[ramicio]

24 bits are 16,777,216 possible values, 16 bits are 65,536.  So you suggest that the other 16,711,680 values are all below 96 dB?  If things are linear, then each decibel of a 24 bit file can be broken down to 1/116,508.  Each decibel in a 16-bit file can be broken down to 1/682.  So yes, there is more resolution, and it's not all about dynamic range.  I can't stand the view that 16/44.1 is all anyone will ever need, so let's not progress and stay with an ancient format forever.  It could be the same view that people can survive without needing computers, and people can live without technology, so why bother with it?  People also seem to think there will be no benefit from higher than 16/44.1 if each instrument was only captured in such a "resolution."  I find this false and stupid.  It would only matter if the song is just that one instrument and nothing else.  But you mix other sounds together, how could it not benefit from say 24/192?  Everything is just math here.  In my mathematical opinion a song is going to sound more natural, and a DAC is going to have to do less work trying to reconstruct a waveform at 24/192 than 16/44.1.

 

[xnor]

Quote:

24 bits are 16,777,216 possible values, 16 bits are 65,536.  So you suggest that the other 16,711,680 values are all below 96 dB?
 
If things are linear, then each decibel of a 24 bit file can be broken down to 1/116,508.  Each decibel in a 16-bit file can be broken down to 1/682.  So yes, there is more resolution, and it's not all about dynamic range.
 
[...]
 
Everything is just math here.  In my mathematical opinion a song is going to sound more natural, and a DAC is going to have to do less work trying to reconstruct a waveform at 24/192 than 16/44.1.

1) 20 * log10( 1 / 65536 ) equals guess what? -96... dB. Feel free to fill in all the numbers up to 16777216 into the same formula. Just math, eh?
 
2) Oh, linear decibels again, how hilarious!
 
3) Right, 192000 samples per second with larger sample size are much simpler and faster to process than only 44100, especially in playback devices without much processing power. And the hardware to process it is cheaper too!
 
t_t'

 

[nick_charles]

Quote:

24 bits are 16,777,216 possible values, 16 bits are 65,536.  So you suggest that the other 16,711,680 values are all below 96 dB?  If things are linear, then each decibel of a 24 bit file can be broken down to 1/116,508.  
 
The db scale is not linear it is logarithmic
 
 
Each decibel in a 16-bit file can be broken down to 1/682.  So yes, there is more resolution, and it's not all about dynamic range.  
 
What you have is a voltage that can be split into different numbers of slices. A full scale CD line-out signal is a nominal 2V so you can do the maths on that. At that point you need to demonstrate that this extra subdivision of voltages is audible. With a 2V signal sliced by 16 bits the difference between say 32766 and 32767 is 0.00003V , I wonder how audible this difference is , I'll give you a clue it isn't. It is way below human discrimination.
 
I can't stand the view that 16/44.1 is all anyone will ever need, so let's not progress and stay with an ancient format forever.  
 
But you will need to provide strong evidence for the impact of the limitations of 16/44.1 in controlled listening tests, to date such tests as exist do not support the view that 16/44.1 as a delivery medium is inadequate. Wrt recording and mixing that is different.
 
 
It could be the same view that people can survive without needing computers, and people can live without technology, so why bother with it?  People also seem to think there will be no benefit from higher than 16/44.1 if each instrument was only captured in such a "resolution."  I find this false and stupid.  
 
???????
 
 
It would only matter if the song is just that one instrument and nothing else.  But you mix other sounds together, how could it not benefit from say 24/192?
 
See above , what is mathematically better is not necessarily audibly better.
 
 Everything is just math here.  In my mathematical opinion a song is going to sound more natural
 
Your opinion, but you need to provide empirical evidence for this extra perceived "naturalness", this is where the hard evidence does not exist.
 
 
, and a DAC is going to have to do less work trying to reconstruct a waveform at 24/192 than 16/44.1.
 
Can you explain this with reference to how DACs work and Nyquist ? 

