The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> Noise Figure Vs. total input referred noise voltag
https://designers-guide.org/forum/YaBB.pl?num=1126731358

Message started by rfzingle on Sep 14^th, 2005, 1:55pm

Title: Noise Figure Vs. total input referred noise voltag
Post by rfzingle on Sep 14^th, 2005, 1:55pm

For converting the Noise Figure to total input referred noise voltage I use the following equation..

F = 1 + Vneq^2/(4kTR.delF)...
delF being the noise integration BW and R being the source resistance. Hence Vneq is as follows

Vneq^2 = (F-1)*4kTR. delF

Now if I compare this answer to the simulation result reported by the PSS -Pnoise simulation first with port as the source and later on the vsource as the source , all three results are different. Anyone has a clue as to how the total input referred noise voltage is calculated in spectre Rf . And finally to what point is it referred to if I use the port Vs Vsource and why does it not match my calculation above...

Thanks

Title: Re: Noise Figure Vs. total input referred noise vo
Post by Jess Chen on Sep 15^th, 2005, 11:39pm

If your system has a mixer and you are simulating input referred noise, the mixer can introduce noise folding that SpectreRF will capture but your manual calculation will not. However, if you are only analyzing the LNA by itself, you should not have to worry about noise folding.

As for the probe/vsource issue, I suspect the probe referred noise is 6dB higher than the vsource referred noise. The probe has an input impedance which forms a voltage divider with the LNA input impedance. The internal source inside the probe must be twice as high as the voltage expected across the matched load. The vsource does not have to gain up the noise because it has no source impedance.

If you are only analyzing the LNA, I am not sure why SpectreRF and Spectre would give different noise numbers, unless you are using a baseband equivalent model. Perhaps you could describe your set up in more a little more detail.

-Jess

Title: Re: Noise Figure Vs. total input referred noise vo
Post by Jess Chen on Sep 16^th, 2005, 7:53am

One other thing to consider when working with ports is that they have noise too. When you use (F-1) in the expression for input referred noise, you are neglecting the noise from the port. I suspect Spectre includes it when computing input referred noise. You may want to try replacing F-1 with just F to see if the numbers agree.

What numbers are you starting with? What input resistance are you using for the LNA model? What port resistance are you using? Exactly how are you computing integrated noise with Spectre and SpectreRF? Are you using the waveform display or are you using the noise summary printout?

Also, just for completeness I should ask about temperature. I assume all of your analyses are performed at the same temperature. If I could do it all over again, I would recalibrate the noise sources in my rfLib models to account for the difference in simulation temperature and port temperature. The ports are held at 16.85 degrees C. However, the default simulation temperature is 27 degrees C. So when you specify a particular noise figure, you will only see exactly that noise figure if you change the port temperature to 27 degrees C or the simulation temperature to 16.85 degress. For a noise specified figure of 2dB, you will measure 2.06dB unless you align the temperatures.