24bit vs 16bit, the myth exploded! - Page 31

Well, if this thread was a debate and not a discussion, then everybody lost with that mindset. Head-Fi is first and foremost a discussion forum, not a debate forum. If there were anybody on the planet in possession of all knowledge on a given subject, I would listen to them, but soon even I would tire of listening to the soliloquy eventually, and probably not adsorb everything I read because sometimes I need to see things form different perspectives before they sink in. Such is learning.

I will try to find time to to do this sometime soon. Then, so long as I can be trusted to provide files that are equal outside of the bit depth, dithering, and noise shaping, people can decide for themselves.

I posted files like this once before, and I recall that you did some interesting analysis of the two versions using Cool Edit or something like it. I believe that with the 16 bit version, there was a smaller difference between the loudest and most quiet samples. But even the 24 bit file had a range well within 96 decibels. Could this difference have been due to the noise shaping?

Check out the mastering forum on Gearslutz.com - Powered by vBulletin. There's lots of in depth discussion about this being done there by people who would most certainly not consider themselves audiophiles.

yeah, I'm pretty shocked that such b*tchy personal attacks have been tolerated on this forum

Wow very interesting post, but a question then, say you were to have an album ripped and stored as flac in 24 bit with a sound sample size of 96.000khz.. how could i change this so it is playable on my ipod classic, as I have run into this issue before when experimenting with, mostly "high res" albums before, mostly ripped from vinyl, which I have not tried doing for myself yet.

You'll need something to convert flac to alac, aiff, mp3, or aac and you need something to convert it from 96/24 to 48/16. dbpoweramp can handle the format conversion on pc, max on windows. Which software to do the sample rate conversion is something of a religious debate, so I'll let others chime in.

people tend to forget that everything is relative. Ofcourse you wouldn't get 48db more pressure as you would just turn down the music. However, instead of a larger dynamic range you could just have more resolution in the sound.

I can best describe it by taking the 48bit images and compare them with 32 or 24 bit images. Yes you might not see the difference. But there is more detail in the image which can be used in post processing... or could be used by compressing dynamic range without sacrificing resolution.

ofcourse postprocessing is a really good reason to have 24 bit audio, which is something that is especially active for normal setups which need some post processing to make the audio fit the room.

You haven't read the whole thread. This claim has been dismissed many times so far. Unfortunately, the comparison to the bit depth of pictures does not hold water; it is still true, that this additional dynamic range is useful for sound processing - but for some other reasons (as it has been mentioned.............)

24/96 has been agreed to sound better than 16/44.1 sometimes, but noone really knows why basically

some high quality dithering algorithm to CDDA might sound very similar to 24/96...just like dithering for video works indeed.

16bit dithered to 8 might look just as good as 12bit, it's been more or less proven on doom9.

it does hold water.... if you think it doesn't you should really explain why.

it's been explained repeatedly in this very thread.

The short answer: the bits mean the numbers get bigger, not that the numbers have more details in the middle. The added bits mean it handles more volume below -96db.

Read this: http://www.head-fi.org/forums/f133/b...lained-455688/. I didn't go much into why higher bit-depth is better, because I thought it's obvious. Apparently not...
I guess a few pictures will show the difference.

Musical Fidelity V-DAC, waveform of undithered 1kHz sinewave at –90.31dBFS, CD data (left channel blue, right red).

Musical Fidelity V-DAC, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data (left channel blue, right red).
Bottom of the line, the more bits you have the bettter you can represent the sample amplitude, which is a real number.

Consider how this picture would look with 1 bit depth and then with 16. With 16, the difference between the digital representation and the red sine wave would be invisible to the naked eye at this scale.

well, that's what the thread author seems to have implied.

grawk and others perhaps interested in a bit of ..... clarification.... as to issues concerning the concept of resolution as presented in this (lengthy, and sadly often technically misinformative) thread,

you may wish to take a look at Pohlmann for a not-overly-technical explanation of quantization step size and quantization error:

Principles of Digital Audio
Chapter 2, Fundamentals of Digital Audio
pp31-32: Quantization
pp32-37: Signal-to-error ratio
pp37-39: Quantization distortion

this chapter is available for free preview on google books; depending on the day / cache refresh timing / phases of the moon, various pages may not be included in the preview (on last check, 31-39 were all visible except 37; then later 37 showed up...):

Principles of digital audio - Google Books

hth

It's not a question of whether or not it's more accurate, it obviously is...

... but is it AUDIBLE? All signs point to "NO."

well ofcourse that's a matter of how you use the information.

you can use the bits in 2 ways. The example of Xel'Naga.shows one kind of way (the way i was trying to explain)

However more range is also an option.

it's the same as with normal numbers. You can use 6 digits to represent a distance. for example 100.000 km Now adding digits, you can choose to add digits to the big end or de little end. so adding digits can make your precision better 1.000.000,00 or you can get more range 100.000.000. Now, if you want to represent a distance from any place to any place on a map you could opt for range. But this has little value of your map is only 50 by 50 km. Your maximum distance will be (50^2+50^2)^0.5 so instead of using the digits to represent range you could also use it to represent precision.

further more. This precision can be used when dealing with post processing (DSP and stuff) it will reduce the amount of information that is lost.

To illustrate this i want to give an example.

Just think of having only 1 digit of precision. now plain d/a conversion would be ok. However with a dsp most likely multiplications and divisions will be done. now dividng 9 by 3 will give you a nie round number of 3. However thigns go astray when you decide to divide 2 by 3. we all know that the anser is 0.66666 rounded up it would have been a 1. However due to the 1 digit precision it would be represented as a 0.

To make a long story short. This can be solved in 2 ways. The easy and best way is to indroduce more bits to get rid of those rounding things. The second is to have a dsp which uses a higher number of bits internallY. It will sotre the number 0.6666 and will round it of when it gives output.