How about low noise SMPS fed buck converted DC instead of the AC mains?

[Kokoriantz]

I have seen this in a schematic of TI, first converting the mains into DC switching regulated high voltage, then switching regulated the low voltages to follow the load. The problem, I don't remember where?
Most of the noise in the outputs are the ripples and mains noise. Once cleaned, very little be left because of jitter in the PWM.
 

[Swainster]

This is not straightforward to implement as the bandwidth of linear regulators typically doesn't extend up to the frequency of switch mode noise - Dave did a video on this.

I guess you could get it to work with suitable filtering between the switcher and the linear regulator, but then it may make sense to skip the linear regulator altogether.

That said, I seem to recall that Riden make an adjustable supply which is said to use this technique.

[Kokoriantz]

I thought about linear capacitance multiplier using a MOSFET, With 30v ripple on 470uF 300v at 1A, it will dissipate 15W. May be it is enough, but I am sure I've seen a buck converter do it. To notice that such a AC/DC filtre can be universal to feed any SMPS.

[Mechatrommer]

your psu in other thread does that, rectify AC to 330/120Vdc and then switch to hi freq AC to transformer using mosfets. this is common way today. if what you meant is without the transformer, it can be done but you dont have galvanic isolation.

[Kokoriantz]

Instead of powering my switching power supply by AC220v and ask the regulator to adjust for both the primary high voltage ripple and the variations of the secondary load, I provide a high voltage DC 260v already cleaned from ripples and noise.
Bellow is symbolic analog type, but I have seen TI use a buck converter instead.

[Swainster]

Oops, I didn't read your initial post carefully, and had some other idea in my head. For a buck regulator, it should already be very good at filtering differential noise from the source (due to the switching inductor). For common mode noise then a common mode choke makes more sense.

For 1A 300V then I think EU regs would probably require power factor correction.

[Kokoriantz]

I will built the circuit bellow, I gathered all the components with 2SK2750 MOSFET that I have several of it. The output voltage shown is the start up with the load 0.5A modulated +/-0.3A 10Hz square wave. Thanks to the two diodes, the MOSFET gets the minimum voltage across it at all current loads eliminating the ripple of the reservoir capacitor.
For the common mode noise, I have a problem to resolve. If I add a CM inductor + a pair of capacitors to earth, the ones that are inside the SMPS which are now filtering the common of the DC will shunt the rectifier diodes of my circuit. The size of the CM inductor is also I don't know how to determine. Coilcraft site gives all details but nothing about the size. 

[David Hess]

The line regulation should be plenty high to remove AC ripple from the DC input.  If an active power factor correction stage is providing the DC, then either an LC filter can remove the high frequency ripple, or the PFC stage and regulator can be synchronized, which is the common case now.

[Kokoriantz]

As this SMPS is intended for HIFI and not PA use, it will rarely go to 250W. This means that the linear source follower MOSFET will have to dissipate peak 15W but less than 5W in average for maximum music level. This said, to use a buck converter is not justified. Still I want to see again what TI designers has proposed, if only I can remember which class D amp application was it?   

[Kokoriantz]

I found from TI website the block diagram of the power supply. It is not buck converter but a booster bringing the AC mains to 400v DC non isolated as the DC DC converter is isolated.
https://www.ti.com/seclit/ml/slup382/slup382.pdf