DSP 28335 教程 9.ppt

DSPx m Numerical Concepts Module 8 TMS320C28x? MCU Workshop Learning Objectives Integers and Fractions IEEE-754 Floating-Point IQmath Format Conversion of ADC Results Integer Basics What is Sign Extension? When moving a value from a narrowed width location to a wider width location, the sign bit is extended to fill the width of the destination Sign extension applies to signed numbers only It keeps negative numbers negative! Sign extension controlled by SXM bit in ST0 register; When SXM = 1, sign extension happens automatically Integer Multiplication (signed) Fraction Basics Fraction Multiplication Coding Traditional 16-bit Q15 Fractions C-code example: y = 0.707 * x Integer vs. Fractions Integers grow when you multiply them Fractions have limited range Fractions can still grow when you add them Scaling an application is time consuming IEEE-754 Single Precision Floating-Point Number Line Insight Using Floating-Point Should be using a C28x device with hardware floating-point support! Add the floating-point RTS library(s) to the CCS project standard RTS lib (required) rts2800_fpu32.lib comes with compiler fast RTS lib (optional) C28x_FPU_FastRTS.lib on TI web, #SPRC664 improved performance Strongly Recommended Select ‘fpu32’ support in CCS project options Getting the ADC Result intoFloating-Point Format Floating-Point Pros and Cons Advantages Easy to write code No scaling required Disadvantages Somewhat higher device cost May offer insufficient precision for some calculations due to 23 bit mantissa and the influence of the exponent IQ Fractional Representation Number Line InsightDistributions Both floating-point and IQ formats have 232 possible values on the number line It’s how each distributes these values that differs Traditional 32-bit “Q” Math Approachy = mx + b 32-bit IQmath Approachy = mx + b IQmath ApproachMultiply Operation IQmath ApproachIt looks like floating-point! IQmath ApproachGLOBAL_Q simplification IQmath Provides Compatibility Between Float