The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> pac and pnoise disagreement
https://designers-guide.org/forum/YaBB.pl?num=1094680012

Message started by skiwindsurf on Sep 8^th, 2004, 2:46pm

Title: pac and pnoise disagreement
Post by skiwindsurf on Sep 8^th, 2004, 2:46pm

I have a question here about phase noise simulation in SpectreRF of Cadence. I tried to see the effect of reduced phase noise due to low frequency noise when the waveform is symmetric, as predicted by Ali Hajimiri.

When I run pss+pac on a 7-stage single-ended ring oscillator (about 3GHz in 0.18um CMOS), and inject a low frequency (100KHz) current into one of the nodes, I can see that its first sideband significantly reduced (by 50db) when the PMOS/NMOS ratio is 0.628/0.372 (compared with P/N ratio of 0.7/0.3). But when I run pss+pnoise on the same oscillator, and varied the PMOS/NMOS ratio from 0.5/0.5 to 0.8/0.2, I can't see ANY reduction of the phase noise in the 1/f^3 region.

I read Ken Kundert's paper on JSSC 1999 (and updated in 2001). It seems that both pac and pnoise are pss + small signal simulation. My question is, why can't I see the reduction of phase noise in 1/f^3 region using pss+pnoise? What is the difference between pnoise and pac simulation?

Title: Re: pac and pnoise disagreement
Post by Ken Kundert on Sep 8^th, 2004, 6:09pm

In the paper on this subject that I have looked at most recently (Jitter and Phase Noise in Ring Oscillators, JSSC 6/99), Hajimiri treats flicker noise as a noise current with a 1/f spectrum that is injected into the drain. This is really not an accurate portrayal of the situation. The flicker noise current is modulated by the time varying g_m of the device. Thus, his comment that it is the DC component of the ISF that matters is flawed.

To model this correctly, one should really consider flicker noise as being injected by a voltage source in series with the gate. Thus, you should perform you PAC analysis using a voltage source in series with the gate as your input and observe the +1 sideband.

I believe that with this more accurate model of flicker noise injection, Hajimiri's advice concerning symmetry breaks down.

-Ken