I have some curious results using a spectrum analyzer (TinySA Ultra) to see the spectral content of the generator's signal.
The interesting part is comparing it with another generator. I have assembled a quick and dirty pulse generator, or rather a pulse shaper, made of a single
1EDN7512 high-speed MOSFET driver.
I wasn't going to use it in this role, but I noticed that it had about 300 ps rise time (at a 0..15V transition) in spice simulation without load, so I thought hmm why don't I make a pulser out of it and see what happens. I don't expect it to behave exactly the same in reality, but it must still be pretty fast: a few ns turn-on time with a 1nF load is quite something.
Its typical output resistance is specified at below 1 Ohm, so I simply placed two parallel 100 Ohm resistors in series with the output, followed by an SMA connector. All mounted tightly on a small SOT-23-5 breakout board.
Now, the spectra I see in both cases don't look quite like what the textbooks say, however, I do think that I see the "second break" point on one of them, that's the point where the frequency is supposed to be equal to 1/(pi*Tr) (assuming Tr=Tf), and it meets the expectations I think.
Here's the 1EDN7512:
A very clear roll-off after about 470 MHz (~680 ps transition time), and it's dropping down pretty much into noise* after 600 MHz.
*Not sure if it can be called noise, because the actual noise floor (when there's no input) is at or slightly below -90 dBm.
But here's what the 4 x SN74LVC1G04 pulser produces:
Barely noticeable hump at ~510 MHz (that would be ~620 ps transition time) and a gradual further decline with no obvious steps, with the power levels at those higher frequencies staying above those of the other pulser.
Both had their fundamental frequency equal and set to 1024 kHz. Output levels were also the same. There was a 20 dB attenuator between the generator and the SA input in both cases.
Not sure what to make of it other than that the second pulser apparently has more energy in higher frequencies and that the first one's transition time can very likely be derived from the spectrum.