Virtual ground circuit

[aheid]

I was looking for a way to split the PSU rails, ie introduce a virtual ground for my circuit, and stumbled over this design using an LM317 and LM337 pair: http://www.goldpt.com/virtual_ground_circuit.html

I have a couple of questions regarding the circuit:

1. Why the 1 Ohm resistors on the output of the regulators?
The page says these resistors can be lowered by reducing the current through the 2.5V reference, but I don't get their purpose or their connection to the voltage reference current.

2. What about the minimum load the LM317/LM337 requires to maintain regulation?
If the load is asymmetric, then one of the regulators will essentially not be needed and thus, as far as I can see, the circuit is unable to ensure the minimum load current for the "idle" regulator. What will happen? I don't have any 2.5V voltage references handy, or I'd wire it up and see.

For my project I was hoping to be able to use a 19-20V laptop PSU, boost to 36V, and then split to +/- 18V rails capable of handling up to 1A. If I could lower the 1 Ohm resistors and the minimum load current is not an issue (or can be handled) then this looks like it should be capable of that. If not, any tips for alternative, or better, circuits?

[bitbanger]

This is far from a technical answer and admittedly more of a hunch, so take this for what it's worth.

I suspect the 1ohms are what's sometime referred to as "ballast", or ballancing resistors. The amount of regulation error you get out of a zener is related to the ratio of current you pass through it, serially, and the current that is passed off to the 'load'. You want the serial current to be an border of magnitude larger, so that load current fluxuations don't influence the regulation. The adust pins on the LMs have a certain amount of error current passing into them, and so based on the external zener current, it may influence the "regulation" of that zener and subsequently the current into the ADJ. In any case if the ADJ currents were slightly mismatched, and the two out pins were tied directly together, the regulators would fight one another to maintain regulation. The 1ohm, especially considering the wattage spec'd, is intended to ballance any disparity at the output by dropping it across one or both of those 1ohms.

Would welcome some technical insight by someone who may know a little more detail. Again, on my phone, off the cuff hunch.

Best of luck.

[spec]

Hi aheid

I was looking for a way to split the PSU rails, ie introduce a virtual ground for my circuit, and stumbled over this design using an LM317 and LM337 pair: http://www.goldpt.com/virtual_ground_circuit.html

I have a couple of questions regarding the circuit:

1. Why the 1 Ohm resistors on the output of the regulators?
The page says these resistors can be lowered by reducing the current through the 2.5V reference, but I don't get their purpose or their connection to the voltage reference current.

2. What about the minimum load the LM317/LM337 requires to maintain regulation?
If the load is asymmetric, then one of the regulators will essentially not be needed and thus, as far as I can see, the circuit is unable to ensure the minimum load current for the "idle" regulator. What will happen? I don't have any 2.5V voltage references handy, or I'd wire it up and see.

For my project I was hoping to be able to use a 19-20V laptop PSU, boost to 36V, and then split to +/- 18V rails capable of handling up to 1A. If I could lower the 1 Ohm resistors and the minimum load current is not an issue (or can be handled) then this looks like it should be capable of that. If not, any tips for alternative, or better, circuits?

Quite a clever rail splitter you have found. You ask what the two 1R resistors are for: here is my explanation.

The rail splitter works by setting the reference input of the LM317 and LM337 to VCC/2 -1V25 and VCC/2 + 1V25 respectively. This is the essence of the design, and it is also a critical area because there is a voltage tolerance issue concerning LM137 Vref (Vref1), LM337 Vref (Vref2) and the LM336 Vref (Vref3). The  circuit will only work if Vref3 is equal to or greater than Vref1 + (-1 * Vref2). If this condition is not met, there will be insufficient voltage between the LM317/LM337 output pins and reference pins, so that the LM317/LM337s will be in an indeterminate state.

The first order tolerances taken from the datasheets are:
Vref1 and Vref2: 1V2 to 1V3 (combined 2V4 to 2V6)
Vref3: 2V39 to 2V59

So you can see that the designer has been careful to match Vref3 (LM336) to the combined voltages of the LM317/LM337 Vrefs: just 10mV off at the worst case extremes of all three voltages.

