Consider the time scale on Leo's plots and the response of the scope he's using which is what you are seeing in the calibration plot. I can't find my plot right now as my bench is undergoing at upgrade and everything is a huge mess. But IIRC the entire length of the step response Leo provides is less than the sample rate of anything most of us are likely to own.
I have put the BNC version on a new MSOX3104T. 436 pS rise time and 7% overshoot. I discussed with Keysight support which confirmed what I was seeing. A 750 MHz low pass filter reduced the overshoot to well less than 1%. But I could not inline the filter.
I then had an RTM3104 on demo. It arrived with 350 pS rise time and 3% overshoot. If I applied a 1 GHz LPF I had no visible overshoot. I was ecstatic. But for some legal reason the K18 option is not available in North America and the available FFT was completely useless. It was suggested that I install the 1.300 FW update. After that it had 10% overshoot. Restoring 1.100 did not restore the original response.
I've looked at 4 other scopes with Leo's pulser. I have no doubt that the waveform I saw is the true step response of the instrument. Now if you have one of the new Keysight 256 GSa/S 111 GHz DSOs, you probably need a better signal source. I suspect it would require one to give an accurate picture of the step response of Leo's pulser.
There are a large number of plots of the pulser output made on a wide variety of scopes in this thread.
My big problem at the moment is he's developed a version that produces 100pS wide pulses. I'm trying to come up with a justification for buying one.
I was amused that there were two posts in succession about Leo's pulser.
