Until recently, hobby-grade digital oscilloscopes were mostly, at most, 8-bit sampling. However, newer devices offer 12-bit conversion. Does it matter? Depends. [Kiss Analog] shows where a 12-bit scope may outperform an 8-bit one.
It may seem obvious, of course. When you store data in 8-bit resolution and zoom in on it, you simply have less resolution. However, seeing the difference on real data is enlightening.
To perform the test, he used three scopes to freeze on a fairly benign wave. Then he cranked up the vertical scale and zoomed in horizontally. The 8-bit scopes reveal a jagged line where the digitizer is off randomly by a bit or so. The 12-bit was able to zoom in on a smooth waveform.
Of course, if you set the scope to zoom in in real time, you don’t have that problem as much, because you divide a smaller range by 256 (the number of slices in 8 bits). However, if you have that once-in-a-blue-moon waveform captured, you might appreciate not having to try to capture it again with different settings.
A scope doesn’t have to be physically large to do a 12-bit sample. Digital sampling for scopes has come a long way.
I would expect the place 12 bits will shine, other than just being nicer to look at, is in any scope that does some DSP. It should drastic reduce sampling artifacts and aliases. Among them is the sampling noise front the size of the steps. DSP doesn’t like steps because they are impossible.
Some cheaper scopes don’t have the hardware to do all possible attenuation steps, so the hardware attenuator is actually the same for different scales, and the rest is done in software, which further reduces the resolution. I don’t recall which brand we observed this on, it was about 5 years ago, I don’t know if they still do that
From HMCAD1511 datasheet:
1X to 50X Digital Gain, no Missing Codes up to 32X
1X Gain: 49.8 dB SNR. 10X Gain: 48 dB SNR
Internal 1 to 50X digital coarse gain with enoB > 7.5 up to 16X gain, allows digital implementation of
oscilloscope gain settings. Internal digital fine gain can be set separately for each ADC to calibrate for gain errors.
(8 good bits is SNR of 48.16 dB for reference.)
Hmm, also note that for every 4x oversampling you get an extra bit, if there is some uncorrelated noise. So in the first example there was plenty of oversampling going on and visible noise that would accommodate a low-pass filter to narrow the line down quite a bit. Not sure if the scope has a button for that.