Designer’s Guide Community

Roky

New Member

Offline

Posts: 5

PSRR of VCO simulation in spectreRF
Aug 17^th, 2004, 4:11am

Hi all,

� � �I want to simulate the PSRR of a VCO circuit. For this purpose I use the PSS analysis in spectreRF together with an PAC analysis. The PAC source with magnitude of 1 I�ve placed in the ground connection. For the start up of the VCO I use the node set in convergence aids in cadence analog environment (or a pulse source as an kicker � doesn�t matter the result is the same). Ok, and here is my problem: The PAC result shows a DC gain (f=0) of 90dB, which is an absolutely unrealistic result. I found out that the DC gain value depends on the node set settings (or the amplitude of my kicker)� The problem is that I have to use the node set settings for the start up of the VCO.

Please could someone help me�.

� � �I�ve tried to simulate the PSRR of the cadence oscillator (from the rf examples Lib) and the PAC effect was approx. the same.

Thx a lot Roky

Back to top

« Last Edit: Aug 17^th, 2004, 7:32am by Roky »

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: PSRR of VCO simulation in spectreRF
Reply #1 - Aug 18^th, 2004, 12:35am

Your results are not very surprising if your f=0 is really a delta_f=0. The PAC gain should in fact go towards infinity for delta_f->0, see section 3 of http://www.designers-guide.com/Analysis/PLLnoise+jitter.pdf.

I currently have no explanation as to why the result would depend on the initial conditions for the oscillator, except perhaps numerical issues related to the very large gain. Are the differences very significant?

Back to top

« Last Edit: Aug 18^th, 2004, 5:17am by Frank Wiedmann »

IP Logged

Roky

New Member

Offline

Posts: 5

Re: PSRR of VCO simulation in spectreRF
Reply #2 - Aug 19^th, 2004, 7:13am

Hi Frank,

Thank u a lot for your answer. I�ve red the article from Ken. My problem is that the initial conditions shift the result up and down (in the desired freq range too). I�ve also tried the PXF analysis, as cadence recommends, and the result is approx. the same. But I can�t understand the dc gain of 90dB. This result would predict that the VCO amplifies ground disturbances in range of (0 to few MHz), and I can�t believe it. In the PAC form I�ve checked out the absolute sweep type. Another point is that start up conditions should not have an influence to my PAC result, but they have.

Only what I need is a transfer function from my disturbance source in the ground to output, but the result should be realistic.

Back to top

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: PSRR of VCO simulation in spectreRF
Reply #3 - Aug 19^th, 2004, 8:34am

First of all, you should probably find out why your oscillation frequency depends on the initial conditions (I hope I understood you correctly here). You might want to try setting the Accuracy Defaults to conservative. If the results of the PSS analysis are different, it is not very surprising that the results of the following PAC analysis differ as well.

The fact that you have a very large PAC gain from DC might indicate that your oscillation frequency depends on the power supply voltage or that your PAC source is modifying your effective tuning voltage because the tuning voltage source has a separate ground connection in your simulation setup.

Back to top

« Last Edit: Aug 19^th, 2004, 11:17pm by Frank Wiedmann »

IP Logged

Jitter Man Community Member Offline Am I? Or am I so sane that u just blew your mind? Posts: 50	Re: PSRR of VCO simulation in spectreRF Reply #4 - Aug 21^st, 2004, 2:25pm The phase noise results from early versions of SpectreRF would occasionally vary as a function of either initial condition or t_stab. I recommend that you upgrade to a more recent version. Jitter Man
Back to top	IP Logged

Roky

New Member

Offline

Posts: 5

Re: PSRR of VCO simulation in spectreRF
Reply #5 - Aug 24^th, 2004, 5:38am

Hi Frank,

My oscilation frequency doesn�t change with initial conditions. The pss result remains the same with different initial conditions, only what changes is the PAC result. Frank, would u be so kind and explain me why should the PAC result reach infinity @ f=0.

Hi Jitter man,
I tried it with, SpectreRF 4, and with SpectreRF5, the result is the same�

Back to top

IP Logged

Frank Wiedmann Community Fellow Offline Posts: 678 Munich, Germany	Re: PSRR of VCO simulation in spectreRF Reply #6 - Aug 24^th, 2004, 7:22am If your PAC source changes the oscillation frequency, you will get very large phase deviations for low PAC input frequencies because the phase difference accumulates over a long time. For f=0, you will get an infinite phase deviation because the phase difference accumulates over an infinite time.
Back to top	IP Logged