I have used the very same approach to power a couple of projects that require mains->DC. I have bought cheap laptop adapters on ebay, snapped their out of plastic cases open, and piut them inside my own case, also used them in series to obtain 19.5V+19.5V.
However I do not do that anymore because:
1) the quality of the output is crap. there is a lot of noise due to a Y2 capacitor between primary and secondary of switching transformer.
2) the output is dangerous : plug the scope on it and you will see wave forms of about 100V-180V peak, each side. I got jolts off of them, and they ALL do it, even the more expensive SONY VAIO adaptors (but less).
3) most adaptors on ebay, to fix problem number 2 above, connect the mains lead earth to the V- -- that means that want it or not, your V- is grounded. Try to put them in series or in other configurations and there will be fireworks. In my case I purposely disconnected the earth lead, but then you have problem number 2 above.
I presume there must be SMPS mains-DC converters out there that do not suffer all the above, but I have bought literally dozens and tested even those that came with the laptops, and they all do it, some more, some less, but unacceptable for anything other than the most crude circuits.
I noticed the Y2 ceramic between primary ground and secondary ground, and left it there. It was 2 kV rated, IIRC. One could perhaps add an X7R across the main secondary electrolytic to reduce some of the HF hash.
Hmm, this has got to be a faulty adapter or something. Sure, the primary side will have spikes (at least on the drain/collector of the switch) of 500-600V. But it should not appear on the secondary side, regardless. But maybe the Y2 cap has something to do with it, I'll check it out later.
This can also probably be fixed by putting a toroid on the DC out leads and winding a few turns through it.
That's a problem, but the ones I'm using are 2-lead, L+N only.
Thanks for the tip, I'll check them out in due course. In my case, I have large electrolytics (4700uF) on each rail in addition to the 470uF at each secondary, and that's probably why I can't see or hear anything offensive.
I removed all decoupling caps. As a matter of fact, even in a simple chip amp with the LM3886, I do not use local decoupling. I do use snubbers though. Decoupling caps seem to add coloration to the sound, a bit fuzzier sound. All this balls down to the return current flow in the ground which flows into the caps and increases interaction with the output.
I have used switching power supplies with great satisfactory. Also just got a design through the required Taiwan BSMI certification recently with them. Modules are commercially available but it took some effort to integrate them with the amplifier for best sonic results and still satisfy the certification requirements. Did test them with the MyRefs, but never tried to optimise the integration.
This result probably indicates that you implemented the local decoupling in the wrong way, or in the wrong location.
No.
I said
Note, I am referring to "you".
It seems that "you" have implemented the local decoupling incorrectly.
Actually my first discovery of the adverse effects of decoupling caps on the LM3886 was using the Twisted pear version of MyRef revC, it is mentioned in the thread.
and yet you ignore National's datasheet recommendation to install four decoupling caps on the 3886 implementation !
This implies the Twisted Pear layout, earthing arrangements & PSU wiring need sorting out.
Does this have separate Clean (signal & feedback only) & Dirty (decoupling & other yucky stuff) Earths?
If you decouple to a Clean Earth, you make it Dirty.
There are further subtleties in the PSU and capacitor charging currents.
I agree that ground arrangement is very critical, and also the PSU and capacitor charging currents. I do not see much options in tha various LM3886 applications, and I doubt anyone has done much experiment with grounding in these applications. I have done a few, but limited to the constraints of the boards in particular application. I wonder what the true experience of others are whom use decoupling. Size of capacitance? Type of capacitor? Relation with sound? Consistent with different designs?
The MyRef designs have a ground layout that doesn't conform to TI data sheet recommendations. TI place the star ground between the bypass capacitors. The MyRef designs place star ground at the PSU (between the 10000 uF capacitors) with a single trace to the 4 bypass capacitors. Can only guess what the current along that trace must look like.
I am curious, in a design with many devices requiring decoupling caps, can the TI recommendation be realistically implemented? They have to be connected to PS ground, and there is always current flowing through the ground, the only way you decouple the decoupling caps from the PSU ground is using snubbing resistors. Hmm, just got an idea.