You are on page 1of 2

Quantization in Sound Processing The process of converting, or digitizing, the almost infinitely variable amplitude of an analog waveform to one

of a finite series of discrete levels. In video compression, quantization is a process that attempts to determine what information can be discarded safely without a significant loss in visual fidelity. The quantization process is inherently lossy because of the many-to-one mapping process
quantization (2012) Webopedia. [Online]. Available from:http://www.webopedia.com/TERM/Q/quantization.html [Accessed 18/01/2012].

What is Quantization in Sound Processing?

Quantization in sound processing is mostly the same stage as if it would be in any DSP system. Once the sampling is completed, the resulting series of values; which are usually float values with infinite decimal length are mapped into a more manageable numeric space. Quantizing is defined by the depth and numeric space of quantizing. For instance an 8 bit quantization has 256 values that the sampled wave must be matched to. Compact disks on the other hand use 16bit quantization resulting in 65,536 levels to which the sampled wave is to be mapped to. Why is quantizing necessary

1. The sampling phase outputs values which are somewhat boundless and not of any numeric range. Therefore, they should be brought within a manageable numeric range 2. The values output from the sampling phase usually contain lots of decimal values. If it were to store those values as decimal, it would take a lot of disk space. Therefore, it should be rounded to the closest possible value that fits within out numeric space. The Quantizing error is the difference between the original sampled waveform and the quantized series of values. In most cases, the quantizing error results in a small noise that is uniformly added to the resulting set of values. It is important to note that the higher the bit depth of the conversion, the lesser the noise percentage is. What is dithering?

Dithering is a common technique used in improving the quantization of a slowly varying signal by adding a small amount of random noise. This is most relevant when the signal strength is very low and is only a fraction of the sampling bandwidth. The added noise makes the digitized signal toggle between adjacent quantizing levels giving more definition to the digitized signal.

Example of Quantizing

In the above graph 2, the sampled curve has values ranging for +3V to 3.025v in decimal form. In the quantized graph, the voltages are rounded and mapped to values from 3000 to 2025 in integer form. The above conversion creates a small quantizing error which can be seen in the below graph and is less than 1% of the actual range.

Example of Dithering

As mentioned earlier, the small noise added during the quantizing stage had given a more definition to the quantized curve.
References http://www.audiosonica.com/en/course/post/218/Digital_Audio-Quantization http://www.dspguide.com/ch3/1.htm http://logix4u.net/tutorials/20-dsp-tutorial

You might also like