A quick question, and hopefully there is a short answer.
Is 4.7uF too big for "decoupling" the power supply of an OP Amp when the cap is soldered directly across the power pins of the OP Amp (on the underside)?
Some background - it has been my routine to solder an X7R 4.7uF cap between the power pins of a DIP Op Amp, which may be used in a preamp or the output stage of a DAC. In most posts I've read, people talk about using a 0.1uF cap for that purpose. I am just wondering if there is any downside to using an X7R 4.7uF cap in that position. Too big? If so, why?
Thanks,
Kurt
Is 4.7uF too big for "decoupling" the power supply of an OP Amp when the cap is soldered directly across the power pins of the OP Amp (on the underside)?
Some background - it has been my routine to solder an X7R 4.7uF cap between the power pins of a DIP Op Amp, which may be used in a preamp or the output stage of a DAC. In most posts I've read, people talk about using a 0.1uF cap for that purpose. I am just wondering if there is any downside to using an X7R 4.7uF cap in that position. Too big? If so, why?
Thanks,
Kurt
Last edited:
None I can think of, except when you use old-fashioned voltage regulators that go unstable with such a load; some brands of LM317 can start to oscillate when the main part of their load capacitance comes from ceramic or film capacitors rather than from cheap aluminium electrolytic capacitors.
Besides, the self-resonant frequency of a 4.7 uF capacitor is sqrt(47) times lower than the self-resonant frequency of a 0.1 uF capacitor of the same size, but that usually doesn't matter much; the impedance at high frequencies (well above resonance) is the same in both cases, because it is determined by the size of the capacitor and the length of the connecting wires and tracks, not by the value of the capacitor.
Besides, the self-resonant frequency of a 4.7 uF capacitor is sqrt(47) times lower than the self-resonant frequency of a 0.1 uF capacitor of the same size, but that usually doesn't matter much; the impedance at high frequencies (well above resonance) is the same in both cases, because it is determined by the size of the capacitor and the length of the connecting wires and tracks, not by the value of the capacitor.
For the frequencie of interest to your opamp, a single smd X7R 4.7uF is actually a good choice. You don't risk much resonance effects, its highish value means it's actually useful at audio frequencies and the inductance of the package is low enough for it to be effective at highish frequencies.
See in the video at 27:16.
Now, if you were dealing with very high bandwidth opamps dealing with mhz signals, you might want to refine things (as per 28:29 in the video) but you're pretty much safe as is.
Btw, decoupling rail to rail might not be as effective as rail to ground depending on the situation. But that's a big topic.
edit: slow to post... I second what MarcelvdG wrote above.
See in the video at 27:16.
Now, if you were dealing with very high bandwidth opamps dealing with mhz signals, you might want to refine things (as per 28:29 in the video) but you're pretty much safe as is.
Btw, decoupling rail to rail might not be as effective as rail to ground depending on the situation. But that's a big topic.
edit: slow to post... I second what MarcelvdG wrote above.
It should be fine for most op amps you're likely to use for audio.
100n is generally given as the minimum recommended value. It's also a 'suspiciously neat' value
and is often used by the bucketload to decouple all sorts of digital and logic where 10n or less would suffice.All things being equal a smaller cap will be effective upto a higher frequency but with less effect in general.
So it's not necessary but if you have them as your standard solution you don't need to change if you choose not to.
Also SMT is better - minimal inductance from no leads.
X2Y type capacitors are also worth a look...
Last edited:
Actually, the voltage dependence of the capacitance of an SMD X7R capacitor was the reason why I asked my question in the first place. I was wondering if such voltage dependence of the decoupling capacitor could be injected into the output of the OP Amp as a form of distortion. My guess is that it should not be a significant problem because (1) the X7R capacitor should have about the same voltage across it even as the cap charges and discharges a small amount during operation, and (2) the OP Amp (e.g., LM4562 or LME49710) tends to have a very good PSRR, so the voltage-dependent capacitance should have fairly negligible effects on the output of the OP Amp.
Kurt
Kurt
Secondary point to be made here - if you have regulators on each rail, each already 'seeing' some rail bypass capacitance to ground - with such a large cap between rails, you are also cross-coupling the regulator outputs.
