Here consider a power divider that consists of three equal resistors with R=50/3.

When I simulate NF of this power divider using SP-ana and HB-ana,

I get following results.

For SP-ana

nf(1)=nf(2)=nf(3)=3.010dB

Sij=0.5 for i != j. Sii=0.0

For HB_NOISE-ana

nf(1)=0.0dB, nf(2)=6.021dB, nf(3)=8.451dB

Here PortZ(1)=PortZ(2)=PortZ(3)=50ohm.

NF result from SP-ana can be explained by a definition of NF

for generalized multiple ports described in the following.

http://eesof.tm.agilent.com/docs/adsdoc ... index.html

S-Parameter Simulation>Noise Analysis>Calculating Noise Figure

According to a definition of generalized multiple ports NF,

denominator of 3-ports NF is twice compared to

conventional 2-ports NF. So NF result from SP-ana is 3dB not 6dB.

On the other hand, how can I understand results of NF from HB_NOISE ana ?

nf(2)=6.021dB is correct NF as 2-ports NF.

But what do nf(1) and nf(3) mean ?

In MDS's manual, a sentence of "If mixing occurs, the Skj matrix is the large-signal,

single sideband, frequency conversion S-matrix and the same definition

of noise figure may be used." existed.

So if MDS's manual is correct, all nf(k) in HB_NOISE must be consistent with SP-ana.

But there is no such sentence in ADS's manual.

Following is ADS's netlist I used.

Options ResourceUsage=yes UseNutmegFormat=no \

TopDesignName="C:\users\default\ADS\Mytest_prj\networks\power_divider"

Port:Term1 _net165 0 Num=1 Z=50 Ohm P[1]=polar(dbmtow(0),0) \

Freq[1]=1 GHz Noise=yes Pac=polar(dbmtow(0),0)

Port:Term2 out 0 Num=2 Z=50 Ohm Noise=yes

Port:Term3 _net167 0 Num=3 Z=50 Ohm Noise=yes

R:R1 _net165 _net123 R=50/3 Ohm Noise=yes

R:R2 _net123 out R=50/3 Ohm Noise=yes

R:R3 _net123 _net167 R=50/3 Ohm Noise=yes

S_Param:SP1 CalcS=yes CalcY=no CalcZ=no GroupDelayAperture=1e-4 \

FreqConversion=no FreqConversionPort=1 StatusLevel=2 CalcNoise=yes \

SortNoise=0 BandwidthForNoise=1.0 Hz Freq=100 MHz DevOpPtLevel=0 \

NoiseInputPort=1 NoiseOutputPort=2 \

OutputPlan="SP1_Output"

OutputPlan:SP1_Output Type="Output" UseEquationNestLevel=yes \

EquationNestLevel=2

HB:HB1 MaxOrder=4 Freq[1]=1.0 GHz Order[1]=3 StatusLevel=2 FundOversample=1 \

MaxIters=10 SamanskiiConstant=2 \

Restart=no ArcLevelMaxStep=0.0 MaxStepRatio=100 MaxShrinkage=1.0e-5 \

ArcMaxStep=0.0 UseAllSS_Freqs=yes OutputBudgetIV=no UseKrylov=0 \

DevOpPtLevel=0 UseInFile=no UseOutFile=no KrylovPrec=1 ConvMode=2 \

InputFreq=noisefreq FreqForNoise=100 MHz NoiseInputPort=1 NoiseOutputPort=2 \

PhaseNoise=no SortNoise=0 IncludePortNoise=yes NoisyTwoPort=yes \

BandwidthForNoise=1.0 Hz OutputPlan="HB1_Output"

OutputPlan:HB1_Output Type="Output" UseNodeNestLevel=yes NodeNestLevel=2 \

UseEquationNestLevel=yes EquationNestLevel=2

Options:Options1 Temp=16.85 Tnom=25 TopologyCheck=yes \

ForceS_Params=yes MaxSpectralSize=512 V_RelTol=1e-6 I_RelTol=1e-6 \

GiveAllWarnings=yes MaxWarnings=10 SaveBranchCurrents=no \

OutputInternalNodes=no ForceM_Params=yes \

DC_ReadInitialGuess=no DC_WriteFinalSolution=no

When I simulate NF of this power divider using SP-ana and HB-ana,

I get following results.

