I employed those strategies when I made the PCB for the the 'nextdsp' board.
There's PCB files on this web-site in case you want to see the approaches that I took.
Noise is pretty low, perhaps approaching -85dB or better?
http://sites.google.com/site/nextaudiodsp/As the others pointed out these are the main points for low-noise PCB design:
1) Separate ground planes for analog and digital sections and joined them one point - usually at the shared source for power/ground.
2) Separate power planes for analog and digital sections - approach same as above
3) Keep the common points for the power and ground (where analog and digital sections are joined) close to each other
4) Bypass caps at the power pins for each IC; use a couple (I use three) for each power pin; values of 100nF and 10uF when using two or values of 10nF, 330nF and 10uF
5) Avoid digital and analog traces crossing paths
That should get you there unless something else is generating the noise. Could it be something else?
1) My board uses a CODEC and therefor the A/D conversion is not on the same chip as the micro-controller. Are you using on-board A/D's of the PIC?
2) Is the analog circuitry for the A/D inputs using a clean power source? Op-amps have good tolerance for power supplies that are a little noisy (supply rejection).
3) Does the analog circuitry rely on biasing or a virtual ground? The signal conditioning circuits are usually built using transistors or op-amps powered via one supply and therefore they operate in class-A. If a bias voltage is used (or a virtual ground) then it need's to be really clean since it contributes to the input signal being sampled.
3) I'm reaching on this one; If the on-board A/D's are used then make sure they're sampling synchronously so that the sampling period is constant. The action of sampling should be driven by a clock rather than source code that may be interrupted or something like that. Sampling jitter will cause noise.
That's all I can think of at the moment. Anyone else have more thoughts?
