Power-Window Control

1
Power-Window Control

• A design exploration using Felix VCC
• Claudio Pinello
• EE249 Fall 1998 Prof. Alberto Sangiovanni-Vincente
  lli
• Mentor Dr. Alberto Ferrari

2
Outline

• Motivation
• Design description
• Possible architectures
• Simulation results
• Conclusions

3
Motivation

• Test the Felix design methodology on a real
  industrial example
• Evaluate different architectures and define the
  level of details needed to carry out the
  estimation correctly
• communication refinement
• interrupt handling

4
Power Window Control

• Window actuated by electric motor
• programmable logic de-couples motor commands form
  input switches
• automatic stop of motor at the end of the run
• conflicts resolution for multiple inputs
• glitches filtration (de-bouncing)

5
Behavioral Description
6
Controller Hierarchy
7
Controller Policy
8
Candidate Architectures
9
Candidate Architectures
Remote
Local
This is obviously a superset of the previous
one more electronics, less wiring (especially
for symmetry)
10
A First Mapping
No events lost low cpu utilization
11
Load Models

• CPU load model
• one asynchronous task, constant load (Poisson
  mean inter.time 8ms, load 700 cycles)
• one periodic task, random load (period 4ms, load
  normal distr. mean700 var. 100 cycles)
• CAN-BUS load model
• two Poisson streams of integers (mean inter.time
  30ms and 40ms)

12
Load Models
13
Gantt Charts
Behavior I_25 has higher priority but preemption
is disabled
14
Gantt Charts
Behavior I_25 has higher priority and preemption
is enabled
15
Communication Refinements
Interrupt handler overload
16
Simulation Diagram (single mP)
17
Simulation Diagram (double mP)
18
Simulation Results
Processor model Motorola HC11 at 10Mhz clock speed
No difference in cpu utilization in 15 seconds
the number of switches operations is very
limited However the time to react to a command
increases due to the communication over the
CAN-BUS (125000bps)
19
Future Work

• Relate reaction time to actual position change
• Evaluate different inputs reading techniques
• polling vs. interrupt
• Add functionality to the controller
• interaction with power door-locks
• interaction with alarm and air-conditioning
  systems

20
Conclusions

• A model of the cpu and bus load has been
  developed
• Two mappings
• the centralized architecture has a shorter
  reaction time
• the distributed architecture reduces greatly the
  wiring across the car