The next thing I tried was to put a tiny 10R SMD resistor across the 50R termination. This brings the effective source impedance down to about 7 ohms. This should really reduce the droop at 500MHz.
When I do this with the simulation, it predicts 1dB droop by 500MHz when the system is loaded by the 10073C probe.
When I did it for real, with the real test system, it gave the result below with the 10073C probe. This also shows about 1dB droop. This is a really good result I think!
It would be nice to do this with a better scope. My Tek scope doesn't have a great front end design. It's noisy and uses cheap SMD technology on a regular PCB for the front end. It does offer 1GHz BW but only above 5mV/div. Some of the ripple and noise in the plot below will be in my noisy old Tek scope. Some of the ripple will be in the 10073C probe, but at the moment it's hard to know (for certain) what causes the subtle ripple in the plot below. The simulation predicts a fairly smooth 1dB decline by 500MHz with none of the ripple seen in the plot below.
