Op amp power decoupling... again...

Lucien

Member
2007-11-14 7:48 pm
So I know there are quite a few threads on this topic already, with quite a bit of different schemes attempted. But my question is has anyone been able to do any actual experiments (measurements), or know of any results that lead to a recommended bypass solution?
 
It's very easy in one way, and very difficult in another.

CMOS OpApms have very high slew rates, so a good bypassing is essential to avoid feedback and noises, with danger of oscillations.

Bipolar and JFET op amps have much less slew rate, but a powerful cross conduction at its outputs, so a also a good decoupling is mandatory.

As a rule of the thrumb, a good ceramic cap (say .1 to .22µF) and a good electrolytic (12 to 22µF) between VCC and ground and or ground to VEE is sufficient. It sometimes a 22R ohms in series with both power lines cooperates with such a decoupling between stages, and in case of a fault, they may act as a fuse.

When choicing electrolytic, please avoid the stupidity of use 50 or 63V units at 15V, use in place, 16 or 25 ones.
 
Last edited:
Hi Guys

Electrolytic capacitors are formed by application of voltage and remain healthy through continued application of voltage. if there is only 15V applied the cap will gradually reform to withstand ONLY 15V regardless of its initial rating and forming.

So, why not just use the properly rated part? It is called "working voltage" for a reason.

With respect to the decoupling: It is both good and bad to decouple the rails to ground, as noise can be injected into the ground reference through these paths. With many opamps, it is better to place the HF decoupler from rail-to-rail directly at or as close to the supply pins of the opamp as possible. The electrolytics and related decoupling resistors provide local filtering and charge storage for signal surges. I find the usual values much too low inasmuch as the RC constant is within the audio band in most cases, so I use much larger Cs. The Rs have to be kept low to not interfere with load driving

Have fun
 
Last edited:
Electrolytic capacitors are formed by application of voltage and remain healthy through continued application of voltage. if there is only 15V applied the cap will gradually reform to withstand ONLY 15V regardless of its initial rating and forming.

So, why not just use the properly rated part? It is called "working voltage" for a reason.

110% agree ;-D
 
There is another technical reason to use the nearest voltage value. As the capacitor volume is increased (the case of higher capacity and or voltage values), the internal parasitic inductance also increases. Then the frequency of resonance (of the self inductance and capacitance) of it decreases. If you use this cap. above this parallel resonant, the capacitor no longer behaves as a cap, it does as an inductance, increasing their reactance with frequency. Then, higher cap insulation voltage, worse high frequency performance.

Clear like water.
 
Hi Guys

Electrolytic capacitors are formed by application of voltage and remain healthy through continued application of voltage. if there is only 15V applied the cap will gradually reform to withstand ONLY 15V regardless of its initial rating and forming.

So, why not just use the properly rated part? It is called "working voltage" for a reason.

With respect to the decoupling: It is both good and bad to decouple the rails to ground, as noise can be injected into the ground reference through these paths. With many opamps, it is better to place the HF decoupler from rail-to-rail directly at or as close to the supply pins of the opamp as possible. The electrolytics and related decoupling resistors provide local filtering and charge storage for signal surges. I find the usual values much too low inasmuch as the RC constant is within the audio band in most cases, so I use much larger Cs. The Rs have to be kept low to not interfere with load driving

Have fun

Hi,
Can you elaborate a bit more on this -
"it is better to place the HF decoupler from rail-to-rail directly ".

Between the +ve rail and ground, -ve rail and ground and one between +ve and -ve rails, is this correct ?

Thanks
 
Last edited:

Lucien

Member
2007-11-14 7:48 pm

Thanks, I've actually already read that. It's helpful but also a bit limited in breadth of examples. I'm also wondering if there's been much change in components since 2009 (when it was published). The "old" advice for op-amp decoupling was a 0.1 µF ceramic in parallel with a 4.7 µF tantalum. But one old datasheet for tantalum caps put the ESR at about 4.7 ohms! Quite a lot compared to even modern Al electrolytics.

In contrast, the LME49990 datasheet says a 0.47μF ceramic, 2.2μF ceramic and a 10μF tantalum together were found to work well for that chip. And the new OPA1688 and OPA1622 simply recommend a 0.1 µF ceramic per rail, though I suspect they mean that as a minimum.

It's very easy in one way, and very difficult in another.

CMOS OpApms have very high slew rates, so a good bypassing is essential to avoid feedback and noises, with danger of oscillations.

Bipolar and JFET op amps have much less slew rate, but a powerful cross conduction at its outputs, so a also a good decoupling is mandatory.

As a rule of the thrumb, a good ceramic cap (say .1 to .22µF) and a good electrolytic (12 to 22µF) between VCC and ground and or ground to VEE is sufficient. It sometimes a 22R ohms in series with both power lines cooperates with such a decoupling between stages, and in case of a fault, they may act as a fuse.

At one point I was considering something similar, but placing a resistor damper in series with the larger valued cap. So:

(0.1 µF ceramic) in parallel with (4.7 - 10 µF polymer or other low ESR cap in series with 2.2 - 4.7 ohm resistor)

Plus a ferrite bead between between the supply and the capacitors. But the Murata ones I looked at have very low DC resistance while the impedance peaks at over 100 MHz. I'm not sure which, if any of the many options would do the job.
 
You may do such, but at audio frequencies, a ferrite bead (10µHy) is useless. It is a short.

Also, as an example, a 1µF 50V standard electro has a resonance at about 100KHz. So, for AF applications, is more than sufficient.

The data published in the datasheets, is not a must. Your needs may be different depending the application and the PCB layout, mainly. According to the needs of the circuit, the decoupling may or may not be that they suggest. This, in that to decoupling refers. But always be under the maximum ratings for the device. This do is a must‼
 
Last edited:

raul_77

Member
2001-05-12 11:28 am
Spain
When choicing electrolytic, please avoid the stupidity of use 50 or 63V units at 15V, use in place, 16 or 25 ones

There is a good reason to prefer electrolytic capacitors with voltages close to 50V: distord less.

Electrolytic below 50V and above 63V, typically have thinner dielectrics for reasons of standard of encapsulated. For this reason 50V typically have thicker dielectrics, it favors them the package size, and this reduces distortion. Always considering the same brand and series of capacitor.

This has been measured and recommended by Cyril Bateman in his magnificent series of articles.

If you want to use lower voltages, perhaps it is best to use electrolytic type 'for Audio' as Nichicon Muse, Elna Cerafine, etc. Usually have thicker dielectrics and therefore distort less (and are larger), also they tend to be better damped internally, to avoid problems of microphonics and self modulation by the passage of the alternating signal.

About ceramic, use only NPO (COG) the 'normal' distord bad, this are also measured by Cyril Bateman. Avoid the 'normal' ceramics, use instead at least poliester, best MKP or Styroflex.

And maintain a good voltage margin is a good idea. For example, if you see darTZeel schemes for power supplyes of 55V using electrolytics of 100V, not of 63V.

Keep components under conditions of easy work improves circuit performance because the less the components to be forced less distorting. Can you do a test: You can do the same circuit with capacitors of 16V and resistors of 1/4 W, and electrolytics of 50V and resistors of 2W ... surely find that the latter sounds better