Attempting to design a simple dual rail fixed supply

[kizzap]

Hey all, I'm in the process of designing a project and I am being stumped about the design for the power supply.

My issue is thus: The board I am making involves quite a number of op-amps, powered by a dual rail supply. However, to keep things simple, I want to power it from a single AC or DC plug pack. So I need to convert the single rail to a dual rail supply.

I've attached a schematic of what I have drawn up so far. (I know there are no values, I haven't calculated anything yet.) My concern is in the use of the virtual ground in the centre of the two regulators. Would it be suitable for being able to supply up to 500mA on each rail?

Secondly would I need to use anything like a LTE2426 Virtual Ground Reference? or would it just be a waste of time?

Thanks.

-kizzap

[Apothus]

I have seen a voltage divider used as a simple virtual ground for a headphone amplifier before however I cannot comment on how effective that may be for the system you are working on or how efficient it needs to be. What are you driving with the op-amps?

[kizzap]

Well the circuit is powering 16 different op-amp bandpass filters, which are multiplexed into another op-amp which then drives an LM3915/6 combo. (In retrospect I should have said that in the original post.)

My concern is that due to the nature of the LM3915/6 combo, that the current drawn
won't be uniform and that it might cause an issue.

-kizzap

[Thor-Arne]

Google is you friend.

A quick google turned up with this site.

[grumpydoc]

Would it be suitable for being able to supply up to 500mA on each rail?

Not as it stands.

It will sort of work if the load on the supply rails is equal because they will form a voltage divider and give you a virtual ground which is about halfway between the rails. However this won't really be a good reference point.

If you want to demonstrate that it won't work ask yourself what happens if you connect a single 500mA load to one rail - where is the return current path?

I would do a single rail regulator at 2x the individual rails that you want, then split that in two. There's a handy virtual ground circuit in this thread https://www.eevblog.com/forum/beginners/perfboard-prototyping/ - 8th post.

[kizzap]

Thanks grumpydoc. I had a niggling thought that something was amiss. I had a look at Florian's post and I think I get what is going on there, please correct me if I am wrong:

The op-amp sets a virtual ground through the resistor divider connected to the +ive input. The output is controlled by the -ive input which detects the difference in loading on the rails, which in turn switches either the NPN or PNP transistor on, giving the extra current from the rail with the higher loading a path to complete the circuit.

Assuming that is correct, would the attached schematic be somewhat correct?

-kizzap

[c4757p]

Is that input always AC, or is the rectifier there just to make it compatible with either one? You can get two rails quite easily out of an AC input by using two half-wave rectifiers, one positive and one negative, like this.

You'll need the capacitors to be twice as big, but who cares? You're eliminating all that virtual ground stuff in the process.

[kizzap]

yeah the idea is that I can use DC and AC.

-kizzap

[c4757p]

Then your second schematic looks like the right idea - but you will need some frequency compensation on that op amp. I suspect it will be unstable.

[grumpydoc]

Assuming that is correct, would the attached schematic be somewhat correct?

Yes, that looks much better.

Personally I would put the regulation first, then the virtual ground as you only need one regulator that way but what you have should be OK.

[kizzap]

I took on the single rail regulation advice, and redrew another basic schematic, Might give this one a go on the breadboard tomorrow when I get a chance. Hopefully it works.

As for the frequency compensation, I'm going to have to do some reading about it...

If anyone else can see any issues feel free to comment.

-kizzap

[c4757p]

As for the frequency compensation, I'm going to have to do some reading about it...

You're in for some long reading if you just go about it in a general way.

Unless I'm missing something in the circuit design, there's one very easy, "trial and error" way you can do it with this circuit. Put a resistor R between the inverting input and the virtual ground (1k-10k is a good guess). Then build it like that, and if it oscillates, at a frequency f, choose a capacitor C such that 1/f = 2*pi*RC. If you get a ridiculous value for C, choose a different R and try again. Put that capacitor directly between the inverting input and the output of the op amp.

Note that the values may be layout-dependent, especially if this is on a breadboard.

Another way is to put resistors in series with the output capacitors, calculated using a similar method.