Importance of the quality of the DC-blocking feedback capacitor?

[DF96]

I am very serious, I sold it now, but I had a super nice power amp of 100 watts with 0.001%thd. Change the feedback amount by 1%, you will hear the difference, change anything inside a feedback loop (well designed) of over 22db, you wont hear anything, because it is precisely the function of the loop to correct any component deviation in the audible frequencies.

There are at least two things the feedback loop cannot correct:
1. distortion in the feedback components
2. common-mode distortion at the summing point

Item 1 is what this thread is about.

This point of the circuit is a virtual ground and is very important use a good bipolar capacitor and bypass it and choose correct grounding node ( signal ground ).

Good cap: yes. Bypass: no. Correct ground: yes yes!

 

[DF96]

Since you can't have electronics to work without current loops, I can't see why it is wrong to generalize using the term signal.

The problem is that 90% of the people who like using the term 'signal path' don't know enough electronics to correctly identify it in a circuit. About the most common mistake is to omit shunt components (e.g. CCS), which are as much in the signal path as series components.

 

[dmills]

Since you can't have electronics to work without current loops, I can't see why it is wrong to generalize using the term signal.

Well you can, but usually people use that term and then look at less then half the diagram, and manage to miss things like feedback paths and where the charge come from (and goes to) as the output changes quadrants.

The fact that opamps are generally drawn as THREE terminal devices (not five) does not help here, and yes it matters more then you would think.

Regards, Dan.

 

[jan.didden]

Lets be sensible here, the signal path is the path (or paths) carrying the signal. This is not hard to understand. The supply rails are not the signal path, a constant current source is not the signal path (but a current mirror might be).

The feedback path is definitely a signal path, and you'll realize that the signal can go round a looped path if you think a bit.

Things get more complicated sometimes, such as the splitting of the signal in two in a class-B output section.

And as for the DC blocking cap in the feedback its important that this is linear enough - which means having only small voltage across it, which means using a high value cap, perhaps 1000µF kind of range, so that the roll-off frequency it defines is way below the audio band (1Hz or lower). Then the signal voltages across it are measured in mV and µV, and any distortion is very small and very insignificant compared to the larger voltages across the feedback resistors.

Capacitor distortion in electrolytics is very real and the only defence is to keep the ac voltages small.

+100

Jan

 

[jan.didden]

Although multi-stage feedback topologies are typically associated with inverting applications, the following non-inverting variation can be useful for attenuating DC without the use of high capacitance values (IOW electrolytics) or DC servos.

In the right figure, the addition of NFB network #2 with its high resistance value (hundreds of kohms, or megaohms) allows the capacitance value to be reduced to a range where film capacitors will be practical.

Apologies for the small text size - hope that blowing up the drawing will make it readable!

hth, jonathan

Very nice Jonathan, I learned something today!

Jan

 

[wiseoldtech]

What are you talking about?

They are VERY MUCH in the signal path.
In fact NFB ***corrects*** signal through amplifier.

To be even more precise, amplifier does not actually amplify input signal , but the difference between input signal and a sample from the output signal.

Guess how we get that sample and what network does it go through?

An important part of which is ... ummmm .... that electrolytic cap.

Still think it´s not in the signal path?

EDIT

Sorry but you are wrong.
It does all the difference a capacitor can make when straight in the signal path: see above.
Any distortion, phase shift, nonlinearity, anything you can detect or measure across it will be directly injected to the - input of the amplifier.
Even more, amplifier will *try* to faithfully follow signal dropped across that very capacitor and the resistor to ground, go figure.

EDIT 2:


I never bring this up out of courtesy but since you mention it, being an audio video service technician does not mean by itself a high theoretical level of understanding.

Some people speak of the "Internet University" education.
That is a mixed bag, there is some useful data, lots of iffy or wrong type, it´s often hard to tell one from the other, not exactly the best education out there.
The MAIN problem is that it can not go very deep, so information tends to be superficial, and many times anecdotic.

NO NO NO NO NO: NFB signal is "practically" exact same level as input signal; amplifier uses the minute difference between them.
The principle behind NFB lies precisely upon the amplifier trying to make that difference exact ZERO, go figure.

