DC Coupled Single Supply Op Amp Circuits With Virtual Earth

[grapefruit]

When using a virtual earth from a half rail IC like the TLE2426 or an op amp output biased at half V+....
I believe that when going from the output of one op amp into another that has unity gain it is ok to DC couple the connection, and let the output of the first op amp bias the input of the buffer at 1/2V+. My question is, if you want to put a pot as a voltage divider between the two op amps, is it ok to DC couple it, but return the bottom of the pot to 1/2V+ instead of ground? Or would you have to put a coupling cap on the input and output of the pot, take the bottom of the pot to ground, and apply 1/2V+ to the input of the buffer?

I've looked through a heap of app notes on single supply circuits but can't see this done anywhere...

Cheers,
Stew.

[R.G.]

Your speculation is correct. Return it to your reference voltage and you don't need capacitors.

Be aware of the buildup of offset voltages. Opamps don't perfectly reproduce the DC voltage on their (+) pin at the output pin. Even in gain-of-one circuits, there's a small offset, generally on the order of 5-10mV. If you use noninverting stages, this builds up at each stage, and if there's gain there, it multiplies the offset as well. Inverting stages have this same problem, but it alternates from side to side of the reference voltage. Noninverting stages it builds up.

The way around this is to make sure that gain stages have unity DC gain with a capacitor in the (-) input side, and periodically stick a capacitor in the output line to break the build up of offsets. Generally this is not needed for only a few opamps, but in a high gain or many-staged circuit it could be a problem.

[grapefruit]

Thanks R.G. That's what I wanted to hear.

It's a bit late to go down to my workshop to try it out now but I wanted to check that the theory was right. It'll only be going through two stages so it should be fine.

Also there's a blend circuit that I've used in the past. I think it may be called a H circuit. The two inputs come in on (say 47k) resistors and each goes to either side of a pot (say 50k). The wiper of the pot goes to ground. A resistor comes off each side of the pot and goes into a summing inverting op amp. It works really well as a blend when you want to totally cancel out one input when the other is on full, and you can comensate for differences in the levels coming into the circuit by adjusting the resistor values into the op amp. I guess I could take the wiper of the pot to 1/2V+ and avoid a coupling cap into the op amp, but I'll still need coupling caps on the inputs.

I'll have to sort out some web space so I can post diagrams. Might make this easier

Also, does crosstalk become a problem, if you have many points going to the virtual earth. It's low impedance so I'm guessing it doesn't cause a problem as long as you don't exceed the current rating of the rail splitter/op amp...

Stew.

[R.G.]

I'm familiar with that circuit, no diagram needed. Yeah, you're going to need caps at the front of the two inputs. See "Panning for Fun" at GEO, and "Designing bias networks to provide Vbias in effects" in "Circuit Sweepings" for some more commentary on panning and bias networks.

Crosstalk is generally not a problem with active splitters like the TLxxx, an opamp, or an LM386. Their output impedance is quite low. It can be with passive splitters like the resistor-resistor-capacitor ones, because the cap is the only AC impedance lower-er and it's hard to get one low enough. But the passive works in most simple, small cases.