Creating a Virtual Power Supply Ground.

Thread Starter


Joined Mar 24, 2008
Many times it is necessary to have a dual power supply when all you actually have is one. It is possible to synthesize a pseudo ground using techniques similar to normal power supplies, but with important differences. I'm going to assume a 36V ungrounded power supply for most of my models when it is necessary to have a number.

The most basic and simplest technique is a simple voltage divider. Add capacitors as shown and the AC characteristics become very low impedance, though this circuit will have a definite frequency response. The DC stability is atrocious, but for many applications this doesn't matter as much as you might think.

................Figure 1

The capacitors also can be used as bypass caps, since they are fundimentally being used that way for the virtual ground. The 3db point for the RC network is approximately

F = 1 / (6.28 R C) or 0.724 Hz

Multiply the answer by 10 (as a rule of thumb) and this virtual ground will be good to 7Hz.

Figure 2 shows how this ground might be used for an op amp inverting amplifier. Note that the input and output of the op amp circuit can be treated as if they were AC signals, since the power supply is presumed isolated (as a battery would be) the DC values will not interact with the outside world.

..............................Figure 2

Another old technique that works well within the limits of a zener, the current is limited by the resistor, exceed the current and the voltage folds back, yet the other polarity continues to function.

........Figure 3

You design this as you would any zener regulator, design information can be found in the AAC eBook in Vol. III, Chapter 3, Section 11.

The fact this works so well might encourage you to make a leap to solid state regulators, to use something similar to these configurations.

............................Figure 4

...........................Figure 5

This won't work. The reason is simple, if the plus side is loaded and the minus side is floating (no load) where does the excess current go? With the zeners the answer is into the second zener, but solid state regulators are not designed to absorb current, only to be a source. This point can be illustrated with a simple one transistor one zener regulator.

.....Figure 6

Note that if the voltage on the emitter exceeds the base voltage the base emitter junction is back biased, and the transistor is turned off. What is needed is more than the series pass transistor, you need a transistor or device that can swing both ways, absorbing as well as sourcing current.

Thread Starter


Joined Mar 24, 2008
Op Amps can do this, some are quite high current devices. Take the following configuration.

..............Figure 7

I have seen recommendations for LM675 for high power applications, and KA334 or L272 op amps for lesser applications. The data sheets for all of these devices include methods for making a pseudo ground. The reason this works so well is the op amps active gain actually reduces the apparent output impedance to millohms, even though the current drive may be limited to the op amps specs. Even a weak op amp or one with really low current output specs will appear to have low output impedance, until it burns out. It is possible that you don't have access to these devices, or don't want to use op amps at all for some reason, so we'll explorer other ideas.

A BJT in voltage follower configuration is one of the most basic voltage regulator configurations. You can read more about this in Vol. III, Chapter 4, Section 6 in the AAC eBook. Basically the output impedance of this regulator is the base resistance divided by the gain of the transistor, this is directly equivalent to the internal resistance of a battery. The following illustration shows the configurations and math behind an emitter follower, and will be important for future concepts.

.......................................Figure 8

We can use this concept to create the following circuit, which is two emitter followers back to back.

.....Figure 9

You start with R2 at its maximum value, then slowly adjust down until you see a current surge, then back it off. It will work, but has several problems. It will have a small dead space where neither transistor is conducting, and if the power supply changes (specifically drops in voltage) the voltage drop across R2 also decreases and may allow the two transistors to conduct in an uncontrolled fashion, burning them out. Single transistors will have a fairly high output impedance, depending on the values selected, but this can be improved using Darlington transistors. It will amplify the problems as well as reducing the output impedance, but it will work.

You can reduce the sensitivity to the source voltage by using devices whose junctions react similar to the transistor base emitter junction, a power diode. If you want to really match the curves put the diodes next to the transistors so they are the same temperature. Here is the schematic of this arrangement.

....Figure 10

While these concept will work they suffer from a the big disadvantage that their output impedance is higher than it has to be. As mentioned earlier, op amps with their active gain can bring the apparent impedance down to almost zero. Given these characteristics it isn't too surprising their used in voltage regulators. The virtual ground is also a voltage regulator of a sort, from the outside it appears to be a tracking dual power supply.

To merge all of the concepts that have been put forward looks something like this.

.....................Figure 11

This will be an op amp with really high current output. You can use Darlington transistors to reduce the load on the op amp, but it will probably not be necessary.

Sgt. Wookie came up with an interesting variation in the following schematic. This will create a regulated virtual ground, but with minor modifications can be used to create a straight dual power supply, such as +12VDC and +5VDC.

