24bit vs 16bit, the myth exploded!
Jun 10, 2011 at 4:37 PM Post #737 of 7,175
It doesn't take a rocket scientist to see that control is very important to Apple. To put it bluntly, they are control-freaks and also suffering from NIH syndrome.
 
Jun 10, 2011 at 5:41 PM Post #739 of 7,175
If anyone's still interested in the original topic of this thread, here's the spectrum of an unprocessed drum recording (recorded with a bit depth of 24 bits, average RMS amplitude: -32 dB !!!):
 

 
The green line at the bottom shows added noise due to re-quantization to 15 bits + 1 bit simple dithering, no noise shaping was used.
 
High end DACs like the WM8742 manage to reach a signal-to-noise ratio of about 120 dB under ideal conditions.
 
To quote Arnold from HA: "While 24 bits is an insane overkill format, 16 bits is an overkill format." 
cool.gif

 
Jun 10, 2011 at 7:02 PM Post #740 of 7,175
for playback, 16 bits is plenty, for recording dynamic music, 24 bits lets you leave lots of headroom to prevent problems.
 
As to why others don't use ALAC, I know some do and some don't.  I don't really care why, as I only use apple devices.
 
Jul 5, 2011 at 11:17 AM Post #741 of 7,175
 
[size=medium]
Hi All, I see this conversation has kind of shifted directions. Nevertheless, I have written the below in response to the original post regarding 24-bit/16-bit music:

The technical description by GREGORIO is pretty good and accurate. However,  the value, or interpretation, of the facts is a little misguided. The number of bits does indeed define the dynamic range, or difference between softest sounds and loudest sounds, expressed in decibels. Each bit adds 6dB of dynamic range. Humans have the capability of hearing over a 100dB range (30dB to 130dB). 16-bit encoding has a 96dB dynamic range, and 24-bit has 144dB. So while you can argue that 24-bit is overkill, 16-bit is just adequate.

The two points missing are:

1.  Physical parts and circuits are not perfect. Digital-to-analogue converter chips never meet their theoretical capability due to resistance, capacitance, and inductance in the chip and in the circuit. 16-bit circuits were often 14-bit and sometimes as bad as 12-bit.  24-bit circuits are even more difficult and seldom are accurate beyond 20-bits. But almost everyone (audio engineers that is) thinks 20 bits is sufficient  - that is 120dB dynamic range.

2. Tests performed at the dawn of digital technology proved that with analogue recordings, there was useful signal below the 'noise floor' that could be perceived by listeners. So even though a good LP only had a signal to noise ratio of only 70dB, there was perceivable signal 15dB below that level.
 
With digital, when you run out of bits you're done. Any signal below the cut-off point is, well, cut-off/missing. This is the big rap against 'digital sound'.  A lot of information is missing never to return. This is why having the extra margin of 20 bit+ recording important.  It allows the recording engineer to have 'margin'. Because with live music the loudness can never be exactly predicted, recording engineers always leave a 6 dB (2-bits) available for something unexpectedly loud. They also use 'compressors' to be sure maximum '0dB' is never exceeded, but the compressor adds its own distortion (though not as bad as going above 0dB in a digital recorder).

So what music is down at these very quiet levels? Mostly, the dying reverb of an instrument and environmental or ambient sound of the recording space.

With many pop recordings there is no recording space as all instruments are plugged into the mixing board and vocals have artificial reverb added.  But acoustic music usually is recorded in a venue. The low level sounds tell the listener about that space – we can tell a large room from a small room by sound alone. So systems with wide dynamic range have 'air' and 'space' as well as all the instrumental nuance that define the difference between a Guarneri and Stradivarius violin or a Gibson and a Fender guitar. Sure, this is subtle, but it is the fine details that are the difference between good and excellent.

The argument that recorded music has a very narrow dynamic range is again accurate but misleading. Sometimes it is narrow for artistic reasons, and sometimes it is narrow due to technical restrictions. A live symphony orchestra experience does cover a 100dB range from quiet to full orchestral crescendo. Artists and engineers will exploit the improved dynamic range capability of the medium to make more lifelike recordings in the future.
 
