So I was playing with the FFT function( fast Fourier transform, it should be a good proxy for a spectrum analyzer so I tried one kHz sine waves from my small function generator we've got the following wave form



then I change the function to a square wave to see the harmonics.



I was wondering if there are any tricks to improve the resolution of this display?

 

Change the window to Hamming, adjust the center closer to the fundamental and decrease the Hz/div. It's one of those things that takes a bit of tweaking to find a sweet spot.

 

What is the difference between hamming and rectangle?
Hereis the display at 10Khz

 

I think that is the limits of your scope. Here is what I get with my SDS1204X+ after tinkering a bit with the start, stop, and span of the FFT. It tries to tag what it thinks are harmonics but appears to simply be noise as they move around quite a bit rapidly. It's so far down in dB as to be insignificant. Signal is 1kHz @ 4Vpp

 

Here is a close up plus showing the FFT parameters of the span. I don't see anything that looks like a harmonic to me.

 

What is the difference between hamming and rectangle?
Hereis the display at 10Khz

View attachment 303659 View attachment 303660 View attachment 303661

A Hamming window will smoothly truncate the ideal impulse response and reduce high sidelobes. It's just a user preference.
https://download.ni.com/evaluation/pxi/Understanding FFTs and Windowing.pdf

The Hamming and Hann window functions both have a sinusoidal shape. Both windows result in a wide peak but low side lobes. However, the Hann window touches zero at both ends eliminating all discontinuity. The Hamming window doesn’t quite reach zero and thus still has a slight discontinuity in the signal. Because of this difference, the Hamming window does a better job of cancelling the nearest side lobe but a poorer job of canceling any others. These window functions are useful for noise measurements where better frequency resolution than some of the other windows is wanted but moderate side lobes do not present a problem.

Try turning the scan sweep speed way down (like ~500ms) so you have a lot more samples for the FFT to bin.

 

@SamR
I think I liked your scope more than I do mine. Interesting to see what that difference a little time next.

I just need to get better at setting this thing up. I'm thinking of doing some distortion tests on unknown sine waves. I was hoping this would improve that.

 

@tautech should be able to tell us if the Siglent 1202X-E scope can do better with FFT? My bench is a mess with some stuff I'm working on or I'd drag out my 1102CML+ to see what it does. It is or was their basic beginner scope but does FFT also.

 

@SamR
I think I liked your scope more than I do mine. Interesting to see what that difference a little time next.

I just need to get better at setting this thing up. I'm thinking of doing some distortion tests on unknown sine waves. I was hoping this would improve that.

The trick is to set the timebase to display lots of periods and all posts I have seen above are too few.
If needed I can post some examples, just ask.

 

Frequency resolution is the inverse of the record length. In other words, collect the longest amount of data for better resolution.

 

Quick example using the 4ch SDS1104X-E and FFT of the 1kHz square wave probe compensation signal.
Here we know the fundamental frequency from the scopes counter and set the center frequency and Hz/div in order to show the fundamental and its harmonics. Using a Max Hold setting to diminish the noise floor as in this case there is nothing down there we need to see.
A couple of screenshots for study.

 

More pictures I tried all three screens on my scope and did not see anything that would control the number of samples:

with suitable filters and white noise source I'm thinking this might make a good signal to noise setup.

 

So this looks like it might be the count?

​

 

Sweep speed 2ms per

Slower speed: 20ms per

Hamming Window

 

What were you trying to show me?

 

Yup, my suggestion was to vary the sweep and see what happens but @nsaspook beat me to it. Doesn't seem to change the # of points collected on the MATH graph but does change their span frequency focus. They increased the sampling rate to 2GSamples/sec on their SDS2000X+ series (4X) so it gives a bit more definition. Plus, I see some actual harmonics that I wasn't getting from my Siglent AWG on my scope?

 

What were you trying to show me?

How FFT works on the scope. The thing that increases frequency resolution is, time.

 

So where is the sample rate and how do you set it?

 

So where is the sample rate and how do you set it?

Here it's the number of total samples of X period waveform that are important. To increase the number of samples you decrease the time base sweep speed (increasing time) so the ADC signal buffer gets more complete signal samples (cycles) at the set scope chosen sample rate to increase resolution.

 

There is nothing to gain by increasing the sampling frequency so long as you are above the Nyquist limit.

You want to decrease the sampling frequency, i.e. longer TIME/DIV in order to increase the total sample time.

FFT frequency resolution = 1 / total time

For example, for 10Hz resolution you need a record length of 0.1 seconds or 100ms. If TIME/DIV = 10ms and your sampling frequency is 1Msps you would need 10k data storage.