When to simulate Noise Figure, the noise source is usually a port, in which the default impedance is 50 Ohm. More importantly, source noise comes from the impedance of the port. So a larger source impedance produces larger source noises. With respect to the Noise Factor equation, N=(No^2-Nl^2)/(Ns^2), larger source noises will make a smaller Noise Factor.
Q1. In a Noise Figure simulation, such as in design of LNA or Mixer, what value of the port impedance should be?
In my opinion, Noise Factor is under the assumption that the transeiver system has a input impedance of 50 Ohm, so is its output impedance. Every module in such a system should consider a 50 Ohm noise source. Otherwise, one cannot get the following equation:
NF = 1 + (NF1-1) + (NF2-1)/G1 + ...
Q2. Am I correct? If anything is wrong, please point it out.
Thanks in advance.
Larger noise sources should not make a smaller noise figure. If the noise (at the output) from the input source increases, so will the noise at the output (No in your equation), and so the increase will cancel out. That said, there will be an influence - the altered matching will impact the noise figure, and potentially the noise from the source from other sidebands will contribute to the output noise - so it's not that simple.
The port impedance should generally be the impedance that the input will be driven by, as far as I'm aware. I don't think there is an assumption that everything needs to be 50 ohms.
Thanks for your reply.
As mentioned in Virtuoso® Spectre® Circuit Simulator Reference, F = (No^2 - Nl^2)/Ns^2.
On the other hand, (No^2 - Nl^2) = Na^2+Ns^2, in which Ns^2 is the output noise due to source noise and Na^2 is the output noise added by the circuit under test(DUT). So we get F = (Na^2+Ns^2)/(Ns^2) = (Na^2/Ns^2)+1, In this equation, it is seen that F is dependent on Ns^2. And, Ns comes from the impedance of Port source. So I think N is dependent on the impedance of Port source.
By the way, No, Nl, Ns and Na are all represented in V/sqrt(Hz).
if there is anything wrong, please point it out. Thanks a lot!
You're right. Sorry, my mistake - I didn't think this through. You can see this quite clearly if you change the noise temperature of the noise source - if it truly was independent of the source impedance (other than due to matching), then changing the noise temperature would also have no effect.
This is discussed (to some level) in RF Microelectronics by Behzad Razavi, 1997 (pages 39-45). I'm sure it's covered in many other books too. I think the primary requirement on the kind of combination of noise figure you describe is that the impedances are consistent.