First try FFT:
Sine- & Squarewave
Have you noticed that the FFT setting might not be optimal?
You’re using a long 2 Mpts FFT for analyzing a LF signal. This is pretty slow and still provides only a rather poor frequency resolution of ~24 Hz, which in turn will result in a RBW (resolution bandwidth) of some 50 to 100 Hz, depending on the window function used.
Speaking of window functions, you should select it carefully. Use Flattop whenever amplitude accuracy is important and blackman for a slightly better dynamic range and narrower resolution bandwidth otherwise.
Your FFT uses 50 MSa/s sample rate, which results in an FFT bandwidth of 25 MHz. This is overkill to watch a 1 kHz sine wave and consequently you have to use an extreme zoom for the FFT display. If you limit the FFT bandwidth to e.g. 100 kHz (by limiting the FFT length to 16 kpts) you avoid generating a bunch of unused data and get a faster update of the spectrum. With this setting, you could still watch all the harmonics up to n=100
The first attached screenshot shows this situation with your timebase setting of 10 ms/div. This also means that the FFT has a lower bandwidth limit of 10 Hz. Frequency step is 12,21 Hz and I used the Flattop window where you nearly have to quadruple the frequency step in order to determine the RBW. So the resolution bandwidth in this example is slightly less than 50 Hz.
The second attached screenshot shows the same situation with a timebase setting of 50 ms/div. To get the same FFT bandwidth, FFT length had to be adjusted to 64 kpts. This slows acquisition down, hence also the FFT. The FFT now has a lower bandwidth limit of 2 Hz. Frequency step is 3,05 Hz and Flattop window is used again. The resolution bandwidth in this example is now slightly less than 12 Hz and you can see this clearly in the graph.
In both examples above the view is zoomed into the first 10 kHz, because as can be seen the harmonics of my sine wave (coming from an SDG6000X) are just too low to be properly measured and there is only noise above 10 kHz anyway.
For the square wave it is a bit more complex. As we all know, (especially) narrow pulses can have a fairly constant spectrum up to very high frequencies. So my advice here is to first inspect a full bandwidth spectrum (i.e. with an FFT sample rate of 1 GSa/s for full 500 MHz FFT bandwidth) to determine the bandwidth of the input signal and then chose the final FFT bandwidth accordingly.
EDIT: if you don't follow this advice, you might get very confusing spurious lines coming from aliasing in the FFT, even though there is no aliasing in the Y-t plot.
The third screenshot shows a triangle wave, which has a spectrum with reasonably fast decay. Analysis bandwidth is 100 kHz again with the full bandwidth displayed. Without the zoom, the spectral lines are nice thin and sharp, so the visual resolution leaves nothing to be desired and a slower timebase (together with a longer FFT) is not required here.
Home
Help
Search
Login
Register