For SP-ana

nf(1)=nf(2)=nf(3)=3.010dB

Sij=0.5 for i != j. Sii=0.0

For HB_NOISE-ana

nf(1)=0.0dB, nf(2)=6.021dB, nf(3)=8.451dB

Here PortZ(1)=PortZ(2)=PortZ(3)=50ohm.

NF result from SP-ana can be explained by a definition of NF

for generalized multiple ports described in the following.

http://eesof.tm.agilent.com/docs/adsdoc ... index.html

S-Parameter Simulation>Noise Analysis>Calculating Noise Figure

According to a definition of generalized multiple ports NF,

denominator of 3-ports NF is twice compared to

conventional 2-ports NF. So NF result from SP-ana is 3dB not 6dB.

On the other hand, how can I understand results of NF from HB_NOISE ana ?

nf(2)=6.021dB is correct NF as 2-ports NF.

But what do nf(1) and nf(3) mean ?

In MDS's manual, a sentence of "If mixing occurs, the Skj matrix is the large-signal,

single sideband, frequency conversion S-matrix and the same definition

of noise figure may be used." existed.

So if MDS's manual is correct, all nf(k) in HB_NOISE must be consistent with SP-ana.

But there is no such sentence in ADS's manual.

Following is ADS's netlist I used.

Options ResourceUsage=yes UseNutmegFormat=no \

TopDesignName="C:\users\default\ADS\Mytest_prj\networks\power_divider"

Port:Term1 _net165 0 Num=1 Z=50 Ohm P[1]=polar(dbmtow(0),0) \

Freq[1]=1 GHz Noise=yes Pac=polar(dbmtow(0),0)

Port:Term2 out 0 Num=2 Z=50 Ohm Noise=yes

Port:Term3 _net167 0 Num=3 Z=50 Ohm Noise=yes

R:R1 _net165 _net123 R=50/3 Ohm Noise=yes

R:R2 _net123 out R=50/3 Ohm Noise=yes

R:R3 _net123 _net167 R=50/3 Ohm Noise=yes

S_Param:SP1 CalcS=yes CalcY=no CalcZ=no GroupDelayAperture=1e-4 \

FreqConversion=no FreqConversionPort=1 StatusLevel=2 CalcNoise=yes \

SortNoise=0 BandwidthForNoise=1.0 Hz Freq=100 MHz DevOpPtLevel=0 \

NoiseInputPort=1 NoiseOutputPort=2 \

OutputPlan="SP1_Output"

OutputPlan:SP1_Output Type="Output" UseEquationNestLevel=yes \

EquationNestLevel=2

HB:HB1 MaxOrder=4 Freq[1]=1.0 GHz Order[1]=3 StatusLevel=2 FundOversample=1 \

MaxIters=10 SamanskiiConstant=2 \

Restart=no ArcLevelMaxStep=0.0 MaxStepRatio=100 MaxShrinkage=1.0e-5 \

ArcMaxStep=0.0 UseAllSS_Freqs=yes OutputBudgetIV=no UseKrylov=0 \

DevOpPtLevel=0 UseInFile=no UseOutFile=no KrylovPrec=1 ConvMode=2 \

InputFreq=noisefreq FreqForNoise=100 MHz NoiseInputPort=1 NoiseOutputPort=2 \

PhaseNoise=no SortNoise=0 IncludePortNoise=yes NoisyTwoPort=yes \

BandwidthForNoise=1.0 Hz OutputPlan="HB1_Output"

OutputPlan:HB1_Output Type="Output" UseNodeNestLevel=yes NodeNestLevel=2 \

UseEquationNestLevel=yes EquationNestLevel=2

Options:Options1 Temp=16.85 Tnom=25 TopologyCheck=yes \

ForceS_Params=yes MaxSpectralSize=512 V_RelTol=1e-6 I_RelTol=1e-6 \

GiveAllWarnings=yes MaxWarnings=10 SaveBranchCurrents=no \

OutputInternalNodes=no ForceM_Params=yes \

DC_ReadInitialGuess=no DC_WriteFinalSolution=no

Change:

Port:Term3 _net167 0 Num=3 Z=50 Ohm Noise=yes

To:

R:R4 _net167 0 R=50 Ohm Noise=yes

The other noise figures in this case nf(1) and nf(3) really do not mean much. You should only consider the output port you specify when setting up the noise analysis. These other nf() are there I think as legacy paramters, since they were included way back in the MDS days.