I have got a project in mind that involves high speed AC and digital signals, So I am starting to document myself on the topic, However I don't seem to clearly understand, why the following is clearly stated in all the documentation I have checked:
Because each plane can have different potentials, currents can flow between them, and they occupy a different adjacent region of space. Actually, two reference planes that are collinear, as in two solid layers of a PCB, would not be very much of a dipole, because they occupy more or less the same region of space (compared with one wavelength of interest). They are talking about separate regions of the board.
it is basically a sensing application in which I got multiple capacitive sensors (Talking about a few Khz for excitation signals), and I also have a few peripherals attached to the same MCU, such as two SPI buses and probably a TWI, So the main goal is to properly understand how to avoid electrical noise on my analog signals from the digital side of the design.
As I understand grounding is critical is this case as High Speed digital signals may add noise to my measurements on the analog side.
So I am just trying to get my head around it, so that my design will be optimal.
A few kHz is not considered high speed for signalling.
The Arduino's ADC is slowish at 9.6 ksps maximum.
In contrast, the I2C default speed is at 100 kHz; SPI default is 4 MHz (for a 16 MHz Arduino).
If you're really worried at those signals at your analog inputs (they're probably too high frequency to really affect your readings) you can add a 100p-1n cap on the input. Alternative: make your Arduino sleep during the analog reading, to guaranteed have no communication on those buses.
A few kHz is 4 to 5 orders of magnitude less than your groundplane acting as a dipole. Not a problem.
The high speed logic noise is what can trigger the board to radiate as an antenna in the UHF and microwave bands. To produce a commercial product you have to prove the design complies with EMI emission limits
which is done by measuring it's RFI in an RF anechoic chamber - costs money.
