About op-amps use.

[ginetto61]

Hi !
reading around i found this intriguing comment about using high value of uF very close to op-amps.

... Here's some of what have learned about op-amps.... Put 6800uF at every power pin of the op-amps, all the op-amps. This will make your op-amps disappear, no more noise, no more distortion! Bass, as good as the circuit can deliver...

Ok .. maybe 6800 uF is a little on the high side.
But i see very few uF used normally. I have listen to a preamp based completely on opa604 with 2x1000 uF close to the opa.
It sounded both clean and not particularly solid-statish ... i would say very good indeed. Clean as solid state can be but also not hard or harsh.
What do you use normally ?
Thanks a lot, gino

 

[abraxalito]

How much to use depends on the particular opamps, their loading and the circuit toplogy. Run balanced with very light loading so the output stages don't come anywhere near leaving classA and you probably won't need that 6800uF. But it might come in useful for filtering a noisy reg, in conjunction with a series L.

 

[ginetto61]

How much to use depends on the particular opamps, their loading and the circuit toplogy.
Run balanced with very light loading so the output stages don't come anywhere near leaving classA and you probably won't need that 6800uF.
But it might come in useful for filtering a noisy reg, in conjunction with a series L.

Hi and thanks a lot for the helpful advice. I think i understand now.
If the opamp works within class A it draws always the same amount of current.
I guess that bias current is what in the datasheet is called idle/quiescent current ?
How can i be sure that this current is not overcome ?
For instance i see an opa209 here with 2.5 mA of quiescent current. This is nothing ... it is clear that it will work always above that level.
Instead the AD811 that is considered an "hot" opamp is around 15mA.
No opamp around 40-50 mA ??? just to be sure ...
and in that case for a preamp no reason to go too high with uF if i understand well.
However .... is it possible to increase bias current for an op-amp ?
or is it fixed ?
Now i start to understand the reason of class A.
Thanks again, gino

 

[abraxalito]

Its not necessarily the case that if an opamp stays in classA that it'll always draw a constant current from the supplies - that's because the load current has to go somewhere. Hence my stipulation of running balanced which cancels the load current effect.

How you can be sure - short of measuring the current taken from the supplies in the real-world circuit, I don't know how to be sure. I think its fairly safe to assume that a 50uA output current isn't going to cause the output stage to leave classA. The quiescent current spec'd in the DS isn't all going to flow in the OPS, only a smallish proportion does.

When I want more current from an opamp OPS than 50uA I'll put in a discrete classA buffer to handle it or alternatively hang a CCS from the opamp output in the case of an opamp with decent drive capability (not a TL072!).

 

[DF96]

If the opamp works within class A it draws always the same amount of current.

No. Class A means the current averaged over a complete AC cycle remains unchanged, whatever the signal amplitude. The signal current may appear (often will appear) on the supply rails, but as it is AC it will average to zero.

Local supply rail decoupling works if it provides a lower impedance for AC than the supply rail itself. At higher frequencies supply rail inductance means that small caps can help. At lower frequencies the resistance dominates, but this will be small so you would need a rather large cap very close to the chip to make any difference - but a large cap will bring its own problems, as it has inductance too. No point whatsoever in putting more than a few uF at the supply pins of a normal opamp - to do its job the cap must be physically small.

 

[ginetto61]

Its not necessarily the case that if an opamp stays in classA that it'll always draw a constant current from the supplies - that's because the load current has to go somewhere. Hence my stipulation of running balanced which cancels the load current effect.
How you can be sure - short of measuring the current taken from the supplies in the real-world circuit, I don't know how to be sure. I think its fairly safe to assume that a 50uA output current isn't going to cause the output stage to leave classA. The quiescent current spec'd in the DS isn't all going to flow in the OPS, only a smallish proportion does.
When I want more current from an opamp OPS than 50uA I'll put in a discrete classA buffer to handle it or alternatively hang a CCS from the opamp output in the case of an opamp with decent drive capability (not a TL072!).

Hi and thanks a lot for the very interesting and a little difficult for me to understand explanation.
But i have two very trivial questions:
1) what are the drawbacks of putting more uF close to the op-amps ? value like 10-20 milliF i mean. Will there be problems ?
2) i see in the dataheet a value of max current often from 20 mA up. Is it not enough for a line stage ? and for an headphone stage ?
Thanks a lot again. Gino

 

[ginetto61]

No. Class A means the current averaged over a complete AC cycle remains unchanged, whatever the signal amplitude. The signal current may appear (often will appear) on the supply rails, but as it is AC it will average to zero.
Local supply rail decoupling works if it provides a lower impedance for AC than the supply rail itself. At higher frequencies supply rail inductance means that small caps can help. At lower frequencies the resistance dominates, but this will be small so you would need a rather large cap very close to the chip to make any difference - but a large cap will bring its own problems, as it has inductance too. No point whatsoever in putting more than a few uF at the supply pins of a normal opamp - to do its job the cap must be physically small.