But there is another effect of the above reference voltage tolerances: the LM317 and LM337 will each be trying to hold their output pins to different voltages. Taking the worst case from the first order values above, the difference between the LM317/LM337 outputs will be 2V59- 2V4 = 0.19V. In that case the current flowing through the two 1R resistors will be 0.19V/2 = 0.095A. Without the two 1R resistors the LM317 and LM337 would fight with each other, poring in more and more current, until they reached a current or thermal limit, when the output voltage would be indeterminate. So that is what the two 1R resistors are for.

http://www.ti.com/lit/ds/slvs044x/slvs044x.pdf

http://www.ti.com/lit/ds/symlink/lm336-2.5-n.pdf

[David Hess]

I was looking for a way to split the PSU rails, ie introduce a virtual ground for my circuit, and stumbled over this design using an LM317 and LM337 pair: http://www.goldpt.com/virtual_ground_circuit.html

That is clever.  DC accuracy at high currents will be poor but a shift of 1 volt/amp is acceptable for many circuits.  This circuit might be thought of as a class-ab output stage using voltage regulators instead of transistors.

1. Why the 1 Ohm resistors on the output of the regulators?

The output resistors limit the current when the reference voltage does not quite match the regulator voltages.

2. What about the minimum load the LM317/LM337 requires to maintain regulation?

Without the minimum load, the output of the regulator will rise but that is not a problem here with the other regulator ready to pull it down through the output resistors.  With a floating regulator like the 317/337, a resistor can always be placed between the output and adjustment pin to provide a constant minimum load current if there is no requirement for low adjustment bias.

For my project I was hoping to be able to use a 19-20V laptop PSU, boost to 36V, and then split to +/- 18V rails capable of handling up to 1A. If I could lower the 1 Ohm resistors and the minimum load current is not an issue (or can be handled) then this looks like it should be capable of that. If not, any tips for alternative, or better, circuits?

If the reference voltage could be trimmed, then the value of output resistors could be lowered.  The reference could be replaced with another LM317 configured as a constant current source and an adjustable resistor but there is something good to be said about circuit which cannot be misadjusted.

Perhaps better would be to add a single operational amplifier to provide feedback and hold the output exactly where desired.  Then the value of the output resistors becomes less important.

This circuit will be difficult to beat from a price/performance perspective at high currents.  Unlike a power transistors boosted operational amplifier design, the integrated regulators provide thermal protection and current limiting.  Note that integrated class-ab audio amplifiers make pretty good high current rail splitters.

I am not a believer in the audio qualities of rail splitters for the simple reason that if the rail splitter matters, then the audio amplifier has something wrong with it.  For the same reason, the power stages of an audio amplifier should not require a regulated low noise power supply.

[Audioguru]

Usually the (+) input of an opamp or power amplifier is at 0VDC and draws VERY low current. Then the rail splitter does not need high power ICs, instead the rail splitter is two medium value series resistors across the supply and their junction is 0V.

[strawberry]

It could be done with OP, TLE2426

[spec]

For my project I was hoping to be able to use a 19-20V laptop PSU, boost to 36V, and then split to +/- 18V rails capable of handling up to 1A. If I could lower the 1 Ohm resistors and the minimum load current is not an issue (or can be handled) then this looks like it should be capable of that. If not, any tips for alternative, or better, circuits?

Although the LM317/LM337 rail splitter is clever, I would not recommend using it for the reasons outlined above and, in general, rail splitters of any kind are best avoided if possible. Remember that the 0V line is the most important voltage for the correct operation of any circuit and, for that reason, the 0V line is sacrosanct, especially for audio amplifiers.

Instead of using a rail splitter, I would suggest generating a -20V rail from the 20V rail from the laptop PSU.  You would do this by using an inverting SMPS circuit.

Alternatively, why not just buy a PSU that provides the +- voltage rails that you want: they are cheap enough.

[aheid]

Thank you all for your input, very much appreciated!

I thought about mismatch between the regulators, but I wasn't sure about the implications.

Instead of using a rail splitter, I would suggest generating a -20V rail from the 20V rail from the laptop PSU.  You would do this by using an inverting SMPS circuit.

Alternatively, why not just buy a PSU that provides the +- voltage rails that you want: they are cheap enough.

