Problem: I have 5V input from USB, I need to split it into four rails with +/-5V and +/-15V.

Currently done: I have built the following circuit in a breadboard.

Note: Cuirciut for splillter taken from https://tangentsoft.net/elec/vgrounds.html.

I have measured the outputs with a multimeter, and it is worked fine until I have tried to connect virtual grounds together for +/-15V and +/-5V. Then the readings in +/-5V changed to - 4.5V and -14.5V.

So the problem I have is combining two virtuals grounds, if anyone could give me any solutions how I can solve it, please let me know.

  • 1
    \$\begingroup\$ Can you include all other gnd pins of the different elements: input gnd (associated with your 5v), of your 2 BB IC and how you have them connected? Layout snapshot of the board might help too. \$\endgroup\$ – vrleboss Apr 20 '17 at 15:26
  • \$\begingroup\$ Do you see voltage drops between the various gnd pins of the IC's ? That could indicate impedance limitation of your gnd traces. \$\endgroup\$ – vrleboss Apr 20 '17 at 15:28
  • \$\begingroup\$ Can you also confirm that the 30V and 10V levels are unchanged when combining the grounds. \$\endgroup\$ – vrleboss Apr 20 '17 at 15:32

Alternative solution:

I generally think virtual grounding adds more issues than it resolves. As such I highly recommend a different power splitting approach that maintains the original ground.

For example....

enter image description here

Of course the above is only one example.

How you actually break the rails up depends on how much current you need from each supply. If the negative rails need higher currents then using separate boosters may be required.

  • \$\begingroup\$ @IvanShorokhov I think you mean "get rid of" not "get reed off" :) \$\endgroup\$ – pericynthion Apr 20 '17 at 17:29
  • \$\begingroup\$ @Trevor, could you please suggest me an example of the capacitive inverter with minimum 280mA output, I spent at least 2 hours to find a good example, but most of them were lower than 50mA... \$\endgroup\$ – Random Data Apr 24 '17 at 15:16
  • \$\begingroup\$ @IvanShorokhov oh that's a lot for a capacitive, you may need to use an inductive for that much. You did not mention your currents in the original question. \$\endgroup\$ – Trevor_G Apr 24 '17 at 15:34
  • \$\begingroup\$ Perhaps these cds.linear.com/docs/en/datasheet/8027fd.pdf \$\endgroup\$ – Trevor_G Apr 24 '17 at 15:37

If you re label your outputs as {0, +15, +30} for the 30V splitter and {0,+5,+10} for the 10V splitter then the problem becomes obvious.

You could use an isolating converter for the step up to 30V, at which point the problem goes away (possibly to be replaced by an EMC headache), or forgo the rail splitters and generate the appropriate negative rails with some buck/boost converters instead.

I do not personally like the flying cap stuff on a mixed signal board, because it tends to be a noise nightmare when fed from a voltage source, and IMHO a high frequency buck/boost or two to generate the negative rails and a boost to make the +15V is probably the way to go. It is one more switcher, but it does away with the mess of passives around those rail splitters and should be more efficient to boot, size will probably be reduced.

Push comes to shove, a two transistor Royer set up as a forward converter plus a pot core and some magnet wire....

I would note that the presence of +-15V hints at opamps and that these often have very poor CMRR at switchmode supply frequencies, think about that when considering your layout.


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.