The transmission-dependent convolution subtraction (TDCS) technique is a promising technique in quantitative SPECT to subtract the scatter components from emission images. Usually a 20% photo-peak energy window is used in SPECT acquisitions. So far no investigation has been done to investigate the effect of energy windows to subtract the scatter components from emission images with the TDCS technique. To evaluate the energy dependency of the TDCS technique, SPECT acquisitions were performed for the photo-peak energy windows of 5%, 10%, 15%, 20%, 25%, 30% and 35% by acquiring the data of nonuniform and uniform phantom with 99mTc radionuclide. The parameters of the scatter fraction to be used in the TDCS technique were estimated with the point sources placed in the water phantom. The estimated parameters were used in the TDCS technique to subtract scatter components from the data of corresponding energy windows of non-uniform and uniform phantom. Poisson noises were added to each projection's data of uniform phantom before scatter corrections. All data were reconstructed with the OSEM algorithm. The energy window dependence of the TDCS technique was verified by comparing the coefficients of variance (COV) of the data of the uniform phantom of each energy window. The non-uniform phantom results showed that the estimated parameters effectively estimated the scatter component from the projection images for each energy window. The COV for 5%, 10%, 15%, 20%, 25%, 30% and 35% energy window were 2.10%, 1.76%, 1.44%, 1.39%, 1.34%, 1.37% and 1.61%, respectively. The COV in the 25% window was lower than the COV of other energy windows. In the TDCS technique the photo-peak energy window 25% or 30% may be more effective to obtain emission images for higher statistics of counts, which may make scan time shorter.