Super Regulator for a Gainclone

A basic idea around the Gainclone concept is to keep it simple so using a LM338 like carlosfm has done is right but what if an even better regulator (note, claimed for being better) is used instead? Will there be any further improvements?

Let's say we design a super regulator good for 10 A and otherwise designed for a load such as LM3886 max but of course LM3875 also be OK.

Try to make this regulator work with AD825 but this is not important if it turns out that this opamp is a bit too fast. Of course there will be room for regulator DIL08 opamp.

Magnify all parts which has to do with the higher current.

Add overtemperature protection which shuts down both voltages at the same time.

The heatsinks will be on the pcb.

Maybe add current limiting but this may be not necessary.

Design this regulator as a complete unit (rectifiers, smoothing, snubbers, regulators, heatsink, protections, connectors), just add low voltage AC, two windings or center tapped. The reason for this is to eliminate all wiring which is extremely sensitive.
there's no problem with the AD825 (or AD817), it's the AD797 which caused problems -- Walt Jung pointed out (Audio Electronics 4/00) that the AD797 was somewhat prone to noise pickup in one of his aps -- the '797 can pick up low millivolts of RFI causing it to stray from linearity.

but there's a more significant problem -- the error amplifier itself may be operating too close to the top-end of its supply range in the super-regulator arrangement. for the AD817 this is 36V. solution is to pre-regulate (which as Jung pointed out, reduces the stress on the super-reg).

if you are really going to do it correctly, each gc should have its own regulator-- and the "sense" circuits should be tagged right onto the Vcc/Vee pins.

quite frankly, if regulating in this application I would prefer a compensation network over the super-reg. but what the heck try it, then test it with this:


when I tested some GC's -- LM3875 and LM4780 based -- I could see no performance difference between a regulated and non-regulated supply. subjectivists might differ in their opinion.
jackinnj said:
there's no problem with the AD825...
No, not at small currents but what happens when you scale up? Just assume that the regulator will loose in speed => potential stability problems. I may be wrong but I count on this.

At this point I see no problem with the voltage stress of the opamp. I'm sure this problem is possible to solve. One simple solution is to limit the output voltage to 30-35 volts.

2006-06-14 10:24 pm
How about this for a gainclone power supply...


More information HERE.

This PSU was developed for a headphone amplifier! For this application it seems way over the top so I think it can be considered high end enough (a super regulator?). It is an all discrete design, no opamps. Works on high enough voltage for a power amp and the pass transistors are MOSFETs rated at 17A each.

Best part is the schematic is open source! Excellent!

I'm interested to hear what you make of it, P-A.

2006-06-14 10:24 pm
Hi, Andrew. I think it can be modified to be more like a super regulator as you describe but it will need more parts. I guess it just needs the MOSFETs to conduct a little bit so that the regulator can start up and take over. Like how it is done with the Jung super reg.

It looks like it is the job of Q1 to just improve the PSRR. So perhaps Q1 can be omitted if the reg is powered from the output. So maybe it is possible to make it with less complexity and better performance. I'm not sure how to implement it though.


2005-01-17 5:29 am
AndrewT said:
if c7 an q1 are fed from the regulated output then it can try to qualify as a super-regulator.

The op amp in the first Jung Super Regulator drew power from the input. The op amp in the second Jung Super Regulator drew power from the regulated output.

So, if CR1 and D5 are fed from the regulated output then it could possibly still be considered a "super" regulator, whatever that may mean. An improvement would involve using the regulated output as the power source for the error amp as Andrew stated.