Vector VN8970 User Manual download pdf (Page 55)

7.2 The test

0.1 ms - for the better. However, halting the ECU during a measurement is not a

feasible option since this will have a too great effect on the real time properties.

An option could be to read a larger amount of data each loop, since the JTAG

does not seem to be the bottleneck, and sort out the different variables locally in

the dll. There is, however, a problem with this approach as well, since the data of

interest is spread out over an area of roughly 1.5 kB in the memory.

In order to try and locate where the time is consumed an attempt was done

by installing the Trace32 application on to the Vector real time device; the VT-

system more speciﬁcally since the VN8970 does not support remote desktop. It was

possible to install the application and drivers for the Lauterbach hardware on to the

real time system and it is therefore possible to run the solution entirely stand alone

since the debugger then is connected to the VT-system; the user PC does not do any

of the real time tasks. However, the time of reading the memory of the target CPU

was not shortened by a noticeable amount.

7.2 The test

The test performed on the ECU - with the application software on board, see Section

2.5 - was a test of the control software where the debug variables were extracted

from the CPU via JTAG and integrated from Trace32 in CANoe. The test consists

of several sub-tests where, for example, a certain variable or couple of variables

are tested so that it behaves as it should and it was proven that roughly 12% of the

sub-tests were able to pass. Some of the tests does not always pass when tested the

conventional way with monitor software on the ECU, which is why these speciﬁc

tests are somewhat disregarded when evaluating the validity of this solution.

There are, however, tests that consistently fails when using this solution, for ex-

ample some of the ”Quality Signal Test Cases”, see Figure 7.1. These tests have in

common a number of boolean variables that appear to ﬂuctuate rather randomly be-

tween true and false during some phases of the tests. This behaviour is inconsistent

when comparing with performing the tests the conventional way. Some research led

to the conclusion that the cause of this behaviour lies in the nature of how the fault

control of these variables is implemented in the software of the ECU. To give an ex-

ample one of the variables is called which is a quality ﬂag

of all the wheel speeds and the ﬂow of the fault control is basically implemented as

follows.

• Read wheel speeds

• Set to true.

• Fault control of the wheel speeds, set to false if fault

detected.

1 2 ... 50 51 52 53 54 55 56 57 58 59 60 ... 66 67

No comments

Vector VN8970 User Manual Page 55