Need help w/ lo-noise precision shunt Volt regulator

I'm trying to decide which approach of two different shunt voltage
regulation schemes will result in the lowest noise for supplying a
LM3875 based home audio amplifier to drive high efficiency(100-107 db)
horn speakers. Granted, the LM3875 already has a PSRR and CMRR of 120
dB, but I'd like to take out the noise before it hits the chip.

The amp can draw considerable current(4-6.5 A) for short periods to
2-5 mS which is the reason I'm choosing shunt over normal in-line
reuglation. I don't want the max current rating of the regulator to be
the limiting factor in performance or have to add banks of capacitors
after regulation to make up for that shortcoming. Experimentation has
shown that 1000uf at the pins of the LM3875 prevent oscillation and
provide the best performance without making the sound sluggish.

To make up for that, I'll be using a a battery supply to avoid AC line
noise and provide the needed capacitance, but this has it's own set of
noise problems which can be viewed at the bottom of this page:
http://www.tnt-audio.com/clinica/regulators_noise4_e.html

Anyway, on to shunt regulation and noise filtering:

1. The Two-transistor Shunt Regulator can be found at the top of the
same page above and boasts an average noise level of -100dB across the
normal audio frequency range. It seems fairly simple and doesn't need
any tweaking.

Or

2. A precison hi current shunt regulator using TI's TL1431 "Precision
Programmable Reference" followed by LM833/2N4401 noise shunt which
boasts a 40dB noise reduction with careful construction.

The particular circuit "Precision High-Current Series Regulator"
can be seen in Figure 18. on page 13 of the datasheet at:
http://www-s.ti.com/sc/ds/tl1431.pdf

Ahich of thee two approaches seem best?

And, I'm hoping someone can tell me what rating resistors are best for
low-noise supply? I'm assuming 1 watt 5% wirewound will do, although
something tells me that 1% might be best in the critical Vref spots.

Also, are electrolytic caps best in these circuits and what mx voltage
rating?

Any advice greatly appreciated,

 

That "regulator" shown inthe first reference has a lot of design
problems.
First, it has temperature problems, due to the fact that the Vbe of Q1
changes over temperature.
Second, unless a goodly current passes thru that zener, it will
*generate* a lot of ugly noise with a large 1/F component.
Frankly, it sucks.
The various "band-gap" shunt regulators (one of which referred to by
the second link) is far better.
Granted, one has the problem of integrating them with a power
transistor to get similar high current results, but figure 20 on page 13
is an excellent start.
Adding a capacitor from collector to base of the transistor will help
decrease the impedance at the high frequency end.
You must realize that the maximm voltage Vo for most of these band gap
references is 6V, and not the 10V that it seems you would like.
I see that the high current TL1431 can go to 35V, which is better than
the low power ones that i have been investigating; good choice.