2.3 GHz lna design sparameter vs netlist

Hi. I am trying to design an LNA at 2.3 ghz with infineon's BFP 740 ultra low noise transistors.
On their page there is an axamle at the same frequency. When i use sparamets i get 0.18 db NF
but when i use netlist with the same matching circuit I get nothing so i changed the bias and now I am getting 0.7 db NF at best. So which one is wiser to use netlist or sparameters ?

Describe your situations correctly.

What do you mean by netlist ?

Do you mean netlist is a S-parameter simulation based on Gummel-Poon-Model ?

Is NF=0.18dB NFmin ?
On the hand, NF=0.7dB is not.

Describe your situations correctly.

http://www.infineon.com/dgdl/Infineo...142740c849070c

The situation is this. In infineon website for bfp 740 http://www.infineon.com/cms/en/produ...77ce0#ispnTab7
you can download SimModel with symbol and footprint AWR MWO v1.0.zip file. Inside that file there is an example at 2.3 GHz. The example uses both S parameters and Netlist(mdl file) of bfp740. When you look at the design with Sparameter data set, It shows 0.1639 db NF. They have the same simulation with mdl file of bfp740 but,Unfortunately their design with mdl file of transistor is not correctly simulated. But when i used the same configuration they have on the example I get around 0.68 dB noise. at 2.3 Ghz which is close to their datasheets. SO i tought their Sparameter files might be wrong since datasheet shows 0.55 dB noise at 1.8 GHz at best.

I don't have AWR MWO.
However I can see contents of "BFP740H6327XTSA1/Simulation Data".

Which one do you use in your simulation
(1) BFP740.net (SPICE Netlist using Gummel-Poon-Model)
(2) bfp740_spar.mdf (Small Signal Table Model written as Keysight EEsof MDIF Format)

We don't call (2) "Netlist".

I think you can not understand simulation model, NF and NFmin at all.

What file do you mean by "S-Parameter" ?

I couldn't see 0.1639 dB NF in s-parameters data-set but anyway..
NF is NFmin what you see in the s-parameters data-set so that the transistor is matched (input) to Zopt to get the NFmin.You should do the same thing ( Zopt should have been supplied) to obtain the similar result.

This one is the simulation based on mdf format by using bfp740_spar.mdf done by infineon http://imgur.com/HFrMxef. This one is the one that I did by using BFP740.net (SPICE Netlist using Gummel-Poon-Model) http://imgur.com/vLuJM2T . Both are NF not NFmin.

(2) "bfp740_spar.mdf" gives NF=0.1645dB at 2.3GHz
(1) "BFP740.net" gives NF=0.65dB at 2.3GHz

Surely do you set same bias points(Vce, Ic) between (1) and (2) ?

I believe (2) rather than (1).

See the followings.
http://www.designers-guide.org/Forum...73409182/11#11
http://www.designers-guide.org/Forum...73409182/13#13

That's exactly what I'm saying.The simulator takes NFmin value while it is simulating the transistor ( because it's been supplied in s-parameters data-set).But when you do this simulation with GP model even the bias conditions are exactly same, the NF will not be NFmin, instead NF will be "50 terminated Noise Figure".That's the difference.

I don't think so.

Even in simulation using (2) "bfp740_spar.mdf", I think impedance of ports are 50ohm.

Here is the Noise Figure simulations of all 3 circuits. I have s parameter simulation, GP model simulation with same circuit as S parameter(Just a DC voltage source of 3.3V added for Bias didn't add anything to it, had the bias network in Spar circuit.) And lastly my own derivation of the same circuit with rogers 5880 and 2V DC source for lower Collector current and better Noise figure.

Simulation using (2) does not require bias at all.
However simlation using (1) require bias.
How do you feed bias for base and collector in Simulation using Gummel-Poon-Model ?

What values do you set Vce and Ic ?

I suspect you don't match same bias points between (1) and (2).

And Title of this thread should be changed to "2.3 GHz lna design, Small-Signal-Model v.s. Large-Signal-Model".

http://www.designers-guide.org/Forum...num=1212376329

as you can see Spar model and GP model has the same bias which is Vc=1V Ic=3mA. But results are different from each other.

So, how does a s-parameter simulation uses NF information ? Is there any additional Noise Figure data in s-parameters data-set ? No.
It computes NF by using existing measured NF data in order to extract the actual NF data, right ?
If we know NFmin,Rn,Ys we can find F ( Microwave and RF Engineering-Roberto Sorrentino,Giovanni Bianchi.Page 429,Eq-11.32)

Right.

You can not understand noise parameters in S-parameter and NF calculation by using them at all.

NF is calculated by {s11, s12, s21, s22} and {NFmin, Rn, Gamma_opt}.
Maybe your Ys is equivalent to Gamma_opt.

So if Zsource is not Zopt, NF is not NFmin.

Touchstone format is a subset of MDIF.
{s11, s12, s21, s22} and {NFmin, Rn, Gamma_opt} are listed in MDIF file.
http://cp.literature.agilent.com/lit...3.html#1112521

I don't think they are same.

Former is (Vce, Ic)=(1.0V, 3.0mA)
Latter is (Vce, Ic)=(1.05V, 3.79mA)

Current discrepancy is fairly large.

Compare NF with adjusting latter (Vce, Ic) as (1.0V, 3.0mA).

However I believe Small-Signal-Model not Large-Signal-Model since former is a raw data without approximation due to fitting.

Yeah.. Only you understand ALL, we know nothing..
Can you show us Highness where the s-parameters are here ?

You often append wrong things elsewhere.

Sorry, {s11,..,s22} are not used.

{s11,..,s22} are used in NF calculation, if s-parameters are cascaded like two stage amplifier.

However, if Zsource is not Zopt, NF is not NFmin.
So all NF in this thread are not NFmin.

This message board surely wishes to attract and keep our expert daily regular posters, and that is the aim of forum rules: