1. The 1024x averaging now reveals spurious signals that we would never see otherwise. Take note of the strongest spur at around 622 kHz with a level of about -142 dBV. Of course this is really nothing. -140 dBV is already only 100 nVrms. The lowest visible spurs are just -162 dBV, so we’re able to measure signals below 10 nVrms…
2. Some folks put high hopes in resolution enhancement techniques, but a higher resolution does not reduce the frontend noise. Any noise reduction effect comes from the lowpass filter effect – but this obviously won’t help at low frequencies.
Some notes (what you know of course but perhaps there is some reader who do not look this enough carefully... and then believe something is possible what is not possible in real...
).
With time domain trace averaging we can reveal spurious and other signals only when these are synchronized to acquisition (normally we talk trigged). If not, averaging attenuate these instead of pick up them from under noise. Example this 622kHz system generated very low level spur in your image.
...so we’re able to measure signals below 10 nVrms…
Yes IF... and this IF is big.
IF we want measure very low signals using FFT and what need time domain trace averaging for pick up this signal from noise we need tight lock to this signal frequency.
If we have only this weak signal available... game is over because oscilloscope can not trig to this weak signal, trigger engine do not know it is there. So wehen it is "random" it is attenuated just as random noise.
Here is example.
I have here ~455kHz -137dBm (-150dBV, 31.6nVrms) carrier. I have measured here its level and it is also nice amount (~30dB) over noise average.
Without perfect trigger it totally disappear. (of course) It display only this locked signal. Not example neighbour carrier if it exist but is not locked.
How this is possible at all. In this example, same signal but higher level, roughly -17dBm go to channel 1 for triger (lock to this signal).
Time domain trage Average 1024 for reduce random signals/noise level (without this, this signal is totally under noise.)
FFT is from channel 4 where is same signal but attenuated to -137dBm level (-150dBV). FFT Average also 1024 for get noise average. (in this case signal can measure also without this FFT average because signal is so much over noise peaks. Also I have checked that cross talk from Ch1 is not affecting any detectable amount. With -137dBm signal disconnected result is just this -180dBV (1nVrms, 2.83nVpp) average. (naturally all this is impossible if system do not have enough random noise. One ADC step is here roughly 1000nV)
So this note shortly: Very low level signals (under reliable trigger level) can only measure and detect like this IF acquisition can lock (trig) way or other to this signal.
Possible system internal spurs what are inside system somehow locked to acquisition they of course can see, as is case with example this 622kHz and some other internal spurs but internally generated spurs what are not sync, they of course disappear due to to time domain averaging, example some internal SMPS spurs)
2.
Any noise reduction effect comes from the lowpass filter effect
Yes but also noise reduction effect comes from what ever filter what reduce frequency band width - low pass, high pass and band pass filters all reduce freq BW and so also noise.
(in theory fBW/10 drops noise 10dB.) Not only low pass.
In this oscilloscope one good example is FRA where band pass filter reduce noise (and signals out of DUT in freq) including also very low freq noise. (but not so much because filter is not very narrow. But proportionally to oscilloscope bandwidth it is still very narrow. Example between 10Hz and 200Hz this "freq. selectivity" filter -3dB width roughly 7Hz, -6dB 9Hz, -60dB 14Hz, -70dB 15Hz and of course this filter center freq is moving syncronous with sweep. )