@Hero999
That's why those cheap AM transmitters schematics based on the 555, and logic chips are all aberrations. =P
Jimb. The wave is not overmodulated nor undersampled, take a look at the carrier frequency about 66% higher than the sidebands. (At the first pic).
As I used FFT to calculate the spectrums, I have to sample (I sampled at 10 MHz -> 25x Nyquist Freq. ), use a number of points for FFT (16384), and use a window (in this case was a rectangular window). All of this allegory causes the spectrum to be shown that way.
Think about it, a rectangular window in time is going to be a Sinc-shaped window in frequency domain, plus the sample pulses and FFT resolution will give us a lot of Sinc-shape forms at the frequency domain. (And if you see those Scopes that measure the spectrum based on FFT display spectrum like those I posted (noisy)).
But, if you observe, the ratio between a sideband (Magnitude = 0.6) and the highest noise peak at 208 kHz is about -10 dB.
If I use another window, the spectrum becomes clearer, but I have some losses, take a look, this is the same spectrum using a Tukey window, with a rolloff = 0.55. (0 = rectangular, 1 = Hanning).
You are able to see the carrier, the sidebands (with degraded amplitude), with much less noise: