Re: How to explain how negative feedback lowers noise?


Tom Lee
 

I partially agree with Goran's statement, but only partially because, as written, it neglects an important subtlety.

It is certainly true that a large first-stage gain suppresses the noise contributions of subsequent stages, so that's all good and intuitively satisfying.

However, once you've succeeded at that endeavor, the first stage's noise figure dominates. Somewhat counterintuitively, matching impedance at the input to maximize gain does not necessarily minimize noise figure. So, it is possible (even probable) to degrade NF by focussing only on maximizing gain.

The reason, in a nutshell, is that the minimum noise figure for an amplifier occurs for a source impedance that is the ratio of the equivalent input noise voltage to the noise current (I'm neglecting possible correlations between the two to make the argument simple). There's no fundamental connection between that ratio and the actual input impedance. You match to the former for best NF, and to the latter for maximum gain. So maximizing the gain of the first stage is not guaranteed to lead to overall best NF.

--Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 3/25/2021 16:27, Ed Breya via groups.io wrote:
Goran wrote: "If you wish to design a multistage amplifier with low phase noise factor you should maximize the first amplifier gain and the noise factor contribution from the following stages will be small.

Does anyone want to comment on this?"

My comment is that I agree. The front-end of any system is the most critical, and especially when the nature of the signal needs to be changed into an electric circuit signal. This happens all the time, whether it's receiving radio waves, light, sound, physical quantities, and so on. Electric to electric is the most common and straightforward - it's what goes on inside circuits. Anything else to electric has more issues, due to needing a transducer of some sort.

Ed



Join TekScopes@groups.io to automatically receive all group messages.