Why "poor"? How exactly would you build it if you're an engineer who's been told to design a 100MHz, 4-channel oscilloscope which retails for $399?
For all you know it might be an extremely ingenious design, given that constraint.
Very sorry, I chose the word "poorly". I did not intend insult; in fact I have to agree, the design is quite ingenious in many respects! The discrete 2N3904/6 front end truly warms my heart.
No, rather than call it "poor" I should instead say they appear to have overlooked a very important functional principle, and thereby missed an opportunity.
First, it could add no more than a few cents to the BOM to include a passive 150 MHz filter (simple or fancy) ahead of the ADC. Even a switchable filter section or two, presuming one wanted different compensation at different ranges, would be no more bother than they evidently found acceptable for software-gelding the low-end models. I could sketch you something in a few minutes on DaveCad
TM and even tell you the BOM cost if you want, but again, I know Rigol could/would do a much better job.
The reason should be obvious, but to beat it to death: all test instruments have limits, and bandwidth is a fundamental one. Input voltage range should be a scalar, not a tensor.
A very typical 'scope application is checking signal integrity on either side of an attenuator, filter or amplifier. Invisibly calling up a factor of two different instrument bandwidth when you simply flip input ranges to account for the attenuation (gain) means you will see artifacts at the input (output) that are not seen at the output (input). You will make false conclusions about your DUT. Suppose you look with, and then without, a x10 probe. Is your probe bad? Were you loading the circuit? Or did simply dialing up the range for the probe factor just introduce a whole new set of sampling artifacts??
This is particularly egregious on a sampling instrument that has no protection against aliasing! The measured 3dB point at 5V/div evidently
exceeds the sampling rate (not just Nyquist) when 3 or 4 channels are selected!! Even at 1 Gs/sec, in the 5V/div picture above, there are less than 2 samples in the pulse transition. The apparent pre- (!) and post- over/undershoot/ringing are obvious figments of sinc interpolation.
Back in the day Tektronix worked very hard to match attenuator bandwidth within 5 or 10% from top to bottom (I don't know what they do now). This is a fundamental property of a scope front end, it directly affects functional utility as a test instrument. It is not "negotiable" for a few cents of profit margin.
They really did an amazing job with this machine, and I do really like it (as an old analog geezer I never thought I'd say such a thing!). It's such a shame to miss this trick; I hope they (and you) take my comment constructively for next time :~)
Cheers,
Zike