Via Redundancy

[BuffaloEngineer]

If I have a 4 layer board with 2 vias that go both go to GND close to each other, is it ever better to keep them separate? For example, if one GND via is connected to a bypass cap and another is connected to and MCU GND pin?

Close meaning less than 1mm apart.

 

[RadioRon]

Such spacing may have some significance in a 5GHz microwave circuit, but otherwise, I think not. Look at the situation as if it was a circuit. You might have current flowing through one via and you wonder how much of it can couple to the other via? The source via is like a voltage or current source with a significant source impedance depending on the signal source driving the current through that via. Perhaps the source Z is, let's guess for fun, oh 40 ohms. Now, whats the reactance of the coupling between the two vias. Thats a strong function of frequency. I might guess it is, oh, maybe 100 ohms at 5 GHz, perhaps 1K ohms at 1 GHz and perhaps 10K at 100 MHz. Just guesses for now. The remaining via is the load and it has an impedance that it presents to the coupled signal. The impedance would be very low because this is just a tiny little stub sticking out of ground plane, so I'm guessing it is less than 0.01 ohms at 100 MHz, perhaps .1 ohms at 1GHz, and maybe on the order of 1 ohm at 5 GHz. Now you have a series circuit and you can calculate the amount of signal that couples across. The loss from via to via might be, oh, -120dB and the effect will be that this coupled signal will superimpose itself on the ground current of the destination circuit (the MCU or whatever the bypass cap is bypassing). But -120 dB means virtually nothing gets across. Gets worse as frequency goes up though.