The flattening happened because of noise conversion matrix singularity close to carrier. The closer to carrier the more the matrix become ill-defined, and finally it becomes singular.
To resolve the numerical issue Harbec uses a simple phase noise "polishing" technique, developed in ADS oscillator analysis, which actually adds a complex impedance Zpolish in serial with OscPort (for 1st harmonic of the osc frequency, disconnected for all other spectrals), and value of Zpolish is optimized to maximize voltage amplitude Vosc at the OSCPORT node (minimize small signal Spectral Gain = 1/|Vosc|).
In steady state regime the OSCPORT has no currrent (disconnected from oscillator), and as result the oscillator phase noise does not depend on Zpolish, but it helps to improve the condition number of conversion matrix (when this technique works), and allows calculate phase noise close to its carrier.
Output noise square voltage = sum of square voltages of all circuit noise contributors, which are primary noise sources, as thermal noise of circuit resistors, shot and flicker noise of noise sources, included in nonliner device models.
Harbec saves independently values of each noise contributor, sorting them at lowest noise frequency, to show which noise mostly defines circuit noise at the specified circuit output.
NCName[nNC]- array of names of noise contributors
\N1\2N5109\Q1:lrb:shot:*white - means that this noise contributor is a shot current source of base resistor (Irb), of Part Q1 (bjt transistor), included in subcircuit model "2N5109", used in design N1, and this is is a nonlinear ("*") "white" noise source.
NCValue[nNC][nFnoise] - swept array(for noise frequency) of noise contributors square voltages at noise output node(s), [dBV]
On graph of NCValue the noise contributors names are included to the highlighted label of graph trace, pointed by mouse cursor.
The flattening happened because of noise conversion matrix singularity close to carrier. The closer to carrier the more the matrix become ill-defined, and finally it becomes singular.
To resolve the numerical issue Harbec uses a simple phase noise "polishing" technique, developed in ADS oscillator analysis, which actually adds a complex impedance Zpolish in serial with OscPort (for 1st harmonic of the osc frequency, disconnected for all other spectrals), and value of Zpolish is optimized to maximize voltage amplitude Vosc at the OSCPORT node (minimize small signal Spectral Gain = 1/|Vosc|).
In steady state regime the OSCPORT has no currrent (disconnected from oscillator), and as result the oscillator phase noise does not depend on Zpolish, but it helps to improve the condition number of conversion matrix (when this technique works), and allows calculate phase noise close to its carrier.
Sergey