Since we are so far off topic, I guess it's okay for me to ask an incidental question. While all this conversation is quite interesting and instructive, I'm curious how many people actually use or need bandwidth or sampling beyond 100 MHz and 312.5 S/s? Presumably, anyone working in comms, but who else? My thought process when buying the DHO800 was that even with 4 channels operating, I still have enough sampling to adequately cover the 100 MHz capability of the scope with the included 150 MHz probes being well clear of having any attenuation effect at 100 MHz.
For example, here are some waveforms from a buck convertor I designed and built this week running at 100 kHz. (Note the oscillations on switching are mostly due to using the clip-on scope probes...when I use the spring and tip, the yellow gate drive signal is clean and the overshoots on the cyan and magenta traces halve.)
My question is....for the kind of work I'm doing (hobbyist audio and SMPS in the 100 W to 5 kW range), what would be the benefits, if any, in me having a more capable scope? (Please excuse my ignorance, as my professional background is 50 Hz HV & EHV transmission)
Back in the late 70s and early 80s 100MHz was suitable for TTL and LSTTL logic. Since then logic speeds have increased "a bit", so the required bandwidth has increased "a bit".
Note that the inputs/outputs of logic gates are
analogue signals that are
interpreted by the receiver as digital signals. In order for the interpretation to be correct, the analogue waveforms have to meet the specifications, e.g. min/max voltages and over/undershoots, timing (the hold time, t
H, is particularly tight), clock edge rates and monotonicity, etc. Ensuring all that is in order is called "signal itegrity", and the required bandwidth depends
only on the transition time; a clock/signal period is
completely irrelevant.
As others have noted, at such speeds the probes become an integral part of the circuit and measurement.
Start by working out the input impedance of a typical 10Mohm/15pF probe at 100MHz; rather than 10000kohms, it is more like 0.1kohms
Then add the inductance of a 6"/150mm ground lead, and calculate the resonant frequency with the 15pF tip capacitance.
Nowadays
jellybean logic has edge rates of 1ns, and faster. That translates to 350MHz, and higher.