Samplerates & Bitdepths

   

Sample rates and bit depths have been heavily debatted subjects during the last years. But it´s probably not all who fully undertands the consequence it has when choosing a higher samplerate or a greater bit depth.

44.1 kHz and 16 bits are keywords in this context, and most people with an interest in digital audio have probably heard of these numbers and terms before. This is the most common sample rate and bit depth, which is also the standard for most commercial audio CD´s. It could change during the next few years as many older CD players are being replaced by new DVD/DVD-R machines often capable of handling higher audio relsolution.

44.1 kHz
The matematician Harry Nyqvist discovered that in order to record a sound at acceptable quality, you have to sample it at twice the frequency. The technical explanation is: in order to sample a signal, the sampling frequency must be greater than twice the bandwidth of the input signal in order to reconstruct the original perfectly from the sampled version.

Since most of us are capable of hearing from approx. 15 Hz to 20 kHz, 44.1 kHz was chosen as a suitable standard. This means that the audio signal is being sampled 44.100 times every second. It would have been enough to sample 40.000 times as we can hear frequencies going up to 20 kHz, but it is necessary to apply a cut-off filter for signals exceeding the max value; that´s the reason for this extra "headroom".

44.1 kHz is a fine and well-defined resolution, that we all know as "CD Audio Quality". Of course, a sound sampled at 96 kHz contains much more information, and will deliver an even more detailed result.

16 bits
As explained the samplerate determins how often we sample a sound. The bit depth determins how accurate a scale we use. Let´s say that I want to buy a new car. Now I would like to find out if it will fit into my garage. As I don´t want to fiddle around with an "analogue" measurement-tool, so I have borrowed a digital measurement-thingy that uses a lightbeam to measure distances. If this was set to a low resolution, for example to the nearest foot, a 14½ feet wall would be measured to just 14 feet, or maybe 15 feet; not very accurate. If the resolution had been set to nearest inch or higher we would get a much more accurate result.

Bits (BInary digiTS) are simply the digits 0 and 1. A 2-bit system would represent just 4 values: 00, 01, 11 and 10. A 4-bit system: 0000, 0001, 001, 0111, 1000, 1100, 1110 etc. The combination posibilities raises exponentially, so in total there is 16 different values.  

When it comes to digital audio this bit depth is very closely related to the dynamic range. If we take the 2-bit system with just 4 values it would translate into just 4 different dynamic levels. Not very accurate, especially considering that 2 of those 4 levels would be "max volume" and "complete silence". The lowest audible level would be a quarter of the full level!

8-bit resolution, which was an acceptable standard in the early days of digital recording, gives 256 dynamic levels. However, this bit depth translates into a dynamic range of just 48dB. The 16-bit Audio CD standard offers 65.536 different levels, which gives a dynamic range of 96 dB. This is much better, but knowing that the human hearing covers 140 dB of dynamic range there is still room for improvements.

If we take it up to 24 bits we will get 16.777.216 different values, offering a dynamic range of 144 dB, which should be enough for even the most dynamic recordings such as acoustic or classical music.    

 

 

But it ends up in 16 bits anyway !
What do you gain from a 24 bit recording if it has to be dithered to 16 bits anyway? It is correct that todays standard is 44.1kHz/16bits, which means you will have do downgrade in the end. Let´s return to the 16 bit recording for a moment. This resolution offers 65.536 different dynamic values, so a signal representing an output value of, say, 65.600 will be cut down to the maximum 65.536 value. This results in a socalled "Digital Clipping", which sounds like pops and crackles in a most un-musical way. Nothing like subtle analogue saturation which (under the right circumstances) can be both nice and musical.

To avoid digital clipping you will have to set the gain very carefully, and leave headroom for any transient peaks. This method has 2 major drawbacks. First of all you don´t use all 16 bits, more likely just 13 or 14 of them. This will reduce the actual dynamic range to 78-84 dB rather than 96 dB. Secondly, by setting a lower gain, you will raise the background noise, when normalizing the audio file after the recording in order to compensate for the lost gain.

This gives at least three good reasons to record in 24 bit:

1) After dithering to 16 bits, you actually use the full dynamic range of 96 dB.

2) You don´t have to watch your peak meters like a hawk while recording.

3) You will bring down the total noise floor considerably.

Tip: As there are good arguments for larger bit depths we recommend recording in 24 bits. The sample rate is more tricky. You will get a more detailed sound, but you will not benefit as with the bit depth. A 96 kHz sample rate converted to 44.1 is not sounding better than one recorded at 44.1 kHz.  

 

Drawbacks of high resolutions
Unfortunately there are drawbacks you need to consider before recording in a higher resolution. For one thing your audio files will grow quite a bit. A 24 bit audio file takes up 50% more space than similar a 16 bit file, and by doubling the sample rate you will also double the size of the file. A 44.1 kHz / 16 bit file takes up approx. 10 MB per minute. Raise the bit depth to 24 and it takes up 15 MB, and if you raise the sample rate to 96 kHz it will take up more than 30 MB per minute.  

However, with todays massive hard disks this is not a very big problem. It´s worse that files 3 times larger will decrease the number of  simultaneous audio tracks, and that you will require 3 times more RAM in order to perform at the same level. Finally you will require a powerful computer CPU if you want to run many plug-ins and virtual instruments simultaneously.

At the end og the day we think it´s nice that hardware and software producers constantly strive for the best possible audio quality. Even though the trend seems to go in the opposite direction in many other ways in the music industry. Mp3, OggVorbis, WMA,  AAC etc. But that´s a totally different story.....