[/size]

 
Jul 5, 2011 at 11:43 AM Post #742 of 7,175
 
[size=medium]
Hi All, I see this conversation has kind of shifted directions. Nevertheless, I have written the below in response to the original post regarding 24-bit/16-bit music:

The technical description by GREGORIO is pretty good and accurate. However,  the value, or interpretation, of the facts is a little misguided. The number of bits does indeed define the dynamic range, or difference between softest sounds and loudest sounds, expressed in decibels. Each bit adds 6dB of dynamic range. Humans have the capability of hearing over a 100dB range (30dB to 130dB). 16-bit encoding has a 96dB dynamic range, and 24-bit has 144dB. So while you can argue that 24-bit is overkill, 16-bit is just adequate.

The two points missing are:

1.  Physical parts and circuits are not perfect. Digital-to-analogue converter chips never meet their theoretical capability due to resistance, capacitance, and inductance in the chip and in the circuit. 16-bit circuits were often 14-bit and sometimes as bad as 12-bit.  24-bit circuits are even more difficult and seldom are accurate beyond 20-bits. But almost everyone (audio engineers that is) thinks 20 bits is sufficient  - that is 120dB dynamic range.

2. Tests performed at the dawn of digital technology proved that with analogue recordings, there was useful signal below the 'noise floor' that could be perceived by listeners. So even though a good LP only had a signal to noise ratio of only 70dB, there was perceivable signal 15dB below that level.
 
With digital, when you run out of bits you're done. Any signal below the cut-off point is, well, cut-off/missing. This is the big rap against 'digital sound'.  A lot of information is missing never to return. This is why having the extra margin of 20 bit+ recording important.  It allows the recording engineer to have 'margin'. Because with live music the loudness can never be exactly predicted, recording engineers always leave a 6 dB (2-bits) available for something unexpectedly loud. They also use 'compressors' to be sure maximum '0dB' is never exceeded, but the compressor adds its own distortion (though not as bad as going above 0dB in a digital recorder).

So what music is down at these very quiet levels? Mostly, the dying reverb of an instrument and environmental or ambient sound of the recording space.

With many pop recordings there is no recording space as all instruments are plugged into the mixing board and vocals have artificial reverb added.  But acoustic music usually is recorded in a venue. The low level sounds tell the listener about that space – we can tell a large room from a small room by sound alone. So systems with wide dynamic range have 'air' and 'space' as well as all the instrumental nuance that define the difference between a Guarneri and Stradivarius violin or a Gibson and a Fender guitar. Sure, this is subtle, but it is the fine details that are the difference between good and excellent.

The argument that recorded music has a very narrow dynamic range is again accurate but misleading. Sometimes it is narrow for artistic reasons, and sometimes it is narrow due to technical restrictions. A live symphony orchestra experience does cover a 100dB range from quiet to full orchestral crescendo. Artists and engineers will exploit the improved dynamic range capability of the medium to make more lifelike recordings in the future.
 
[/size]


You forget that no one advocates for recording at 16 bit, what we are taking about is playback in domestic venues and if 16 bit + dithering is enough. Personal opinion: YES!

And about FLAC, GPL and viral contamination? Guess what, the libraries and the codec specifications are under BSD, which means absolutely no contamination effect! :cool:
 
Jul 5, 2011 at 12:18 PM Post #743 of 7,175
Quote:
1.  Physical parts and circuits are not perfect. Digital-to-analogue converter chips never meet their theoretical capability due to resistance, capacitance, and inductance in the chip and in the circuit. [...]


What you miss here is that the recording equipment, the entire chain from the mics to the A/D conversion, usually has a lower signal to noise ratio than what 16-bit D/A converters can achieve.
To quote Ethan: "You'll be lucky to have 60 dB s/n and really lucky to have 70." (mic + preamp + adc, even in a professional recording studio)
 
Quote:
2. Tests performed at the dawn of digital technology proved that with analogue recordings, there was useful signal below the 'noise floor' that could be perceived by listeners. So even though a good LP only had a signal to noise ratio of only 70dB, there was perceivable signal 15dB below that level.

