Each sample is then briefly stored and all the information on the signal between samples is then lost. The system which performs this task is called a sample and hold circuit, this is because it samples the original waveform at a given point, and holds that level until the next time that signal is sampled. (See figure 3).
Figure 4.
.
Taking exact measurements of the level of each sample is very difficult as the device used for measuring is not infinitely precise. This becomes a problem as it is rare for the sample level from the sample and hold circuit to land exactly on one of the levels in the measuring system (see figure 5). To overcome this, the measuring circuit rounds the level of each sample to the nearest value it knows, this procedure is called Quantisation. (See figure 6).
Figure 5 Figure 6 .
.
Since the values of the sample level are being rounded off, a quantity of error is being introduced and therefore the detail of the analogue waveform suffers. This is known as Quantisation error and is represented in the output signal as noise or distortion (Quantisation noise). In order to limit these errors more degrees of quantisation is essential to enhance the accuracy of rounding off, for example a 16 bit system has a total of 65,536 (216) quantisation levels available whereas an 8 bit system has only 256 therefore the 16 bit system will have less quantisation noise.
Dither is another process which is used for eliminating quantization error and it does it by adding white noise to the audio signal with a level that is half the Lowest Significant Bite (LSB). A random, constant, and low-level noise is then generated which effectively eliminates the distortion that are derived from low-level signals. The idea of adding noise to a signal which is affected by distortion seems like a counterproductive method to solve a problem.