AN3288

Firmware design

Doc ID 18011 Rev 2

11/27

During the last state, the MCU varies the PWM signal duty cycle only once the event occurs

(braking effort higher than activation level), and the detected deceleration value is higher

than the activation threshold and bounded in a threshold range (see Section 3.3).

The state machine timing has a fundamental importance for synchronizing the data

acquisition from the accelerometer with the data elaboration process. In fact, the state

machine clock has been tailored to acquire a new value, calculate the average with the

previous deceleration values and actuate the light response, all in the same clock cycle. As

the time between two subsequent acquisitions represents a limiting condition for the clock

setting, the state machine clock rate is fixed to 40 Hz by the accelerometer acquisition rate.

Starting from this assumption, the clock rate has been fixed to come to an arrangement

between the MCU performances and the acquisition rate, considering that for a higher

sensor output data rate there is a higher noise introduced in the system.

3.2

Digital filtering

As anticipated, the filtering process is necessary to clean the deceleration signal captured

by the accelerometer AIS326DQ and overlapped to the vibrational noise produced by the

oscillations of a vehicle's mechanical parts. The choice of a suitable digital filter has been

made considering also that it inserts latency in the data elaboration process and this latency

depends on the complexity level of the filter.

A moving (running)-average filter is optimal for this task, mainly for its simplicity to be

implemented and integrated in the data processing flow; the latency of this filter is relatively

low in comparison with other FIR filters, and depends on the samples number to be

averaged out. Before designing the filter, it's important to have a complete modeling of the

vehicle’s dynamic characteristics to understand how the noise affects the acceleration signal

and identify the right number of samples for processing.

After a preliminary analysis on the test vehicle characteristics, a moving (running)-average

filter able to elaborate 10 samples per cycle has been designed. For each cycle, it computes

the average value in an array of 10 elements, subtracting the maximum and the minimum

values.

for( k=0; k<10; ++k)

{

if(k!=9)

Sample_Vector[k] = Sample_Vector[k+1];

else

{

...

/* when new data available, copying in the buffer vector */

Sample_Vector[9] = AIS326DQ_Read_1Axis(OUTX_L);

}

subtotal += Sample_Vector[k];

/* calculation of the max and the min in the buffer vector */

if( Sample_Vector[k] < min ) min = Sample_Vector[k];