 
So approximately 85 dB (with mostly noise down there I guess) which is still inferior to 16-bit digital audio (96 dB). Even if you had a magnificent vinyl setup you'd only need about 14-bits in a digital system which would still be a lot more accurate in various other aspects (no clicks or pops, no large frequency deviations, no wear, etc).
 
 
Quote:
With digital, when you run out of bits you're done. Any signal below the cut-off point is, well, cut-off/missing. This is the big rap against 'digital sound'.  A lot of information is missing never to return.

 
And with analog there is no clipping nor a noise floor and you can store indefinitely silent signals, right? Sorry for the rant, but that's nonsense and, in fact, you get more headroom with digital.
More information is missing (never to return!) with analogue recordings. And see the previous paragraph, you kinda contradict yourself here.
 
 
Quote:
This is why having the extra margin of 20 bit+ recording important.  It allows the recording engineer to have 'margin'. Because with live music the loudness can never be exactly predicted, recording engineers always leave a 6 dB (2-bits) available for something unexpectedly loud.

 
Even if that were true it wouldn't matter in a playback system with a 16-bit format.
 
 
Quote:
The argument that recorded music has a very narrow dynamic range is again accurate but misleading. Sometimes it is narrow for artistic reasons, and sometimes it is narrow due to technical restrictions. A live symphony orchestra experience does cover a 100dB range from quiet to full orchestral crescendo. Artists and engineers will exploit the improved dynamic range capability of the medium to make more lifelike recordings in the future.

 
Any references? The Handbook of Recording Engineering (John Eagle, 2005) says that the dynamic range in a concert hall normally doesn't exceed 80 dB.
Also, please feel free to post 24-bit samples (10 to 30 seconds should do) of orchestral recordings that you think exploit the additional 8 bits. I'm very curious. I don't listen to classical music but I've taken a closer look at some free, supposedly audiophile-quality 24-bit recordings and found that the dynamic range didn't exceed that of what 16 bits are capable of storing.
 
 
 
 
Jul 5, 2011 at 12:54 PM Post #744 of 7,175


Quote:
 
Any references? The Handbook of Recording Engineering (John Eagle, 2005) says that the dynamic range in a concert hall normally doesn't exceed 80 dB.
Also, please feel free to post 24-bit samples (10 to 30 seconds should do) of orchestral recordings that you think exploit the additional 8 bits. I'm very curious. I don't listen to classical music but I've taken a closer look at some free, supposedly audiophile-quality 24-bit recordings and found that the dynamic range didn't exceed that of what 16 bits are capable of storing.
 
 
 


I have about 700 Classical CDs , not one of them has a dynamic range anywhere close to CD limits - the CD I have with the biggest range Solti/CSO Mahler 1 (DDD) notches up about 62db  (measured)  it peaks at 0db in several places
 
Jul 5, 2011 at 2:35 PM Post #745 of 7,175
Quote:
I have about 700 Classical CDs , not one of them has a dynamic range anywhere close to CD limits - the CD I have with the biggest range Solti/CSO Mahler 1 (DDD) notches up about 62db  (measured)  it peaks at 0db in several places


Thanks nick_charles. 60-ish dB max also lines up with what I've analyzed.
 
 
Aug 18, 2011 at 6:53 PM Post #746 of 7,175
After more than 2 years, I thought I'd pop back here briefly to see how it was going and to thank those of you for the kind emails you sent. I started this thread to provide information to help you, the consumers, to avoid the huge amount of snake oil present in the audio and HiFi industry. I'm glad I was of help to many of you. I'm not so glad that the industry continues using deception as a marketing tool. If 24bit is a waste of time (and space) for playback and 96kHz sampling rate is already higher than the optimum rate for Analogue to Digital Converters, the latest ADCs boast 32bit and 384kHz! We are getting further and further away from the optimal design, therefore resulting in less accurate conversion, while the marketing is trying to convince you of exactly the opposite. This is a real shame, instead of developing better quality filters, clocking mechanisms and analogue components in 96kHz converters to give us better quality audio, the industry has decided it's cheaper and easier to develop poorer quality converters (at 192kHz and 384kHz) and take us for fools by marketing them as better. It's deceptive and unethical and makes me pretty mad as a consumer myself.
 
