Measuring ultrafast laser pulses

n0_3sc

I have an optical pulse train at 40MHz. The bright pulses are very broad (~25ns) and the dark regions are ~700fs - people usually refer to this as a "dark pulse train". I need ideas on measuring this 700fs 'dark pulse' ie. measuring the shape/dip of the dark section. My extinction ratio is very poor (~3dB) so nonlinear methods (eg. FROG/SPIDER/Intensity autocorrelation) is out of the question.

Thoughts?