I got some MC34063's inbound, might try to create a voltage inverter with one of those. Though ideally I was hoping to avoid rolling my own SMPS for now (never done it before). From what I understand layout is important, breadboarding is a no-go, and I don't have the equipment for making custom PCBs yet. Can I do it by using prototyping PCBs?

As for the cheap dual rail PSUs, any pointers for what to look for? The 19V 65W laptop bricks are readily available and cheap here, but haven't found anything dual rail that's available and not considerably more expensive.

[capt bullshot]

e cheap dual rail PSUs, any pointers for what to look for? The 19V 65W laptop bricks are readily available and cheap here, but haven't found anything dual rail that's available and not considerably more expensive.

Just use two of them. Ensure their outputs aren't connected to earth (easiest to achieve with a brick that has a two-wire AC input).

[Zero999]

e cheap dual rail PSUs, any pointers for what to look for? The 19V 65W laptop bricks are readily available and cheap here, but haven't found anything dual rail that's available and not considerably more expensive.

Just use two of them. Ensure their outputs aren't connected to earth (easiest to achieve with a brick that has a two-wire AC input).

That will work but beware, the leakage current will double and might exceed the maximum allowed by the safety standards. The centre tap could be connected to earth, either directly or via a suitable Y capacitor, to make it safe.

[David Hess]

I thought about mismatch between the regulators, but I wasn't sure about the implications.

There are improved replacements for the 317/337 from companies like Linear Technology (now Analog Devices) which would allow lower valued output resistors because of greater accuracy but they come with a premium price.  One of the old Linear Technology application notes discusses being able to use their regulators in parallel with just the circuit board resistance for current sharing.

[spec]

I got some MC34063's inbound, might try to create a voltage inverter with one of those. Though ideally I was hoping to avoid rolling my own SMPS for now (never done it before). From what I understand layout is important, breadboarding is a no-go, and I don't have the equipment for making custom PCBs yet. Can I do it by using prototyping PCBs?

Very wise. Yes, because of the relatively low power involved, with care, you can make a decent SMPS with a prototype construction. 

One approach is to use a piece of copper clad fiberglass PCB and outline the unwanted copper with a scalpel. Then, any copper you wish to remove, just heat with the tip of a soldering iron and peel off. Conversely you can add traces/ground planes, by  aralditing  tinned copper wire (TCC) or copper foil to the PCB.

But, as you imply, it would be better to buy a ready made PSU.

[soldar]

I cannot see the need (in general) for two OpAmps or that much complication. In its simplest form the attached circuit will do. If you need higher current sourcing or sinking then emmitter follower amplifier can follow.

[spec]

I cannot see the need (in general) for two OpAmps or that much complication. In its simplest form the attached circuit will do. If you need higher current sourcing or sinking then emmitter follower amplifier can follow.

There are a few issues:

Most opamps will not be happy driving large capacitors.

The opamp maximum supply voltage and power dissipation limits the applicability.

None these rail splutters, provide a solid 0V. Just imagine using a class AB audio power amplifier with a rail splitter.

I am not sure how the frequency stability would work out if you did an analysis.

It is hard enough getting your 0V connections correct, even without the added complication of a rail splitter.

[soldar]

There are a few issues:

Most opamps will not be happy driving large capacitors.

The opamp maximum supply voltage and power dissipation limits the applicability.

None these rail splutters, provide a solid 0V. Just imagine using a class AB audio power amplifier with a rail splitter.

I am not sure how the frequency stability would work out if you did an analysis.

It is hard enough getting your 0V connections correct, even without the added complication of a rail splitter.

Well, I agree that it is better to have a split power supply and not need a rail splitter but this thread is about using a splitter for whatever reasons.

[Zero999]

An audio amplifier IC can be used as a very good rail splitter. Refer to the thread linked below:
https://www.eevblog.com/forum/projects/to220-rail-splitter/msg2077699/#msg2077699

I wonder how well a class D amplifier would work? That would keep losses/power dissipation to a minimum.

[spec]

There are a few issues:

Most opamps will not be happy driving large capacitors.

The opamp maximum supply voltage and power dissipation limits the applicability.

None these rail splutters, provide a solid 0V. Just imagine using a class AB audio power amplifier with a rail splitter.

I am not sure how the frequency stability would work out if you did an analysis.

