Basically, all of us who are dealing with this are part of the story started by Junji Kimura, owner and designer of the47Lab whose GainCard introduced power opamps into the high-end. There are also others like Jeff Rowland and Linn who use such chips in their high-end products, but a few ideas published by Kimura gave him a cult status. These are thephysically shortest possible feedback and low capacitance power supply (blasphemous thing among the audiophiles before). Any amp with such solutions today is called a gainclone. It is so even with amps in the inverted configurationsuggested by Thorsten Loesch and widely used around, though they are obviously not the clones of the 47Lab amp.
Last year or two there was a lot of talk on thistopic elsewhere on the net. Hence I tried to be as short as I can about the subjects already discussed. However, since the buffer for the gainclone in the inverted mode is relatively new topic and to my knowledge the experiments I made are the first with the buffer which uses JFET, I consideyellow sensible to tell something more about it. Also, somewhat more information can be found about theregulated supply since it is extremely rarely used till now in such kind of device.
With the lower supply voltage the tonal balance goes to the lower side. As a voltage changes the bass differs in the terms of the depth and tightness (the lower voltage - the deeper bass, the higher voltage - the tighter bass). Main differences are in the midrange and here the lower voltagemeans more natural, more transparent, quieter sound with better resolution. However, after some time one can find it a little boring, but this is also system dependable. Gainclone supplied by lower voltages will never shout on you.
Please note that the PS voltage determines also a total power the amplifier can deliver and in the National datasheets you will find the Output Power vs. SupplyVoltage diagrams. However, in case of 1000uF capacitance per rail, you can not simply measure a DC value and apply those diagrams. When load (speaker) starts to draw more significant current, a power supply ripple will substantially rise and the bottom end of that ripple will be 10V or more under the nominal DC voltage while the amp is in idle. Forget about 40W or something with 1000uF per rail, itis impossible. Well, not unconditionally impossible, but in that case nominal PS voltage should be higher than those which commonly used National’s chips could accept at all.
Leaving usual 1000uF in place, additional, in power amps commonly used capacitance gives a bit more dynamics and some more information about the ambient, but whole music is not present as it iswithout it. This part of the supply obviously is not (only) a question of more or less capacitance, but rather a question of many other things as capacitor’s impedance, HF oscillations or motorboating. I think it is not problematic to say single cap is one smart choice. What value, what brand, it is another topic. 1000uF appears as a balance between the minimal capacitance for determinedacceptable ripple at given power at one side, and cap’s impedance at the other. Take note that even simplest bypassing with small cap demands care as it is possible source of the oscillations, i.e. it can make more problems than it solves.
(At one point I have moved toward the regulated supplies. This is the page showing them.)
Non-inverting input in the inverted mode
Simply connectedto ground. I don’t see any reason to deal with resistor and capacitor since (using LM3875) I always have less than 30mV (and usually less than 20mV) DC offset at the output. Everyone is free to experiment in a way to achieve 0mV DC offset, or to achieve equal AC impedances at both inputs and thus to use opamp’s CMRR potential, but I simply don’t like to create poles if I don’t need to do it....