True, but this added to your earlier statements proves you have a superficial knowledge about how NFB works.
Not really WRONG, there´s some shreds of truth in it, but still showing misunderstanding so you don´t realize some basic errors.

Again, not way below at all.

Sorry, but most of what you've mentioned only indicates to me your level of understanding due to no professional schooling by certified instructors.

And that's ok, it's fine, not everyone has degrees in electronics.

 

[wiseoldtech]

No. In most SS amps the feedback signal is about the same amplitude as the input signal, and it is more in the signal path than the main gain circuit.


No. In most SS amps the feedback is in phase with the signal but it is applied to a different point. Either you slept through the amp design and feedback parts of your training or your teacher was confused.

LOL!
I can see you're stubborness and desire to argue with written and documented facts.
What I don't understand is why.

 

[knutn]

Well you can, but usually people use that term and then look at less then half the diagram, and manage to miss things like feedback paths and where the charge come from (and goes to) as the output changes quadrants.

The fact that opamps are generally drawn as THREE terminal devices (not five) does not help here, and yes it matters more then you would think.

Regards, Dan.

Then I agree.

I came to think of digital electronics: Take a simple AND-gate; how many think of current flow looking at the symbol?

 

[DF96]

I can see you're stubborness and desire to argue with written and documented facts.
What I don't understand is why.

And I thought it was you being stubborn!

Anyway, you seem to agree that it is facts that I put forward in correcting your misunderstandings. The reason I do this is a peculiar habit of mine: I like people to believe the truth instead of error.

By the way, JMFahey is usually right. He may be a bit blunt but he understands electronics. You should listen to him.

 

[JMFahey]

Sorry, but most of what you've mentioned only indicates to me your level of understanding due to no professional schooling by certified instructors.

And that's ok, it's fine, not everyone has degrees in electronics.

Maybe that´s why AES invites me to Lecture on Pro Audio hardware design and Manufacturing at their Conventions

Maybe because I have been successfully Designing and Manufacturing Amplifiers and Speakers for 50 years now.

 

[jan.didden]

But it is in Spanish, so it can't be true!

Wiseoldtech, you made the classical mistake: judging someone against your own opinion before checking him out.

Think about it: the amp actually amplifies the difference between the input signal and the feedback signal. It amplifies it with its open loop gain. So the difference between the input signal and the feedback signal is Vout/Aol. The feedback signal has exactly the same impact on the output as the input signal, so any error or distortion in the feedback network distorts the output.

It may not be literally in the signal path if you define that path as the direct path from input to output, but that's just semantics.

Jan

 

[DPH]

I think the most important thing for people to take note of in this whole discussion is to work from the fundamentals rather than blindly apply shortcuts. Certainly the concept of "signal path" is not a classically taught.

Otherwise several folks nailed it as a bypass cap in function, so make it a decent quality bipolar electrolytic and make it big. Depending on the opamp/amp topology, the stacked feedback method can incur some healthy low frequency current noise penalties.

Not that I've done it, but if one is doing active crossovers, then this cap provides a nice place to apply a filter rather than adding an additional sallen key/mfb stage.

 

[dmills]

I came to think of digital electronics: Take a simple AND-gate; how many think of current flow looking at the symbol?

The guys dealing with the "analog" bits of digital (Like say the people concerned with SI and EMC mitigation surely do), as do the folks dealing with the power planes and such once you get above the near DC stuff below 100MHz or so.

Frequency does not matter so much with modern logic, edge rates do, and you better believe they are quick enough to be a serious issue in design. There is a reason you seldom see a two layer board in anything digital these days.

Consider an opamp, what is the Vas integrator capacitor referenced to? Where does the output current return to? Draw the loop. In an AB stage driving a reactive load, how does this vary? Does the Vas integrator share an impedance with the output stage loop? Why might this be an issue in that class AB stage?

Yep, some of my job interview questions, I have a bit of a reputation with our usual recruiting company (Apparently I make people cry)!

Regards, Dan.

 

[scottjoplin]

The feedback signal has exactly the same impact on the output as the input signal, so any error or distortion in the feedback network distorts the output.

