Elimination of PWM noise on LED

Hello all!

Here the situation: I have all mi pieces wired up fine and working, my project drives an RGBW LED, high power (e.g. Luxeon or similar) using a NPN transistor wired in the classical way. This LED ha to work on the same circuit of a "wave shield" I have optimized and miniaturized that with a custom SD reader, an MCP4921 and a TDA7052 amplifier wired in the classical way reported also on the datasheet for 1W amplifier like so: TDA7052 datasheet and circuit Portable Audio | Schematic Power Amplifier and Layout

Now the problem: sending PWM signal under 255 values makes the Speaker catch some noise from the PWM signal. I tried filtering the Amplifier but did note work, plus I had a loss of volume. I would like to avoid this solution, what I am seeking is how to filter the positive side of the wired LED (or add something on the transistor) to get rid of the interference. At the moment I can lower the sound putting an electrolitic capacitor of 470?F from positive THE to the negative battery. But it is not enough to turn off the sound. Puting higher values does not makes the trick, I already tried. What I need is a filter that takes the square waves of the PWM exiting from the LED and take them to a flat level that does not make interference with the speaker.
How is that possible?

I put a picture of the partial schematic if you need more specific ones I will have to provide it later. Tahnks for any help or hint you can give me!

You need more supply decoupling, like the last circuit on this page.
http://www.thebox.myzen.co.uk/Tutorial/De-coupling.html
Also a good layout is important.

Thanks, the "last" is the one with the inductor am I right?

For your worries about the layout, as I said my picture was simplified I usally draw schematics and PCBs with dedicated software, but what I needed was an hint like yours not a complete re-draw of the schematic that is what usually happens when you give details of the projects.

I will study supply decouping, thanks again!

Perhaps look into raising the pwm rate to a level that's not audible?

the "last" is the one with the inductor am I right?

Yes that's it.

For your worries about the layout, as I said my picture was simplified

No it was the physical layout of your circuit not your schematic that is a worry. Short wires and common point star grounds will minimise the interference as well as not running the cables carrying heavy current close to the processor.