Reducing DC offset on a NIGC?

[Optical]

Hi,
I have built this exact circuit with an LM3886, (well actually its got a 770 instead of a 680R resistor and a 1K input resistor but they will be changed/removed tomorrow..) but im not sure whether i have a safe dc offset or not..

I have built 2 of them and get 59mV on one and 72mV on the other side..
Im have some expensive speakers so i have only hooked the amp up to a cheap subwoofer, it works well but i cant comment on the tone quality as yet

So are there any 'tricks' to lower the dc offset to a few mV or so?
Help and advice appreciated!

[peranders]

The offset values you have got aren't anything to worry about if you are concerend about the speakers. My rule is <50 mV is good, bad over 100 mV. This is just my opinion.

The speaker has a limit when you will get increased dist and a second one when it become damaged (which is several volts).

[Optical]

Hmm, ok well thats good to know..
But before i put it in a box and forget about it..
Is there any components that directly relate to dc offset in that circuit that i can have a play with to try and bring it down into the 'good zone'??
Or is it more of a black art and i shouldn really bother?

[ashok]

It looks like the chip has no other terminals that you can use.
One method might be to connect your 680 ohms to ground through a large capacitor..... I guess a bad thing to do - polar or non polar! But that would drop the dc gain to 1 and hopefully reduce the error. There can still be inherent errors that you can only remove ( like if you want 1mV ?) by injecting additional voltage to one of the inputs. Something that I would VERY strongly advise against.
It would be better to leave things as they are. Anything else added on to drop the dc offset might just degrade the sound.
You might also open a Pandora's Box ! Your chips are alive now. Let them stay that way .

Can't you keep your hands away from an already good performance ? Just enjoy the music .
Cheers.

[peranders]

Originally posted by ashok
It looks like the chip has no other terminals that you can use.
One method might be to connect your 680 ohms to ground through a large capacitor..... I guess a bad thing to do - polar or non polar! But that would drop the dc gain to 1 and hopefully reduce the error. There can still be inherent errors that you can only remove ( like if you want 1mV ?) by injecting additional voltage to one of the inputs. Something that I would VERY strongly advise against.

It's no harm to trim away this offset. The only thing you have to think of is to inject a clean current so you won't get more noise and hum. Just create a voltage divider 12 V -> 10 mV (or so). Make sure this 12 V is well filtered and stabilized. This voltage divider should go via a trimpot, so you can adjust.

[runebivrin]

Break the connection between R5 and ground, and insert a capacitor in the break. Since the resistor values are rather small you'll need a large capacitor to avoid rolloff in the bass, so you might consider bumping up the resistor values by a factor 10 or so.

Then again, I don't think the offset is much to worry about in practice.

Rune

[Optical]

Thanks for the speedy replies guys!
....but i have another problem......

Ok i thought everything was fine, but when i changed the input resistor (R2 = 22k) to a 22k pot my DC offset went from -30mV on mute to +200mV fully open....!!!

I havent tried putting any caps from the 680R resistor to ground as yet..
I also have an input cap placed before the variable resistor, is this the correct place to have it or is it better to have it after the pot from middle pin to the non inverting input of the chip?

[peranders]

Optical, you can calculate the offset. For this you must know:

Input voltage offset
Input bias offset
Used resistor values in feedback and at the input.

If you don't know how to calculate, draw a schematics and I (or somebody else ) calculate it for you. Offset is natural but you can reduce it... or increase it. Your choice.

[GregGC]

Originally posted by Optical
Thanks for the speedy replies guys!
....but i have another problem......

Ok i thought everything was fine, but when i changed the input resistor (R2 = 22k) to a 22k pot my DC offset went from -30mV on mute to +200mV fully open....!!!

I havent tried putting any caps from the 680R resistor to ground as yet..
I also have an input cap placed before the variable resistor, is this the correct place to have it or is it better to have it after the pot from middle pin to the non inverting input of the chip?

The sch you use is good for OPA541 where Ibias are very small and will not be affected by changing value of the pot.

That's a bit better.
http://www.diyaudio.com/forums/showt...500#post259500

But I would do this:
http://www.diyaudio.com/forums/showt...525#post259525

This circuit will give you steady DC offset at the out, no matter the position on the pot.
If you want to experiment:
1. omit R1/C1 (HF filter, I'd ceep just in case)
2. Check different res values (keep the ratio the same) like 6.8k/220k and see what the output offset will be. It should be fine.

Mind you I did it and it sounds very good:
Another GC project

[peranders]

I have made some calculations:

R3 22k
R5 680 R
R2 22 k

=> Pot at min, 33-333 mV offset

=> Pot at max, AC coupled signal source, 170 mV to over 1 V.

The big problem is that you have a DC-connected amp (no cap between R5 and ground.

I say that your amp can't be good unless:

Insert a cap in the feedback
DC-servo
Offset trim by a pot.

I suggest that you take a good look at page 1 and typcial application in the datasheet of LM3875 (or maybe LM3386 since you used it). Check also AN-1192 from National.

EDIT: I totally agree with ashok below that you should avoid current paths, and sensitive ones through the pot. Isolate the + input with a cap or have at least a pull down resistor. In case of a bad pot you can get bad behaviour of the LM IC. The IC MUST have current to it's inputs.