Who tried Class D only to return to S/S or Tube

Ait, Perhaps I'm missing something.  Imagine rectifier charging capacitors to 50V with 5mV ripple present. 

http://i.stack.imgur.com/BLGnf.gif

Capacitors are charged ONLY during voltage increase on capacitor.  This period for 5mV ripple (1% of 50V) is:

Arccos(0.999) x 16.6666ms / 360deg=0.12ms  

Charging current has to be huge (hundreds of amperes) during such short time to meet average current draw.  In addition to this huge current spikes rectifiers are not fast enough to turn off at the peak of the voltage conducting current for a moment in opposite direction to snap back to zero.  That creates extremely short spikes that large inductive caps won't filter and that will couple to any LC circuit.  Using "softer" diodes (faster switching - slower snap back) helps to reduce this noise (makes pulse longer).  There is a lot of switching noise from Linear PS polluting from 120Hz to tens of kHz

In addition to noise created by linear power supply huge spikes of current have also huge rms/average ratio causing big losses in transformer windings while high frequency component causes losses in the core (eddy currents, hysteresis).  You might filter these spikes with chokes but capacitor itself can be only charged during very short time (fraction of ms).

Ait, I understand that choke helps storing and releasing energy, but it won't affect charging of capacitors that has to happen in very short time.

I showed the current waveforms through the first PS Capacitor in each case. It's clear that in one case there are sharp spikes through the cap, in the other there are not. 
The rectifier is not directly charging the cap, it's charging the inductor, which stores the energy in its magnetic field, then the inductor output charges the cap. Filter topology is LCLC.

Your link shows a cap input supply, by the way. Put a choke between the rectifier and cap and you get a choke input supply.

Ait, I think it is possible to place very large coil in front of capacitors in order to obtain constant charging current in form of sinewave.  The problem I can see with this scheme is size of the coil (huge inductance with low resistance) and the fact that voltage on the capacitor will depend on the load.  Voltage with load should be roughly 2/3 of peak voltage, but without any load it could climb up to 100% (peak value). I guess it should be possible in class AB amps to provide minimum load with additional resistor (if necessary).   Interesting concept, but I suspect seldom used because of the cost and size.  On the other hand high end audio is never practical.  Thanks for the info - I will find some articles to read on this subject.