Phase Noise vs frequency graph problem 

Hello

I am a newbie engineer starting my career in IC design and working on a  high frequency oscillator (38 GHz). I am using Cadence IC6.1.5-64b.500 version and using spectre simulator for the schemtic design and simulations.

 I am having doubts about my 'pnoise' analysis. All the research papers I have referred show the graph of phase noise (dbc) vs. frequency which have values of phase noise(PN) in negative i.e. around -90 to -100 dbc (the graph should be like the flicker noise graph) but I am getting positive  values of PN(with similar type of curve). I have used the 'relative harmonic as 1 and all the settings seem to be as described in the cadence simulation guide. 

 I have referred most of the  community forum in EDAforum,designer-guide and even the Cadence community...there are problems related to this..but there is no definite answer as to how i can get the correct graph as shown in most of the research papers...In one of the community one guy gave an explanation of the equation spectre uses to simulate pn...My question is whether i need to make changes in the circuit or some settings in cadence...

Kindly help me as it is my first major project and need to do well...thanks in advance.. 

Have you already seen http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395 and http://support.cadence.com/wps/mypoc/cos?uri=deeplinkmin:ViewSolution;solutionNumber=1832229? Are you comparing the phase noise values at similar offset frequencies?

Thanks Frank for replying. Yes, I have seen that post-it explains the equation which spectre uses to find the pnoise...but it doesnt help me to get the right graph with negative values of pn.

I am doing the pss-pnoise simulation using the following settings:

PSS : Harmonic Balance

beat freq - 37.93G

  no of harmonics -10 

accuracy defaults - liberal 

Pnoise : sweeptype - relative

relative harmonic - 1 

start stop from 10K to 10 M

sweeptype - log & points per decade -10

max sideband - 10

positive o/p node -  q+

negative o/p node  -  q-

input source - none

noise type - sources 

And there is no augmented option in the convergence parameters in options....I think Andrew mentined it in his post that augmented is used in the new version by default... 

It's very difficult to tell what the reason for your problem is without being able to examine your circuit in detail. I don't see any obvious errors, but you should try if selecting conservative accuracy defaults changes anything and should make sure that you are using the oscillator mode of the pss analysis and a recent version of Spectre.

If that doesn't help, ask an experienced colleague to examine your circuit or submit it to the Cadence Support. The problem could be anywhere: from unrealistic noise parameters in your transistor models to problems with the sizing of your circuit components, leading to incorrect operating points.

Hello Tawna,

 I am very thankful for your reply...Read a lot of posts by you and Andrew in various topics and a fan of your knowledge and work..

I got the correct graph yesterday when I changed errpreset to 'moderate' and putting tstab as 60n..but the values of phase noise was aroung -140dbc/hz at 1Mhz offset...I was surprised as to how it could be so low...

Today,after reading your post,I changed errpreset to 'conservative' & the simulation is failing..the output logs  had few warnings-

C0.dsx.dnwx:breakdown voltage exceeded 

C0.dsx.dnwx:junction current density exceeds 'imax'

and after that the error came-

 maximum number of iterations reached.Simulations failed to converge.you can try to increase tstab and harms.

 I increased the tstab from 60n to 100n and the harmonics from 10 to 20...the simulation still fails...

Also, thanks for the urls you posted...it helps just knowing that people are there to help amateurs like me...

Hello Frank,

Really appreciate your help..I have put my latest schematic in this post..if you have time to have a look...I haven't used the oscillator mode in pss analysis..

I have a few other problems too which i will be posting separately.. 

I am  having trouble getting quadrature output..i.e. the I and Q signals run in phase right now...Earlier the pmos transistors in my circuit were having w/l as 100u/60n and the nmos were 25u/60n ..at that time  i was getting good quadrature phase....but now i reduced the size of all trans (which my supervisor suggested was very big and told me to increase L and C values) to almost 45u/60n....

The pmos transistors are in cutoff and are just used for coupling to lock the 2nd subharmonic & generate quadrature op...so the transient waveform has a dc level at 1.5 V(which is the supply voltage)..I am not sure if this is correct even though the signal swing I am getting is nearly 1 V or even more..I know it shouldn't be this high..also if i connect a port to i+ and i- terminal with resistance of 100ohm , the swing just reduces a lot..i know the circuit needs a buffer..will that solve this prob..

I am having the same problem after completing the layout and extraction....The pnoise analysis doesnt start because the pss fails to converge...

Can anyone help ??