I expect this sort of shortcoming to be documented as part of the specifications.
How would they word that, exactly?
Tektronix did decades ago when bandwidth commonly decreased at the highest sensitivities; they simply specified the bandwidth where it was different for each range. Many oscilloscopes had a small drop in bandwidth at their highest vertical sensitivity for the reason I mentioned and it was sometimes considered a feature if there was not. (1) One of the most extreme examples that I know of is the 11A52 which was unusual for a 600MHz vertical amplifier for having a sensitivity down to 1mV/V:
>=10mV/div 600MHz
5 to 9.95mV/div 400MHz
2 to 4.98mV/div 250MHz
1 to 1.99mV/div 200MHz
I think D.Hess is referring to this image:
Something about the input amplifier overloading a tiny bit when the DS1054Z clicks down below 500Mv mode and you then fine tune the range up to to 400mV manually (ie. towards the limits of the low-range input amplifier).
And if that is what he's referring to, how would it be worded in a manual so that he wouldn't feel "deliberately mislead"?
Some instruments might specify full power bandwidth but I have never seen it applied to an oscilloscope. (2) Such a limitation is just broken. If they had specified full power bandwidth, then people would have asked embarrassing questions. I would not want to advertise such a limitation but I sure would not want to mislead people either.
Note that there are actually two separate problems here. TurboTom's test results
here show a significant change in small signal bandwidth with different vertical sensitivities like some oscilloscopes which I mentioned. The overload leading to non-linearity I spotted (3) produces a change in full power bandwidth which depends on both frequency and amplitude. Measuring just bandwidth at high signal levels (5 vertical divisions or greater ) will show contributions from both.
(1) The 7A13 listed constant bandwidth as a feature but it was unusual for having 1mV/div sensitivity at 100MHz at a time when most vertical inputs were limited to 5mV/div or 10mV/div where they often had a slightly lower bandwidth. My guess is that someone had a similar 1mV/div "100MHz" amplifier which was actually significantly slower; Tektronix might even have been comparing it to one of their earlier vertical amplifiers.
(2) Many analog CRT oscilloscopes actually do specify bandwidth under specific large signal conditions but this is due to non-linearity of the CRT itself. This is typically about 5 divisions (the same amplitude used for rise and fall time measurement) except for some very old oscilloscopes which did not even have 5 vertical divisions. This shows up in the low frequency trace compression test used verify proper vertical amplifier operation and CRT geometry.
(3) This is not a transient response problem because it disappears at lower signal levels.