to get noticed very much.

Quote:



Um, the Balinese Gamelan has harmonics above the noise floor extending way up beyond human audibility.

Quote:



No, it does not work like that, the composite frequency is never higher than the higher of the two components , the waveform just becomes more complex but Shannon-Nyquest operates for any arbitrarily complex signal not just sine waves.

I'm trying to understand this - trying to visualize it.

the wave gets 'more complex' but isn't that adding more harmonics than were there with just the 2 initial sound sources?

and again, the thought experiment about recording 2 audio sources separately and playing them back at the same time vs combining them and playing back the composite. isn't the 2 going to be more accurate than one?

another analogy: s-video vs composite vs component. if summing was non-destructive, as you say, then why do you get better fidelity with the separation of components, carrying them separately and them combining them at the final destination? composite video is 'a mess' and s-video breaks out Y/C into 2 wires; and component breaks it down even further.

it just seems that the 'divide and conquer' gets you better accuracy in tracking complex waveforms than simply combining them and hoping you have enough 'total bw' to carry them without loss.

Quote:



I was not referring to the the theoretical noise floor of CD but to the actual noise floor when recording or replaying the signal and the actual practicalities of the equipment. IE. Good quality studio condensers (U87, etc), have a frequency cutoff around 20kHz. There are one or two which go up to 40kHz but again their roll off is usually quite severe above 20kHz. It's only usually laboratory mics that capture anything much above 30kHz and these are not suitable for studio use. And I haven't even mentioned mic-pres, EQ units, etc. Most producers and engineers would not want to capture anything above 20kHz anyway. Because these frequencies cannot be heard, engineers can't mix them with any degree of accuracy. So although there are the odd instruments (like Gamelan and some cymbals) which do produce harmonics above the theoretical noise floor of digital audio, they are not above the noise floor of the practical equipment or environment (including Brownian motion) in a recording and playback chain.

Also, I was referring to sample rates of 192kFs/s, which gives us a theoretical audio frequency maximum of 96kHz. So the argument is, what is between 48kHz and 96kHz in a music performance that in practice can either be recorded or replayed? The answer is nothing! When we start to look at what humans are capable of hearing the figures start to get even more silly. 44.1kFs/s can encode pretty much to the limit of human hearing. 96kFs/s more than doubles the human limits and even exceeds the hearing limits of a dog. 192kFs/s goes more than double the hearing limitations of a dog and starts to approach the limits of what a bat can hear.

G

Quote:




Quote:




Quote:





Quote:



No, these are separate issues and I do not disagree with you on the pragmatics, but to say something is below the noise floor you now have to say what the noise floor is. If it is -50db then Gamelan harmonics are in play

Quote:



What is the ball park noise floor for a decent acoustic recording set-up ?. The ~ 20K cymbals harmonics I recorded from a 16/44.1 sample played back on an average CD player and recorded on a low end ADC were 30db above CD noise floor which is a lot , they may be above recording room noise floor ?.

In the Oohashi study the Gamelan had harmonics at 50khz that were at -50db wrt the dominant tone and harmonics at 20K that were much much higher. The Oohashi study is highly flawed but I do not doubt that part.

Quote:



You're so easy to refute -- I like it!



Here's a 96-kHz recording (below) and the same recording downsampled to 44.1 kHz, then upsampled to 96 kHz again (above) -- for comparability. Below you can see signals with a frequency content in the high 40 kHz, whereas the downsampled version is smoothed by the 22-kHz limitation. There are a lot of such signals in this recording («Dragon Boats», linked by Oldschool).


Quote:



Maybe it's not so important to hear it (–> ultrasonics), but rather to have a better preserved signal shape without distortion by the filter in proximity to the audible range (–> ringing). After all I do hear a difference in favor of hi-rez (also in the case at hand).


Quote:



So you're really sure? I'm (almost) about to believe you, since your reference is quite impressive (trained as an orchestral musician at conservertoire and then played professionally with symphony orchestras for some years... own recording studio and probably been analytically listening to orchestral recordings since the late '70s...). But what kills your image is your attitude to display your knowledge and experience in an obtrusive manner. Apart from your absolutisms.


