sorry, forgot the √3 divisor in the velocity! BTW, that is just a guess, you need to design the impedance of that stub (microstrip), and the actual εr varies with line width. You WILL get better results with a Rogers type board--lower loss.
I did not read the full paper, just looked at the schematic. The SRD is a device that conducts current in the forward direction forever, but conducts current in the reverse direction just momentarily (until the electrons are swept out of its I region). They are probably countin on this effect to let the rising edge of the pulse to go through, but make the falling edge of the pulse happen sooner than the input would suggest. I am not sure exactly how well this circuit would work...
As far as the mmic amplifier following it, to maintain the pulse shape, you will want a very broadband amplifier, as a 1st step. The 5th harmonic of 1/300 ps period is 10 GHz, not 3 ghz.