What kind of Modulation causes "drops" in FFT Power Spectrum

I am analyzing a signal with FFT and I found an effect, that I do not understand. I am expecting to have an amplitude modulated signal. As far as I know, if you compute the power of the FFT of an amplitude-modulated signal, you observe sidebands to the left and right of the carrier frequency.

E.g., if you have a sinusoidal carrier at frequency $$w_c$$, and you modulate the carrier's amplitude with frequency $$w_m$$, you see sidebands appearing at $$w_c \pm w_m$$.

However, in my signal's power spectrum, I see the following. Note, that the x-axis shows not the absolute frequency, but the delta from the carrier.

At the frequency, where I would expect peaks in the spectrum, I see drops. How can that be? What does that tell me about the signal? If I look at the signal in the time domain. I can see, that amplitude modulation is actually happening!

• Why is your axis "frequency deviation"? To do what you describe in your question, you need to plot using plain frequency.
– MBaz
Commented Mar 13, 2020 at 13:45
• Can you post a snippet of your code? Commented Mar 13, 2020 at 13:50
• Can you tell us what signal you're amplitude modulating, and maybe show it to us in the time-domain? If it's supposed to be an AM-modulated sine wave, something is definitely wrong. Commented Mar 13, 2020 at 13:54
• Is this a real world signal or simulated in MATLAB? Reason I am asking is, when you generate a real world carrier signal cos(omega_c), it can start the Local Oscillator with random phase. So x_c = cos(omega_c + phi). It may happen that phi=180 degrees. So your x_c and its modulating signal will be out of phase (cos (theta+pi) = -cos(theta) ) which will negatively affect the total amplitude. Commented Mar 13, 2020 at 17:34