LM317 Dual Shunt Regulator

[thyself]

I have recently discovered these schematics showing the use of LM317 as a Shunt Regulator.
I haven't tested it yet, but I wonder If anyone here have used this approach and would like to comment?

[jwet]

That's a strange circuit.  Any time I see 470 uF caps on low level signals, its a clue that this is not very conventional.  I can't see any special benefit to the circuit.  The first 317's, are set up as a current source and the downstream 317's are shunt regulators basically multiplying the output current of a zener diodes but really just acting like a protected transistor.  The overall accuracy will be no better than these zeners.  It might offer some benefits in cross regulation but a straight 7815, 7915 would probably be better overall.

[inse]

The circuit is also known as rail splitter to generate a virtual ground.
I have this as add-on to my lab supply, slightly different to deal with variable voltage.

[MadScientist]

That's a strange circuit.  Any time I see 470 uF caps on low level signals, its a clue that this is not very conventional.  I can't see any special benefit to the circuit.  The first 317's, are set up as a current source and the downstream 317's are shunt regulators basically multiplying the output current of a zener diodes but really just acting like a protected transistor.  The overall accuracy will be no better than these zeners.  It might offer some benefits in cross regulation but a straight 7815, 7915 would probably be better overall.

With respect your analysis is incorrect , this is quite a standard application and has appeared in many an application note

[jwet]

Never seen in it and I was an FAE and Apps manager at Maxim for 25 years!  I'm talking specifically about the first cirucit with the cross coupling between the two regs.  The other circuits are pretty conventional.   Can you point me to an app note number or a reference circuit A?.  Maybe I'm just not recognizing it in the way its drawn- doesn't look familiar or especially useful to me?

[wraper]

The circuit is also known as rail splitter to generate a virtual ground.
I have this as add-on to my lab supply, slightly different to deal with variable voltage.

The problem is that first circuit has a secondary winding with a center tap which is connected to the virtual ground output  Second circuit is similar but ground is not actually virtual at all.

[inse]

Of course you’re right, I concentrated on the last circuit…

[thyself]

I believe the third circuit is incorrect. It is missing the source resistor or some kind of a CCS. It should be like that:

[thyself]

I understand the first LM317 is configured as a CCS to provide the source current for the zeners.
I don't understand why the zeners are not referred to the ground?
instead to the opposite rail.

[inse]

This is the rail splitter I built (courtesy of goldpt.com).
https://www.goldpt.com/virtual%20ground%20goldpoint%20basic.jpg

[thyself]

I understand the rail splitter concept. What I don't understand about the use of 317 as a shunt regulator is what function it performs, especially  if we already have a zenner regulator (even with CCS on top) a front of it? The zenners connected to the adjust pin just lift the ref voltage up, what else is there?

[bdunham7]

I don't understand why the zeners are not referred to the ground?
instead to the opposite rail.

Consider that the LM3x7 is working 'backwards'.  Its basic function is to maintain it's OUT pin at 1.25V above (or below for the LM337) the ADJ pin.  It does this by conducting if that voltage is below 1.25V and not conducting if it is above.  Now look at the circuit and imagine the voltages are such that the ADJ pin is either more or less than 1.25V below the OUT pin (of the LM317) and see what happens in each case.

I understand the rail splitter concept. What I don't understand about the use of 317 as a shunt regulator is what function it performs if we already have a zenner regulator (even with CCS on top) a front of it?

In the first circuit, the CCS will be putting out ~300mA.  If you just used a 15V zener as a shunt regulator, it would dissipate 4.5W. 

The first two circuits seem pointless (as a whole) and I'm not sure what the reason would be for using a shunt regulator where you have anything other than an isolated DC source to deal with.  They look like they were designed by an audiophool with OCD, especially the first.  10mF filter caps!   

[thyself]

Does that mean the 317 as a shunt after the zeners will "straighten" the residual ripple of the zeners?

[thyself]

I am asking because currently I have a project where I need to bring 40V after capacitance multiplier down to 16V for powering 20 OpAmps, and I wonder which route to take: zener (simple, cascaded, bridge, etc.), transistor (serial, shunt, etc.), LinReg (LM317-723 etc.), OpAmp+Mosfet, LDO, SMPS, etc...!?

