In Quadrature second phase transition occurs in middle of first phase transition

Phase 2
------------| |--------------| |
|_____________| |__________|

t1 t2 t3 t4 t5 t6 t7 t8

phase 1
-----| |---------------| |---
|___________| |_____________|


|||| --->vibration or jittering

will occur around point t1 ,t3,t5,t7 etc for phase 1
But for phase it is around t2,t4,t6

Physical distance(liner or anglular) in encoder correcponding to t1-t2 , t2-t3 , t3-t4 should be greater then movement due to vibrations
considering encoder is producing 10 pulses per revolution and its circumference is 2 cm (20 mm)
then distance for/between t1 to t2 will be 0.5 mm
(wheel has to be divi into 20 zone , 1 zone = 1 mm
t1-t2 = half of zone )
Vibration could not go beyond 0.1 mm in worst case and if goes beyond that then encoder will less pulses/rev (or per mm) could be used.

Ans to Mr.Benjamin timed read
I have already suggested software based counting
ie reading pulse state after definite time interval (which will be much shorter then max pulse time interval).If Pulse max freq is say 1 khz (t = 1 msec between two low to high trans ) then scanning can be at the interval of say 250 usec.
even it read time = 4 * trans time then also there will not be any loss of range(i could not understand this) as transition time will be order of 100ns to 500 nsec or max i usec
If software counting is slow then 8051 t0,t1counter can be used, they scan state at t0,t1 inputs after each instruction cycle for updating counters and can provide good debouncing for very high freq noise/jittering pulses

Anil