The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> Noise Figure Definition
https://designers-guide.org/forum/YaBB.pl?num=1179205882

Message started by aaron_do on May 14^th, 2007, 10:11pm

Title: Noise Figure Definition
Post by aaron_do on May 14^th, 2007, 10:11pm

Hi all,

for the definition of noise figure, we assume the input noise power is kTB. However, if the input resistance is not matched, am i right in saying that the input noise power will be

kTB(1-S11^2)

and the input noise voltage will be

kTB.RS.(1+S11)^2

?

thanks,
Aaron

Title: Re: Noise Figure Definition
Post by mg777 on May 15^th, 2007, 12:16am

Your reasoning is essentially correct - the signal and noise could be shaped by different functions. It just so happens that the input thermal noise and the output (excess or channel) noise are often shaped by the same frequency function. If your input match is bandpass, then both signal and noise will be bandpass and the |H(jω)|² will cancel. A similar argument should hold for channel (excess noise), but if you have multistage then your reasoning will kick in.

At any rate, I consider noise figure and input referred noise to be historical baggage. I can understand the import of the Friis Law without invoking the concept of noise figure. My life is simpler without these fake parameters which are susceptible to specmanship games. Where I am forced to use them I do so under protest.

M.G.Rajan
www.eecalc.com

Title: Re: Noise Figure Definition
Post by Eugene on May 15^th, 2007, 7:57am

mg777,

Your response intrigues me. How would you like to see noise specified, measured, and simulated?

Title: Re: Noise Figure Definition
Post by mg777 on May 15^th, 2007, 10:57pm

I prefer noise to be presented in its bare physical form i.e; a set of samples. Depending on the systems context I know what to make of this raw data, including assumptions about its statistics. What I dislike is the artificial partitioning of systems and the use of archaic catch all contrivances & 'figures of merit'.

For example, I do not like to use an instrument that calls itself 'noise figure meter'. I prefer a more versatile measurement and signal processing platform where I can determine what I want to measure. Over the past decade, T&M equipment has been evolving precisely in this direction.

So, my fellow Amer...er, designers, it IS the software.

M.G.Rajan
www.eecalc.com

Title: Re: Noise Figure Definition
Post by aaron_do on May 15^th, 2007, 11:35pm

Hi M.G. Rajan,

your responses are always really insightful.

Anyway, are you saying that, for example, rather than have a noise figure meter, you'd want a powerful oscilloscope/spectrum analyzer...i.e. versatility over simplicity? That is understandable if you are someone who really knows how to interpret raw data. I think though that many engineers are simply given a set of specs that they need to meet without knowledge of the system.

Also systems are partioned in the way we design them and different circuits in the same system may need completely different inputs and specs. For example and LNA may need the desired FSK signal as well as a bunch of unwanted signals of specific frequencies and amplitudes as its input, while the input to a VGA might be much cleaner. FOMs like NF, Phase Noise, IIP3 and image rejection ratio allow us to separate the general problem of unwanted random signals into a base set of problems which can be individually tackled.

cheers,
Aaron

Title: Re: Noise Figure Definition
Post by mg777 on May 16^th, 2007, 4:47am

Aaron,

You have raised valid points, so I won't insist on my preferences.

A T&M instrument can at once be simple and versatile. For example, the data acquisition can be kept 'simple' while the signal processing of the samples can be made 'versatile'. The acquisition is analog (physics) and the signal processing is digital (math). The math has Moore's Law behind it, so it's a good strategy to keep the physics simple.

As for partitioning a design - it's a sensible & conservative approach, no doubt. However, it should not divert us from synergies between the math and the physics. This is a powerful reason to adopt a 'top down' design approach, even though today we shake hands with our digital counterparts only across the miry trenches of GDS-II.

M.G.Rajan

Title: Re: Noise Figure Definition
Post by RFICDUDE on May 16^th, 2007, 9:03pm

Oh yee of little Johnson-Nyquist noise faith.

Noise is not samples nor can you find a sampler with enough resolution to measure thermal noise. As digital as we may be you still must have a LNA and automatic gain controls to servo a signal to within the maximum resolution of a data converter.

Altough, constructs such as NF may seem contrived, they are much more physical than samples and it is the only way to deal with maximizing SNR in a system. SNR is what we care about and NF is inherently defined by SNR.

Therefore I am content with my faith in the measurement.

My question to you would be ...
How would you design a cascaded system which maximizes SNR given additive noise constraints for every electronic and resistive component in the system?

On the flip side, we do live in the digital world too. T&M equipment is much more versitile because of it and it is improving all the time. We are spoiled considering the theoretical experiemental validation of all communications theory was originally validated in analog instrumentation.