I don't think it reflects negatively on peoples' English skills, but I do think that many of the remarks in this thread -- "rolloff is not infinitely steep" and "some margin for the anti-alias filters" understate the role of rolloff.
It's typical for modern scopes to devote half (or more!) of their Nyquist band to rolloff rather than sticker bandwidth (i.e. samp rate >= 4x BW). Gentle rolloffs make for
better step responses, so Oscilloscope designers prefer them when possible, but the tradeoff is that a huge part of your Nyquist band must then be dedicated to rolloff in order to promise decent attenuation of aliases. Not 10% or 20% of your Nyquist band, 50% or 60% of your Nyquist band, and even then the alias attenuation isn't spectacular. For the front end on my personal scope, a Rigol DS4014 hacked to 500MHz (see "Rigol MSO4024-500" line on the graph below), here is the relationship between sample rate and the alias attenuation they can promise given the gentle rolloff of their front end filter:
SampleRate | AliasAttenuation |
2BW | -0.5dB (!) |
2.5BW | -2dB (!) |
4BW | -13dB (actual sample rate, 4 channels) |
8BW | ~30dB (actual sample rate, 2 channels) |
I am quite pleased they did not try to sample at 2.5*BW. The benefit of going from 4*BW to 8*BW sampling (-13dB aliases to -30dB aliases) is more debatable, but -13dB attenuation isn't actually all that great -- it's still 20% of the original voltage amplitude, which is a considerable amount of worst case waveform fuzziness. It's comforting to have the option to bump up to 8*BW sampling (aliases at most ~3% of original amplitude).
With credit to Altemir:
Of course, we are discussing the GDS2204E, not the Rigol DS4000 series, but I could not find frequency response plots for the GDS2204E so it would be difficult to comment on the tradeoffs made by GW Instek. Here is the Bode plot for the DS1054Z (credit to ankerwolf) which might be more representative: