During near field testing I encountered this thing I've never seen neither on the wild nor on the book. The board has an STM32F107 with an 8MHz crystal and no ferrites or choke whatsoever (I'm already crying.) The only other oscillating thing I could think of is the 24V to 5V switching regulator at about 40kHz. From what they tell me the 8MHz is multiplied with the STM32 PLL to obtain the main 72MHz clock.
With a magnetic field probe on a cable bundle from that board I see substantially a spectrum comb from about 4MHz to about 36MHz, spaced at about 105kHz (see image.)
As you can see it's quite regular and, by itself, it's not an issue (the 32-36MHz overtones however are horrible but that's another story, maybe the APB clock leaks, too.) What kind of phenomena could cause that? The most similar thing I could think of would be intermodulation but it would be so wide and regular in amplitude.
SIMPLE AND REALLY STUPID THEORY: from signal theory a Dirac comb signal gives a Dirac comb spectrum.
So what if that's simply a 105kHz really short pulse that expand up to the about 6-36MHz? That could be somehow the switching regulator with really low load (since it's so old it has no pulse skipping or other modern power saving techniques). Tomorrow I'll try to load it, if the pulse widens the upper harmonics should go down.
