Department of Electronics and Telecommunication Engineering

Report of Data Signal Processing

Activity: Lecture by Prof. Ranjushree Pal

Title: Data Signal Processing Seminar

Date: 21st September, 2018.

Number of students participated: 55-60

Year: T.E.


As a student of Electronics and Telecommunication Engineering, one can play with signals which are analog or digital in nature.

A lecture on ‘Architecture of advanced DSP processors and implementation of a few final year projects in MATLAB and Simulink’ was conducted on 21st of September 2018 from 3:30 to 5:30 P.M. by Mrs. Ranjushree Pal.

Digital signal is a signal that is being used to represent data as a sequence of discrete values; at any given time it can only take on one of a finite number of values. This contrasts with an analog signal, which represents continuous values; at any given time it represents a real number within a continuous range of values. Digital signal processing (DSP) refers to various techniques for improving the accuracy and reliability of digital communications. The theory behind DSP is quite complex. Basically, DSP works by clarifying, or standardizing, the levels or states of a digital signal. ADSP circuit is able to differentiate between human-made signals,which are orderly, and noise, which is inherently chaotic. A discussion on DSP processors led to the conclusion that it is a processor designed specifically for digital signal processing. Further, these are very important for Embedded Systems (E. S.), since E.S. are targeted for ‘situated computing’ i.e. an E.S. is situated in an external environment and sensors provide input about external environment and input signal is processed by the E.S. Most of the task performed by DSP processors require constant and real time processing along with a high memory bandwidth. Hence, they must perform these tasks while minimizing cost, power consumption and memory usage. Some common tasks include time-frequency transformations and vice-versa, filtering of noise and data correlation.

This lecture informed us that over the years, many processor architectures have been developed like the Von Neumann architecture and the Harvard architecture. The Harvard architecture is based on the Von Neumann architecture but it also includes separate memory and bus for data and program. It also makes instruction pipelining easy. Hence it is faster at performing tasks. Because of the computational complexity, most digital signal processors have a MAC (Multiply and accumulate) instruction to increase the speed of theprocess. It is further improved by introducing a MACD instruction which performs all the task of MAC along with data move.

Discussions upon memory access in DSP processors led to a conclusion that memory is main reason for the slow speed of the data computation. Since majority of the Signal Processing operations require multiple memory access, research is carried on as to how to increase the number of memory access per clock cycle. This is done by using Multiple Access Memory and Multi-ported memory. Both are interfaced with the processor using Harvard architecture to speed up the memory access.

Another type of architecture discussed is Very Large Instruction Word Architecture in TMS320C6X. These DSPs have a number of functional units like ALUs, MAC units, shifters etc. The VLIW instruction is accessed from the memory and is used to specify the operands and operations to be performed by each functional unit.

Instruction pipelining helps to increase processing speed by executing multiple instructions in the same clock cycle. The number of instructions that can be processed simultaneously in the CPU is known as the depth of pipeline. e.g. TMS320C67XX has a pipeline depth of 11.

DSP processors possess certain special addressing modes. The two addressing modes that are different than general purpose microprocessors are Circular Addressing mode and Bit ReversedAddressing mode. Bit reversed addressing mode is very useful to compute the FFT of a given signal.

Applications of DSP processors include but are not limited to

  1. Audio coding, decoding and decompression

  2. Conversion of speech signal into digital data for transmission

  3. Robotics and disk drive control

  4. Data processing of Bio-medical sensors

  5. RADAR and SONAR data processing

Implementation of some projects in MATLAB and Simulink based on Digital Signal Processing was discussed in detail:

  1. Automatic Speaker Recognition

  2. Music Synthesis with audio effects

  3. Low frequency Noise Cancellation

  4. Unknown Channel Identification

  5. DTMF detection using Correlation, FFT

  6. Encoded Voice Signal in BPSK

  7. Time Frequency Analysis of Signals with spectrogram

  8. Acoustic direction Tracker

Automatic Speaker Recognition

Encoded Voice Signal using BPSK

Photographs of events: