DC Offset Question

[dsavitsk]

Sorry for another dc offset question. Search either gives too much or too little.

I built the BrianGT Rev 2 amp. with no source connected, I get 68mV and 102mV offset. With a source connected, it drops to 18mV and 4mV. Is this normal, or is there something I should look into?

Thanks,

-d

[peranders]

This is normal. You have a current flowing into the inputs. Take this current and let it go through the input resistor. This will create a voltage.

Datasheet says 0.2-1 uA. Take this current through 22 kohms => 22 mV times your gain which is 30 => 660 mV!

You seems to have around the lower limit in the current.

[dsavitsk]

The connected source was off, and unplugged. I don't really understand your answer, so I am not sure if this matters.

Anyway, with a different source (my preamp this time), I am now back up to about 70mV. Is this a reasonable amount to run into speakers (B&W)?

-d

[carlosfm]

Originally posted by dsavitsk
Anyway, with a different source (my preamp this time), I am now back up to about 70mV.

With a source connected you will have the 22k input resistor to ground paralleled with the source's output resistor to ground.
So, depending on the source you connect to the amp you will have different DC-offset.
With an input coupling cap that wouldn't happen.
Your high(ish) DC-offset is also because the datasheet-recommended (and copied) resistor values are not a very wise choice.

Originally posted by dsavitsk
Is this a reasonable amount to run into speakers (B&W)?

A little high, but OK.

[dsavitsk]

Originally posted by carlosfm
So, depending on the source you connect to the amp you will have different DC-offset.

The preamp (a modified version* of Nuuk's discrete buffer) has an cap (4.7uF) on the output and the pre shows no offset.

Originally posted by carlosfm
With an input coupling cap that wouldn't happen.
Your high(ish) DC-offset is also because the datasheet-recommended (and copied) resistor values are not a very wise choice.



A little high, but OK.

Where, and of what value, would an input coupling cap go -- just on the input bofore the boards?

Additionally, can you point me to a thread or a link with more appropriate resistor values?

* the modification was to add a 1uF cap and a 1K resistor on the input (after the pot). It seems that there is DC offset on the input, and when the volume is all the way down is shorts the circuit. These mods fixed it, however.

Thanks,

-d

[peranders]

Originally posted by dsavitsk
The connected source was off, and unplugged. I don't really understand your answer, so I am not sure if this matters.

Anyway, with a different source (my preamp this time), I am now back up to about 70mV. Is this a reasonable amount to run into speakers (B&W)?

Your problem is "input bias current" mainly. If you want to have an independent offset you must have a coupling capacitor or a DC servo. People thinks that a DC servo is pretty unneeded, it _is_ statisticly but some people will experience trouble with their LM3875 LM3886 designs because those IC's are not precision devices.

Unfortunately I haven't got the time to look up for material, nor make it myself but Carlos, Peter D, Brian etc are all experts in LM3875, LM3886 and offset trimming techniques and how you in theory should optimize the design. It's also pretty easy.

Besides trimming and inject current you should always strive to have the same DC resistance from both inputs, which is not the case when the amp is DC connected and you will use different signal sources.

Remember though, is this really a problem? When you have the unconnected, then 100 mV and a silent amp can't be a problem?

[jackinnj]

Originally posted by peranders
This is normal. You have a current flowing into the inputs. Take this current and let it go through the input resistor. This will create a voltage.

Datasheet says 0.2-1 uA. Take this current through 22 kohms => 22 mV times your gain which is 30 => 660 mV!

You seems to have around the lower limit in the current.

660mV -- consider this -- 1 watt into a nominal 8 ohms requires 2.63 VRMS -- most of our listening is going to be at low levels...

hmmmm, I think you mean to say that you have a pair of bias currents flowing FROM the opamp (LM3XXX) inputs through the source resistor causing a POTENTIAL between ground and the Opamp. The difference between the POTENTIALs on the + and - inputs is multiplied by the gain of the amplifier resulting in a DC OFFSET voltage expressed at the output.

precisely matching the resistors is a first step toward reducing DC OFFSET.

[dsavitsk]

Originally posted by peranders
Remember though, is this really a problem? When you have the unconnected, then 100 mV and a silent amp can't be a problem?

Well, even connected it is still 80mV. Really, the problem is that the amp hums, quietly at first, but the longer it is on, the louder it hums, and within 30-40 minutes is really unacceptable. Perhaps this is not DC offset causing the problem, but then what is it?

Originally posted by jackinnj
precisely matching the resistors is a first step toward reducing DC OFFSET.

Resistors were precisely matched (they are mostly PRPs with a Kiwame for the feedback.) What do I look at next?

[jackinnj]

Originally posted by dsavitsk


Well, even connected it is still 80mV. Really, the problem is that the amp hums, quietly at first, but the longer it is on, the louder it hums, and within 30-40 minutes is really unacceptable. Perhaps this is not DC offset causing the problem, but then what is it?

Could be heat --

[janneman]

Your hum points to a mechanically humming transformer. Could that be the cause? mech noise from the transformer rather than hum from the speakers?

Jan Didden