The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> Why noise figure doens't match
https://designers-guide.org/forum/YaBB.pl?num=1083264800

Message started by tinaxie on Apr 29^th, 2004, 11:53am

Title: Why noise figure doens't match
Post by tinaxie on Apr 29^th, 2004, 11:53am

I do Front_End noise simulation, use print noise contribution, I know that RF port noise contribution percent is 50%, that is to say Noise figure equals to 10*log10(1/0.5)=3dB, however ploted noise figure is 5dB, much higher. What's the problem.

When I do LNA noise simulation, ploted noise figure match very well with calculated one according to noise summary.

Title: Re: Why noise figure doens't match
Post by Eugene on Apr 30^th, 2004, 7:46am

What simulator are you using? Spectre? SpectreRF? Are there any mixers in your front end or are you only looking at the LNA? If there are mixers, are they converting to IF or baseband? And how do you know the port accounts for 50% of the noise?

Title: Re: Why noise figure doens't match
Post by tinaxie on Apr 30^th, 2004, 9:34am

thanks very much.
I use SpectreRF, The front_end includes LNA and mixer, mixer convert RF signal to high IF. I use print noise contribution to see that RF port contribute 50% of total noise.

Title: Re: Why noise figure doens't match
Post by Ken Kundert on Apr 30^th, 2004, 11:29am

How much of the noise at the output is contributed by the load? It is ignored in the NF calculation but not when calculating the percentages in the output nosie.

-Ken

Title: Re: Why noise figure doens't match
Post by tinaxie on Apr 30^th, 2004, 2:15pm

Only 0.44% noise is contributed by loading.

Title: Re: Why noise figure doens't match
Post by Ken Kundert on May 1^st, 2004, 8:15am

Are you performing the computation using voltages or powers? Things will not add up right if you use voltages (V/rt(Hz)); you must use powers (V²/Hz).

-Ken

Title: Re: Why noise figure doens't match
Post by Eugene on May 4^th, 2004, 9:05am

If the question is still unresolved, you might try replacing your circuit models with behavioral models (i.e. VerilogA). Since the behavioral models simulate only what you include, you can surgically remove noise sources to isolate the offending source. The SpectreRF user's manual describes several behavioral models in detail that might serve as a starting point. That chapter also discusses test circuits, including noise test circuits, that might shed some light on the issue.