Dithering is the process of adding a low level noise to audio, almost a hiss. This is used to
slightly vibrate the audio so as to allow less significant sounds to still be heard. When i say less
significant i mean smaller softer sounds with very small amplitude. the 16 bit recording process was
a good standard when media was expensive and data storage was not as readily available.
Supposedly 16 bits has a 96 dB range which is all well and good, if you could use all 16 bits all the time,
but you cant or you will experience another type of distortion. the problem is that when you record a song
at 16 bits, it lines up so that the highest peaks, of amplitude, are what hits 16 bits which is what needs to
happen, otherwise you would start to experience clipping and foldback distortion etc.
So now we have a digital recording of our song in 16 bits lined up with the peaks at 16 bits. Now
the smaller sounds have much less effect and the smaller they get the more distorted they become. so
if the loudest sounds are played at 16 bits, then the smallest could be 1, when you think of this on a
wave that turns it into a square wave which will not even slightly resemble the original wave. this is where
dithering comes into play, when a very low level noise is added its like adding random up and down
movements along the wave. this smaller, usually less than 1 bit, noise can slightly vibrate the lower noises
allowing them to go from 1 to maybe 2 bits in some places or maybe 0 in others. this slight vibration
allows the human ear to pick up these changes much easier, which allows us a clearer perception of the
sounds.
The dithering algorithm can be changed in many ways, the most common of which is having the
noise fluctuate along the waveform, adding more noise when needed. this process keeps the noise out of
areas where its not necessary and puts it where it is needed.
Dithering is mainly used and only really needed for compressing audio from a higher to lower bitrate.
the reason for this is that when compressing audio from say 24 to 16 bits, sounds that used to be at say 5
bits will now be at like 4 but many of the pieces at 4 will remain at 4 so we lose our differentiation. But when
you dither the audio you spread this distortion out ofer the majority of the file. the noise basically helps the
sounds that are now 4 bits instead of 5 differentiate themselves from the 4 bit sounds that are supposed to
be 4 bits.
The main reason for dithering is when you record from an analog signal, virtually an unlimited bitrate,
to say a 24 or 32 bit file, you have decreased the bitrate. The sound is then less accurate than it was and has
been sort of melted together or rounded off. rounding off is one of the worst ways to distort any audio stream,
sounds just sort of collapse on each other. Gain is also a huge culprit to the need for dithering, gain multiplies
the signal, or divides, whichever direction you are going in to raise or lower the volume. when you multiply two
16 bit values you come out with a 32 bit value, but alot of the time there are slight remainders which do not
show through, hence the need for dithering.
One of the big mistakes made with dithering is that people try to normalize each track as they put them
together, this leads to excess distortion, and the need to redither. you must wait until you have finished production
until you dither.
Dithering is not entirely necesary, it is merely something we need because of the medium we rely on.
In a 24 bit system, 32 bit floating point, there is no need to dither because at 24 bits, "the least-significant bit is
equivalent to the amplitude of noise at the atomic level?no sense going further."
(http://www.earlevel.com/Digital%20Audio/Dither.html)
24 bit recording has become the industry standard in any type of recording. 16 bits was good but being
over a 20 year old technology its just not good enough anymore. an audio waveform is defined by amplitude over
time, amplitude being the bit number that it peaks at, and frequency being the pitch.16 bits was the outcome from
negotiations between audio quality and storage space. the average cd hold around 650-700 MB, this allows for 74-80
minutes of 16 bit audio. this is where the industry decided to standardize audio production and playback, a
minimization of cost while maximizing audio quality and duration on a storage medium. the problem is that the 16 bit
recording doesnt quite capture all of the detail of an analog instrument. Say an instrument is recorded at a low volume,
this restricts the number of bits that can be used to represent it, in a 16 bit system a sound that is around say 3 bits,
1 would be silent, 2 would be audible, and 3 would be as loud as it can get. this is much too much quantization, and
would sound like a square wave distortion. this is once again where dithering comes into play.
dithering can be though of as an analytic integration which smooths out the squares. it is like adding a dynamic
second derivative to the equation, instead of having a steady slope in between two point on the audio map, the dither
creates a sort of gradual slope change, giving the audio a much better resolution.
http://www.24bitfaq.org/#Q0_1_1
(http://www.earlevel.com/Digital%20Audio/Dither.html)
