Effect of S11 and S22 of LNA on Noise Figure of system

Hi,

I am designing a TV Rx ckt (circuit). I am having LNA at front end. The LNA is having a S11 of 5dB and S22 of 5dB. Overall Noise figure of system is 6dB.
So i want to know if i change Input impedance matching ckt of LNA to improve S11, will the noise figure of system improve?
Also if i change output impedance matching ckt of LNA to improve S22, will the noise figure of system improve?
Which one of the above two is effective in terms of improving noise figure?

Regards
Vinod

Normally there is a tradeoff between nf and s11 . As you improve the s11, the noise figure srarts to rise up.

---------- Post added at 10:12 ---------- Previous post was at 10:02 ----------

As an example, consider a common gate amplifer. The input impedance is 1/gm. Therefore the s11 is (1-gmRs)/(1+gmRs). The noise figure is simply is ~1+gamma/(gmRs).
As you increase the gm, the noise figure goes down and the s11 becomes worse. This is a fundamental trade-off.

Hi all,

I am looking for answer in a system level. I am a board design engineer. So i understand little about IC design stuffs. So basically i am trying to understand if S11 is very bad (high mismatch loss) compared to noise figure of the LNA, should i consider S11 in Friss Formula for calculating noise figure of the system. By improving S11 so that mismatch loss is less than Noise figure of LNA, then how to calculate the Noise figure of the system. Does Bad S11 (high mismatch loss) or Noise figure of LNA will influence the overall system noise figure.

Regards
Vinod A H

Milad-D said:

Normally there is a tradeoff between nf and s11 . As you improve the s11, the noise figure srarts to rise up.

---------- Post added at 10:12 ---------- Previous post was at 10:02 ----------

As an example, consider a common gate amplifer. The input impedance is 1/gm. Therefore the s11 is (1-gmRs)/(1+gmRs). The noise figure is simply is ~1+gamma/(gmRs).
As you increase the gm, the noise figure goes down and the s11 becomes worse. This is a fundamental trade-off.

Click to expand...

ah.vinod said:

Hi all,

I am looking for answer in a system level. I am a board design engineer. So i understand little about IC design stuffs. So basically i am trying to understand if S11 is very bad (high mismatch loss) compared to noise figure of the LNA, should i consider S11 in Friss Formula for calculating noise figure of the system. By improving S11 so that mismatch loss is less than Noise figure of LNA, then how to calculate the Noise figure of the system. Does Bad S11 (high mismatch loss) or Noise figure of LNA will influence the overall system noise figure.

Regards
Vinod A H

Click to expand...

I wrestled with the same issue of NF vs S11 a while back, and came to the conclusion that you simply have to decide which is more important, which is a completely application-dependent issue. If your biggest concern is noise, then you want the best noise figure, and therefore a noise match, even if that gives you a bad S11. If your concern is overall gain, then you want to impedance match, and get a good S11, and you'll end up with a sub optimal noise figure.

There are techniques that allow you to obtain simultaneous noise and impedance match, but I'm not really qualified to explain them.

S11=-5dB for LNA, I think it's not bad.

S11 of -5 db is bad if you plan to have an input filter terminated.

There is a lot missing from your requirements. TV requires high intermod performance. You likely need AGC on the LNA.

Output is alway matched conjugately (best gain, best S22). You may include a ballast loading resistor on output to help stabilize at expense of giving up some gain. Depending on type of device used there are ways to better align NF match and maximum gain match (improved S11). These methods give up some gain also. Filters are not matched to, they are terminated per their design criteria. If you change the termination due to poor vswr you change the filter response.

If you have AGC then the challenge is to avoid too much S11, S22 degradation with gain cutback. Cascode designs are generally better in this aspect.

Why don't you plot NF circles and Input Gamma Circles and find a optimum point between these circles ???
Standart procedure starts with these circles and nobody wants to plot them. ...Interesting..
In ADS takes few seconds..