Minimizing dc offset in simple amplifiers

[Tekko]

Most of the time really simple amplifiers can have a huge DC offset of up to 0.1 volts or even higher depending on brand and age which is not very good at all, but with some fine tweaking, dc offset on these amplifiers can be brought down to picovolt levels. Observe that the LTP transistors should be tightly coupled together face to face, preferably with a screw based clamp, but a tight zip tie works ok, with some thermal grease to minimize offset drift due to unequal temperature.

Here is such an example:
http://i.imgur.com/8zimX.png

Of cource in reality this is gonna drift wildly over time, possibly several millivolts up and down due to air currents over components and VLF signals picked up at the input.

However minimum possible dc offset is something that should ALWAYS be strived for as it offers the best possible sound quality with minimum assymetrical distortion.

[AndrewT]

Agreed. Except for the picoVolt exaggeration.

[Tekko]

I have a commercial amplifier that has so low dc offset that it doesent even register on a digital multimeter.

That is what i use as my reference.

Here DC offset is minimized on a blameless design:
http://i.imgur.com/eqe2u.png

[AndrewT]

I am surprised that you get consistent <0.05mVdc output offset with that "blameless".
But well done in however you have achieved that.
Can I ask why you have ~1V across R16 and R21, since both are 47r ( ~ 20mA bias)?

[AKSA]

Tekko,

DC offset up to about 150mV has no effect on sound quality except to displace the speaker cone slightly to one side, very slighly increasing H2 and H4. You wouldn't notice it.

Clamping LTPs together is not necessary to achieve quite adequate offset control.

Other non-linearities elsewhere in the circuit, particularly at the voltage amplifier, are far more significant than any concerns about picovolt offset control.

The big one is distortion profile.

I would not normally comment, but you seem pretty adamant about this....

Hugh

[Samuel Groner]

Getting low offset is trivial--just add one or the other sort of trimmer (countless options here). Getting low drift is the true difficulty.

Samuel

[Tekko]

In my eyes, anything above 30 millivolts offset is wrong and possibly a sign of a badly built amplifier.

The lower the offset, the better.

[AndrewT]

If you can achieve less than +-15mVdc of output offset (total range of 30mV) for the full range of operating conditions you will be doing well.
Neither |pV| nor |nV| from a simulation gives little clue to how all the device parameters will interact and affect output offset, as voltage and temperature changes.

[sawreyrw]

There is a fundamental problem simulating circuits like this. This is that all instances of a particular type of device are identical. This means that if the are operating at the same currents and Vce, the base - emitter voltages will be idsentical, and thus the circuit appears to has zero offsett. The solution to this is to use similar transistor but intelligently change some of the key parameters, like Bf and Is. In LTspice you can use the AKO (a kind of) feature.

[Tekko]

Never heard of that feature.