Fix issue introduced by 'Fix frequency computation when there is no more signal'
It happens that frequency was measured at 0 whereas signal was present.
It is due to the fact that rollover is not abnormal:
timer is never reset to avoid missing some cpu cycles in the computation.
Loss of signal is detected by 2 consecutive rollover without rising edge.
Fix issue introduced by 'Fix frequency computation when there is no more signal'
It happens that frequency was measured at 0 whereas signal was present. It is due to the fact that rollover is not abnormal: timer is never reset to avoid missing some cpu cycles in the computation. Loss of signal is detected by 2 consecutive rollover without rising edge.