 

[JerryLove]

Quote:

Originally Posted by nick_charles 
 
The db scale is not linear it is logarithmic

 
In regards to power yes, in regards to audiability no. 3db (thereabouts) is the threshold of audiability: whether we are discussing the difference between 75 and 78 db or 105 and 108db.
 
Since hearing aberrations is linear in db (if not, we need to toss out those useless FR charts now), it makes sense that you would set resolution linear in db.
 
Mind you, you could also cheat and set a variable resolution, after all the different values below the noise floor are irrelevent, as is any noise below the hearing threshold: but things would get a bit complex. Simpler to use enough bits for the desired dynamic range at the desired resolution from end to end.
 
 
Quote:

What you have is a voltage that can be split into different numbers of slices. A full scale CD line-out signal is a nominal 2V so you can do the maths on that. At that point you need to demonstrate that this extra subdivision of voltages is audible. With a 2V signal sliced by 16 bits the difference between say 32766 and 32767 is 0.00003V , I wonder how audible this difference is , I'll give you a clue it isn't. It is way below human discrimination.

 
Here I agree: though I think we actually have two questions.
 
Dealing solely with end-user listening: the resolution exceeding the ear is the standard. When we start discussing something which may undergo more processing: working with as high a resolution as you can get away with helps minimize problems in the end product.
 
I think too we agree on the rest. Higher resolution is closer to reality: but generally not in a useful way. If anything, I'd think you might want to play with the 44Khz, to allow for VHF harmonics with minimal distortion: but I'm just guessing that it might matter.

 

[wyager]

I'm a little confused by the OP-It said that each bit represented a 6dB increase in maximum range-but doesn't this assume the same scale for all recordings?
Let's say you are playing a song on your ipod, and the output reaches a peak voltage of ±1.28v (just as an example). The total voltage range is 2.56 volts.
If you have 8-bit audio, there are 256 discreet voltage levels that the DAC can produce. This means that every voltage "step" is .01v. This would probably sound pretty grainy. Here is an accurate graph of a 100hz audio signal played at 8-bit resolution (assume that y=voltage and x=time in seconds).

As you can see, the audio is visibly "choppy", and this would probably not sound so great.
Now, if I understand what the OP is saying, it means to tell me that  16-bit audio looks like this (graph scale is adjusted along time axis to make steps visible):

where the range is exponentially wider, but scale stays the same... I have to be misunderstanding this part. Because if the OP really says what I think it does, then it says that the number of bits determine the maximum volume, which is just plain wrong...
The way I think of it, if you have 16-bit audio then there are 65536 discreet voltage levels between the same two peak voltages, and it looks more like this:

ie. smoother and less "digital".
 
Am I misunderstanding how this works?

 

[xabu]

Ah, well, the thing is ... the membran of a speaker can only do two things ... move into one direction or into the reverse direction ... in between there's continous motion of that membran ... so there is no such thing as digital sound ... sound is always analog and continous.
What you're visualizing is the side of an ADC.

 

[wyager]

Quote:

Ah, well, the thing is ... the membran of a speaker can only do two things ... move into one direction or into the reverse direction ... in between there's continous motion of that membran ... so there is no such thing as digital sound ... sound is always analog and continous.
What you're visualizing is the side of an ADC.

So by that logic, square wave sound sounds just the same as sine wave sound? That's not the case... The response time of a small speaker is so fast that relying on it to perform mechanical low-pass filtering and/or mechanical signal smoothing is stupid (which is why we need high-resolution non-PWM ADCs).

 

[xnor]

No, see http://en.wikipedia.org/wiki/Anti-aliasing_filter

 

[wyager]

Quote:

No, see http://en.wikipedia.org/wiki/Anti-aliasing_filter

Are you responding to me, or Xabu?
Can you explain what you mean? I don't understand what you are negating with the link to the AA filter article...

