doing the practical example on my compensation signal (if I remember correctly coupled AC maybe, but it is not important ..)
I have reduced the x axis of FFt to 10Khz with 1Khz hz / div.
What can I understand from the 1Khz compensation signal?
The peaks in the FFT graph correspond to each ramp of the signal; I have 1 peak every signal period; in the FFt graph each peak occurs at 1Khz, then 3Khz, 5 Khz etc. etc. or at each multiple frequency of the original frequency, except for the positive frequencies; I can also see that each peak gradually decreases in value on the Y axis (0 / -10 / -18 / -20 etc ..)
Now I have to rearrange my ideas and understand what this teaches me!
(yes I saw the sum of signals in the video, now I'm analyzing only 1 signal, it is obvious that the FFt graph can also represent a sum of signals)
It's not difficult to understand, sometimes a bit more difficult to explain
![Smiley :)](https://www.eevblog.com/forum/Smileys/default/smiley.gif)
The signals we see in the real world, and I mean not only electrical signals, but sounds, are complex. These "complex" signals can be represented like a sum of components, each component having the waveform of a sinusoid. You could call a sinusoid a "pure" signal.
So, let's say you are listening to music. You have an amplifier with the typical "bass" and "treble" tone controls. If you turn down the treble you will notice that the sound changes. You are reducing the level of the high frequency components.
For periodic signals you can still measure a frequency (for a music sound that would give you the note) which is the lowest frequency component. That's called the "fundamental". And most of the complex signals you will find (again, imagine the sound of a flute although some evil instruments such as bells are different) are a sum of a fundamental and a series of components that have a special property: their frequency is a multiple of that of the fundamental. They are called harmonics
So, if the fundamental (hence, the signal) has a frequency f, the harmonics will have frequencies 2*f, 3*f, 4*f, 5*f, etc.
The funny thing is, by summing those harmonics you don't perceive different signals being added together, but "sound texture" changes. Imagine a trumpet, a pipe organ and a flute playing the same note. The frequency is the same but the levels of the harmonics are different.
By decomposing a signal into its components you can find out a lot about it. First, many waveforms are quite similar despite having a very different sound. Decomposing the signal into its components makes it easier to study.
And that by examining the composition of the signal, which is called the spectrum, you can learn a lot about it and it can help you diagnose some faults. The Fourier transform is the mathematical method to obtain the spectrum of a signal, and the FFT (Fast Fourier Transform) is an especially optimized method to compute the Fourier transform so that it runs much faster, hence being more efficient.
Now you can see why the spectrum of the square wave in your calibration output looks like that. It is a sum of a signal at frequency f and harmonics with odd mutiple frequency (3*f, 5*f, 7*f, 9*f...) each with an amplitude 1/3, 1/5, 1/7, 1/9...
So it would be (WARNING, NOT CORRECT EQUATION BUT TRYING TO MAKE IT MORE INTUITIVE)
sin(f*t) + 1/3 sin (3f*t) + 1/5 sin(5f*t) + 1/7 sin(7f*t)...
You can try a function generator and compare the spectrums of square, sawtooth and triangular signals for example.
Also, do you know why telephone voice sounds so bad? Yes, the narrow bandwidth transmits very few of the signal components.
I forgot, for an intuitive treatment of the matter, especially good if you like music,
"The Science of Musical Sound" by John R Pierce.
And it's a really deep subject. It is also related to the amount of information you can send over a transmission channel, go figure!