Increasing linearity of the op-amp

[Junus2012]

Hello,

one of the advantages of the closed-loop negative feedback is to improve the amplifier linearity. Therefore, most of the designers I see when they design the op-amp working in closed loop feedback they usually don't care or mention about the op-linearity.

However, I am wondering, and assuming that the negative feedback can not be ideal, hence one at least pay little attention to not completely rely on the feedback performance, for example, fast solution is to design the differential pair transistors with higher overdrive voltage.

Thank you very much
I am looking forward to hear your thaughts

Regards

 

[danadakk]

Why do you consider the feedback non linear, its normally passive components
used ? Or are you referring to the non linearities inherent in the OpAmp architecture
and their effects where negative feedback does not fix a specific non linearity, like
the slew rate problem in OpAmps where the control loop temporarily "opens" and
the feedback loop is effectively broken ?

Here is an interesting discussion that may be useful -

https://linearaudio.nl/sites/linearaudio.net/files/Are%20Op%20Amps%20Really%20Linear.pdf

Page 5 of this another recent problem when RRIO parts introduced -

**broken link removed**


Regards, Dana.

 

[KlausST]

Hi,

Even feedbacked linearity is not ideal.....maybe no electronics circuit is ideal.

I know we are in the "IC design" section, but you may have a look at Opamp datasheets.
They usually show the linearity (distorton) at various configurations.
It usually is < 0.x % ... now it depends on your requirement whether this is good enough or not.

I can't remember the exact formula but isn't it something like:
Distortion_feedbacked = distortion_nonfeedbacked x closed_loop_gain / open_loop_gain.

Klaus

 

[Dominik Przyborowski]

Linearity of passive elements is one thing and usually used poly resistors are good (they resistance doesn't change too much with voltage drop).
They have obviously capacitance to substrate, so for high frequency one would start to consider effect of impedance difference between elements in feedback network, and compensates it by use of time constant ratio instead of R ratio (but this seems to be extreme case).

The simplest way is boosting open loop gain - diff pair transfer function is ~tanh(Vin/2nVt), so for 1V signal amplitude and 60dB gain, Vin is ≈1mV, so one can see distortion caused by diff-pair. For very high linearity, one would eliminate effects of V_DS on transconductor and current sources - so cascoding everything in opamp.
Non-linear nature of devices result with some "sweet-spots" in operating points. However, it has been shown, that biasing conditions providing such sweet-spot for single mosfet CS amplifier, doesn't applies to diff-pair (which has not such sweet spots). So, for multi-transistor design finding it might be impossible and such design would be very PVT sensitive.
It was a few papers for the last two years about THD for MOSFET amplifiers and how to find proper biasing point. I think one was published by C. Enz in 2019.

So, if you need THD<-N dB (N whatever spec), you need to go for opamp gain. If you need some extreme numbers, feedback element might be needed to check. However, IMO there is no realistic linearity spec to look on passives linearity in feedback.

 

[Junus2012]

Dear friends,

Thank you very much for your contribution,

You made the non-linearity of the feedback elements very clear to me.

Coming to the op-amp part, as Dominik mentioned, this requires high gain, in my opinion, I prefer to say high loop gain.

The other quantity I more interested in is the overdrive voltage of the differential pair of the amplifier. In gm-C filter, the amplifier integrates the signal in an open loop, no negative feedback mechanism, the differential input signal will then appear as it is between the op-amp inputs, so in order to increase the differential range, one try to maximize the overdrive voltage of the differential pair by using longer transistors, high biasing current, using degenerative resistors, and or using different linearization techniques. Indeed I have some work in this direction.

But if we talk about closed-loop op-amp amplfier, and assume for now ideal feedback element (to only blame the opamp), and also assume high gain amplifier, now with the negative feedback principle, vin+=Vin-, hence we don't need to maximize the differential input overdrive voltage (unless for different purposes). The only difference that appears at the inputs is as Mrs. Dana said, at the time of slewing, however, designing op-amp with sufficient slew rate will avoid this distortion.

I found this paper where the authors increased the overdrive voltage of the input stage toward increasing the linearity of the op-amp used to construct Sallen key filter (stated in the second part of the introduction)

Request Rejected

and here I cut this part for your further discussion please