Split SMPS

[N-Brock]

Hello all,
A bit of background first. I built an amplifier using the TDA2040 and the schematic found on page six of this pdf: http://www.datasheetcatalog.org/data...onics/1460.pdf (1.83MB)
For a power supply I decided to use two matching SMPS to get the split voltage I need to run this type of amp. The top schematic of the first picture shows how it is set up as of now. When all is hooked up, a loud tone is created. Clearly something is wrong and I think it is the grounding. the second and third schematics (pic 1) are what I think might fix the issue and further improve he sound (smoothing caps?). The input, output, and power grounds from both channels are all connected though the heatsink/chassis. Could that be the problem? The datasheet says that isolation between the heatsink and chip is not needed but it might be the problem as well.

Anyone know what this might be caused by?

Edit: the first picture went below and is titled "split psu.bmp"

[cocolino]

The datasheet says that isolation between the heatsink and chip is not needed but it might be the problem as well.

I´ve been to lazy to search where in the datasheet it`s stated that chip isolation is not needed in this case but as I see it this can only be true for a 'single voltage supply' not a 'bipolar supply' (aka split-power-supply). Maybe an error in the data-sheet? There is already one: Figure-12 is refered to as "Amplifier with Split Power Supply" whereas what is shown is clearly not "Split".

The datasheet says tab of device is connected to pin-3 which is V-.
With a 'split power-supply' with it`s V- connected to pin-3, You`ve grounded it (short circuited) it through the unisolated tab of the chip.

I`d say isolating the chip from the case / heatsink should fix Your problem (provided Your SMPS are short circuit proved and the V- power supply part survived).

[N-Brock]

Thanks cocolino, I may have to call this the "moody amp"
Yesterday, I isolated the chips from the 'sink as suggested and they worked great, I played them for over an hour before power down to make some improvements. I added a mesh cage and gave the power supply a new home (pics below). Well, today I fired them back up and once again I got the harsh buzzing and various tones from them. I was not pleased, but also not discouraged. I have decided that the turn on time of the SMPS's causes the sounds which continue afterward. If I wait ~4 seconds before connecting the amp, everything works pretty well. It is still a bit finicky but this seems to have fixed most of my problems.

Does anyone know if the below modifications will help or hurt the performance?
Grounding through a low value resistor or thermistor
Adding smoothing caps to the PSU (did I even draw that right?)