The paper I linked to previously (http://www.lavryengineering.com/documents/Sampling_Theory.pdf) is as true today as it was the day it was written. Computing power has obviously dramatically increased since 2004 but there is a finite speed limit to how quickly a capacitor can be charged and amps can settle, no amount of computing power is going to change the laws of physics. I was pleased to see at least one other manufacturer come out recently and tell the truth about sampling rates. My estimation of them as a company has been greatly enhanced: Benchmark official statement 96kHz vs.192kHz
 
I might pop back here occasionally (time permitting), to answer questions where I'm able and to help those others here like me trying to combat the marketing fraud being perpetrated. Accurate information is the best weapon in my opinion. If the highly offensive personal attacks from shills and those entrenched in their ignorance starts up again though I'll just call it a day again.
 
Regards,
 
G
 
 
 
Aug 19, 2011 at 5:42 PM Post #747 of 7,175


Quote:
After more than 2 years, I thought I'd pop back here briefly to see how it was going and to thank those of you for the kind emails you sent. I started this thread to provide information to help you, the consumers, to avoid the huge amount of snake oil present in the audio and HiFi industry. I'm glad I was of help to many of you. I'm not so glad that the industry continues using deception as a marketing tool. If 24bit is a waste of time (and space) for playback and 96kHz sampling rate is already higher than the optimum rate for Analogue to Digital Converters, the latest ADCs boast 32bit and 384kHz! We are getting further and further away from the optimal design, therefore resulting in less accurate conversion, while the marketing is trying to convince you of exactly the opposite. This is a real shame, instead of developing better quality filters, clocking mechanisms and analogue components in 96kHz converters to give us better quality audio, the industry has decided it's cheaper and easier to develop poorer quality converters (at 192kHz and 384kHz) and take us for fools by marketing them as better. It's deceptive and unethical and makes me pretty mad as a consumer myself.
 
The paper I linked to previously (http://www.lavryengineering.com/documents/Sampling_Theory.pdf) is as true today as it was the day it was written. Computing power has obviously dramatically increased since 2004 but there is a finite speed limit to how quickly a capacitor can be charged and amps can settle, no amount of computing power is going to change the laws of physics. I was pleased to see at least one other manufacturer come out recently and tell the truth about sampling rates. My estimation of them as a company has been greatly enhanced: Benchmark official statement 96kHz vs.192kHz
 
I might pop back here occasionally (time permitting), to answer questions where I'm able and to help those others here like me trying to combat the marketing fraud being perpetrated. Accurate information is the best weapon in my opinion. If the highly offensive personal attacks from shills and those entrenched in their ignorance starts up again though I'll just call it a day again.
 
Regards,
 
G
 
 

Thanks for your honesty, we need more people like you.
beerchug.gif

 
 
 
Aug 19, 2011 at 9:50 PM Post #748 of 7,175
There's only one thing that confuses me about the first post. 
 
If what people hear from 16-bit systems is the recording + noise from Dither, then doesn't that mean that with larger bits, we would hear less noise and dithering would be unnecessary? 
 
But Gregorio goes on to state that most recordings do not even play back a dynamic range of 60db and so even a 16bit system is overkill. And Nick_Charles stated earlier that the most he's been able to measure is roughly in the same range (and confirmed by Xnor). So does that mean that what we really need is a 12bit system? If that is true, then dithering is completely unnecessary because it is something used with 16bit systems for improvements, but clearly a 12 bit system is more than enough (based on classical music -generally agreed upon with having a greater dynamic range-having roughly a 60db dynamic range). So that would mean that Dithering should do absolutely nothing in a 16 bit system, and no one should be able to perceive it. Is that the case? Based on the first post, it appears that dithering is commonly used because it is beneficial and individuals can perceive its effects. 
 
