I'm designing a two layer PCB for a keyboard where most of the board space is taken up by the key switches and backlight LEDs. The switches have a traditional switch matrix configuration with diodes and the LEDs are charlieplexed. Thus, none of the current for the switches and LEDs will actually return through ground.
Of course, there is also an ARM MCU with a crystal oscillator, I2C between the MCU and the LED matrix chip. The ARM MCU will be placed wherever it will fit since the board is jam packed with switches and LEDs and this means I will have to run USB traces from that part of the board to the rear where the connector will be.
My initial idea would be to have as many signal traces as possible on the top and a ground fill/plane on the bottom. I would then only take short jumps to the bottom layer where needed in order to keep the ground plane as contiguous as possible. However, since this is a switch and LED matrix, traces crossing each other is inherent to the design. If I could use both layers for the matrix, that would greatly reduce the number of vias needed and simplify routing.
So the question is, is there any downside to limiting the ground plane only to the area around the MCU and LED matrix controller chip? As I said above, none of the return paths for the matrix goes through ground. What would be purpose of ground be in the matrix part?
Also, since USB has to be routed from the MCU to the rear, should I just put a wide "ground strip" under the USB lines? How wide would that be?
Of course... I could also got to a 4 layer board and all of this would be solved but that sounds excessive for a simple keyboard.