There has been a debate about measuring a weak noise signal (from the vague description, I had to assume 600 µVpp) and have demonstrated how that would be perfectly possible with an SDS2000X Plus. Since the problem appeared to be very low frequency noise, I had to concentrate on that specifically.
I've also said that the SDS2000X HD would be an even better match for tasks like that. And as always, I like to back up my claims with some practical demonstration.
The SDS2000X HD is even better at analyzing weak signals, because it has:
• Genuine 12 bit analog to digital converters for higher resolution and lower granular noise
• Full speed hardware accelerated ERES and Average acquisition modes
• Digital brick wall filters as math functions
First screenshot shows the noise level of the SDS2504X HD itself. An assumed spurious signal with 600 µVpp amplitude would be shown as -70,46 dBV. As can be seen, even at only 100 Hz the noise level of the DSO is just -110.9 dBV, which gives plenty of headroom.
SDS2504X HD_Noise_50_10ms_100kHz
In the previous test I haven't even gone all out to get the best possible result: neither did I take advantage on the long memory, nor did I use the digital brick wall filters in order to minimize aliasing.
So this time, I've used long memory (200 Mpts) in order to preserve a high primary sample rate of 1 GSa/s. Together with the 20 MHz bandwidth limiter, this gives a high protection against aliasing in the acquired raw data. Then I calculate an 800kHz low pass filter in one math channel and then the FFT on that.
I did not only want to show the noise floor but also a low level low frequency signal (100 µVrms = 283 µVpp @ 120 Hz) to demonstrate how it can be accurately measured. We can clearly see that signal in the time domain as well as its filtered version (math channel F2) with twice the amplitude at 500 µV/div.
This screenshot also demonstrates how to work around any problems coming from the axis labes interfering with markers at the very left of the screen. It's easy: just start at one division later (use an appropriate negative start frequency):
SDS2504X HD_Sig120Hz-80dBV_50_LP800kHz_20ms_90kHz
If you look at the table, you'll notice that the 120 Hz signal is accurately measured, yet the noise level at 100 Hz is surprisingly high in this measurement. This is because 100 Hz is just too close to 120 Hz, so that we don't actually get the noise level at that frequency, see last screen shot, which shows a zoom view on the previous measurement.
SDS2504X HD_Sig120Hz-80dBV_50_LP800kHz_20ms_1kHz
We can also see the measurement of the noise level at higher frequencies, such as -149.47 dBV at 100 kHz: 33.61 nVrms = 95.07 nVpp for 22.65 Hz bandwidth. That results in a noise density of 7.06 nV/√Hz. Of course we can get even lower noise down to about 2 nV/√Hz at the high frequencies, but at 100 kHz the 1/f noise of the MOSFET semiconductors takes its toll already.