........................................................................Figure 12
Last edited:

Thread Starter


Joined Mar 24, 2008
Another idea, shown below, was submitted by EBLC1388. It uses an interesting alternative to Darlington transistors, a Sziklai pair. The drive on this configuration is extemely high, and while it is slightly dependent on power supply voltage, this has been reduced by a significant degree. Adjust R2 until the circuit is drawing around 20ma, and it is in its linear region. Values are not critical, R6 and R7 can be reduced to as low as 0.2Ω for large current draw. Connect an op amp output between CR1 and CR2 and it will amplify the op amp the same way as shown in figure 11. It even makes a dandy output driver for an audio amp.

.......................................Figure 13
Last edited:


Joined Jul 17, 2007
Hi Bill,
Just tried simulating the very first circuit. Basically, significant loads on either side will throw the circuit a good bit out of whack.

I tried simulating it using a 15v source, and a pair of 1N751 5.1v Zener diodes. R1 needed to be about 9 Ohms (2.8W) to assure reasonable regulation near 5v, at nearly 550mA current.

A 100 Ohm load on the - side caused the circuit to become unbalanced by an average of 140mV.

Thread Starter


Joined Mar 24, 2008
Didn't say it was perfect, but I bet you got the zeners voltage across each zener. This is a very old cookbook design, not mine, that I came across in college.


Joined Aug 13, 2008
Cool, I've been thinking of making a variable power supply, this gives me more options to think about rather than switch mode power supply.


Joined Nov 9, 2007
I remember some years ago I was asked by a model railway enthusiast for a simple control. the circuit was essentially as Bill / Wookie propose.

He wanted one knob, with a centre detent, so that if he turned the knob one way the train moved forwards faster as he turned it up. If he turned the knob the other way the train slowed and eventually moved backwards as the position passed the detent.

He possessed the standard Hornby train power supply.

This was easily achieved with a 741 driving TIP 4amp power transistors configured as a complementary emitter follower. The circuit was a modification of one originally due to Texas Instruments which provided a split power supply for an audio amp from a single ended one.

The extra oomph was provided by reconfiguring the original rectifier circuit.

A 741 is adequate in this case as we are talking low frequency.


Joined Dec 20, 2007
If the input of an opamp is biased properly with two resistors then it works fine from a single polarity supply. The input and output signals might need coupling capacitors.

But a lousy 40 years old 741 opamp is designed for a 30V supply and many won't work with only 9V as its supply. There are many newer opamps that will work fine from a single polarity 9V supply.

Thread Starter


Joined Mar 24, 2008
If the input of an opamp is biased properly with two resistors then it works fine from a single polarity supply. The input and output signals might need coupling capacitors.

But a lousy 40 years old 741 opamp is designed for a 30V supply and many won't work with only 9V as its supply. There are many newer opamps that will work fine from a single polarity 9V supply.
Their are many applications where a voltage reference, that two resistors provide, is all that's needed. Their are other applications, such as active filters, where you must have a low impedance ground to have them preform as predicted and desired. Dual power supplies aren't always available.

Just curious, as much as you loath the old 741, what did you cut your teeth on with op amps? I found it interesting that the data sheet doesn't specify any minimums. I checked after you said something similar on another thread. Later data sheets do spec a minimum voltage for other parts.

Personally I find the limitations of the 741 good for students. You don't have to try very hard to run into them, and they are pretty typical for all op amps, they just happen sooner. This means you can use a low frequency scope (or other equipment) to see what slew rate or frequency response is. This could actually be the secret to their continued use.


Joined Nov 25, 2008
Interesting and practical topic, I hope it has legs and some new and useful ideas come out.

I struggled searching for so called rail-splitter ideas a couple of years ago when I needed a source of negative voltage for a RF microwatt meter I was building. The measurement range was -70dbm to +15dbm and I wanted to display the values directly in units using a 3 1/2 digit DC +/- 200 millivolt LCD module. If anyone is ever interested in such a meter, building one has been made simple with the Analog Devices AD8307 log amp, a tremendous chip:

I tried to breadboard a few of the simpler rail splitting methods but was not real happy with either their stability or efficiencies. A nice DIP packaged 5vdc isolated DC to DC converter would have worked OK but they are not common nor inexpensive. I ended up just using two 9vdc batteries followed by small +/- 5 volt regulators. As the instrument is a low usage item, I think the batteries usage won't be much higher then their normal shelf life and I'm still on the original

If I was to be designing it today I might consider trying one of those MAX232 RS232/TTL converter chips. Not sure how much negative current one could draw from such a chip but many times not much current is needed for simple op-amp measurement circuits.