It is hard enough getting your 0V connections correct, even without the added complication of a rail splitter.

Well, I agree that it is better to have a split power supply and not need a rail splitter but this thread is about using a splitter for whatever reasons.

Point taken. Afraid that my fear and loathing of all rail splitters may have colored my response to your post.  But to be pedantic, the OPs question was how does the LM317/337 rail splitter work and is it any good, and having answered that, hopefully we have moved on.

[spec]

As for the cheap dual rail PSUs, any pointers for what to look for? The 19V 65W laptop bricks are readily available and cheap here, but haven't found anything dual rail that's available and not considerably more expensive.

Hmm, after a quick look, neither have I. But will investigate further and list any approaches that may be of interest.

[1]  SINGLE MAINS SUPPLY WITH +- 18V RAIL OUTPUTS
With this approach, the mains PSU provides stabilized +-18V supply rails directly from the mains.

This is the board that I originally had in mind, but it is only generates +-15V at 500mA:

https://www.ebay.co.uk/itm/HIFI-Single-voltage-to-positive-negative-Regulator-powerFinished-board-Low-noise/253615314877?_trkparms=aid%3D222007%26algo%3DSIM.MBE%26ao%3D1%26asc%3D20170920101022%26meid%3D8b93c439f71147b7ad5c5d353e5d1c86%26pid%3D100010%26rk%3D3%26rkt%3D3%26sd%3D262381275345%26itm%3D253615314877&_trksid=p2047675.c100010.m2109

After a good search, I could not find any low-cost +-18v,  at 1A or greater, SMPSUs, so it looks like this avenue is a dead end as far as your application is concerned.

[2]  TWIN MAINS SUPPLIES TO PROVIDE +- 18V RAILS
As capt bullshot says in reply #8, two isolating mains supplies can be connected in series to generate the +-18V stabilized supply lines. This is a really neat approach and provides a good performance.

Most, if not all laptop PSUs are isolating and seem to have a fairly standard topology. I suspect that many of the laptop PSU guts are made by the same company.

If you wanted to make a tidy single PSU, you can remove the plastic cases by running a hot knife down the seams to expose the guts, which are quite compact. Then just put the two PCBs in a nice little metal case for good screening (ensure adequate heat-sinking/ventilation). With a bit of work, you could make a really professional job of it.

This laptop PSU is low cost and has an adjustable output voltage:
https://www.banggood.com/96W-Multi-Function-Universal-Notebook-Laptop-AC-DC-Power-Adapter-Charger-p-981672.html?gmcCountry=GB&currency=GBP&createTmp=1&utm_source=googleshopping&utm_medium=cpc_bgs&utm_content=frank&utm_campaign=pla-uk-elc2-brand-pc&gclid=Cj0KCQiAmuHhBRD0ARIsAFWyPwgEjaMVK15etZSdrOWzIJcf2mH17DD_pm715GJglHzqy1JiPcMQ4i8aArdxEALw_wcB&cur_warehouse=CN

But you are not limited to laptop SMPSUs. There are other suitable PSUs, including, board-level, wall-wart, chassis, and cased.

Here is an example from the many 18V SMPSUs available:
Output current: 4A maximum
Mains supply: international
Size: 110mmx79mmx36mm
Make: Velika
Cost: 91.84 NOK ($10.84US) delivered (free) to Norway

https://www.aliexpress.com/item/Small-Volume-Single-Output-18V-4A-72W-Switching-power-supply-Source-Transformer-100-240V-AC-DC18V/32837291241.html?ws_ab_test=searchweb0_0,searchweb201602_5_10065_10068_10130_10547_319_317_10548_10696_10192_10190_453_10084_454_10083_10618_431_10307_10820_10301_10821_10303_537_536_10059_10884_10887_100031_321_322_10103,searchweb201603_51,ppcSwitch_0&algo_expid=d54973ac-b655-42ea-9574-477d33b908af-20&algo_pvid=d54973ac-b655-42ea-9574-477d33b908af&transAbTest=ae803_5

[3}  TWIN 24V OP MAINS SMPSUs AND +-18V LINEAR REGULATORS
Here twin mains SMPSU provides +-24V which is regulated down to +-18V by a pair of complementary three-terminal linear regulators (LM317/LM337 etc). This arrangement can be made suitable for-high end audio pre amplification, but filtering out the switching hash, from the two SMPS to negligible levels, would take some doing. And the 0V line would be problematic.

