Is the 22uf C2 (feedback cap) in LM1875 optional

[mudihan]

In the LM3875's datasheet, the feedback cap Ci is marked optional (as does the one in LM3886, I think). But in LM1875's datasheet, the same feedback cap C2 is not. Is there a reason that the National engineers treat the cap in LM1875's datasheet differently? Or were they just over cautious at the beginning?

I of course want to omit it.

 

[mudihan]

I know that I should measure to the the DC offset at the output if I choose not use the feedback cap in a LM1875 gainclone. But I just wonder whether the engineers at National intentionally want to treat LM1875 differently in regard of the cap for other good reasons.

 

[Dxvideo]

I have done a lot of chipamps with LM3886, LM3875, LM4780 and LM1875.. And as I measured;
For overture series (3886,3875,4780) <=20K Rb (input shunt resistor) the DC offset value is not more than 50mV which is acceptable.. And for LM1875 its better (~20mV).
So I don't see any reason to put Ci both for 1875 and the others.. Of course, you must consider the sources DC offset first (if your source has 10mV offset then with a 21X gain amplifier you will have 210mV on speakers!)..

 

[Stream]

But if we use input capacitor (up to 10uF), does it means that chip shouldn't has dc offset?

 

[AndrewT]

Hi,
make the amplifier DC coupled and put in the protections to ensure reliability of the system.
Alternatively make an amplifier that is AC coupled.

Do not build an amplifier that is mixed AC and DC coupled.
Fitting one of the DC blocking caps and omitting the other is mixed AC and DC coupling. Don't.

 

[Dxvideo]

If an amplifier is AC coupled then its AC coupled. Means you should use both input DC stopper (Cin) and NFB DC stopper (Ci)..
And if its DC coupled then you dont need neither..
In real life, I have tried all combinations for these capacitors;
No Ci but various MKP Cin
No Cin but Elna Silmic II, Elna Cerafine and Nichicon Muse bypassed with low value MKP as Ci
MKP Cin + Elna Silmic II, Elna Cerafine and Nichicon Muse bypassed with low value MKP as Ci
No Cin, No Ci
And I can say the best sounding one is no cap = DC coupled.. But the possible DC offset problems should be considered as mentioned.
But for many modern sources like CDP etc.. this is not an issue.. Most of them has an output cap or a dc servo unit for output..

 

[AndrewT]

But for many modern sources like CDP etc.. this is not an issue.. Most of them has an output cap...

This is an issue.
The DC blocking cap in the output of the source acts just like the DC blocking cap at the input of the receiver. You end up with mixed DC and AC coupling. Don't.

 

[Dxvideo]

So we have to make AC coupled amps always!
Because we can't know which source has an output cap and which has a dc servo!

 

[AndrewT]

do the homework.
Be informed before you make the decision.

There is little advantage in guessing.

 

[Dxvideo]

Ok... What about, if there will be an input selector before the amp? In this case that means I am planning to use multiple kind of sources, some with output cap and some without...

 

[jackinnj]

So we have to make AC coupled amps always!
Because we can't know which source has an output cap and which has a dc servo!

When you take the output from a computer source (particularly bad were old laptops) you don't know what kind of DC you are going to see on the output. If you are taking the output from an old analog source like a tuner, the DC will most likely be de minimus.

DC on the output of any opamp arises from current flowing FROM the input pins, amplified by the ratio of (1+Rf)/Rg for an non-inverting orientation.

 

[mudihan]

So the general consensus is that the C2 in LM1875's datasheet and Ci in LM3875's datasheet have exactly the same functions and could be omitted for the exact same reasons, despite the fact that one is marked optional while the other is not on the datasheet.

 

[pacificblue]

So the general consensus is that the C2 in LM1875's datasheet and Ci in LM3875's datasheet have exactly the same functions

Not a consensus, but common knowledge. An amplifier design based on consensus is not a promising idea. When humans tried to design a horse based on consensus, they came up with the camel.

Is there a reason that the National engineers treat the cap in LM1875's datasheet differently? Or were they just over cautious at the beginning?

The reason is, today they are promoting DC servos as better sounding. Whether they tell the truth or simply try to sell more op amps is everybody's own to assess. One thing is for certain, you have to deal with DC somehow, and AC-coupling is the easiest way to do it. If you don't use boutique capacitors, it will also be the least expensive way.

