Noise affect after zeroing the offset voltage

[Junus2012]

Dear friends,

If I am zeroing the input offset voltage by any kind of technique, let say I make it zero as an ideal case.

I am thinking that input offset voltage is combination of multisource like CMRR, PSRR, and noise. So if I made the offset voltage zero this means I made the noise zero
right ?

By this way, the dynmaic techniques like chopper of auto-zero are reducing the offset voltage, and the noise consequently

Thank you in advance

 

[KlausST]

Hi,

You talk about offset, but offset "of what"? ADC, OPAMP, DAC, COMPARATOR....

So if I made the offset voltage zero this means I made the noise zero

No. Offset and noise are completely different things, they don't relate to each other.

Klaus

 

[FvM]

The question is a bit vague. Static offset cancellation, e.g. by trimming, does not affect noise directly. Dynamic auto-zero can however cut low frequency 1/f noise down to white noise level.

 

[Junus2012]

Thank you for your reply

I am talking about the offset of the operational amplifier,

for sure noise and offset voltage are different things, but both of them are voltage signals.

If I could make ideally the input offset voltage is equal to zero, then can you please tell me where the noise voltage is ?

let us only consider the low frequency or DC condition.

Thanks

 

[barry]

As previously stated, noise and offset are two different things. If it helps you understand it, think of offset as a DC voltage and noise as an AC voltage.

 

[KlausST]

Hi,

Noise is not (only) low frequency or DC.

Klaus

 

[dick_freebird]

There will be noise embedded in the correction term
as with anything else. But you're talking microvolts
to get rid of millivolts. That's often a trade worth
taking despite the complexity.

Chopping will pretty much get rid of RTN, and it can
"fold" low frequency noise over to just above/below
the chop clock, to be post-filtered away.

 

[sutapanaki]

Offset itself is a DC quantity. Well, it drifts over time, but let's assume it is DC. Noise on the other hand is not DC. You have 1/f noise and thermal noise. You can compensate for low-frequency noise, but not completely. If you do chopping with a chopping frequency higher than the 1/f corner, then you can bring the noise level in-band to about the thermal noise floor with just a little bit of overhead.
If you do auto-zeroing or CDS, then the cancellation depend on the loop-gain of your auto-zero loop, the speed of auto-zeroing and since it is more akin to sampling, you will get some high-frequency noise folding.

 

[KlausST]

Hi,

You can compensate for low-frequency noise,

One need to specify what "low frequency" means.
For audio the range of 20Hz ... 200Hz may be considered as "low frequency".
In this range I doubt a low_noise_chopper_stabilized Opamp can have a noise improvement against a low_noise_non_chopper_stabilized Opamp.

Your statement maybe becomes true for frequencies (much) below 1Hz.

Two points:
* noise is random, it does not depend on aging, temperature and supply voltage the same way like the drift of offset voltage. ... thus it can't be canceled out the same way.
* when I see the noise chart of low noise chopper stabilized Opamps (like AD8628) it shows much more noise than non_chopper_stabilized Opamps, even at low frequencies. The values say 20nV/sqrt(Hz) whereas the others go below 1nV/sqrt(Hz) (both at 1kHz).

Klaus

 

[Junus2012]

Thank you friends for your nice explanation,

"The offset of the operational amplifier can be viewed as a noise having very low frequency. We, therefore, expect that periodic offset cancellation can potentially reduce the low-frequency noise (1/f) as well"...Behzad Rezavi

So I would remark two things based also on your answers
1. Noise is an ac having frequency from DC to high frequency while offset voltage is only DC component
2. If we assume that reducing offset voltage would reduce the noise, this only would be true if we are talking about low frequency noice (mean near to DC ) and secondly, the offset correction should be performed periodically not as static trimming appraoch

 

[sutapanaki]

Hi,

One need to specify what "low frequency" means.
For audio the range of 20Hz ... 200Hz may be considered as "low frequency".
In this range I doubt a low_noise_chopper_stabilized Opamp can have a noise improvement against a low_noise_non_chopper_stabilized Opamp.

