The following excerpt was taken from section 12-306.3 of the 1995 Idaho Transportation Department Traffic Manual.
Cycle length is composed of the total signal time to serve all of the signal phases including the green time plus any change interval. Longer cycles will accommodate more vehicles per hour but that will also produce higher average delays.
The best way is to use the shortest practical cycle length that will serve the traffic demand. Vehicles at a signal installation do not instantaneously enter the intersection. Early studies by Greenshields found that the first vehicle had a starting delay of 3.7 seconds to enter the intersection with subsequent vehicles requiring an average of 2.1 seconds each. Generally, vehicles will pass over an approach detector with a headway of 2 to 2.5 seconds. For general calculation purposes, an average time of 2.5 seconds per vehicle to enter the intersection is a conservative value. This value can be used to estimate signal timing for planning purposes.
The cycle length includes the green time plus the vehicle signal change interval for each phase totaled to include all signal phases. A number of methods have been used to determine cycle lengths as outlined in the Highway Capacity Manual, ITE Manual on Traffic Signal Design, and ITE Transportation and Traffic Engineering Handbook. Webster provided the basic empirical formula that would minimize intersection delay as follows:C = (1.5*L + 5)/(1.0 - SYi)) Where:
C = optimum cycle length in seconds adjusted usually to the next highest 5 second interval. Cycle lengths in the range of 0.75C to 1.5C do not significantly increase delay.
L = Unusable time per cycle in seconds usually taken as a sum of the vehicle signal change intervals.
SYi = critical lane volume each phase/saturation flow
The saturation flow will be between 1500 and 1800 vehicles per hour. Refer to Highway Capacity Manual. The "Y" value should be computed for each phase and totaled to arrive at SYi for all phases.
Note: The traffic volumes used should be the predicted volumes at time of signal turn-on. The volumes should also be the peak hour or peak fifteen-minute period for the cycle determination.
When the cycle length has been determined the vehicle signal changes are deducted giving the total cycle green time which can be proportioned to each signal phase on the basis of critical lane volumes. The individual signal phase times are then the proportioned time plus the vehicle change interval on each phase.
To ensure that critical lane volumes are adequately served, a capacity check should be computed for each green interval.