Quote:



And so on.
.

Quote:



It's impossible to say what the noise floor of an average studio is. The recording environments of the top studios can be as low as 30dB but for much of the equipment in the chain figures beyond 20kHz are quite difficult to obtain. The big commercial studios may have a low noise floor but their live rooms are probably big enough to start introducing attenuation of higher frequencies through air absorption. +60dB is about the maximum that mic-pres go. The closer to this maximum one gets though, the more distortion is introduced. The real problem though is the mics, for the vast majority of mics -50dB @ 30kHz is just not possible. All dynamic mics are out of the question and all the commonly used studio condensers. Try EQ boosting above 20kHz and all that is present in the signal is noise. Also, cymbals and to an extent Gamelan in the higher frequencies are so full of both odd and even harmonics, the result is indistinguishable from white noise anyway and this is even in the audible frequency band. In other words, in practice we are (perceptually) adding to the noise floor rather than being able to distinguish sounds separate from the noise floor.

I accept that for a limited few instruments there may be harmonic information present at 30kHz which would register above the theoretical noise floor of digital audio and even above the noise floor of some recording environments. But when we sum these two together and factor in mic response and the effects of everything else in the recording and playback chains, I still don't believe there would be much at 30kHz which could be captured. Regardless of this though, there is still no justification for people believing there is something to be gained by encoding audio frequencies of 96kHz.

G

Quote:



If you want some reliable proof that anyone can hear beyond 22kHz. You can check out the SPL Phonitor...I was able to listen to the Phonitor and compare my previous Ultrasone PL650 to the K702. Check it out if you get a chance, you just might be able to hear something above the normal human range. If not, you can still hang on to your theory.

Now, I get the impression that you don't really have experience working in HR sound environment...just theoretical concepts. Even for a theoretical concept dude whose beliefs are stronger than his experiences, you still believe that the leading expert at ADAM and Yamaha and SPL and all the other manufacturers have just created this range out of thin air to use as a marketing tool.

Quote:



1. As far as I'm concerned, I am willing to admit that it may be possible to perceive some sonic content above 20kHz, although there is only anecdotal evidence for it. However, if there is anything to be heard it is definitely lower than 40kHz, simply because there are no studio mics that can pick up anything above this point. A 40kHz signal can be encoded using a sample rate of 96kFs/s. There is however absolutely no advantage to a sample rate of 192kHz as there simply is nothing between 48kHz and 96kHz which can be recorded. If someone thinks they hear a difference with 192kFs/s, it cannot be anything to do with the recording, because no studio mics can record above 40kHz, so the only thing which can be in these higher frequencies on a recording is noise.

2. Sorry but you are way off the mark with your second paragraph. I have been recording and mixing exclusively in higher than 16bit since the end of 1992, not long after the technology was first available (Yamaha DMR8 + DRU8). I switched over from 20bit to 24bit in about 1995, when multi-channel 24bit converters first became available (DigiDesign 888). So, there can't be that many engineers who have a longer practical experience than me in working with >16bit. Also, I've used higher resolution recording technology not just for music recording and production but also quite extensively for film and TV sound too. My understanding of the theory side of HR has come from a fair bit of research over the years and particular thanks need to go to Nika Aldrich who gave many hours of his time online to help iron out many of my misunderstandings.

In fact, there is not much in the digital audio chain that I haven't thoroughly tested. Take dither for example, to start with only TDPF (Triangular Probability Density Functions) were available but I've used extensively Sony Super-Bitmapping, UV22, POWr, Waves L2 and DigiDesign. I've gone through or tested countless mics, mic-pres, cables, ADCs, DACs and speakers. Since the early '90s I must have spent around $500,000 on equipment and acoustics. I've also done work in many of this country's (UK) top studios and dubbing theatres as well as my own of course.

G

Nevermind, I'm a retard.