I found this schematic and it looks interesting like Sigma11 by AMB. But it doesn't need that hard to find current regulator diodes. What's more, it can be simplified and you can leave out short-circuit protection if you wish.
 
Here's a link, but I don't know who the author was: http://www.extremecircuits.net/2010/06/ultra-low-drop-linear-voltage-regulator_14.html .
 
I copy it here just in case.
 

 
Here's the original text:
 
This circuit is a Mosfet-based linear voltage regulator with a voltage drop of as low as 60mV at 1A. The circuit uses a 15V-0-15V transformer and employs an IRF540 N-channel Mosfet (Q1) to deliver the regulated 12V output. The gate drive voltage required for the Mosfet is generated using a voltage doubler circuit consisting of diodes D1 & D2 and capacitors C1 & C2. To turn the Mosfet fully on, the gate terminal should be around 10V above the source terminal which is connected to the DC output. The voltage doubler feeds this voltage to the gate via resistor R3. IC2, a TL431 adjustable shunt regulator, is used as the error amplifier. It dynamically adjusts the gate voltage to maintain the regulation at the output. With an adequate heatsink for the Mosfet, the circuit can provide up to 3A output at slightly elevated minimum voltage drop.
 
Trimpot VR1 is used for fine adjustment of the output voltage. The RC network consisting of R5 and C6 provides error-amplifier compensation. The circuit is provided with short-circuit crowbar protection to guard against an accidental short at the output. This crowbar protection works as follows: under normal working conditions, the voltage across capacitor C5 will be 6.3V and diode D5 will be reverse-biased by the output voltage of 12V. However, during output short-circuit conditions, the output will momentarily drop, causing D5 to conduct. This triggers the MOC3021 Triac optocoupler (IC1) which in turn pulls the gate voltage to ground. This limits the output current. The circuit will remain latched in this state and the input voltage has to be switched off to reset the circuit.
 
 
I tested this circuit today and it works but I didn't have irf540 but irf520. So i needed to tweak resistors r6, r7 and r8. And I left out short-circuit protection: r1, r2, zd1, c5, ic1, d5 and r4. And off cource you can use diode bridge rectifier, transformer with single secondary winding and I would use c3 and c4 low esr with higher value.
 
Hope I will have irf3205, irfz44n, irfz48n or irf540 yesterday and be testing more. They have less Rds(on) like irlz24n in Sigma11.
 
I don't know how good is this voltage regulator in comparison with Sigma11. I tested it with my mosfet amp and I noticed a bit strange colored sound. But I guess that is because that high Rds(on) off irf520. Or maybe it is how it should sound. I had 78s12 regulator and now I'm using lm317t which is a bit better and I believe with a bit more bass.
 
Edit: I'm currently using this regulator with 12V transformer and works great. Vin drops to 15,9V, Vout is set to 15V (!) and Iout is 0,5A. So no need for huge Vin-Vout voltage as with 7812, LM317T or other regulators.

So I have irf540 and it is absolutely the same. And that voltage doubler with c1, c2, d1 and d2 doubles voltage only when it isn't connected to r3. Otherwise it has almost the same voltage as on mosfet's drain. So I could connect r3 directly to drain as on other schematics. Crazy.
 
Edit: I discovered that sometimes is connecting just gate with drain via resistor enough. It obviously works for some input and output voltages. And that voltage doubler actually works even if I'm unable to meassure it correctly.
 
I'd be still interested how good is output impedance in comparison with sigma11.
 
I've just found interesting reading:
 
http://www.irf.com/technical-info/appnotes/an-970.pdf
http://www.edacafe.com/books/PSpice/san63267_ch06.pdf

Open loop gain of sigma11 error amplifier is twice as gain of TL431, so sigma11 i s definetely better (better PSRR). I've done some testing and i found, that usinig OA (and filtered zener reference) is better than any PSU with TL431 (PSRR with common NE5534 was cca 100dB).
Impedance of this MOSFET PSU wil be quite low, but TL341 will be degrading PSU perfomance a bit.

I was doing A/B testing for lm317 vs mosfet and frankly I was unable to hear any difference at all. But lm317 is better than 78s12. I'll stick with this mosfet just because I'm a mosfet fan and it's something new for me.
 
I'm not sure but I think that when my amp needs constant current and it actually doesn't "amplify" (voltage follower) I can't hear any difference. But why 78s12 sounded slightly different I don't have a clue.
 
Here is modified schematic if someone will be interested:
 

 
With higher input voltage is possible to remove c1,2 and d1,2 and r3 connect directly to input voltage on drain. I added c8 because I have right part of the schematic around 10cm further and it oscillated without it. I know that it isn't a good idea but I had a hole in heatsink and space in my case is somewhat limited.
 
Resistors r6 and r8 determine output voltage: (1+r6/r8)x2,5. For example: (1+5k6/1k5)x2,5=11,83. But it depends on real part values and adjust voltage of tl431. Now I have (1+7k5/1k5)x2,5=15 and 15,03 in reality.

akgfan said:

I was doing A/B testing for lm317 vs mosfet and frankly I was unable to hear any difference at all. But lm317 is better than 78s12. I'll stick with this mosfet just because I'm a mosfet fan and it's something new for me.

Click to expand...

It's not surprising that you didn't hear difference, LM317 has quite low noise, so noise from PSU was not present (was present, but under hearing threshol), PSRR of LM317 is also quite good and if your amp is all class A, then current drawn from PSU is constant, so high imedance of PSU doesn't (can't) degrade sound quality.
 
I would recommend to place 10 to 100nF caps in parallel to all six diodes to reduce noise since you used doides with high Qrr.

My P2P construction of this voltage regulator (Vin=15,9V; Vout=15V). Now I can run my amp on 15V which is great.
 
 

Something similar is used inside BURSON Conductor Virtuoso 2+: 4 x BJT transistors, 1 x TL432 and one IRF610 (or 9610) per rail. Input voltage is +/-35V, output is +/-21V. Initial AC ripple is 40mV RMS (8200uF/35V) and after regulation gets below 1mV RMS.

Thanks for sharing the schematic!
BTW, audiophonics.fr has a dual PSU based on this schematic, in case anyone's interested; it's PCB board fully populated, but without the case and wires. I assume eBay is also full of such modules.