Continuous Op amp offset cancellation

[branred]

Dear all,

I am designing an op amp and offset is an issue to me. My op amp has to operate continuously thus any clocking method of offset cancellation is not a good choice. eg auto-zeroing. Is there any other method that does not require clocking? If possible attach the link.

Thank you so much in advance.

 

[eklikeroomys]

DC blocking capacitor?

 

[erikl]

... any clocking method of offset cancellation is not a good choice. eg auto-zeroing.

What's your max. offset requirement?
Could you possibly live with a discontinous offset cancellation scheme, say once a second (or minute)?

 

[threekingtiger]

Maybe fully differential OPAMP and optimizing the layout to reduce the mismatch are preferred. Choose larger devices, decrease the overdrive but make sure devices DO NOT enter subthreshold region and sufficient margin should be reserved.

However, the drift of the Vos is hard to avoid. The lower the Vos, the smaller the TC of Vos.

 

[branred]

erikl,

it's in the range of few milivolts. 10mV or so.
I guess discontinuity is not allowed because this block must be always on. when the system is in idle, the oscillator/clock is switched off but the voltage regulator must remained on.

thank you so much

 

[LvW]

Dear all,
I am designing an op amp and offset is an issue to me. My op amp has to operate continuously thus any clocking method of offset cancellation is not a good choice. eg auto-zeroing. Is there any other method that does not require clocking? If possible attach the link.
Thank you so much in advance.

At first, offset "cancellation" is not possible. The only goal you can achieve is offset reduction.
Secondly, I suppose your concern is not "offset" but "consequences of input offset" as seen at the output, correct?
In this case, a simple method is to provide low dc gain (preferrable: unity gain) and at the same time a signal gain as required using an appropriate frequency dependent feedback network.

 

[erikl]

erikl,

it's in the range of few milivolts. 10mV or so.

branred,

< 10mV offset should be easily attainable with a big differential input stage. How "big" depends on your process.

 

[checkmate]

You can provide a trimming circuit. Either factory trimmed, or on-the-fly self trimming upon power up.

 

[dick_freebird]

Can you tolerate a lesser regulation accuracy in
"idle" mode? One would think so. Perhaps chopping
can be modal.

Chopping wants a post-filter with f/10 or lower BW,
depending on how much residual noise you can stand.
A swap-chop front end can take you from >10mV to
<1mV provided the clock is nicely symmetric, etc.
but I see some residual square wave on the amplifier
output if I look close enough. I'm OK with it because
there's a much bigger filter downstream.

erikl,

it's in the range of few milivolts. 10mV or so.
I guess discontinuity is not allowed because this block must be always on. when the system is in idle, the oscillator/clock is switched off but the voltage regulator must remained on.

thank you so much

 

[leo_o2]

I agree with erikl.
I know Texas Instrument uses very big size transistor (eg. w=5000um in total) to achieve low offset.

erikl,

it's in the range of few milivolts. 10mV or so.
I guess discontinuity is not allowed because this block must be always on. when the system is in idle, the oscillator/clock is switched off but the voltage regulator must remained on.

thank you so much