Hi,

I'm simulating phase noise for a pretty standard LC VCO and I get the following warning from the pnoise analysis:


**********************************************************************
Periodic Noise Analysis `pnoise': freq = 3.03189 GHz + (1 kHz -> 1 MHz)
***********************************************************************
Using the operating-point information generated by PSS analysis `pss'.

Warning from spectre at freq = 1 kHz during PNoise analysis `pnoise'.
   Relative residual of linear system is significantly above tolerance.
       Results may be inaccurate.

   pnoise: freq = 1 kHz          (0 %), step = 0 Hz            (0 %)

The pss analysis seems to complete fine and the results seem reasonable.  I have tried tightening simulator tolerances, increasing maxacfreq, and the standard pnoise solver and oscsolver modes (instead of turbo).  I still get the warnings at almost every frequency step out to about 1MHz.

I'm not sure if I can trust these results or not because of the warning.  Can anyone recommend a course of action to correct the warning?  The message seems to suggest I need to increase tolerances, but I don't want inaccurate simulation results.

Thanks,
nate

Hi Ken,

Thanks for the tip.  Unfortunately, I get the same result in reverse.  The first couple of points at larger offsets (>1MHz), have no warnings.  However, pnoise still generates the same warnings from 1MHz on down.

I tried more points per decade and starting from 10MHz and 1MHz, all with the same result.  Any other possibilities I might try?  This one's a tough one because the warning doesn't give any clear path to resolution, at least for me.

Thanks again for your help.

nate

Sounds good Ken.  I'm also pursuing that avenue but I thought I would post here in the meantime, just in case it was a simple fix.

Thanks,
nate