The correlations among three important parameters in SPIDER, the time delay τ between the test pulse replicas, the frequency shear Ω and the second-order dispersion φ determined from a stretcher, are investigated. The results show that there are different optimum time delays τ for different test pulses with a fixed φ. The optimum time delays τ will increase as the duration of test pulse gets longer. However, the measurement may fail if the frequency shear Ω, which is direct proportional to the time delay τ, becomes too large to meet the condition of the sum-frequency. Therefore, for different widths of pulses, the magnitude of φ should be properly adjusted for selecting the optimum τ.