Dear pancho_hideboo, Thank you for your feedback.
I have several questions in your reply.
First,
Andrew Beckett wrote on Apr 14th, 2010, 4:28pm:You should use pnoise with a probe as the output, pointing at the port. The whole point is that the port is part of the measurement equipment (think of how you'd do the measurement on the bench), and so the noise produced by the termination is not part of the noise in the circuit.
Similarly, the noise produced by the input port is at whatever temperature is set on the port via the noisetemp parameter - rather than what the simulation temperature is. That's because the input port is not on the chip, and so wouldn't be at the same temperature as the chip - it should be producing a known amount of noise..
Using a voltage as the output for pnoise is fine if you are computing noise figure, and there's nothing in the circuit which represents the loading of the "probe" - maybe you have a voltage rather than power output. It's also OK if you're not computing noise figure, but (say) output noise.
Not really a design issue - would have made more sense in the RF simulation board.
Hope that helps,
Regards,
Andrew.
Andrew's opion makes sense I think. And if you set the noise temperature. The noise of the port is settled, not relative to the Rs of the port. So I do a test, use a attenuator, input port (Rs=200) seriesly connect to 2 resistor, both are 50 ohm, output port(Rl=50) paralle with the last resistor . However, I found , the NF is not 6dB but 8.88dB. And I try to fill the input port noise temperature, and only this noise temperature is 16.9, the NF=8.88dB. I am confusing.....
Second, it is correct you mention in another discussed issue
NF is calculated like following.
NF=1+Na/(Gain2*Nin)
here Na=Square of intrinsic output voltage noise from DUT [Vrms2/Hz],
Nin=Square of input voltage noise from source(=4*k*T*Rs) [Vrms2/Hz],
Gain=Voltage Conversion Gain which is the sum of sidebands you have interests in.
I calculated use these equations, I just found the Nin is not equal to (=4*k*T*Rs) in all my LNA designs with matching input and un-matching input.
It seems to be different matching conditions, the source noise power transferred differently to the output.
When Z11 is much larger than input source ressistor, Nin=200z, as you know, 4KTRs(290K 1.38*e-23 50ohm) is 800z.
And in matching conditions, I use (Noise Factor) /(output noise input reference)=Nin=400.4z,output noise input reference is the total noise of output divide by the Gain**2,equivalent to input side, still not 800z. I wonder whether spectre caculated it in these condition(290K 1.38*e-23 50ohm). I did set my simulator under the 27 temperature in ADE, but did not set temerature in port.
Please be free to discuss. Thank you~~
ps You know, when you use port to do ac analysis, port with 1v ac magnitude,the input of your circuit could get -∞ to 6dB gain. This is because port deliver 1 ac magnitude in matching condition, so total is 2v ac magnitude provide by port. So when Zin big enough to Rs, all 2V is delivered to Zin, so 20log2=6dB. When Zin=0 input only get -∞ gain.
So I mean, using port do analysis, whether it is mathcing,is very important and quite differrent to use vdc (with 1v ac magnitude ) in ac analysis.
Does it the same when do noise? Actually, the noise results in noise analysis is always the same with sp analysis under same condition.