15-0-15 supply slightly unbalanced

[Yurk]

Just curious. I've built 5-0-5 and 12-0-12 supplies but with virtual grounds. The rails were always off by 0.10VDC between pos/neg. I thought it was just because the virtual ground circuit I was using was off somehow.

Now I've got a 15-0-15 bread-boarded and it's using a real ground (grounded center tap). The rails are still off by 0.10VDC. I this normal, or have I got something wrong? I'm building almost straight from the datasheet. Circuit attached. Bolded voltage values were measured all within a minute of eachother.

[sawreyrw]

A tenth of a volt is no big deal in this circuit. However, if the plus and minus load currents are significantly different, you will have a big imbalance. I wouldn't recommend this circuit for serious use; use a real ground instead.

[rrrremus]

You must measure the voltages while a load is applied to the PSU.
I've build a very precise power supply that is always spot on with load, but without any load it can drift away by even 0,3 volt.
Regarding your schematic, you might want to use same value and type capacitors for both rails ( C7 and C8 ) and also same type of voltage regulators. Also , between C1 and C3 ( resp C2 and C5 ) i think you intended to add a low value resistor that you might have missed or if not use the same type of caps.

[benb]

With your virtual ground circuits, using 1 percent resistors in the divider can easily give an output difference of 0.1V over + and - 12V outputs.

With the fixed regulators, the datasheet surely has some output voltage tolerance. Here we are, look toward the bottom of page 2 for the 7915:
http://www.national.com/ds/LM/LM7905.pdf
It shows miminum of -14.4V and maximum of -15.6V so the regulated output could be anywhere between those two values.

This isn't the best tolerance, but a couple of tenths of a volt difference on the supples won't make a significant difference. The circuits in op-amps that these voltages drive are in most cases constant current sources and sinks, and it doesn't matter what voltage they're operated at. The exception is the output, and this imbalance might make one polarity clip a tenth of a volt lower than the other polarity, and I don't see where that's a problem.

If you're still concerned, use LM317 and LM337 adjustable regulators, and use better-than-1-percent resistors to set the voltages within 0.1V.

[gmphadte]

If u want proper tracking, u will not get it if u use two regulators having their tolerances.
Use one regulator for positive and use another discrete/opamp based circuit for negative polarity.

Gajanan Phadte

[AndrewT]

the accuracy of matching R1 to R2 affect the centreing of the virtual ground between the two input voltages..
The leakage matching of C1+C3 to C2+C5 affect the centreing of the virtual ground between the two input voltages.

The output voltages bear no relation to any of the input component matching.

[Yurk]

Thanks for the replies, I'll not worry too much about being off then. This is just going to drive a bunch of opamps.

Just a note, this circuit is grounded to mains ground, not a virtual ground.