 

[xabu]

Quote:

Quote:

Ah, well, the thing is ... the membran of a speaker can only do two things ... move into one direction or into the reverse direction ... in between there's continous motion of that membran ... so there is no such thing as digital sound ... sound is always analog and continous.
What you're visualizing is the side of an ADC.

So by that logic, square wave sound sounds just the same as sine wave sound? That's not the case... The response time of a small speaker is so fast that relying on it to perform mechanical low-pass filtering and/or mechanical signal smoothing is stupid (which is why we need high-resolution non-PWM ADCs).

A membran is actually not able to produce the motion of air equivalent of an ideal square wave ... accoustical square waves are physically impossible ... you only get approximations.
 
But thats not what I wanted to get at. I just doubt that a speaker is able to reproduce the signal which your first graph shows.
 
 
 
... and I'm still trying to get the picture of the relation between a range of possible volume levels depending on the bit depth and the bit dependency of the "resolution" of a specific wave at a specific volume level ...
 

 

[wyager]

Quote:

Quote:

Quote:

Ah, well, the thing is ... the membran of a speaker can only do two things ... move into one direction or into the reverse direction ... in between there's continous motion of that membran ... so there is no such thing as digital sound ... sound is always analog and continous.
What you're visualizing is the side of an ADC.

So by that logic, square wave sound sounds just the same as sine wave sound? That's not the case... The response time of a small speaker is so fast that relying on it to perform mechanical low-pass filtering and/or mechanical signal smoothing is stupid (which is why we need high-resolution non-PWM ADCs).

A membran is actually not able to produce the motion of air equivalent of an ideal square wave ... accoustical square waves are physically impossible ... you only get approximations.
 
But thats not what I wanted to get at. I just doubt that a speaker is able to reproduce the signal which your first graph shows.
 
 
 
... and I'm still trying to get the picture of the relation between a range of possible volume levels depending on the bit depth and the bit dependency of the "resolution" of a specific wave at a specific volume level ...
 

Ah, OK, I understand now.
Yes, I recognize an acoustical square wave is impossible (for now), but what I was saying is that the response times of small speakers are so fast that you can get VERY close to a true audible square wave. The response times of a small speaker are also so fast, and our ears so sensitive, that the small 1/256 volume jumps are, in fact, audible and decrease overall sound quality. Even if speakers did act as perfect mechanical smoothing filters and square waves at x frequency sounded just like sine waves at x frequency, what happens when you try to overlap multiple waves? Have you ever tried to add separate waves with 1 bit audio resolution (in a non ∆∑ configuration)? It just doesn't work. (If you want me to make a snazzy graph showing why 1-bit wave addition doesn't work, I'd be happy to.) 

 
 
And as for your last sentence, look at this... the OP said "So, 24bit does add more 'resolution' compared to 16bit but this added resolution doesn't mean higher quality, it just means we can encode a larger dynamic range." and "4 = Dynamic range, is the range of volume between the noise floor and the maximum volume."
 
OP seems to be suggesting that more bits=more volume, which is wrong... more bits=more discreet volume levels (a smoother curve).

 

[xabu]

Quote:

Ah, OK, I understand now.
Yes, I recognize an acoustical square wave is impossible (for now), but what I was saying is that the response times of small speakers are so fast that you can get VERY close to a true audible square wave. The response times of a small speaker are also so fast, and our ears so sensitive, that the small 1/256 volume jumps are, in fact, audible and decrease overall sound quality. Even if speakers did act as perfect mechanical smoothing filters and square waves at x frequency sounded just like sine waves at x frequency, what happens when you try to overlap multiple waves? Have you ever tried to add separate waves with 1 bit audio resolution (in a non ∆∑ configuration)? It just doesn't work. (If you want me to make a snazzy graph showing why 1-bit wave addition doesn't work, I'd be happy to.) 

 
 
And as for your last sentence, look at this... the OP said "So, 24bit does add more 'resolution' compared to 16bit but this added resolution doesn't mean higher quality, it just means we can encode a larger dynamic range." and "4 = Dynamic range, is the range of volume between the noise floor and the maximum volume."
 