Your statement maybe becomes true for frequencies (much) below 1Hz.

I think I elaborated on this point in my previous reply but maybe I didn't do it clearly. Low frequency noise in the case of chopping is relative to the chopping frequency. If chipping frequency is higher than flicker noise corner frequency and the LPF BW is below the corner, then after chopping one's left with the thermal noise floor with a little bit of overhead because of thermal and flicker noise being downconverted from noise around harmonics of the chopping frequency.

In the case of autozeroing is a similar story in the sense that low frequency noise, this time relative to the autozero i.e. sampling frequency, is attenuated, while high frequency noise is transferred to the output unchanged. Note, noise is attenuated, not remover completely.

 

[KlausST]

Hi,

I agree,
but still have a problem with the vague description "low frequency".
Maybe call it "philosophical" but I see it more mathematical.

With a 4kHz chopping frequency... 20Hz to 200Hz is way below the frequency where the noise becomes mirrored (2kHz).
It is at least one decade away. But even at two decades (factor of 100) below chopping frequency there is no benefit in noise against a non chopping Opamp.

The noise benefit begins maybe three (3) decades below chopping frequency.

****
Junus said: "near DC".

Is 1Hz more "near" to DC or more near to 4kHz?
Some may clearly say: 1Hz is more near DC.
But when talking about frequency one uses decades or octaves.
How many decades is 1Hz to 4kHz? About 3.5 decades
And how many decades from DC to 1Hz? Infinite decades.
Now it seems 1Hz is infinitely far away from DC but close to 4kHz.

With noise one uses 1/Hz or 1/sqrt(Hz)....0Hz is hard to grasp. ... and impossible to calculate.

This is why I recommend to avoid words like "near DC" and "low frequency" ...and instead use values like "10mHz"....or "about 10mHz".
One can grasp them and one can calculate with them.

Klaus

 

[sutapanaki]

I don't agree with that.. The goal of chopping is not only to eliminate DC offset but also to remove flicker noise. So, from system point of view we should see less noise after chopping in the frequency band where flicker noise was dominant noise source. If you say the amplifier ends up with more noise after chopping than what it would have without chopping then there is something wrong. But then, if the amplifier is an ADI part there is no chance of something wrong happening with it.

In what I wrote before, I really look at F/Fchop. Or if you prefer F/Ffilter where Ffilter is the BW of the LPF after the chopper. This ratio being <1 delimits low and high frequencies in my view.

 

[FvM]

Some real OP data to illustrate the effect of chopper topology on 1/f noise.

A typical CMOS noise density curve (AD8651)


CMOS chopper OP noise density (ADA4522)

 

[sutapanaki]

Yes, exactly. After chopping 1/f coloring of the noise is removed and we are left with slightly increased thermal noise floor.

 

[KlausST]

Hi,

I was referring to: Analog Devices Tutorial MT-055.

https://www.analog.com/media/en/training-seminars/tutorials/MT-055.pdf

On page 6 they compare OP177 with AD8628.
And the noise density of OP177 isn't really the top end of low noise.

Klaus

 

[sutapanaki]

It does remove the flicker noise, though. Bipolar amps have by definition much lower 1/f noise compared to MOS amplifiers. I don't know what technology they used for AD8628. It says the amplifier has a combination of chopping and autozeroing. Autozeroing is known to not cancel noise for higher frequencies - at and above the autozero sampling frequency.

 

[Junus2012]

Dear friends,

Thank you for your contribution,

here is another source stating that " Two commonly used types—autozero amplifiers and choppers—achieve nanovolt-level offsets and extremely low offset drifts due to time and temperature. The amplifier’s 1/f noise is also seen as a dc error, so it is removed as well"

Source:

https://www.analog.com/media/en/analog-dialogue/volume-44/number-1/articles/zero-drift-operational-amps.pdf

in this source he also described the effect of the auto-zero or chopping on the noise performance of the circuit