More on the "55db" dynamic range. Assuimg this number means 20*log(largest rms signal/rms noise floor), which is not really explained anywhere, then the rms noise floor in bits is vn_rms=vmax_rms/10**(55/2) or (4096/(2*sqrt2))/10**(55/2)=2.57bits rms. So there will be 2.57bits of rms noise in the baseline. Assuming the pk-pk value is somewhat betwee 8 and 10 times the rms (very arguable) then using 8 the pk-pk bits of noise seen will be 20.6bits. So, with 20+ bits of pk to pk noise, can you resolve (i.e. "see") a perturbation that is only a few bits in amplitude in a single shot aquisition? I couldn't. If you can't, then the noise is not low enough for a 12bit a/d. QED
My bad; 16:1 is right. However, it's an overstatement to say that a "normal" 8 bit scope has about 40db of SNR because this is a huge function of bandwidth. Many 2 to 4 GHz oscilloscopes have a SNR better than 40db, and when used in high resolution mode that limits the BW down to HRO BW's, can easily equal the 55db of the HRO. Now you have effectively a "12 bit" scope at lower BW and a higher BW scope at "8 bits". Additionally, I think you'd find that 55db SNR for a 12bit system is still kind of marginal; you'd get something like 20+ bits of pk-pk noise in the baseline.
You are absolutely correct that more bits without a better signal path design could be useless. By my calculation 4 extra bits would mean 1/16 of the noise, not 1/32. A factor of 16 is about 24db. A "normal" 8 bit scope has about 40db SNR - although a few are somewhat better. The LeCroy HROs have about 55db SNR. That's just about what you'd expect for 4 extra bits. You should take a look at one... I think you'll be impressed.
There is undoubtedly a certain knee jerk appeal to more bits in an oscilloscope. However, more bits without a comensurate lower noise floor is not useful. Realistically for every bit increase in a digitizer, there ought to be a 2:1 decrease in noise level, so check to see if a 12 bit scope has 1/32 of the noise floor of an 8 bit scope. I think you'll be disappointed....
Hear hear for the PSU designers!
(If you know how to design PSUs and your bosses don't appreciate you then go work for a PSU design company).
More than 12 bits at any speed is a threshold, suddenly you have to consider temperature and component variations. Not everyone needs to go at Croy speeds but they themselves have to so you can trust what their instrument is telling you. Engineers generally work better with instruments and tools they can be proud of - and vice versa.
I am seeing some good value Croys on sale right now, maybe its time for a treat!