wakibaki's virtual ground

I don't think much of virtual ground circuits, but I drew this:-

As drawn it draws just 1.4mA, has an output offset of 6 nanovolts, and an output impedance of ~0.1 ohms. Build it with the LT1097 shown for the best performance, pretty much any complementary NPN/PNP pair will do, the bigger the beta the more current you will be able to sink/source (within the dissipation of the transistor). You could use the 2N3904/6 for modest currents, they have loads of gain and are pretty fast. You can even null the LT1097 offsets with trimmers.

It shows 2 big caps, dimension these at your discretion, you don't want the leakage currents being the biggest current drain. 

The 2 voltage sources on the diagram are just for simulation of the worst-case opamp input offset. The LT1097 is ~$5 a throw, not cheap.

If you build it casually with an LM2904 use 1 ohm buffers instead of the 0R1 shown, you'll get worst case current draw of ~3.4mA. The LM2904 is a dual, ~U$0.40, and you might get away with reducing the buffer resistors to 0R1 if you measure the voltage drop across them, because it's unlikely that the input offsets of 2 amps on the same die will have an adverse coincidence. If you use the 2N2904/6 you can probably build it for less than a dollar.

If you don't care about the current drain as when using a wallwart, use a NE5532, I buy them here, 50 for $10 http://stores.ebay.com/Tayda2009/_i.html?_nkw=NE5532&submit=Search&_sid=872591586. Quiescent current will go up to 10~12mA. Then if you can find a 4A complementary transistor pair with beta of 100, the opamp is good for 40mA.

All round it just blows away pretty much anything else I've seen for headphone-sized (and somewhat larger) currents...

w

No, the ground follows the ripple, but in opposition.

The circuit has negative feedback. The reference point for the opamps is the resistive divider which is connected to the non-inverting input. The opamp outputs whatever voltage is necessary to force the inverting input to the same voltage as the non-inverting input. You need the unfiltered deviation at the feedback takeoff point and the high speed response of the circuit to allow the circuit to respond quickly enough to null out the ground return.

wakibaki vs. Sijosae with 220u caps, same txistors, 200mA @ 1kHz ground return:-

wakibaki vs. Sijosae with 10,000u caps, same txistors, 200mA @ 1kHz ground return:-

Factor of 100 

@20kHz...

Iq; Sijosae 5mA, wakibaki 1.4mA

100n across the opamp rails might be good.

Of course my circuit is expensive, but cheap with LM2904 and still only 3.4 mA.

w

Thanks are due to KT88, who pointed out that the bases of the transistors can be driven too negative (or positive in the case of the PNP) in this arrangement. This is easily solved by the addition of a couple of diodes.

He also suggested going down to one opamp, and moving the feedback takeoff point. This requires the addition of a couple more diodes, and the circuit then looks as below:-

The red and blue traces show the voltage on the bases. The circuit is now just an opamp splitter buffered by discretes, as opposed to an opamp splitter buffered with an IC buffer, nothing a great deal different from the IC buffered circuits proposed previously, although probably still cheaper and less current-hungry than those using IC buffers, and capable of being built to accommodate an arbitrarily large current.

Using a single opamp removes the necessity for a closely controlled input offset, the output offset may move a little closer to one rail or the other, but this is of little consequence.

If a large current is required the next move would be to use darlingtons or some other conventional amplifier output drive stage to increase the input impedance seen by the opamp.

w 

Hi SW

Once you boil it down to what you see in the last schematic, it's not that different from the many existing virtual ground designs. It's easy to get excited over something you come up with in the middle of the night that doesn't seem so extraordinary in the cold light of day. 

w

Better choice than what?

IMO the best choice for both mains powered and battery powered amplifiers is a true dual-rail supply with a regular ground arrangement. That would mean a conventional fullwave rectifier with centre tap connected to ground, a dual-rail halfwave rectifier with one side of the transformer connected as ground, a dual-secondary transformer with dual bridge rectifiers with the two rectified and smoothed outputs stacked, with the common point connected to ground, or a dual-battery supply with the connection between the batteries connected to ground.

A rail-splitter like the one shown above would be most suited where the designer or operator was determined to operate the amplifier from a single (say 9V) battery. It can be built to have a very low quiescent current dependent on the choice of opamp, and the exact choice of diodes and transistors which will also influence the quiescent current in the transistors. The comparatively large capacitance as drawn means that crossover effects are minimized without the expense of an increased quiescent current. It could also be used where a single voltage DC wallwart was envisioned as the source of power. It offers a significant performance improvement over, say, the Sijosae. Comparative simulations can be seen above. It has significantly lower quiescent current than rail-splitters employing an IC buffer and is capable of dealing with larger return currents if suitable transistors are selected.

The circuit has the advantage of being readily constructed from components likely to be on hand in many constructors' junkboxes, or might take advantage of a spare opamp which sometimes occurs in some designs.

w

Yes, but KimLaroux devised the circuit that you're trying to take credit for... which all-in-all is a pretty low way to behave, and 3.5mA is still ~4 times the quiescent current of mine.

Even if you convince the whole world it's yours, you and me and Kim will always know the truth.

w