And so now the next line of questioning would be, if dithering does not matter then technically all audio products in the market are actually over kill. And all arguments against going insanely overkill are quite frankly moot. If I were to use a bat to kill a fly that would be overkill, and using a grenade to kill it would be.. well the same. Both equally stupid, and both equally acceptable as fly killers, especially when no one sells fly swatters. 
 
If the answer is that dithering does matter, then we are capable of perceiving a noise level exceeding the capabilities of CD recording, in which case, it might be beneficial to go past 16 bit. It doesn't necessarily mean we must go to 24 bit, just a few bits more. Take for example HDCDs, which if I remember correctly are 20 bit rather than 24. 
 
So i guess what I'm really asking is, does ditherning matter? Oh and does anyone make 12 bit audio gear? 
 
Aug 19, 2011 at 9:58 PM Post #749 of 7,175
dither matters when something actually uses the dynamic range.  If it's not using it, it'll be over 90 db below the level of the music, so you won't hear it.
 
Aug 20, 2011 at 2:29 AM Post #750 of 7,175
Quote:
There's only one thing that confuses me about the first post. 
 
If what people hear from 16-bit systems is the recording + noise from Dither, then doesn't that mean that with larger bits, we would hear less noise and dithering would be unnecessary? 
 
But Gregorio goes on to state that most recordings do not even play back a dynamic range of 60db and so even a 16bit system is overkill. And Nick_Charles stated earlier that the most he's been able to measure is roughly in the same range (and confirmed by Xnor). So does that mean that what we really need is a 12bit system? If that is true, then dithering is completely unnecessary because it is something used with 16bit systems for improvements, but clearly a 12 bit system is more than enough (based on classical music -generally agreed upon with having a greater dynamic range-having roughly a 60db dynamic range). So that would mean that Dithering should do absolutely nothing in a 16 bit system, and no one should be able to perceive it. Is that the case? Based on the first post, it appears that dithering is commonly used because it is beneficial and individuals can perceive its effects. 
 
And so now the next line of questioning would be, if dithering does not matter then technically all audio products in the market are actually over kill. And all arguments against going insanely overkill are quite frankly moot. If I were to use a bat to kill a fly that would be overkill, and using a grenade to kill it would be.. well the same. Both equally stupid, and both equally acceptable as fly killers, especially when no one sells fly swatters. 
 
If the answer is that dithering does matter, then we are capable of perceiving a noise level exceeding the capabilities of CD recording, in which case, it might be beneficial to go past 16 bit. It doesn't necessarily mean we must go to 24 bit, just a few bits more. Take for example HDCDs, which if I remember correctly are 20 bit rather than 24. 
 
So i guess what I'm really asking is, does ditherning matter? Oh and does anyone make 12 bit audio gear? 


Good questions! You are correct in your assumption that you can't hear the noise and 16bit is already overkill, which was the whole point of my original post, to dispel the market hype about 24bit. Even in a listening environment equivalent to a top class recording studio, the dither noise of 16bit will be below the noise floor of the playback equipment and/or the noise floor in the listening environment. The only error in your reasoning is the consequence of not using dither. Dither is a mathematical process which randomises all the quantisation error, the result is a perfectly encoded (and reconstructed) waveform plus some noise (which is the result of the randomised quantisation error). If we don't apply dither the consequence is potentially correlated (non-randomised) error, which is much more noticeable and unpleasant.
 
The only thing I'd take issue with is that I personally would say that using a grenade to kill a fly is not "equally" but slightly more stupid than using a bat!
beyersmile.png
Generally a quite good analogy and understanding of the situation though. I would not say that 12bit was quite enough, although you theoretically only need 10 bits to encode a 60dB dynamic range. The overkill of 16bit ensures that the dither noise is below the noise of your playback environment and therefore you can forget all about it. Dither noise using 12bit would probably add enough to the noise of a good sound system to be perceivable even at normal hearing levels.
 
Hope this helps,
 
G
 

Users who are viewing this thread

Back
Top