OP seems to be suggesting that more bits=more volume, which is wrong... more bits=more discreet volume levels (a smoother curve).

You may be able to differentiate between more than 256 different volume levels, but not "inside" one wave and at the scale you mentioned ... try to translate that to dB ... you're max. able to differentiate approx. 1 dB differences in volume levels (...hmmm ... actually that would translate to only 140 max. steps before you you experience irreversible damage to your ears ... on the other side, sensitivity varies depending on frequency and volume ... so you may be able to differentiate 0.25 dB in some areas ...)
 
Hm ... and if we take this practical approach with anti aliasing filter into account ... there should practically no problem at all.

 

[wyager]

Quote:

Quote:

Ah, OK, I understand now.
Yes, I recognize an acoustical square wave is impossible (for now), but what I was saying is that the response times of small speakers are so fast that you can get VERY close to a true audible square wave. The response times of a small speaker are also so fast, and our ears so sensitive, that the small 1/256 volume jumps are, in fact, audible and decrease overall sound quality. Even if speakers did act as perfect mechanical smoothing filters and square waves at x frequency sounded just like sine waves at x frequency, what happens when you try to overlap multiple waves? Have you ever tried to add separate waves with 1 bit audio resolution (in a non ∆∑ configuration)? It just doesn't work. (If you want me to make a snazzy graph showing why 1-bit wave addition doesn't work, I'd be happy to.) 

 
 
And as for your last sentence, look at this... the OP said "So, 24bit does add more 'resolution' compared to 16bit but this added resolution doesn't mean higher quality, it just means we can encode a larger dynamic range." and "4 = Dynamic range, is the range of volume between the noise floor and the maximum volume."
 
OP seems to be suggesting that more bits=more volume, which is wrong... more bits=more discreet volume levels (a smoother curve).

You may be able to differentiate between more than 256 different volume levels, but not "inside" one wave and at the scale you mentioned ... try to translate that to dB ... you're max. able to differentiate approx. 1 dB differences in volume levels (...hmmm ... actually that would translate to only 140 max. steps before you you experience irreversible damage to your ears ... on the other side, sensitivity varies depending on frequency and volume ... so you may be able to differentiate 0.25 dB in some areas ...)
 
Hm ... and if we take this practical approach with anti aliasing filter into account ... there should practically no problem at all.

I beg to differ... I just edited an app I had that generated a sine wave-I set it to 100 hertz. Then I added a line that effectively made the audio 8 bit, and the difference was VERY obvious to my ears, even with just a single wave. And I'm not sure what you're saying about an AA filter, all an AA filter does is remove frequencies above the Nyquist frequency (usually 22050hz) during sampling I believe... it has nothing to do with smoothing the audio.

 

[khaos974]

What is usually done when a * bit signal is generated is that the signal is actually dithered down to 8 bit, ie quantization error is replaced by noise.
So a higher bit depth indeed means a greater dynamic range.

 

[wyager]

Quote:

What is usually done when a * bit signal is generated is that the signal is actually dithered down to 8 bit, ie quantization error is replaced by noise.
So a higher bit depth indeed means a greater dynamic range.

 
Umm, turning something from >8bit to 8bit doesn't replace quantization error, it creates it... LOL. From the wiki article on quantization error-"This error is either due to rounding or truncation." When you turn audio from >8bit to 8bit, you either round or truncate it. If you mean the other way around, then true, higher bit depths tend to have more noise, but saying that quantization error is better than noise is like saying that having a calculator that does integer-only math with 100% minimum accuracy is better than a calculator that does 10-digit floating point math with 99% minumum accuracy. I'm not sure if that is what you are saying, as text doesn't convey tones of voice properly, but if so then I might as well respond to that argument.
 
And again, I'm not sure why you say it means a greater dynamic range... the maximum volume stays the same. The computer just scales the voltage value of a single bit.