DC coupling amplifier stages

You can reduce the offset problem with a large amount of feedback, the trick is to shunt out the AC component in the feedback path so you get a decent gain.

Split the nfb resistor into 2, the first resistor is just large enough to not load the output down as its AC shorted to ground by a capacitor, the C/R junction goes also to the second half of the nfb resistor - the series total of which you have calculated to give your reduction in offset.

Is the only issue with DC coupling op amp stages the offset error? If so, in general, is this a big problem with modern op amps?

How well does offset nulling work?

Are op amps without nulling capabilities meant for AC coupling?

Is the issue mainly a problem with high gain? Obviously high gain will increase the DC offset but is it something that should be worried about for low gain too?

Does the CMRR reduces the DC offset significantly? Higher CMRR's will mean less DC offset? Can I expect by using simple circuit design(stuff found in basic textbooks) that the CMRR will reduce the DC offset by it's rated value? -120dB CMRR will reduce any DC offset by about 120db?

The real issue with DC offsets has to do with the finite voltage swing capabilities? If we amplify the DC offset too much then the op amp cannot amplify the AC signals that may be outside than the rails? So DC offset problems simply reduce the headroom?

Thanks

You can always adjust out the initial DC offset. Then, you make sure it doesn't drift with temperature very much. That drift is what I usually have to watch out for.

boB

That depends on how much gain you want, but mostly.

Which ones? There are lots of op amps available, for lots of reasons. Some are made to be small, some are made to be cheap, some are made to be fast, some are made to use vanishingly small amounts of power, etc.

In general, you need to check and do your math.

It's better to design your circuit so you don't need it. In general, leaving tweaks in your circuit costs more than buying more expensive components (or changing your system architecture) to render the tweaks unnecessary.

No.

You should always pay attention to offset voltages, at least long enough to verify that they're not a problem. Offset voltages can be a big problem if they're bigger than your allowable error, even if you're just building a voltage follower.

Yes and no. IIRC there are things you can do when designing an op-amp to reduce both offset and CMRR. But you can't look at the CMRR on a data sheet and divide the data sheet offset by that amount -- they're both measured with the op-amp working in closed loop in similar circuits.

CMRR reduces DC offset not at all.

That depends on what you're trying to do. If you're building a circuit that could be AC coupled but is entirely (or largely) DC coupled, then yes, about the only thing that the DC offset is going to do is cause clipping or other distortion. In circuits that really need to preserve DC throughout, then the DC offset is going to cause problems throughout.

Engineering is all tradeoffs. If you're building an audio circuit you don't care about DC offset at all in the end. You do care about low-frequency response, and you may well care about avoiding the use of huge and expensive capacitors, so a DC-coupled circuit can be very attractive. But to make it work you'd need to use very low offset amps at the front end (and make sure that whatever your input is it goes through very low leakage capacitors). Low offset amps may not sound good (I wonder what spurious signals you'd get from a chopper-stabilized amp?) and ones that do sound good may add more to the cost than those hypothetical big capacitors.

So there is never just one 'right' or 'best' answer.

Nice response Tim, I was going to say that if the offset times the gain is a volt or more then... watch out. And I do like big caps. Panasonic make these 10uF film caps for ~$2 ea.

George H.

CMRR is generally a function of the input stage matching, the current source driving the long tail pair, and the output impedance of the long tail pair. Basically if some parametric in the circuit changes as the input stage is yanked around, it effects CMRR.

Circuit tricks to improve CMRR generally lead to reduced common mode range, i.e. cascodes use up wiggle room.

Unity gain buffers generally have higher distortion than a gain of minus one since the unity gain buffer exercises the CMRR of the op amp.

It's better to design your circuit so you don't need it. In general, leaving tweaks in your circuit costs more than buying more expensive components (or changing your system architecture) to render the tweaks unnecessary.

-----

Well, this is the problem. I'm designing some AC coupled stuff for audio but would like to forgo the coupling capacitors for various reasons. I'm worried about the dc offset issue as it could severely limit headroom. Since I've don't have any experience with dc offsets I'm not sure how big of an issue it will be. A few mV isn't a big deal at all but several hundred will be. The amplifier stages are rather low gain(< 50 easy). If a 10mV dc offset exists then that can become 500mV quite quickly.

What I'm more worried about is some type of integration of the DC offset that might creep in, in some way which obviously can magnify the problem significantly. My 10mV dc offset maybe extreme though. In some of the op amp's I've looked at the input offset is typical 1mV but can reach 10mV. With any type of integration problems I could be seeing several volts rather quickly.

I guess it's just something I'm going to have to get down and dirty with and work out.

but

ied

her

and

It would help to view circuitry rather than speak in generalities. For instance, your comments on integration make little sense. If you are truly integrating DC, then any offset will blow up eventually. Without feedback, integrators blow up. If you have integrators in a feedback loop, then the output of the op amp in the loop is just the offset. That is, an integrator in a loop tries to drive its input to zero, where zero is really the offset of the op amp.

Many audio designs use a DC servo technique to reduce offset by creating a filter that is sub-audible. Beware of start up "thumps."

The first step to designing anything is to see if it has been done before. That doesn't mean you copy whatever circuit you find, but you at least look at old solutions and see if they are acceptable, if only as a starting point. What makes you think your design problem is out of the realm of existing circuits?

o but

rried

ve

sue

e.

ather

h and

The OP doesnt say why a half volt output offset would be a problem. Speaker offset, or signal circuit on low v psu? The 2 have different potential solutions. OP also doesnt say why wanting to avoid caps - again solutions differ depending on whether one little coupling cap is ok or not. Show a circuit.

NT

but

ied

her

and

Look at many solid state audio power amplifiers. The feedback network is unity gain at DC but much higher for AC. However this requires a modest size capacitor in the feedback. Is that acceptable? Another common solution is a servo to take out the DC but that also requires a capacitor. These caps are not directly in the signal path but do exist. You could try a digital servo that digitizes the output and generates a correction summed into the input to null out any offsets.

G=B2

An approach that some designers have used, is a servoing system... basically, an automatic integrator / negative-feedback loop which actively nulls out the offset (either around the entire system, or at each stage which requires it).

Basically: take the output from a given amplifier stage, sample and integrate it, and feed the integrated value back to the amplifier's inverting input. You can use a slow/precise/low-offset-voltage op amp for the integrator, and can even stick a passive low-pass filter prior to the integrator to keep most of the desired-signal content from being "seen" and integrated.

I believe that this approach probably works best if the amplifier stages are designed with this sort of servo nulling in mind... simply feeding an op amp's existing offset-nulling pin may or may not work well, depending on the specifics of the op amp's nulling circuitry.

A bit like what I said but with a bit more elaboration and a few posh words.

Eh?