That's what I understood. Am I misunderstanding Rod here: Distortion and Feedback
"Also notable is that any signal that is created within the feedback loop (most commonly noise) is also cancelled by global feedback. Because this generates signals that did not exist at the input, the error amplifier 'sees' any such extraneous signal as a deviation from the input signal, and cancels it to the best of its abilities. Note that input device noise is not cancelled, because the error amplifier cannot differentiate between noise it has created and the input signal"

 

[wiseoldtech]

And I thought it was you being stubborn!

Anyway, you seem to agree that it is facts that I put forward in correcting your misunderstandings. The reason I do this is a peculiar habit of mine: I like people to believe the truth instead of error.

By the way, JMFahey is usually right. He may be a bit blunt but he understands electronics. You should listen to him.

No, I do not agree with your so-called "facts", sorry.

Your statement regarding making people "believe the truth" to me is mildly amusing. Because it only shows me yet another source of "internet misinformation", of which I've seen plenty of already. You simply cannot change fundamental electronic facts to suit your agenda, or pacify your beliefs.

I prefer on not arguing on here about things that are credible facts derived from proper schooling, or listening to remarks that my instructors of such schooling are incompetent idiots. I was educated long before the internet came along, in classrooms, with textbooks and highly regarded instructors, in accredited institutions.

From the fundamental elements of learning Ohm's Laws, up to 1990's technology. I also went to various manufacturers Training Academy Sessions regarding specifics on their line of product servicing. (Sylvania/ECG, Magnavox, RCA, etc.)

Nevertheless, I see online, a substantial amount of misguided information, popular beliefs, and obvious misunderstandings about things. With that, I'm done with this specific thread, as I see I'm either not making things understandable, or others just want to argue their points to infinity.

 

[dmills]

I think the most important thing for people to take note of in this whole discussion is to work from the fundamentals rather than blindly apply shortcuts.

Indeed, shortcuts and rules of thumb can be a great time saver but far more important then knowing them is knowing when NOT to apply them (and when to get your maths on instead).

Not that I've done it, but if one is doing active crossovers, then this cap provides a nice place to apply a filter rather than adding an additional sallen key/mfb stage.

Note that that cap combines with the series resistor to form a pole-zero pair, so the gain ultimately winds up at Av=1 rather then dropping to zero, but yea you can get some low frequency attenuation there (but then that cap has signal voltage across it, so the film part come back out as being the sane option).
You also need to make sure the amp is stable with Av=1 if doing that or you can wind up with low frequency oscillation.

 

[DPH]

Yes, good points, Dan, that I left as an exercise for the reader, that maybe I shouldn't have done.

(Or laziness of typing on my phone)

 

[jan.didden]

I was educated long before the internet came along, in classrooms, with textbooks and highly regarded instructors, in accredited institutions.

From the fundamental elements of learning Ohm's Laws, up to 1990's technology. I also went to various manufacturers Training Academy Sessions regarding specifics on their line of product servicing. (Sylvania/ECG, Magnavox, RCA, etc.)

So, pray tell, what was it that you were told that specifically contradicts my post 53? I think it would be only fair to point out my errors if you feel they are there.

"Think about it: the amp actually amplifies the difference between the input signal and the feedback signal. It amplifies it with its open loop gain. So the difference between the input signal and the feedback signal is Vout/Aol. The feedback signal has exactly the same impact on the output as the input signal, so any error or distortion in the feedback network distorts the output. "

Jan

 

[KCHANG]

Wow! When I went to bed last night there were fewer than 10 posts in this thread, and now there is a whopping 6 pages of posts with some heated debates/arguments. I did not expect my simple, seemingly innocuous, question would be so controversial. I thought that this issue must have been considered, and there must have been a consensus reached, as people have been building solid state amps for half a century by now.

In seriousness, thanks guys for all of your comments. I know whose comments/advise I'll follow, and I recognize those who are true experts, but still I appreciate the inputs of those whose views I do not subscribe to.

One thing I am going to try is to short the DC-blocking feedback e-caps on my LME49810 amp to ground (i.e., "bypassing" each cap with a piece of wire), and see how bad the DC offsets of the amp output would be. If the DC offsets are tolerable, then perhaps I will use the amp that way, with the (maybe false?) peace of mind that there is no cap distortion to pollute the amp output.

 

[ctrlx]

ofcourse, you can use a lower value NP "quality" cap, just scale up the f/b R's.