[bdunham7]

The problem is that first circuit has a secondary winding with a center tap which is connected to the virtual ground output  Second circuit is similar but ground is not actually virtual at all.

The first two are not rail-splitters at all, except maybe the first is sort of by accident.  And it would be interesting to see what happens at the junction of the two huge 10mF filter caps since whoever drew that omitted a connection to ground for that section. 

[bdunham7]

I am asking because currently I have a project where I need to bring 40V after capacitance multiplier down to 16V @200mA for powering 20 OpAmps,

Single ended, not split/bipolar?

[thyself]

I am asking because currently I have a project where I need to bring 40V after capacitance multiplier down to 16V @200mA for powering 20 OpAmps,

Single ended, not split/bipolar?

Bipolar +-40V to +-16V.
I thought of using LM317-337HV, which I have in stock, but I am not sure about PD, i.e. 4.8W (24V drop*0.2A) at least.
In the datasheet the PDmax is missing, it shows just "internally limited"

[bdunham7]

Bipolar +-40V to +-16V.
I thought of using LM317-337HV, which I have in stock, but I am not sure about PD, i.e. 4.8W (24V drop*0.2A) at least.
In the datasheet the PDmax is missing, it shows just "internally limited"

You wouldn't actually need the HV part, but it would give you a bit of safety margin in case of certain malfunctions.  I didn't know there was an LM337HV available in TO-220.

As for the power dissipation, you just need a 100R 5W dropping resistor before each regulator.  That way at full load, the resistor will take 20V of the drop and still leave you 4V for good operation of the regulator.  That gives you 0.8W, which is handled by a small solder-in finned heatsink.  If you want extra filtering, just use 3x 33W 5W resistors in series with a 470uF capacitor at each stage. 

PDmax being internally limited just means that it will fold back if it gets too hot and the actual PDmax is determined by ambient temp and how well you heat sink it.  You don't want to be trying to keep an LM3x7 going with 4.8W dissipated as this would take heroic levels of heat-sinking.  Resistors like that sort of thing, semis don't.

[thyself]

Thank you for the advice bdunham7!
What about a Zener shunt after the 100R/5W as a pre-regulator to the 317?
20V zener will provide 200mA, and 4V remain for the 317.
Would it be beneficial in terms of noise?

[schmitt trigger]

I am going to bring, and perhaps kill, the elephant in the room: would you consider using a SMPS? And perhaps follow it with a linear regulator?

-ducking for cover…

[thyself]

I am going to bring, and perhaps kill, the elephant in the room: would you consider using a SMPS? And perhaps follow it with a linear regulator?

Well it is a PSU of a guitar amplifier, maybe it would bring noise and harmonic artifacts...

[bdunham7]

Thank you for the advice bdunham7!
What about a Zener shunt after the 100R/5W as a pre-regulator to the 317?
20V zener will provide 200mA, and 4V remain for the 317.
Would it be beneficial in terms of noise?

You could do that, although I'd at least use 3x ~6.4V 5V zeners in series to keep them from cooking.  The only way I see it as having any benefit is it would keep your RC filter loaded if you were using the 3x 33R + 3x 470µF I suggested.  With a straight resistor I don't see how it would help.  You could also use the zener(s) directly as a dropper instead of the resistor.
However, in either case I'm dubious of using zeners like this because I've seen examples like this having reliability issues and any noise benefits are questionable, as in unproven.  The LM317 is really, really good at filtering out ripple, especially if you use the recommened additional components like a small cap right at the input and an ADJ pin capacitor.

[thyself]

Is that what you mean?

[bdunham7]

Is that what you mean?

Something like that, except I don't see why you want to use D1 and D2 instead of a ~2.6k resistor?  And I don't think you need C4 unless your layout is unusual. 

I neglected to ask what the exact source of the 40V is or how much ripple it has, nor how quiet you need the output to be.  Many op-amp circuits will have pretty good PSRR anyway.  So the RC thing may be overkill, but it also may not cost that much to implement since you can probably use 33R 2W resistors.  Don't spend extra for non-inductive ones! 

[thyself]

Corrected