DSP: the "quick calculator" of digital signals

DSP: the "quick calculator" of digital signals
3 min read
02 December 2022

DSP (Digital Signal Processing) is a digital signal processing technology. After receiving digitally processed real-world signals such as audio, video, pressure, temperature, and position, it will be processed mathematically. It can perform operations such as addition, subtraction, multiplication, and division very quickly.

You can understand the working principle of DSP in MP3 through the above figure. During the recording process, the analog audio signal is input through a microphone or other input sources, and then converted into a digital signal by the ADC (analog-to-digital converter) Industrial Router/Gateway 
and sent to the DSP. The DSP performs MP3 encoding and stores the file in memory; during playback, the file is converted from It is taken out of memory, decoded by DSP, converted into analog signal by DAC (digital-to-analog converter), and then output by audio system such as speakers.

DSP is unique in that it can quickly collect, transform, filter, compress and other operations on signals. And this is determined by its special hardware and software structure.

1. Harvard structure

As shown in the figure above, in the traditional Von Neumann structure, there is only one memory space through a set of buses (the address bus and data bus are connected to the core), and one multiplication will access the memory 4 times, which takes 4 instruction cycles. The DSP adopts the Harvard structure, which divides the memory space into a program area and a data area. There are two sets of buses connected to the core, allowing them to be accessed at the same time. This method can provide data and instructions for the core at the same time.

2. Pipeline operation

DSP adopts pipeline technology, and each instruction is fetched, decoded, fetched, and executed by multiple functional units in the chip, so as to reduce the execution time without increasing the clock frequency.

3. Zero Overhead Loop

DSP uses specialized hardware for zero-overhead loops. Zero-overhead loops mean that the processor does not spend time checking the value of the loop counter, conditional transfer to the top of the loop, and decrementing the loop counter by one when executing the loop, thereby reducing processing. time.

4. Fixed-point calculation

Most DSPs use fixed-point calculations instead of floating-point calculations; although fixed-point calculations are not as accurate as floating-point calculations, they are cheaper and faster, and DSPs support saturated calculations, rounding and shift.

5. Addressing mode

DSP chips support special addressing methods, which can improve the time of some signal processing algorithms and operations; for example, bit reverse addressing can improve the processing time of FFT (Fast Fourier Transform), Single serial port server and circular addressing can improve the processing time of digital filtering. The device delay line achieves a boosting effect.

6. Dedicated hardware multiplier

As shown in the figure below, for filtering, correlation, matrix and other operations that require a large number of multiplication and accumulation operations, most DSPs use independent multipliers and adders, so that the two operations of multiplication and accumulation can be completed in the same instruction cycle.

In case you have found a mistake in the text, please send a message to the author by selecting the mistake and pressing Ctrl-Enter.
chart loop 2
Ebyte est une entreprise nationale de haute technologie spécialisée dans la recherche et le développement de modules sans fil et de terminaux IoT industriels. L...
Comments (0)

    No comments yet

You must be logged in to comment.

Sign In / Sign Up