How to minimize noise of OTA in unity feedback?

[sharkies]

Let's say I have an opamp... actually is an OTA to be more exact
anyways..
I have a certain output noise, which is proportional to gm,
I get the input referred noise, which is proportional to 1/gm
Therefore, I can lower input referred noise by increasing gm, or burning more power.
What I don't understand is that very often OTA's are configured to have unity gain when used with feedback. In this case, output referred noise is equivalent to input referred noise. Since output noise is proportional to gm, we must lower gm to reduce the overall noise, which is the exact opposite of what I explained above. Is this right? I will be using my OTA in unity feedback, and I'm not sure how to minimize my noise because of this conflicting direction

 

[prcken]

opamp noise help

really a good question! i agree with you, i think you can add gain stages before the unit buffer to minimize input referred noise.

 

[BackerShu]

Re: opamp noise help

I get the input referred noise, which is proportional to 1/gm

You got this because the output noise is divided by the DC gain of the OTA, which is not the case in real application.
Theorically you could reduce gm to reduce noise. However, the other specifictions which are related to gm may be deteriorated, like the DC gain of the OTA which is needed to be high in high linearity application.

In a word, noise cannot be considered seperately,otherwise the circuit may fail to meet other requirement.

 

[tifox]

Let's say I have an opamp... actually is an OTA to be more exact
anyways..
I have a certain output noise, which is proportional to gm,
I get the input referred noise, which is proportional to 1/gm
Therefore, I can lower input referred noise by increasing gm, or burning more power.
What I don't understand is that very often OTA's are configured to have unity gain when used with feedback. In this case, output referred noise is equivalent to input referred noise. Since output noise is proportional to gm, we must lower gm to reduce the overall noise, which is the exact opposite of what I explained above. Is this right? I will be using my OTA in unity feedback, and I'm not sure how to minimize my noise because of this conflicting direction

In unity feedback the output impedance is 1/gm so vn² & gm * (1/gm²) = 1/gm. Maximizing the gm of the input differential pair still lower the noise.
Actually it is equivalent to the input refered noise of the OTA which is a good new : the feedback doesn't degrade Signal to noise performances.

 

[tzg6sa]

There is nothing like input and output referred noise if you short the input and the output. Let's assume that we have some noise injected after the gm stage before the feedback. Control (feedback?/signal flow? - don't know its exact name) theory tell us that this noise will be reduced by the loop gain. And actually this will be the same as the input referred noise in the open loop case. (Assuming loop gain is much higher than unity)