When replacing opamps in some older equipment with newer opamps, I understand that it is good design practice to add decoupling.bypassing capacitors right at the opamps' power supply rail pins.
What exactly does this do?
Is it better to use ONE 100nF cap between the rails, or one from EACH rail to ground, or does it not make a difference either way?
What exactly does this do?
Is it better to use ONE 100nF cap between the rails, or one from EACH rail to ground, or does it not make a difference either way?
It allows for a low inductance (short) path for higher frequency power supply decoupling.
Some op amps work better with a small rail to rail capacitor, and the bulk 10-22uf from rails to ground. Some aren’t that finicky, some (not many) can oscillate without the smaller part right on the pins.
Probably have to try it both ways to see what combination sounds good anyways.
Did you search this first?
Some op amps work better with a small rail to rail capacitor, and the bulk 10-22uf from rails to ground. Some aren’t that finicky, some (not many) can oscillate without the smaller part right on the pins.
Probably have to try it both ways to see what combination sounds good anyways.
Did you search this first?
For instance the NE5532/NE5534A need something like 100nF ceramic very close to the pins to prevent internal high frequency instabilities that degrade its performance. Many modern high-bandwidth opamps have similar requirements.
Without this decoupling the chip may appear to function but fail to live up to its datasheet performance, more distortion, more supply current, more noise, that sort of thing
Without this decoupling the chip may appear to function but fail to live up to its datasheet performance, more distortion, more supply current, more noise, that sort of thing
Yes, I understand that to be true. But I am trying to wrap my head around whether it's better to have one 100nF from each rail to ground (which may be tough to do in some retrofits) or if it's just as good to have one between the rail pins.
It would depend on how the op amp circuit and device in which it as been placed have been designed. The different parts can respond differently to power supply variations. Start with the data sheets, then an exhaustive search for previous efforts, then go try something, and decide which works best.
Best to place the capacitors between both rails and ground. Op-amps operate in class AB. This means that the device is continually switching between rails. Imagine trying to stop one current and getting another one moving.
It all depends. Where are the output currents going? Does the original design impose hygiene between power ground (dirty) and signal ground (clean)?
Not the issue you think it is, opamps run at high impedance, low current and for audio frequencies this puts little stress on the power rails - you can ignore stray inductance. But possible oscillation modes at 10's of MHz are capacitively coupled much more effectively to the rails, and the stray inductance is much more relevant at 10MHz than 10kHz. All the parasitics are 1000 times as relevant in fact!
TL072's for instance run without decoupling perfectly well. (The datasheet does recommend decoupling if the PSU rails are noisy, but that's a separate concern from stability).
Typical regulators used for +/-15V would have output impedances in the sub-ohm region across the audio band, and opamp outputs drive a few mA, so only a few mV are impressed on the rails in audio operation (for volts of signal).
Last edited:
Thanks.
I always solder a 100 nF capacitor directly to the power leads, on the bottom of the board. Creating a low impedance locally along the rails is what makes the chips I use stable.
Electrolytic capacitors from the rail to ground should not be grounded locally to the signal ground. These capacitors can potentially have significant current transients (although not so much in a low current line level circuit) and dumping that current into the local signal ground can introduce noise and distortion. These caps should have a separate path to the power supply or else be grounded to a local "dirty" ground. (This is a technique I read on the internet- probably here- and used successfully. I'm reading Doug Self's books and he talks about this too.)
I usually use just one electrolytic cap, from rail to rail, to locally decouple the rails. Many line level circuits have the signal "pass through" without being locally referenced to ground. To reference the local decoupler capacitors to ground on a board like this is pointless.
So am I to infer from that statement that ONE capacitor between rails is sufficient decoupling in a line-level circuit?
Note the distinction between decoupling for stability, and to reduce rail noise, different things entirely.