Hi and thanks for the valuable advice. The guy seemed quite convinced actually. The only way would be to see the V changes at the caps pins.
If they do not sag during peakes no problems at all.
I understand that for line stage the delivery requirements are quite light.
Low current delivery is needed.
Thanks again, gino

 

[abraxalito]

1) what are the drawbacks of putting more uF close to the op-amps ? value like 10-20 milliF i mean. Will there be problems ?

10mF is going to be a fairly bulky beast so practically impossible to put it up close to an opamp unless you move Muhammed to the mountain (so to speak). Meaning perch the opamp between the pins of the cap, using the cap itself as the mounting of the opamp.

2) i see in the dataheet a value of max current often from 20 mA up. Is it not enough for a line stage ? and for an headphone stage ?

Certainly enough for a line stage, enough for a few higher impedance headphones. But that's a quantitative spec, tells you nothing about the sound quality when delivering that amount of current. Some DSs spec a current limit and a lower figure as 'maximum linear output current' or something similar.

 

[ginetto61]

10mF is going to be a fairly bulky beast so practically impossible to put it up close to an opamp unless you move Muhammed to the mountain (so to speak). Meaning perch the opamp between the pins of the cap, using the cap itself as the mounting of the opamp.

Sorry 😱 silly me. I wanted to say 1-2,2mF.
That can be done quite easily. Sorry i said a stupidity.
I will attach a picture with 1000 uF caps surroundigs some OPAs this evening as an example of what i have in mind.

Certainly enough for a line stage, enough for a few higher impedance headphones. But that's a quantitative spec, tells you nothing about the sound quality when delivering that amount of current. Some DSs spec a current limit and a lower figure as 'maximum linear output current' or something similar.

I understand and thanks a lot for the important advice.
Especially now that op-amps come almost always in smd it is quite clear to me that they are not power parts. They are not even heatsunk.
I have seen op-amp with up to 250-500 mA of current delivery without a metal plate to heatsink them properly.
I have had bad experience with heatsinks ... nowadays everything seems to be designed to stay in a cell phone or in a tablet ...
Thanks a lot again and sorry for the mistake.
I had 1000-2200 uF in mind as max values. No more.
Kind regards, gino

 

[AndrewT]

...................I have seen op-amp with up to 250-500 mA of current delivery without a metal plate to heatsink them properly. ......................

high current IC (opamps and Buffers) will require a heatsink.
This could be using the multiple legs/leadouts to a very big copper plane on the PCB, or could be a contact surface on the bottom side of the chip that requires a big copper plane to dissipate the heat.

Some are required to be soldered to the copper plane.

 

[ginetto61]

high current IC (opamps and Buffers) will require a heatsink. This could be using the multiple legs/leadouts to a very big copper plane on the PCB, or could be a contact surface on the bottom side of the chip that requires a big copper plane to dissipate the heat.
Some are required to be soldered to the copper plane

Hi and thanks a lot for the very helpful reply.
Still i see some high current parts smd ... what is this fashion of smd ?
I can understand for digital circuits ... for pc components ... but for analog ?
parts really need to be so tiny ? to then maybe mount them on adapters ?
i prefer the old big packages by far.
Thanks again, gino

 

[abraxalito]

Looks to me you got the question the wrong way around 🙂 'Do through-hole parts really need to be so big?' And the answer is 'no they don't'. Smaller parts perform better (except for heat dissipation) because their parasitics (primarily lead inductance, capacitance) are lower. SMT makes for better results with high speed analog as well as digital.

 

[ginetto61]

Hi this is the preamp i was mentioning ... the caps surrounding the OPA604 are all 1000uF/63V ( i mean quite big) with other smaller around
Regards, gino

 

[ginetto61]

Looks to me you got the question the wrong way around 🙂 'Do through-hole parts really need to be so big?'
And the answer is 'no they don't'. Smaller parts perform better (except for heat dissipation) because their parasitics (primarily lead inductance, capacitance) are lower.
SMT makes for better results with high speed analog as well as digital.

