Transformation Algorithm

Table of Contents

• Introduction
• Features Supported
• Features Not Supported
• Benchmark Results
• Provided Examples
• Additional References
• API

Introduction

The Transformation module consists of transformation commonly found in motor-control applications.

Features Supported

Transformation module includes:

• Clarke transformation
• Park transformation
• Inverse Park transformation
• Space Vector Generation (SVGEN)
  • Common-mode subtraction approach
• DPWM Generation (Part of SVGEN)
  • Maximum Modulation
  • Minimum Modulation
• SVGEN current reconstruction for single-shunt (SVGENCURRENT)
• Phase voltage reconstruction in overmodulation (VOLTS_RECON)

Features Not Supported

N/A

Benchmark Results

A benchmark on R5F core has been conducted to observe the following results when running the following functions:

Transform Function	Cpu Cycles
Clarke Transform
CLARKE_run_twoInput	16
CLARKE_run_threeInput	18
Park Transform
PARK_run	19*
Inverse Park Transform
IPARK_run	19*
Space Vector Generation
SVGEN_runCom	41
DPWM Generation
SVGEN_runMax	36
SVGEN_runMin	36
Phase voltage reconstruction
VOLREC_calcVolSF	30
VOLREC_run	90

Note

both PARK and IPARK does not include trigonometric calculations of phasor. For R5F core, it's recommened to leverage the fast ti_arm_sincos/FastRTS_sincos function found in mathlib of MCU+ SDK.

• Ran with TI Clang Compiler v3.2.0.LTS, with -Os flag and the benchmarking function placed in TCM memory. Obtained the average result from 600 consecutive loops of running transform functions with DPL CycleCountP, mimicking the usage of a control loop.
• Actual result may vary depending on provided datasets and memory configuration. For R5F, it is recommended for users to map control loops to TCM for the best performance.

Provided Examples

The following examples has been provided to demonstrate transformation module:

• CLARKE-PARK-ICLARKE-SVGEN Test

Additional References

N/A

API

APIs for Transformation Algorithms