Proper oscilloscope probe grounding
The following pictures show a little example of what happens when the grounding on your oscilloscope probe sucks.
This is a 50 MHz signal as measured with a 15 cm (6 in) ground lead on the
discovery board pin header:
This is the same signal, only this time measured with a spring ground lead:
In this last picture you can see that even the traces from the STM32F407 to the pin header have an effect on the signal, this time the signal was measured with the spring ground lead directly at the chip: