The noise characteristics of active devices are often specified with the four "spot" parameters: F(min),G0,B0,RN. I have read how GENESYS uses the "noise correlation matrix" approach for noise modeling.

My question: "Is their a direction implied for GENESYS noise analysis?"

For example, suppose I build an microwave amplifier around a FET. It is straightforward to use GENESYS and spot noise data from the FET manufacturer to calculate "forward" Noise Figure(NF) for the amplifier. -by "forward" I mean from input-to-output.

Would it make sense to use GENESYS to calculate the "backward" noise figure for the amplier (ie. from output to input)?

My question: "Is their a direction implied for GENESYS noise analysis?"

For example, suppose I build an microwave amplifier around a FET. It is straightforward to use GENESYS and spot noise data from the FET manufacturer to calculate "forward" Noise Figure(NF) for the amplifier. -by "forward" I mean from input-to-output.

Would it make sense to use GENESYS to calculate the "backward" noise figure for the amplier (ie. from output to input)?

The noise analysis needs admittance and correlation matrix of each circuit element. For reciprocal elements with thermal nature of noise the correlation matrix is calculated just from its Y-matrix, when for nonreciprocal, or having noise sources of other different natures, it may be get from experimental measurements, or from complicated math model, taken into account nature of all noise sources.

Sure, the parameters, defining noise model of 2-port devices: Fmin, Rn, Gn, Yo are direction dependent. From math viewpoint it does not matter what the port of the 2-port devices is used as input to characterize its noise model the parameters have all needed information to convert them to full noise correlation matrix of the 2-port device. If you have accurate values of the parameters for input and output ports, then convertion them to noise correlation matrix must get identical result.

However, the sensitivity of correlation matrix to errors of the measured noise parameters may be significantly different for each of the measurements, especially for low noise elements.

For example, if you measure the 4 parameters of low noise amplifier, you typically may get Fmin(input) < 1dB, while measured from output port Fmin(output) > 30dB. In this case using the output measured noise parameters for calculation of noise correlation matrix results to too high errors of the low noise device characterization, calculated from input to output port.

To guarantee high accuracy of noise modeling of the amplifier, it are to be characterized from input port, otherwise the errors of the modeling may be unpredictable high. In the same time, if device is almost reciprocal, it may be measured as from input, so from output port, with very close result of the noise modeling.