Op Amp Decoupling with a Virtual Ground

Question

I'm designing a low-frequency op. amp. circuit that operates from a single 10V supply. I'm using a TLE2426 "rail-splitter" to generate a virtual ground, 5V above earth ground, which is used as a reference for the op. amps. Normally, if I had bipolar +5V and -5V supplies, I'd decouple the high and low inputs of the op. amp. to the earth ground. With a single-supply op. amp. design, I've heard to decouple straight from the high supply to the low supply. So, I have the following possibilities:

• 2 capacitors: op. amp. high to virtual ground, op. amp. low to virtual ground
• 1 capacitor: op. amp. high to op. amp. low
• 3 capacitors: both of the previous options

In theory, the first option creates an equivalent of the second option with half the capacitance, so long as the trace impedances are low. I was hoping for a discussion on the pros/cons and if there is a definite recommended approach.

Answer 1

It depends on where the currents are flowing. In general, if current flows from the power supply rails (your +10V/0V) through the load resistor to your virtual ground, I'd use three capacitors, with the one across the 10V supply before the rail splitter and two equal capacitors from the +5 and -5 rails to the virtual ground.

From the datasheet figure, you might want to consider a dummy load resistor to improve the stability if power consumption isn't a big deal to you.

Answer 2

It all depends on what the rest of the circuit uses and how good the "other" devices are at dealing with inevitable power supply noise. As an example, here's the graph of power-supply-rejection for a Linear tech device: -

This graph tells me that noise on the negative rail is 30dB more likely to have a bad influence on the op-amps output compared to noise on the positive rail. Under these circumstances I wouldn't hesitate to use a single capacitor from pseudo mid-rail to the most negative point of the supply. If your power supply is a tad noisy, putting both caps on will share noise equally at the mid-rail point and this would be the wrong choice.

Clearly, the positive rail noise is much more easily dealt with inside the LT1057 so your mid-rail should strive to be stable with respect to the negative rail.

Other devices will/may be different - read the data sheet - if it doesn't quote PSRR for both rails it may be OK but it won't help you make a decision. Also, depending on your operating spectrum, choose your "other" circuits to have decent PSRR (on either rail) at the maximum frequency you might be processing.