Hi and thanks for the advice and i think i start to understand something
It is just that smd has put kits out of my reach.
I have just no hope with smd parts.
So deep in my soul i hate them, even if the circuits with them can be work of art, very very beautiful indeed.
But i still do not know how many mA are needed for instance to drive properly a power amp. 1 ? 10 ? because some power amp have also high input capacitance. They are not purely resistive.
I am confused and then i see also commercial units that confuse me even more 😱
Thanks again, gino

 

[jcx]

cargo-cult design? - pursuing their idee Fixe without an appreciation of other factors, making a reasoned weighting of often competing choices that can't all be "maxed out" at once

the big electro pair per op amp, apparently on PS rails spaced apart: if you really are designing PS distribution impedance, minimizing circuit interaction then +/- power traces want to be close together - "common centroid"


another clue is close spaced gnd? "pour" - not every node in a circuit wants added few pF fringing C to the pour plane - at least the cutout pour regions do seem to be staked with vias so they aren't just EMI antenna


and there really is "good enough" - even in audio - when calculated and measured pssr, crosstalk, ect. signal consequences of say PS distribution design are below hearing thresholds

 

[peranders]

Hi !
reading around i found this intriguing comment about using high value of uF very close to op-amps.



Ok .. maybe 6800 uF is a little on the high side.
But i see very few uF used normally. I have listen to a preamp based completely on opa604 with 2x1000 uF close to the opa.
It sounded both clean and not particularly solid-statish ... i would say very good indeed. Clean as solid state can be but also not hard or harsh.
What do you use normally ?
Thanks a lot, gino

Notice that your findings are not accepted by engineers nor high end audio designers. The basic purpose of the decoupling is to create a low impedance path from the supply pin down to ground.

 

[abraxalito]

But i still do not know how many mA are needed for instance to drive properly a power amp. 1 ? 10 ? because some power amp have also high input capacitance. They are not purely resistive.

A typical power amp will have input capacitance probably of the same order of magnitude as the cable capacitance (the low hundreds of pF likely due to input RF filtering) so only at the top of the audio band will this potentially dominate over the resistive portion of its input impedance. I'd say 500uA will be the most you'll need to source even into the lowest input impedance amps. You'd only need more if you're using long high capacitance cables.

 

[ginetto61]

cargo-cult design? - pursuing their idee Fixe without an appreciation of other factors, making a reasoned weighting of often competing choices that can't all be "maxed out" at once
the big electro pair per op amp, apparently on PS rails spaced apart: if you really are designing PS distribution impedance, minimizing circuit interaction then +/- power traces want to be close together - "common centroid"
another clue is close spaced gnd? "pour" - not every node in a circuit wants added few pF fringing C to the pour plane - at least the cutout pour regions do seem to be staked with vias so they aren't just EMI antenna
and there really is "good enough" - even in audio - when calculated and measured pssr, crosstalk, ect. signal consequences of say PS distribution design are below hearing thresholds

Hi and thanks a lot for your very important advice.
I think that your conclusion is also fundamental. As long as some effects are below human sensitivity everything can sound just right and good.
Actually i have no clue about how this preamp performs on the bench in terms of S/N ration, THD, etc.
The sound is quite ok ... like other sounds by the way.
I was just surprised to see these big 1mF caps around the opa.
Maybe i should try some smaller ones and still i would get something below my hearing threshold ... placed quite high 😱 i have not a bat ear.
Thanks a lot again, gino

 

[ginetto61]

Notice that your findings are not accepted by engineers nor high end audio designers. The basic purpose of the decoupling is to create a low impedance path from the supply pin down to ground.

Hi and thanks a lot for the valuable advice.
So what i was thinking about a " energy buffer effect " is very wrong.
The cap upgrade is very common in audio equipment mods ... very.
Some Companies offer upgraded versions of their power amps for instance.
Different situation now i understand.
For preamps this is useless.
Thanks again for the helpful explanation, gino

 

[ginetto61]

A typical power amp will have input capacitance probably of the same order of magnitude as the cable capacitance (the low hundreds of pF likely due to input RF filtering) so only at the top of the audio band will this potentially dominate over the resistive portion of its input impedance. I'd say 500uA will be the most you'll need to source even into the lowest input impedance amps.
You'd only need more if you're using long high capacitance cables.

Hi and thanks again. This is a very important reference for me.
Let's be generous and make it 1 mA ? i can only say that i was completely wrong. ANY opamp around i guess is able to provide a current that low.
This is very important for me to know.
I have also the feeling that if people do not see them they also accept them. I mean the opamps ... 😉
They are just so handy and cheap that maybe it is difficult to accept that they can sound lovely ? 🙄
Thanks a lot again.
Kindest regards, gino