It's described briefly here and elsewhere:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.385.6164&rep=rep1&type=pdf
It says that if you integrate the 2D time-delay signal plot for SHG FROG over delay, you get a guess of the field. This simply can't be true. The 2D FROG plot (freq-delay) has a peak at 2x the frequency of the pulse you want to measure (pic related). Integrating over delay doesn't magically downgrade the frequency by a factor of 2, regardless of whether or not you do it before or after IFFTing the frequency axis to time.
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.385.6164&rep=rep1&type=pdf
It says that if you integrate the 2D time-delay signal plot for SHG FROG over delay, you get a guess of the field. This simply can't be true. The 2D FROG plot (freq-delay) has a peak at 2x the frequency of the pulse you want to measure (pic related). Integrating over delay doesn't magically downgrade the frequency by a factor of 2, regardless of whether or not you do it before or after IFFTing the frequency axis to time.
