confusions on lna matching problem

[bigworm]

hi,
as is well known that when lna is well matched, you will get either best power delivery or best noise figure.
but now what confuses me is that , a low-input-impedance lna with S11 about -3 is simulated in icfb and I get its gain(gain1) and NF(NF1),
and after I match the lna with S11 <-10, I found that the gain is still more or less the same with gain1. so is the NF.

but I think when the lna is not matched, most power of input signal is reflected and little power is transferred into the lna, so that I think the gain of lna should be much less than the matched one. but the simulation result doesn't support this conclusion.

I wonder whether it is necessary to match the lna for any case?

 

[GVVIN]

Check your simulation/setups.

There may be some error with your simulation or measurement. There should be some improvement when matched than the unmatched condition.

regards
GV

 

[Abhishekabs]

In LNA.
Whne you want to have low noise figure, yo need little/more mismatch at the input of transistor.
Your S21 will not much affect due to input i suppose.
To increase gain you need to match the output of transistor, also make transistor stable(K>>1) by placing reactance in common terminal.
How many stages your LNA has?
WHat is the Gain & NF you want & you are getting.
Also which transistor you are using?

 

[freetofree]

noise impedance mathcing is different from power matching. If it is designed for noise matching, then if the input network resonates, then gain can still be improved because it is QGmVin

 

[RF-OM]

There is always some trade offs between optimal noise matching and maximum power gain matching. But very often the difference is not big and if you do the noise optimized matching you introduce some mismatch losses into circuit. This will reduce the gain and increase noise figure. Sometimes there simply no sense to do optimal noise matching because mismatch losses will shift noise figure back. In such a case it may be better not to do noise matching. Especially it is true for broadband systems. Recently I designed LNA for 3 – 5 GHz band. There was choice to do optimal noise matching and get noise figure 0.72 dB or leave LNA as it was and get noise figure 0.91 dB (all numbers from circuit utility simulation with LINC2 program). But optimal noise matching introduced about 0.3 dB mismatch loss and noise figure become 0.7 + 0.3 = 1 dB. Clearly there was no sense to do optimal noise matching.

Best regards,
RF-OM

 

[RFnewbie]

Hello RF-OM,

I am a little confused by the following sentences in your last post:
"But optimal noise matching introduced about 0.3 dB mismatch loss and noise figure become 0.7 + 0.3 = 1 dB. Clearly there was no sense to do optimal noise matching."
Why are taking into account the loss in the input matching in the above NF calculation? I would expect that this input matching has been designed to get the targeted NF and its loss is already included in the 0.7 NF ... could you help me to understand? I apologize in advance if my question is silly....

thank you.

Bil

 

[RF-OM]

Any loss before LNA counts one-one for NF. The number 0.72 dB in NF was obtained with LINC2 circle utility from S-parameter file. This is NF for LNA. not for LNA+input circuit. Actual NF can be much worse with all the elements into input circuit. For example, one SMA connector can introduce about 0.5 dB loss, but you cannot measure NF or S-parameters without connector. This is why manufacturers always de-embed actual S-parameters from measured. In your real design you need to keep it in mind. Also there should be common sense consideration too. For example, you have an LNA with 0.9 dB NF and after all with input circuit (BPF+transmission line, connector and so on) you get about 3 dB NF. In this case you need to think is it worth to include several extra elements to do the match and get 0.2 dB better NF or leave the design with what you have without matching.

In this particular case I prefer not to do optimal matching.

Best regards,
RF-OM

 

[bigworm]

but, I think , when you design the LNA, (in my case, I use cadence icfb, simulation includes input matching and LNA), considering to some extent the parasitics, the NF you get from simulation is what you will more or less get in the measurement, right?
it is not easy to get pure lna nf since other elements will affect the value.