Embedded Signal Processing with the Micro Signal Architecture

Chapter 6

Real-Time DSP Fundamentals and Implementation Considerations

In Chapter 5, we introduced the architecture of the Blackfin processor and moved data between memories and registers. This chapter introduces important topics on representing data in the fixed-point Blackfin processors, processing and handling digital signals with finite wordlength, segmenting signal samples into blocks for processing, and evaluating the resources (speed, memory, peripherals, and power management) for real-time tasks. We use many examples and hands-on experiments to explain these topics and show a step-by-step approach in designing an embedded signal processing application. This effective embedded system design moves from floating-point simulations to fixed-point representations using MATLAB and then to porting the fixed-point code and data into the Blackfin processor. We also introduce more advanced debugging and profiling features in the VisualDSP++ IDDE to meet real-time requirements.

6.1    NUMBER FORMATS USED IN THE BLACKFIN PROCESSOR

This section defines number representations in digital systems with different formats and performs different arithmetic operations based on these formats. In particular, we use the Blackfin processor that supports 8-, 16-, 32-, and 40-bit fixed-point data to explain different number formats.

6.1.1    Fixed-Point Format

In fixed-point processors, a number is represented with a series of binary digits of 1s and 0s. An unsigned number without sign information ...