How is digital audio processing done?
Digital audio is processed by mathematical operations applied to individual samples of a signal, or to groups of samples of different lengths.
The mathematical operations performed can simulate the work of traditional means of analog processing (mixing of two signals – sum, amplification / attenuation of a signal – multiplication by a constant, modulation – multiplication by a function, etc.), or use alternative methods – for example, decomposition of a signal into a spectrum (Fourier series), correction of individual frequency components, then inverse “assembly” of the signal from the spectrum.
Digital signal processing is subdivided into linear (in real time, on a “live” signal) and non-linear, on a pre-recorded signal. Linear processing requires sufficient speed from the computer system (processor); in some cases it is impossible to combine the required performance and quality, and then simplified processing with reduced quality is used. Non-linear processing is not limited in time, therefore computing facilities of any power can be used and the processing time, especially with high quality, can reach several minutes or even hours.
For processing, both general-purpose processors (Intel 8035, 8051, 80×86, Motorola 68xxx, SPARC) and specialized digital signal processors (Digital Signal Processors, DSP) are used Texas Instruments TMS xxx, Motorola 56xxx, Analog Devices ADSP- xxxx, etc.
The difference between a general-purpose processor and a DSP is that the former focuses on a wide class of tasks: scientific, economic, logical, gaming, etc., and contains a large set of general-purpose instructions, in which common mathematical and logical operations prevail. DSPs are especially focused on signal processing and contain sets of specific operations: limiting addition, vector multiplication, calculation of mathematical series, etc. Implementing even simple audio processing on a universal processor requires significant performance and is far from always possible in real time, whereas even simple DSPs often cope with relatively complex real-time processing, and DSPs powerful are capable of processing high-quality spectrals of several signals at the same time.
Due to their specialization, DSPs are rarely used independently; Most of the time, the processing device has a universal average power processor to control the entire device, receive / transmit information, interact with the user, and one or more DSPs to process the audio signal. For example, to implement reliable and fast signal processing in computer systems, specialized boards with DSP are used, through which the processed signal is passed, while the central processor of the computer has only control and transmission functions. .
