I want to obtain +-12VDC supply from a 24VDC supply where the total load will driver 100mA max. I tried using two power supplies but one of them blew up. I need the split supply for excitation voltage which I mentioned in my previous question. So I decided to make a spit supply from a 24VDC supply. I have some many LM7812, LM7912 or equivalent. I tried the following circuit:

Schematic using LM7812 and LM7912 with common ground

But instead of +-12VDC I'm getting the following output:

Voltage output on the scope

These are the only components I have at the moment.How can I use them in a way to obtain +-12VDC?


Regarding @Trveor's answer:

enter image description here


Discrete solution(?):

enter image description here

  • 3
    \$\begingroup\$ A hint on what went wrong: What are the drop-out voltage specs for LM7812 and LM7912? \$\endgroup\$ – The Photon Oct 4 '17 at 16:10
  • 4
    \$\begingroup\$ That's number 1. (2) You have nothing holding the ground rail steady. It will jump all over the place once current starts to flow. \$\endgroup\$ – Transistor Oct 4 '17 at 16:12
  • \$\begingroup\$ If that power supply is isolated, use a virtual ground circuit instead, one capable of sourcing/sinking say 250mA or more. \$\endgroup\$ – Trevor_G Oct 4 '17 at 16:12
  • \$\begingroup\$ @Trevor I have some LM7812 LM7912 and 24V DC supply. There is only one power supply 24VDC. Can you draw the topology as an answer you meant? Thank you \$\endgroup\$ – user1999 Oct 4 '17 at 16:14
  • 1
    \$\begingroup\$ You can e.g. use a Power-OP-Amp connected as a unity gain amplifier with a 10k/10k network to +Bat/-Bat on input. That would give you a virtual ground with the strength of the OP-Amp output driver. \$\endgroup\$ – Janka Oct 4 '17 at 16:19

If the 24V supply is isolated use a virtual ground instead..

Something like this..

enter image description here

Choose an op-amp capable of sourcing and sinking the currents you need. You may need to heat-sink the op-amp if currents are large.

RE ISOLATION Note the drawing below, both power supplies have their negative outputs ties to mains ground. When wired up this effectively shorts the output of the virtual rail to the negative terminals. As such you need to be sure things are correctly isolated in your circuit and internal to the supply. BTW: Some supplies have a jumper on the terminals that needs to be removed.

enter image description here

  • 2
    \$\begingroup\$ Not a great circuit, C4 will cause the output to be 24 V to GND at initial turn on, then will charge to the mid point. You either put two caps, or no caps ...but not just one on the center point divider. \$\endgroup\$ – Jack Creasey Oct 4 '17 at 16:33
  • \$\begingroup\$ @JackCreasey hmmm riiight. \$\endgroup\$ – Trevor_G Oct 4 '17 at 16:40
  • \$\begingroup\$ @JackCreasey.. corrected \$\endgroup\$ – Trevor_G Oct 4 '17 at 16:42
  • \$\begingroup\$ @Trevor Thanks for this. Btw, I just blew up a power supply(when trying to make split form two identical power supplies) yesterday thinking that it was isolated. What I understand from isolated is VDC- should not be wired to the protection earth, is that correct? This split supply question is related to my previous question.electronics.stackexchange.com/questions/332708/… ..I will also upload your suggestion have one more question about it soon. \$\endgroup\$ – user1999 Oct 4 '17 at 16:50
  • \$\begingroup\$ @newage2000 yes the supply can not be connected to any other ground or rail either internally or in circuit. Though PSU frame may/should be grounded if it is not connected internally to the PSU. \$\endgroup\$ – Trevor_G Oct 4 '17 at 16:53


simulate this circuit – Schematic created using CircuitLab

Figure 1. A dual supply from one.

Try this.

Note that the power dissipated in the 7812 will be \$ (V_S - V_O)I = (24 - 12) 0.1 = 1.2 \ \mathrm {W} \$. You'll need a heatsink.

You could reduce the power dissipation by dropping some of the voltage in a series resistor between 24 V and C1. To drop 6 V at 100 mA, \$ R = \frac {V}{I} = \frac {6}{0.1} = 60 \ \mathrm {\Omega} \$.

Note that the 24 V supply must be isolated from circuit ground.


As I have been reminded in the comments, I was not thinking this through (as I have for this topology on a previous occasion). Adding some load to the negative rail to ensure that it draws more current than the positive rail does is required to ensure regulation as the 7812 will not sink current.

As Janka points out the positive rail will be unregulated.

  • 2
    \$\begingroup\$ This will work in many cases, but not if the load on the +12 V output is greater than the load on the -12 V output. \$\endgroup\$ – The Photon Oct 4 '17 at 16:20
  • \$\begingroup\$ An unregulated dual supply from one, you should add. \$\endgroup\$ – Janka Oct 4 '17 at 16:20
  • \$\begingroup\$ Um? Will that 7812 sink current? \$\endgroup\$ – Trevor_G Oct 4 '17 at 16:21
  • \$\begingroup\$ You're all correct, chaps. See the update. \$\endgroup\$ – Transistor Oct 4 '17 at 16:25
  • \$\begingroup\$ How about this toplogy: tangentsoft.net/elec/bitmaps/vgrounds/sijosae.png I have these discrete componets. Do you think I could obtain around 100mA without heatsink needed here? I cannot see any feed back in this topology. Is this split supply unregulated? \$\endgroup\$ – user1999 Oct 4 '17 at 17:15

Here is a solution which uses two adjustable linear regulators (LM317/LM337) to form a virtual ground. It won't work with fixed regulators, though.

enter image description here

That is just a proof of concept circuit. As mentioned in a comment a TL431 voltage reference would be a better choice than the Zener. 10R resistors will limit the regulation of the V-ground a little bit. Tolerances of the regulator's voltage ref vs. the Zener/TL431 can cause some problems too.

You can read a lot of discussion on this and similar virtual-ground/rail-splitter circuits at these links:



  • \$\begingroup\$ Interesting. I would put TL431 instead of 2.5V zener for better precission. \$\endgroup\$ – Todor Simeonov Oct 4 '17 at 16:45
  • \$\begingroup\$ Read this for the source descriptions: head-fi.org/threads/… \$\endgroup\$ – Jack Creasey Oct 4 '17 at 17:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.