 

[mudihan]

Re: Re: Is the 22uf C2 (feedback cap) in LM1875 optional

Not a consensus, but common knowledge. An amplifier design based on consensus is not a promising idea. When humans tried to design a horse based on consensus, they came up with the camel.


The reason is, today they are promoting DC servos as better sounding. Whether they tell the truth or simply try to sell more op amps is everybody's own to assess. One thing is for certain, you have to deal with DC somehow, and AC-coupling is the easiest way to do it. If you don't use boutique capacitors, it will also be the least expensive way.

Please allow me to indulge myself:

Well, I am posting in a forum for hobbyists, rather than giving a lecture on, or actually planning, an amplifier design (not would I attempt to do either) -- who am I to declare what is common knowledge, and what's the point to make such a declaration.

Besides, the topic is not about how a feedback cap works in an amp design, or how to make LM chips work, but why two such caps in two similar amp designs received different attention from the engineers working in the same company. My intention was simply to see if someone knows something I don't about the two chips that would warrant the different attention for the feedback caps.

As always, I want to thank you for your reply.

 

[Dxvideo]

Re: Re: Re: Is the 22uf C2 (feedback cap) in LM1875 optional

My intention was simply to see if someone knows something I don't about the two chips that would warrant the different attention for the feedback caps.

May be these two datasheets wrote by two different team or engineer that one of them does not believe in AC couplage and capacitors (like most of us) and the other does!

Or may be after some years understanding and perceiving capability of the Application Note team has been changed! (like most of us)

 

[AndrewT]

maybe the APP note writer/team that omitted the DC blocking got sacked!
or, was it the other way round?
The team that used extra opamps to servo out the DC and detect when the servo was incapable of correcting the DC error all got promoted.

 

[mudihan]

By the way, which chip came first? LM1875 or LM3875? And how long were they apart?

 

[Dxvideo]

As I know, LM1875 came first, may be twenty years ago.. LM3875 is the first chip of overture series. I dont exactly know the reason but LM1875 sounds me more musical. But this is my subjective evaluation.

 

[Michael Bean]

My take on this issue is that you not only “can” mix AC and DC coupling, but you “should” do it that way, provided it’s done correctly. Here’s the deal; a feedback stabilized amp shouldn’t be asked to deal with any signal the feedback loop can’t pass, so if you DC couple the input and AC couple the feedback loop, a signal of a lower frequency than the corner frequency of the feedback loop will cause instability and distortion. Like wise, if the whole amp is AC coupled, and the input coupling is able to pass frequencies that the feedback loop can’t deal with, you will have problems. For any fully AC coupled amp the input corner frequency should be at least twice the corner frequency of the feedback loop or preferably higher. Further, consider the case where minimum gain for stability is greater than unity; AC coupling the feedback loop will guarranty problems with any signal that is lower in frequency than the feedback loop. I’ve always thought it odd that National Semi and other manufacturers considered it acceptable to AC couple the feedback loop of chip amps that require greater than unity gain for that very reason. I prefer to AC couple the input, and DC couple the feedback loop to avoid the issue altogether. Many are concerned about DC coupling causing excessive output offset, but I’ve never seen more than about 50mV at most using this method. Just my $.02 worth.

Mike

 

[pacificblue]

I’ve always thought it odd that National Semi and other manufacturers considered it acceptable to AC couple the feedback loop of chip amps

Which leaves you with two options. Either keep on assuming that all those engineers slept through university and don't know, what they are doing. Or find out, why they think AC coupling is the right thing to do.

Many are concerned about DC coupling causing excessive output offset, but I’ve never seen more than about 50mV

The first question is, whether that amount of offset is acceptable for the particular application. The second is, will all chipamps turn out 50 mV and less? Other forum members report 100 mV and more.

Engineers that develop an amplifier must take worst case assumptions into account. The datasheet specifies the worst case input offset with 100 mV. If want to use the maximum gain of 50, you have to take expect a worst case 5 V of DC at the output. That is more than 3 W for an 8 Ohm load and more than 6 W for a 4 Ohm load. Make it AC coupled and the output offset remains the same (acceptable?) 100 mV as the input.