The World Leader in High Performance Signal Processing Solutions

Model-Based Design For Motor Control Development

Paul Crook – Analog Devices

2

Topics

Developing motor control applications using model-based design and automatic code generation

Hardware platform Model-Based Design overview Modeling of motor control systems Simulation and automatic code generation Running generic C-code on embedded platformWhen to use MBD and when to use “traditional” design

methods

3

ADI Demo Platforms

Motor Drive hardware 3 phase PM motor PC with MATLAB®, Simulink®, and IAR Embedded Workbench

4

HW Platform

100V-250V AC input with PFC, 5A Isolated current, position, voltage and diagnostic feedback 3-phase, 0.55kW, permanent magnet synchronous motor

5

ADSP-CM40x Embedded Target

Key PointsARM CORTEX M4F240MHzTwo 16-bit ADCs

380ns conversion time

SINC filter384kB SRAM2MB Flash20 DMA channelsPeripherals for

motor controlCommunication

interfaces

6

MODEL-BASED DESIGN

7

Why Model-Based Design?

• Requirements Development• System Simulation• Automatic Code Generation• Continuous Verification

8

The Traditional Design Process

Requirements Phase

Design Phase

Realization Phase

Testing Phase

9

Traditional Design of a Multi-Domain System

Integration, Test & Certification

Research & Requirements

Software

Requirements

Design

Realization

Testing

Hardware

Requirements

Design

Realization

Testing

Mechanical

Requirements

Design

Realization

Testing

10

The Cost of a Traditional Design Process

Source: ROI for IV&V, NASA

11

Model-Based Design

• Automate regression testing

• Detect design errors

• Parameter tuning

• Support certification and standards

• Generate efficient code

• Explore and optimize implementation tradeoffs

• Model multi-domain systems

• Explore and optimize system behavior in floating point and fixed point

• Collaborate across teams and continents

INTEGRATION

IMPLEMENTATION

DESIGN

TE

ST

& V

ER

IFIC

AT

ION

RESEARCH REQUIREMENTS

ARM

C, C++

Environment Models

Physical Components

Algorithms

12

Model-Based Design for Embedded System Development

Designwith

Simulation

ExecutableSpecifications

ContinuousTest and

Verification

AutomaticCode Generation

Models

Designwith

Simulation

ExecutableSpecifications

ContinuousTest and

Verification

AutomaticCode Generation

ModelsTest with Design- Detects errors earlier

Simulation- Reduces “real” prototypes- Iterative “what-if” analysis

Automatic code generation- Minimizes coding errors

Executable models- Unambiguous- “One Truth”

13

Steps in MBD approach

1. Plant modeling- Motor, load, power electronics etc.

2. Interface modeling- Sensors, device drivers

3. Controller modeling- Field oriented control of 3 phase PM motor

4. Analysis and synthesis- The models created in step 1-3 are used to identify dynamic

characteristics of the plant model.- Tuning and configuration of system

5. Validation and test- Offline simulation and/or real-time simulation- Time response of the dynamic system is investigated

6. Deployment to embedded target- Automatic code generation- Test and verification

14

Steps in MBD approach

Simulation of Controller and Plant model Off-line development of algorithms without access to HW Code generation and deployment to embedded controller Comparison between simulation and actual implementation

PC

Target

Controller Model

Motor/inverterModel

System Model

TestSignals Verify

CodeGeneration

Motor Drive System

System

Embedded Controller

Motor/inverter

HW

15

Model complexity

What is a good model?Represent the states of interestSupports automatic code generationExecute fastUse existing (trusted) librariesReusable across platforms

Too little Too much

?

Right balance

5

16

Model complexity

Understanding strengths and weaknesses of tools is critical when defining model scope

Strength

Weakness

- Solving differential equation- Time domain modeling- Visualization- Target specific setup- Managing system resources- Time scheduling

- Register setup and control- Managing System resources- Time scheduling- Real time control systems- Debugging and test- Visualization

17

Interfacing to Embedded Target

SW architectureGeneric C-code needs to be tied to embedded platformUtilizing strength of multiple environmentsGraphical environment for application code development“Classic” C-code for target specific control

Control Algorithm

Embedded Platform

HW interface layerPl

atfor

m

inde

pend

ent

Platf

orm

de

pend

ent

18

Interfacing Embedded Target

Two platforms – one implementationCommon control algorithm for Simulation and Embedded systemPlatform specific device drivers or behavioral models

Debug algorithm using simulation and test on embedded platform

19

Drive system feedback and control

Feedback

20

Motor Control Application Program

Power Inverter and Motor

Sensors, interfaces and Device Drivers

MC Algorithm

MC ‘C’ codeDevice driversApplication code

CC & LinkMC ApplicationFirmware

Device Driver

Aut

oco

de

21

AC Motor Algorithm (Field Oriented Control)

Shaft position sensor measures rotor magnet (flux) position – velocity is calculated. Two/three motor currents measured (two is enough). Clarke-Park transfer calculate Torque (Iq) and Flux (Id) components of current. Speed loop generates Iq reference; Id reference set to zero for PMSM. Current loops and Field alignment generates a,b,c voltage commands.

22

Field oriented control

23

Simulation

0 1000 2000 3000 4000 5000-250

-200

-150

-100

-50

0

Frequency [kHz]

Mag

nitu

de

[dB

]

0 1000 2000 3000 4000 5000-3000

-2000

-1000

0

Frequency [kHz]

Pha

se [

Deg

rees

]

Simulink scopes Code debugging

Data analysis in MATLAB

24

Generating C-code

25

Building and Linking – IAR EWB

26

Building and Linking

Workbench

Matlab/Simulink

Motor Control- FOC functions

ARM Library- M3/M4 support

Executable

C-code

Device drivers- SW enablement package

Application code- State machines

Scheduling- IRQs/RTOS

C-code C-code

Device Drivers- Functional model

System resources- Allocation & setup

Legacy code- Existing code

CM40x

27

Running target

MATLAB GUI + embedded engineStreaming data back to MATLAB

0 20 40 60 80 100 120 140 160 180 200-1000

-500

0

500

1000

Dut

y-cy

cle

Captured runtime data

0 20 40 60 80 100 120 140 160 180 2002.5

3

3.5

4x 10

4

Pha

se c

urr

en

t

0 20 40 60 80 100 120 140 160 180 2000

2

4

6

8R

oto

r a

ng

le

Sample number

28

Sampling domains

10 kHz

1 kHz

29

Summary

Model-Based DesignSystem modeling and simulationAutomatic code generation

Modeling of motor control systemsElectronics/mechanicalInterfacesControl algorithm

Interfacing generic C-code to embedded targetUse of different environmentsDevice drivers

Debugging and testOff-lineIn real-time

30

Questions