Things can get quite interesting then, even with 'old-fashioned (3pin) voltage regulators'... and expect gross instability into major oscillation if you are trying to use anything with a wider regulator internal bandwidth in the hope of '...better'. An oscilloscope, some measurement savvy, and a bit of compensation / anlaogue-electronic finesse will definitely be required.
Just for completeness' sake: ceramic class 2 capacitors also have a time dependence. The specified value normally applies 1000 hours after the last time the capacitor was heated above its Curie temperature. The capacitance drops with the logarithm of the time since it was last heated above the Curie temperature, usually by a few percent per factor of 10 increase of the time.
Through-hole ceramic class 2 capacitors are also quite nonlinear, though not as badly as small SMDs.
I don't see why voltage dependence would be a big deal for supply decoupling, but if you want virtually no voltage dependence and no ageing, use ceramic class 1 capacitors like NP0 / C0G. Mind you, they are only available up to 220 nF or so (as 1812 SMDs).
I'm using X7R caps to ground on the output of some modern regulators (ADP7142 and ADP7182) that power audio op amps, and they can work without any apparent distortion problems, down to the residual of an APx-555. However, I'm only using the minimum needed to satisfy the 1.5µF requirement of the regulators, and only caps to ground - I see no need to couple the + and - rails with an additional capacitor.
For my uses with 15V rails, I need to use 3x 1µF 1206 parts to get enough capacitance, taking temperature, tolerance, voltage sag, and capacitance decay into account. Even then, not all X7R 1µF caps work well enough - different parts have different amounts of capacitance loss at DC bias levels of 15V. Changing to a single 4.7µF with 30V bias will get you down into the deep regions of "dielectric soakage", and you'll probably get only about 20% of the faceplate capacitance, and you'll end up in a deeply nonlinear region of the capacitor: most of the capacitance domains are used up by DC bias, with little left for AC capacitance.
So, I see no real point to using extra capacitance across the rails, and a slightly large capacitance value used at a high DC bias will provide relatively little extra capacitance, and a lot of nonlinear behavior. I'm not sure that nonlinearity would matter, but it seems to not be worth trying, and with a pair of good regulators, I see no benefit. Again, I use these X7R caps only to keep the regulators happy, and to maintain low impedance at very high frequencies for the amplifiers. However, I'd be happier if the regulators allowed a 100nF cap, since it could be a C0G part, but alas, reality doesn't work that way.
I say skip the caps between the rails entirely. It's just another part that you have to pay to place on the PCB, and you'll have to move stuff around to fit its pads onto the PCB, space that could be used for other things to better effect.
For my uses with 15V rails, I need to use 3x 1µF 1206 parts to get enough capacitance, taking temperature, tolerance, voltage sag, and capacitance decay into account. Even then, not all X7R 1µF caps work well enough - different parts have different amounts of capacitance loss at DC bias levels of 15V. Changing to a single 4.7µF with 30V bias will get you down into the deep regions of "dielectric soakage", and you'll probably get only about 20% of the faceplate capacitance, and you'll end up in a deeply nonlinear region of the capacitor: most of the capacitance domains are used up by DC bias, with little left for AC capacitance.
So, I see no real point to using extra capacitance across the rails, and a slightly large capacitance value used at a high DC bias will provide relatively little extra capacitance, and a lot of nonlinear behavior. I'm not sure that nonlinearity would matter, but it seems to not be worth trying, and with a pair of good regulators, I see no benefit. Again, I use these X7R caps only to keep the regulators happy, and to maintain low impedance at very high frequencies for the amplifiers. However, I'd be happier if the regulators allowed a 100nF cap, since it could be a C0G part, but alas, reality doesn't work that way.
I say skip the caps between the rails entirely. It's just another part that you have to pay to place on the PCB, and you'll have to move stuff around to fit its pads onto the PCB, space that could be used for other things to better effect.
Misinterpretation.
Don't (obviously) use X7R etc where you require signal fidelity.
Like for like sticking some wires onto a capacitor doesn't improve anything unless you want inductance there
Good stuff showing reality always a bit more complex than simple generalisations. Thanks for posting the link.
Misinterpretation.
Don't (obviously) use X7R etc where you require signal fidelity.
Like for like sticking some wires onto a capacitor doesn't improve anything unless you want inductance there
I didn't refer to the x7r topic. Small components inevitably have thermal problems (distortion, noise), as well as mechanical problems (cracks).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.