bypass caps, opamps and two power rails

[exe]

I'm trying to finish one of my old projects, and found an oddity there. On the pic I show how bypassing is arrange for an opamp. Basically, it's a cap between rails. I'm trying to figure out if this was a mistake. I recall I did research on this, but can't remember details. I guess I need to figure this out again :/

So, my question is: why we usually put capacitors between rails and ground? The opamp itself doesn't know what ground is, it has only two power pins and output.

[TimFox]

Typically, the capacitor between and very close to the two power pins is an absolute requirement, unless the manufacturer specifically states otherwise.  Unless otherwise recommended, I add a capacitor from ground to the power pin that is the reference point from which the output voltage is determined.  For a 741, that is the negative power pin, since the output voltage originates at the collector of an NPN transistor after the current mirror from the differential input stage.  Since bypass capacitors are cheap, it probably doesn’t hurt to bypass the positive power pin to ground as well, supplying the suspenders to complement the belt.
Analog Devices has good app notes on this topic.

[Neilm]

I would take it that while the op-amp does not have anything to do with 0 V, the power supplies do. The purpose of the capacitor is decouple the SUPPLY to the chip, not decouple the supply to the CHIP. If there is any  noise on the supply that noise will be referenced to 0 V. If the chip pulls a pulse of current from the supply, the current has to get back to the supply somehow. If the opamp is decoupled by one capacitor across both positive and negative supply, you are involving the second supply in seeing those pulses. In the worst case - I guess you would actually then introduce noise onto the other supply rail

[exe]

Analog Devices has good app notes on this topic.

Is it AN-202? (https://www.analog.com/media/en/technical-documentation/application-notes/AN-202.pdf)

[TimFox]

Yes.  Fig. 2 illustrates my point about bypassing the negative pin.
This kind of bypassing is designed to eliminate instability and oscillation (at relatively high frequency) caused by parasitic series impedance between the pins and the power supply, and is required even if the supply is perfectly clean.

[David Hess]

If exe had not linked that application note from Analog Devices, then I would have.  Figure 3 shows the reason.

So, my question is: why we usually put capacitors between rails and ground? The opamp itself doesn't know what ground is, it has only two power pins and output.

The operational amplifier has no ground connection *unless* as shown in the above example, the load is grounded.  If the load went to one of the supply rails, then only a single decoupling capacitor is required.  An example of this is a single supply audio amplifier where the the speaker is connected to one supply rail and has a large value capacitor in series to block the DC.

The decoupling capacitors should be placed to minimize the loop area between the output stage of the operational amplifier and the load.  If the other end of the load is connected to ground, then a capacitor between each rail and ground is suitable.

[Jan Audio]

Because else you are smoothing garbage with garbage, both has to be smoothed.
If 1 line was already smooth it might work.

[THDplusN_bad]

So, my question is: why we usually put capacitors between rails and ground? The opamp itself doesn't know what ground is, it has only two power pins and output.

Think of your bipolar power supply. Where does the return current from the V+ and V- supplies flow? And how can you "decouple" these currents - in terms of how can you maintain a low-impedance path close to the OpAmp for these currents.

Cheers,

THDplusN_bad