Simulink Fixed Point 6.2

Fixed-point Optimization and Refinement Using Fixed-Point Tool

Learn how to convert a floating-point model into a fixed-point model and obtain an intial scaling for the fixed-point data types.

 

Contents

Demo Overview

The video shows how to apply float- to fixed-point conversions on models created using:

• Simulink®
• Stateflow®
• Embedded MATLAB™ functionality
  

Apply designs for:

• On-target rapid-prototyping
• Embedded systems
• Mass production

Generate code in:

• C
• C++

Target hardware for:

• Microcontrollers
• Digital signal processors
  

Exploring the Model

In a previous recording, we took a fault-tolerant fuel control system model and introduced an initial scaling for it using the Fixed-Point Advisor. We will use the resulting model and further refine and optimize the fixed-point scaling. Note that the Fixed-Point Tool can be applied for any fixed-point model, not just those converted using the Fixed-Point Advisor.

See how we explore the model.

Logging Original Data

We run the simulation once and log the minimums, maximums, and overflows of the model using the initial fixed-point scaling obtained using the Fixed Point Advisor.

See how we log the minimum, maximum, and overflows.

Data Type Override

We can use the data type override feature to convert all signals into floating point and store them as our golden reference results. This will be used to compare with results obtained with our new fixed-point scaling later on.

See how we convert all signals to floating-point and log the golden reference results.

Automatic Scaling

Using the floating-point results, the Fixed Point Tool proposes fixed-point scaling for signals of interest. We can add in safety margins and choose to accept or ignore the proposed scaling.

See how we use automatic scaling to obtain more optimized fixed-point scaling.

Compare New Results

We can plot the comparison between the new fixed-point scaling and the golden reference floating point results. We can adjust the safety margin and play with individual scaling as we go through several more iterations to obtain the desired performance.

See how we plot the new fixed-point results with the original floating-point golden reference results.

Additional Resources

Advanced Fixed Point Tips Document and Models

• Tips for Fixed-Point Modeling and Code Generation for Simulink 7 (R2008a)
  

Technical Evaluation Kit

• Request for the Real-Time Workshop® Technical Kit

Related Video Demos

• Float- to Fixed-Point Conversion Using Fixed-Point Advisor
  

Related Demos

• Simulink Fixed Point™ Demos
• Real-Time Workshop Demos
  
• Real-Time Workshop® Embedded Coder™ Demos
  

Products Used in this Demo

• MATLAB®
• Simulink
• Stateflow®
• Simulink Fixed Point
  
• Real-Time Workshop
• Real-Time Workshop Embedded Coder
  
• Stateflow Coder™