Input offset voltage specs for opamps

[Max999]

So every opamp has some input offset voltage. It is the amount of voltage you need to apply to the + or - input to make the output voltage 0. It is there because of manufacturing tolerances.

I therefore assume that it accounts for some non linearity in the opamp. Perfection is not possible and how much is too much depends on the job it has to do.
The job I have for it is audio. Looking at OPA134 I see a max input offset voltage spec of 2mV at 25 degrees Celsius.

Would this lead to a significant ( 0.1% or so) percentage of distortion?
OPA134 does has input offset voltage trim pins, but if I can I would like to avoid using them. What are your thoughts/advises on this?

[diffeq]

What are your thoughts/advises on this?

In my opinion, input offset voltage only matters in precision circuitry where these few mV get amplified and and cause significant errors at the output.
OPA134 is a low distortion op amp - 0.00008%, regardless of the offset voltage.

[PRR]

> Would this lead to a significant ( 0.1% or so) percentage of distortion?

No. Not in audio.

In precision DC measurements, it is an error, which you can correct for.

[Rob Strand]

I suspect the reason you can't see the distortion is because the open loop gain of an opamp is very high and with feedback present the input voltage is tiny.   There's no easy way to drive the input stage without clipping the output.  I suppose you could use a very high frequency where the loop gain has dropped off - it will be beyond audio frequencies for sure.   

If you build you own diff amp I'm sure you could measure some even order harmonics when the diff-amp has some imbalance.

[ElectricDruid]

I therefore assume that it accounts for some non linearity in the opamp.

No. The offset is a DC bias, and doesn't affect the linearity (or otherwise) of the op-amp. That's a different parameter.

The job I have for it is audio. Looking at OPA134 I see a max input offset voltage spec of 2mV at 25 degrees Celsius.
Would this lead to a significant ( 0.1% or so) percentage of distortion?

No. For audio the input signal will likely be AC-coupled via a capacitor at some point coming in and again going out, so exactly what the DC conditions around the op-amp are doesn't really matter much. Assume we run from a 9V battery and have a 4.5V midpoint voltage. It wouldn't really matter if that midpoint was 4.2V - we'd get a touch less headroom is all. And a 0.3V difference swamps a few millivolts of input offset.

As Paul said, the place where it matters is for DC measurements. Imagine you have a weight sensor with an output of 20mV for 20kg. Now it's a big deal and needs compensating.

HTH,
Tom

[R.G.]

Just to extend the extensions, input offset is why you will sometimes see capacitors inserted between opamp stages in a string of opamps which are all biased at the nominally same reference voltage.

All opamps will have an offset voltage of some amount. the data sheets spec this referred to the input, because it gets amplified by the forward gain of the opamp, and reduced by feedback.

But in stages with DC gain, it gets amplified, and Murphy's Law being what it is, all errors will add, not cancel, and in the worst possible direction too. Caps inserted every few amps in a series string of opamps stop this and are highly recommended for strings of opamps which only need to amplify AC signals.

If you're amplifying DC, you have to deal with accumulated offsets other ways.

[Max999]

Ah I get it now! It is a dc spec. It can only affect the bias a bit in audio application.
Rg I thought opamps in audio don't amplify dc. Are you talking about some kind of dc-dc converter or charge pump?

[PRR]

> I thought opamps in audio don't amplify dc.

They generally DO *until* you do something about it.

Since DC in audio is useless, often distressing, we *design* audio systems with a low-cut. Sometimes a few stages may be DC-coupled, if you do it right.

[Phoenix]

Rg I thought opamps in audio don't amplify dc. Are you talking about some kind of dc-dc converter or charge pump?

Inverting op amps always amplify DC, that is, they will amplify any DC presented to their input, but even if they are AC coupled, they will still amplify their own DC offset - so an inverting stage with 2mV offset and a gain of 100 will have 200mV offset at the output. If you had a series string of inverting stages without AC coupling caps you can see how this could quickly become a problem.
Non-inverting op amps can be arranged to avoid DC gain by AC coupling the feedback path, but some designs will omit this cap if possible to reduce parts count and reduce BOM cost, especially if there are already AC coupling caps between stages which also address the issue. In my experience it's more common to see the feedback AC coupling cap omitted in designs with bipolar rails rather than the single-ended split rail more common to stompboxes.

[Max999]

Paul is it so that I need a high pass before an opamp or is having any size coupling capacitor sufficient for blocking DC?
Phoenix, so if I put a "coupling cap"  in a non inverting feedback loop I stop DC gain?
The biasing of the opamp is only affected by the DC voltage at the input right, not the output?

[Phoenix]

Paul is it so that I need a high pass before an opamp or is having any size coupling capacitor sufficient for blocking DC?
Phoenix, so if I put a "coupling cap"  in a non inverting feedback loop I stop DC gain?
The biasing of the opamp is only affected by the DC voltage at the input right, not the output?

"Needing a high pass before an opamp" is one way to think of it - DC is the lowest possible "frequency" - 0Hz, so any high pass filter will eliminate it. You can select any capacitor that you want, either for your lowest frequency of interest, or for tonal shaping purposes.



The cap I am talking about is Cg. This will reduce the DC gain of a non-inverting op amp to 1, so any input offset will not be increased at the out (though there are other factors that effect output offset like input bias current). Cg and Rg also set the high pass of the op amp, again, they can be chosen for lowest frequency of interest, or for tonal shaping.
Another advantage of Cg is that it can be returned to anywhere convenient and low impedance (and low enough noise), regardless of DC level - so in a typical pedal with 9V, Vref and GND, Cg could be returned to any of these because they are all "AC ground". Ri and Rl must always be returned to the correct bias voltage though, in a pedal that would normally be Vref, or in a bipolar supply it would be ground.
To answer your question about DC voltage at the output, it does not effect the bias of the op amp, but if the output is connected to a different DC level, the opamp will sink or source current to try to match the voltage it is connected to. If the series resistance is low and the voltage difference is great, then the op amp will run out of current quickly, and even if it doesn't, it will reduce output voltage swing capability, and may cause other nasty artifacts.

[Max999]

Ok but what happends when I do this. I made the coupling cap exorbitantly big, 100uF. The -3 dB point of the high pass will be effectively zero.

Does this still block DC?




And would adding c2 like so also stop the input offset multiplication?



In the case of a gain of +1, would this C2 be a smart solution to stop the input offset voltage from reaching the output?

[PRR]

> I made the coupling cap exorbitantly big, 100uF.

WHY??

A typical 100uFd cap make leak 1uA. In 1Meg this is a 1V offset, giving a small but significant change of headroom.

> The -3 dB point of the high pass will be effectively zero.

No, it will be right near 0.001,7 Hz. This may be "zero" for audio; significant for seismic and strain-gauge and other studies.

And again, WHY? Your guitar does not make 0.001,7Hz. Not really 20Hz. You could use a cap 10,000 times smaller, which may be cheaper or smaller. And less leaky.

A further issue. 100uFd and 1Meg is a 100 second time constant. When you turn-on the bias voltage will rise very slowly. Over a minute to get halfway to the goal. Probably many seconds just to get in the op-amp's working zone.

Don't frame a dog-house with 2-foot tree trunks.

[PRR]

> And would adding c2 like so also stop the input offset multiplication?
> ..., would this C2 be a smart solution to stop the input offset voltage from reaching the output?

Both are totally wrong. You have inverted Cause and Effect. Get a basic Opamp Theory book and use your brain.

[Max999]

PRR the 100uF is hypothetical to make the question about dc blockage clearer.

I will look for a good book on opamps, it is sound advice.