It seems that order of transistor and OscPort in netlist
could affect a phase noise(pnmx) result.
In my case, phase noise results are different between the following two cases.
(1) Case-1
OscPort:I6 net044 net14 Z=1.1 NumOctaves=2 Steps=10 FundIndex=1 Harm=1
npnStd:Q0 net20 net044 net37 0
(2) Case-2
npnStd:Q0 net20 net044 net37 0
OscPort:I6 net044 net14 Z=1.1 NumOctaves=2 Steps=10 FundIndex=1 Harm=1
(1) gives about 2dB larger result than (2)
regarding pnmx in very small offset frequency region(less than 5Hz).
Why does order in netlist affect an phase noise(pnmx) result ?
It could cause a confusion for user who use both native ADS schematic and RFDE,
since native ADS and RFDE don't always give same order in netlist.
[Note] pnfm is not affected by order in netlist.
could affect a phase noise(pnmx) result.
In my case, phase noise results are different between the following two cases.
(1) Case-1
OscPort:I6 net044 net14 Z=1.1 NumOctaves=2 Steps=10 FundIndex=1 Harm=1
npnStd:Q0 net20 net044 net37 0
(2) Case-2
npnStd:Q0 net20 net044 net37 0
OscPort:I6 net044 net14 Z=1.1 NumOctaves=2 Steps=10 FundIndex=1 Harm=1
(1) gives about 2dB larger result than (2)
regarding pnmx in very small offset frequency region(less than 5Hz).
Why does order in netlist affect an phase noise(pnmx) result ?
It could cause a confusion for user who use both native ADS schematic and RFDE,
since native ADS and RFDE don't always give same order in netlist.
[Note] pnfm is not affected by order in netlist.
"I haven't seen it before but I'm not surprised. The matrix is created in a way that is dependent on the order of components in the netlist. Because of the numerical conditioning of the matrix, it is not unusual to get slightly different results in any circuit simulation as the component order changes.
The pnmx solution requires computing the inverse of a row of the matrix. In some oscillators, you sometimes see pnmx go flat at small offsets, due to the numerical ill conditioning of the matrix at these small offsets. My guess is that is this is
happening here; if we were to look at the phase noise at very small offsets (0.01 Hz to 1 kHz), they'd both go flat but at slightly different frequencies.
To minimize this, all oscillator phase noise simulations should be done with Order >= 8 and FundOversample >= 4."
Please send the design to EEsof Technical Support in case you're still having problems after making sure the order and oversample are high enough.