It was described before but people seem to not be understanding the situation. Perhaps this is my fault so here it is again, in excruciating detail.
I am using an analog function generator that has a dedicated TTL output, not rated at 50ohm. I have a 50ohm terminator on one side of the T and a BNC coax on the other end. This gives a ringing square wave as the output. It is not meant to be a proper square wave. I was looking for the fastest rise time I could get from the output. It was just by accident that I arrived at this setup.
I am still waiting for my SDG1025 to arrive.
I have a DS1052E upgraded to 100MHz that came with RP3300 probes, old versions, that do not have a compensation adjustment at the BNC.
I have a DS1054Z upgraded to 100MHz that came with RP2200 probes.
I was casually comparing the performance of the two systems. When I started looking at the waveforms that were displayed on both scopes while probing the same signal I noticed that the DS1052E seemed to show more detail on the waveform than the DS1054Z. So thinking that perhaps the upgrade had not worked on the DS1054Z I started to investigate. After some poking around I swapped the probes from one scope to the next (after compensating of course) and the difference I observed moved with the probes.
Both RP3300 probes showed more detail on the waveform than the 4 RP2200 probes I have, and on either scope and on any of the channels. One RP2200 and one RP3300 probe were always connected at the same time to the signal so any difference in loading of the signal from the different probe model is a moot point. I have tried probing the source BNC directly with ground spring clips instead of the alligator clip leads, the same on the end of the original arrangement, at the T adapter, etc....
In each and every case, the same difference in the intentionally ringing waveform appeared. The RP3300 probes show more detail and higher peaks on the ringing than the RP2200, regardless of scope, channel and test method.
When calculated according to the formula presented on Dave's video, a theoretical 100MHz bandwidth scope with theoretical 150MHZ probe will have system bandwidth of around 84.2MHz. The same scope with a 350MHz probe will have a bandwidth of 96.2MHZ. With this in mind, it is a small leap to justify the difference I see with my different probes. It also shows that theoretically the probes limit the performance of the scope, whether it is normal practice to supply the same bandwidth probes for a scope or not.
It has now been demonstrated, at least in one person's case, that the DS1054Z upgraded to 100MHz actually seems to have a bandwidth of between 130MHz and 150MHz. With this in mind the difference I see in the waveform is even easier to rationalize. It is because this affects what effect the probes have on the bandwidth. It is a wider spread in the system bandwidth, namely 106MHZ for the RP2200 vs 138MHz for the RP3300. This is assuming that the scope has 150MHz bandwidth.
So as far as I am concerned the difference that I see is easily explained by these last theoretical calculations and it answers my question about whether or not the probes as supplied affect the overall scope's performance. If the 100MHz bandwidth of the scope is assumed, or lower, then the probes will bring the overall system performance down below the rating of the scope. In the case of the 50MHz bandwidth model, down to a theoretical 44.7MHz. If the bandwidth of the 100MHz model scope is really closer to the 150MHz suggested by another person then the RP2200 probes bring the system bandwidth down to 106MHz and the probes do not affect the rated performance of the scope as a system with the probes.
Another conclusion can be reached to I think. If you do not want your scope measurement system to be degraded by the probes, they should have at least 3.5 times the bandwidth of the scope. At the very least you should be aware of the effects of the bandwidth of the probes in your overall system performance. The last sentence holds true for any system that all components need to be taken into account of course but I just wanted to complete the logical conclusion.