Transmission line effects become important when the line delay exceeds/equals the risetime of the pulse - for shorter lines the rising edge will 'see' the load even as it is rising. It's exactly like an echo - you realize you got an echo only when you're far enough away from the mountain...tain...tain. Too close and it's pretty much the original signal.

So, what is the line delay? Assume all of the EM wave on the PCB track propagates in the fiberglass (FR-4) where the velocity is half that of light. In this worst case (largest delay) a line length 4 cm means 250 ps. In reality part of the wave will go in air, so that's why we took the largest delay.

In your case the 20 MHz clock has a 50 ns period, so a 5 ns risetime is reasonable. So you can size your line for PCB fab (8 mil is typical) and worry more about the load impedance than the line. Whether you want to use a differential signal is a totally different thing...

M.G.Rajan