Circuit board ground plane question
eagletalontim
Posts: 1,399
I am designing a circuit board that will connect to a vehicle. My question is, should I create a ground plane for each of the voltages I have on my board (12V, 5V, and 3.3V), or one for the 12V and one for low voltage? The last board I designed had a 12V ground plane but me being a newb at the whole circuit design process, I did not put a ground plane for the low voltage and the USB would randomly disconnect when the engine RPM reached around 6K to 7K. I want to avoid this and believe the problem was from not having a ground plane and proper decoupling caps in the right areas.
Comments
The earlier design, did it have one ground plane that covered an entire surface (well mostly) of the PCB?
Was there an enclosure that was also connected to ground?
Obviously, a ground plane does not know that it is intended for 12V or 5V or 3.3V. I have never tried multiple ground planes and would not expect any significant benefit. Perhaps other have empirical results to the contrary? Unless you are measuring precise analog inputs, I expect that you are just fine with the single and common ground plane with no further complications there.
My gut feel, though, is that you identified half of the solution with proper decoupling caps. One at each IC power pin is a good starting point.
Proper shielding will help to reduce the stray noise introduced by ignitions wires and such. Signal wires (like USB) should be in a shielded cable, with the shield grounded at one point ... it is not intended to carry circuit current, but rather to give stray fields a path to ground. Similarly, a conductive enclosure should be grounded to provide a Faraday cage.
To give credit, I stole the attached image from here. It is a great visualization of the two most common countermeasures to noise - filter and shield.
Large gnd & power planes can actually be a bad thing as electrons will swirl around everywhere,
placing signal traces as to direct the gnd and power flow to its single entry-point is a good way to handle it.
Or draw power traces as star burst/ tree brances, and just think on how water would flow in a pipe with minimal swirling and speed inertia disturbing the neighboring IC.
So don't daisy-chain power traces if it can go on/off as it will create a "water-hammer" down the line.
Use a automotive rated LDO that can handle surges to 40v, use 200uF cap.
Run a thin wire for Ground all the way back to the cars battery.
On the first design, there was one small ground plane for the 12V area, but all the rest were traces for the ground and power to each component. I also believe the power traces were too small to properly handle the number of IC's on one trace. Without being able to catch the noise with an O-Scope, I believed the problem was created from the USB FTDI IC not having a ground plane to properly ground the required pins and I was pulling 5V from the board's power supply instead of the computer's supply. By changing these 2 things, I believe the random USB disconnect issue will be resolved.
Rule of thumb is not separating voltages but currents. As Tony just indicated, it's the high currents that create bumps in the road so to speak. So, in your case, one ground plane for heavy 12Volt switches, and a second ground plane for control currents (at any voltage).
So that gets the high currents out from the middle of your control circuit. But you've still got to bridge the divide to get the power switches to operate, this is just as important as getting the currents separated ... Electrical isolation (relay, optocoupler, transformer) is preferred solution but does add cost - can be significant extra. Rugged level shifters can be cheaper alternative but it all still costs and takes board space too.
Schematic too as to see inductor and capacitor filtering
The original board had a few simple flaws that were easily fixed except for the USB disconnecting problem which I believe can be fixed with the proper ground plane setup and using the computer's 5V power instead of the circuit board's power. I was also using a switching 5V regulator and I don't think the output was properly being filtered. I have since switched to a LDO linear regulator that can handle 40V input and is much smaller. It works perfectly for another product I created about a year ago.