MATLAB XPC TARGET 4 - IO Manuale Utente Scaricare il pdf (Pagina 24)

1 Introduction

1-18

4 Visit the benchmark page of the xPC Target Product area at

http://www.mathworks.com/support/product/XP/productnews/benchmar

ks.html

. Compare the complexity of the benchmark models to the

complexity of your application model. This enables you to estimate the CPU

class that you need to execute your target application at the given base

sample time.

Based on MathWorks experience, applications can be separated into the

following four distinctive classes concerning the base sample time you need to

achieve.

Base Sample Time Above 1ms

This sample time is usually not a problem. It leaves enough time to execute

even the most complex models with large amounts of I/O on almost any typical

CPU class. For complex hardware-in-the-loop applications with very complex

I/O, you should calculate and check the total I/O latency.

Base Sample Time Between 250 μs and 1ms

This sample time is usually not a problem. Consider the I/O driver latency only

if the application depends on high I/O complexity. Calculate and check that the

total I/O latency is within the relevant sample time. We recommend a

Pentium-class CPU running no slower than 200MHz.

Base Sample Time Between 50 μs and 250 μs

Total I/O latency and model complexity estimation are important. In most

cases, you will need a powerful CPU in the Pentium II/III or AMD Athlon class.

And if the I/O complexity is average to large, you might need to use advanced

I/O boards with low-latency drivers.

Base Sample Time Between 20 μs and 50 μs

Applications that need these very small base sample times are usually

controller applications with moderate I/O complexity. Nevertheless, I/O driver

latency has to be kept to a minimum by using advanced parallel converting

boards and parallel I/O access at the software level. Additionally, the algorithm

(Simulink model) has to be optimal, and the CPU performance should be the

highest available. Depending on the application, you may have to use the

polling execution scheme of the xPC Target kernel to achieve these small

1 2 ... 19 20 21 22 23 24 25 26 27 28 29 ... 51 52

Nessun commento

MATLAB XPC TARGET 4 - IO Manuale Utente Pagina 24