Mains PSU with 24V Output
Once again, there is any number of suitable low-cost 24V mains SMPS. Here is just one example: https://www.banggood.com/4A-To-6A-24V-Switching-Power-Supply-Board-AC-DC-Power-Module-p-969204.html?rmmds=search&cur_warehouse=CN

[4]  MAINS SUPPLY AND +-18V GENERATOR
With this approach, a mains to DC SMPSU powers a board that generates +-18V supply rails.

Mains 4V to12V SMPSU
With the +- supply line generator described below, a mains to 4V to 12V, 52W min PSU would be required. There is an endless choice of suitable PSUs, especially with 12V outputs, which would be the best output voltage. An example would be the 12V, 5A SMPS from here: https://www.ebay.co.uk/itm/DC-12V-24V-5V-10A-12-5A-15A-20A-30A-Universal-Regulated-Switching-Power-Supply-/181881169501

+- Supply Line Generator
Input supply: 4V to 12V
Output supplies: +- 8V to +-18V adjustable @ 1A5

https://www.ebay.com/itm/HIFI-Regulated-Power-Supply-Single-Voltage-To-Positive-And-Negative-Power-Output/253689999076?_trkparms=aid%3D222007%26algo%3DSIM.MBE%26ao%3D1%26asc%3D20131231084308%26meid%3D50b108bfc6f043fe879ece914a932086%26pid%3D100010%26rk%3D2%26rkt%3D5%26sd%3D272417883151%26itm%3D253689999076&_trksid=p2047675.c100010.m2109

[5]  ROLL YOUR OWN +- 18V OUTPUT SMPS
This would be good fun and not too difficult to do, but more of project in its own right rather than a means to an end.

[6]  MAINS TRANSFORMER/ DIODE/ RESERVOIR CAPACITOR WITH OR WITHOUT +- 18V REGULATORS
The title is self explanatory.
This is the Rolls Royce approach that is not difficult to design and build: with the appropriate components and layout, suitable for high-end audio pre amp stages.

CONCLUSION
Which is the best approach? To me one wins hands down.

[David Hess]

I cannot see the need (in general) for two OpAmps or that much complication. In its simplest form the attached circuit will do. If you need higher current sourcing or sinking then emmitter follower amplifier can follow.

With the right operational amplifier and capacitor selection that can work very well but finding the right combination can be tricky and it is limited in power handling capability unless a power part can be found.

[spec]

+ aheid

I am dying to know- what is this PSU for?

[Audioguru]

+ aheid

I am dying to know- what is this PSU for?

I would also like to know the circuit because a dual polarity supply and a high current virtual ground circuit are probably not needed for it.

[spec]

+ aheid

I am dying to know- what is this PSU for?

I would also like to know the circuit because a dual polarity supply and a high current virtual ground circuit are probably not needed for it.

You must mean ' a high current virtual ground circuit are probably may not needed for it'.
There is a lot more to the 0V line than just current handling capacity.

[soldar]

An audio amplifier IC can be used as a very good rail splitter. Refer to the thread linked below:
https://www.eevblog.com/forum/projects/to220-rail-splitter/msg2077699/#msg2077699

An audio amplifier built with discrete components can easily be modified for this purpose. For example  the circuit I posted in another thread.

[spec]

An audio amplifier IC can be used as a very good rail splitter. Refer to the thread linked below:
https://www.eevblog.com/forum/projects/to220-rail-splitter/msg2077699/#msg2077699

An audio amplifier built with discrete components can easily be modified for this purpose. For example  the circuit I posted in another thread.

I can't find any information about a discrete audio amplifier being used as a rail splitter in the thread that you link- have I missed something?

Besides which, are you really suggesting that you dedicate a whole discrete audio amplifier to generating a 0V rail- surely not.

[spec]

Usually the (+) input of an opamp or power amplifier is at 0VDC and draws VERY low current. Then the rail splitter does not need high power ICs, instead the rail splitter is two medium value series resistors across the supply and their junction is 0V.

This is one of your famous sweeping statements AG.