I would be inclined to suspect EMI. Try permutations of 'Y' capacitors. In particular, the one to bridge '-' is probably wrong, but one from there to output may be helpful. Additional filtering (another CMC?), shielding (stuff everything into a box and filter all wires that pass through it) and testing are likely to prove fruitful.
Tim
I tried using the Y capacitors, and it did improved the noise under one "condition":
I was testing my smps with a 400 watt class D amplifier board, which is locally assembled one, I found out that the noise is created by that class D amplifier when I connect it to my smps. Does class D amplifers oscillates to create noise?, but after replacing that with a class AB amplifier, which I made, the noise just vanished! (still audiable when I hold my ears really close, covering the tweeters which I'm happy with)
So the *condition, only applies for the class ABs not Ds. And yes, the Y caps helped a lot
Is it a good idea to use a CMC with .1uf mylar filter caps only for the classD amp does it improves the performance and reduce noise?