sniperwill0
Head-Fier
- Joined
- May 6, 2012
- Posts
- 89
- Likes
- 11
I know it indicates that a digital file is encoded in 24 bits and sampled at 96kHz. However, what does that actually mean? What does that have to do with sound quality? Thanks.
What does 24/96 mean?
- Thread starter sniperwill0
- Start date
mikeaj
Headphoneus Supremus
- Joined
- Jan 10, 2010
- Posts
- 1,639
- Likes
- 110
Most music is encoded as (linear) pulse-code modulation, or formats which decode into LPCM like mp3. That is, the information for each channel is represented by amplitude values that change over time. The resolution of the amplitude values is 24 bits so 2^24 possible values can be represented, and there are 96000 values spaced evenly per second. In other words, you get a value per 1 / 96000 Hz = 10.42 microseconds. This is as opposed to 16-bit resolution at a rate of one per 1 / 44100 Hz = 22.68 microseconds.
With a higher resolution per sample, the quantization error—difference between the actual value and the rounded-off version that you get—is decreased. If properly handled, the error can be handled and processed in such a way that it can be considered just a very small amount of noise, even for 16-bit samples.
With a higher sampling rate, higher frequencies can be represented, as per the sampling theorem. Due to limits of practical implementations and other issues, you can't quite get up to half of the frequency of the sampling rate for playback.
In practical terms, for further processing in the studio, you want 24-bit samples or higher so errors don't keep accumulating in a significant way. For home playback, it takes a very extreme case where 24 bits has an advantage over 16 bits, mostly involving cranking up the volume to highly unrealistic levels and trying to listen for the noise floor while pretty much nothing is going on. Keep in mind that the microphones and recording process are not noiseless, so in 16-bit tracks, there is usually going to be noise above any of these issues anyway. Getting a 96 kHz sampling rate may mean that there was less sample-rate conversion going on between the master and the version you are playing back, but that's not really an issue. I guess the main thing is losing out on any potential ultrasonic frequencies that you can't hear but your dog can. Most mics aren't designed to really capture too much outside of the range that humans hear, so there might not be a lot outside that range anyway. Take that as you will.
Music that tends to be released as 24 / 96 tends to be mastered better, to sound better on high-fidelity playback systems. That's a significant advantage, not really anything to do with the format itself (though many claim otherwise, for various reasons).
With a higher resolution per sample, the quantization error—difference between the actual value and the rounded-off version that you get—is decreased. If properly handled, the error can be handled and processed in such a way that it can be considered just a very small amount of noise, even for 16-bit samples.
With a higher sampling rate, higher frequencies can be represented, as per the sampling theorem. Due to limits of practical implementations and other issues, you can't quite get up to half of the frequency of the sampling rate for playback.
In practical terms, for further processing in the studio, you want 24-bit samples or higher so errors don't keep accumulating in a significant way. For home playback, it takes a very extreme case where 24 bits has an advantage over 16 bits, mostly involving cranking up the volume to highly unrealistic levels and trying to listen for the noise floor while pretty much nothing is going on. Keep in mind that the microphones and recording process are not noiseless, so in 16-bit tracks, there is usually going to be noise above any of these issues anyway. Getting a 96 kHz sampling rate may mean that there was less sample-rate conversion going on between the master and the version you are playing back, but that's not really an issue. I guess the main thing is losing out on any potential ultrasonic frequencies that you can't hear but your dog can. Most mics aren't designed to really capture too much outside of the range that humans hear, so there might not be a lot outside that range anyway. Take that as you will.
Music that tends to be released as 24 / 96 tends to be mastered better, to sound better on high-fidelity playback systems. That's a significant advantage, not really anything to do with the format itself (though many claim otherwise, for various reasons).
xnor
Headphoneus Supremus
- Joined
- May 28, 2009
- Posts
- 4,092
- Likes
- 227
The sampling rate determines the highest possible frequency that can be stored in the signal. For 44.1 kHz (Audio CD) theoretically frequencies up to below 22.05 kHz can be stored, for 96 kHz the limit is 48 kHz (always half the sampling rate).
Humans can hear up to about 20 kHz, so theoretically a sampling rate of 40 kHz would be enough but 44.1 kHz was chosen for audio CDs due to technical reasons.
The bit depth determines the range between the softest and loudest possible sound = dynamic range. 16 bit (Audio CD) with shaped dither yields an effective dynamic range of 120 dB. Only few DACs + amplifiers reach that, let alone the loudspeakers.
There's a thread in sound science somewhere on why bit depth above 16 bits don't make much sense for playback and there's also this article.
edit: Oh mikeaj was quicker and also explained it better.
Humans can hear up to about 20 kHz, so theoretically a sampling rate of 40 kHz would be enough but 44.1 kHz was chosen for audio CDs due to technical reasons.
The bit depth determines the range between the softest and loudest possible sound = dynamic range. 16 bit (Audio CD) with shaped dither yields an effective dynamic range of 120 dB. Only few DACs + amplifiers reach that, let alone the loudspeakers.
There's a thread in sound science somewhere on why bit depth above 16 bits don't make much sense for playback and there's also this article.
edit: Oh mikeaj was quicker and also explained it better.
sniperwill0
Head-Fier
- Joined
- May 6, 2012
- Posts
- 89
- Likes
- 11
Thank you both for your replies! I think I get some of how it works now. Do ya'll have any good analogies that can explain it further?
khaos974
Headphoneus Supremus
- Joined
- Mar 19, 2008
- Posts
- 2,085
- Likes
- 120
Quote:
Another example of sampling rate and bit depth.
- Screen resolution is a 2 dimensional sampling rate, horizontal and vertical, the point is that once you get past the "Apple Retina" resolution (~300 dpi), there's no visible progress any more. It's the same for audio. Peer reviewed academic literature suggest that 44.1 kHz is plenty enough, some audiophile tell anecdotes supporting the it is not.
- Bit depth is the number of colours, for example 16 bit for 65,536 colors, 24 bit for 16.7 million. Likewise there's little need to go over 24 bit in video. Again, peer reviewed academic literature suggest that 44.1 kHz is enough for reproduction purpose.
Another example of sampling rate and bit depth.
- Screen resolution is a 2 dimensional sampling rate, horizontal and vertical, the point is that once you get past the "Apple Retina" resolution (~300 dpi), there's no visible progress any more. It's the same for audio. Peer reviewed academic literature suggest that 44.1 kHz is plenty enough, some audiophile tell anecdotes supporting the it is not.
- Bit depth is the number of colours, for example 16 bit for 65,536 colors, 24 bit for 16.7 million. Likewise there's little need to go over 24 bit in video. Again, peer reviewed academic literature suggest that 44.1 kHz is enough for reproduction purpose.
sniperwill0
Head-Fier
- Joined
- May 6, 2012
- Posts
- 89
- Likes
- 11
Ok it makes more sense now, thanks!
Users who are viewing this thread
