There have been dreams about a 14 bits oscilloscope that can resolve 25 µVpp signals and has at least 100 (better 200) MHz bandwidth. And it has to be cheap, of course, as always.
https://www.eevblog.com/forum/testgear/best-14-bit-digital-oscilloscope/It has been pointed out that this is physically impossible, because a pure 50 ohms resistor alone exhibits already ~60 µVpp noise at 100 MHz bandwidth.
Funny ideas like paralleling channels in order to reduce noise are not a serious approach either – for a number of obvious reasons.
This leaves us with averaging as the universal concept for measuring signals below the noise floor. Averaging also increases the resolution, so we could get even more than just 14 bits. For this to work, we need a copy of the signal strong enough to reliably trigger on.
When we think about the situations where we actually want to measure signals down in the microvolts, it usually has to do with EMI and related topics, i.e. unwanted feedthrough and crosstalk due to insufficient isolating/filtering/shielding. In most of these cases, a strong copy of the signal to be measured is available, such as the mains supply (even available as trigger source) in case of measuring weak residual mains hum, the signal from an antenna in case of RF interference from a nearby transmitter or the switching signal from an SMPS.
If the above requirement is fulfilled, then there is something else to be considered: fast HW-accelerated averaging for getting the results reasonably quick. Not strictly a requirement, yet very convenient to have. It just so happens that the SDS2000X HD averaging is very quick, so let’s have a look…
A square wave signal with ~25 µVpp amplitude at 1 MHz. The bandwidth of the input channel has been limited to 200 MHz for this test. Even on a low noise DSO like this, the signal is not recognizable in normal acquisition mode and noise fills up the entire screen at 50 µV/div:
SDS2504X HD_25µVpp_Square_Normal
256 times averaging is already enough to get a reasonably clear picture at 10 µV/div. The amplitude is measured much too high as 47 µV, just because of the residual noise:
SDS2504X HD_25µVpp_Square_Avg256
With 1024 times averaging, the signal representation should already be very acceptable. Of course, the amplitude measurement is still significantly off at 37 µV.
SDS2504X HD_25µVpp_Square_Avg1024
Finally, we try the 8192 averaging. The noise is almost completely gone, we can clearly see some synchronous interference signal at about 10 MHz. The amplitude level is now measured as 31 µV and this is as good as it gets with the remaining interferences. Yet reasonably close to the nominal value.
SDS2504X HD_25µVpp_Square_Avg8192