The Designer's Guide Community
Forum
Welcome, Guest. Please Login or Register. Please follow the Forum guidelines.
Oct 18th, 2019, 2:14pm
Pages: 1
Send Topic Print
Reg peaking in noise spectrum in pnoise simulation with Spectre (Read 617 times)
ULPAnalog
Community Member
***
Offline



Posts: 97

Reg peaking in noise spectrum in pnoise simulation with Spectre
Jan 25th, 2019, 9:59am
 
Dear experts,

I am trying to simulate a mixer that upconverts flicker noise. Fig(a) in the attachment shows the schematic of the circuit. in,f and in,t are flicker and thermal noise sources modeled in verilogA. fclk is 2.5kHz. Fig(b) is the output voltage noise PSD obtained through pnoise analysis using spectreRF. My question is about the peaking at 10kHz. While I understanding that mixer upconverts flicker noise to all odd harmonics of fclk, I expected the fundamental to be larger than any other harmonic. But it turned out not to be the case, at least in this simulation.

I also used Mentor graphics afs to simulate the circuit with same pnoise settings and obtained Fig(c),  and it does not exhibit such peaking. The spot noise at low frequencies match between spectreRF and afs.

The following are my questions:
1. Is the peaking seen in simulation due to the fact that I do not specify specific points where the analysis needs to be run? I noticed dominant peaking happen at 2.5kHz when I add 2.5kHz and its harmonics as specific points where pnoise needs to be run. This was expected as flicker has infinite power at zero frequency. (Similar explanation was provided at https://community.cadence.com/cadence_technology_forums/f/custom-ic-design/28660...
/regarding-pss-and-pnoise-analysis-getting-a-peak-at-1khz-in-noise-plot)

2. As can be seen, spectreRF and afs spot noise values are also in agreement at 2.5kHz but not at other harmonics of fclk. Is this just a coincidence? (Ideally I would have expected both to give a spot noise fclk that is very large).

Thanks and regards
Back to top
 

mixer_002.png
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2222
Silicon Valley
Re: Reg peaking in noise spectrum in pnoise simulation with Spectre
Reply #1 - Jan 25th, 2019, 2:15pm
 
Flicker noise goes to infinity as frequency goes to 0.  Thus it also goes to infinity at odd multiples of the clock. You run the PAC analysis at a discrete set of frequencies, and the size of the peak will depend on how close the analysis point falls to a multiple of the clock. Do not worry about the apparent amplitude of the peaks. Those are simulation artifacts that depend on your point placement. Instead focus on amplitude a fixed distance from the harmonics.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
ULPAnalog
Community Member
***
Offline



Posts: 97

Re: Reg peaking in noise spectrum in pnoise simulation with Spectre
Reply #2 - Jan 25th, 2019, 4:31pm
 
Thanks Ken, for the quick response. It is really helpful. Another issue that I experienced was the frequency at which the peaking was observed. As mentioned earlier, I used 2.5kHz clk and 10kHz, where the peaking was observed is not an odd multiple of fclk. As in attached Fig(a), one can see peaking at 10kHz with a fclk of 2.5kHz. When I changed fclk to 2kHz (Fig.(b)), I see peaking at 10kHz but this was expected being a odd harmonic. The clk phases are non-overlapping and have nearly 50% duty ratio.

Is there something that I am missing in this sim/setup?

Thanks and regards

PS: I noticed that figures in my previous post were awful. Sorry about that.
Back to top
 

mixer_2.png
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2222
Silicon Valley
Re: Reg peaking in noise spectrum in pnoise simulation with Spectre
Reply #3 - Jan 26th, 2019, 5:55am
 
The flicker noise up conversion is only suppressed if the circuit is perfectly symmetric. To you it may be, but the simulator will introduce extremely small asymmetries because of the way it solves its equations (it can only do one thing at a time, and so it will need to process one side of a symmetric pair before the other). That, combined with the fact that flicker noise is going to infinity, means that there will be peaks at all harmonics of the clock. The peaks at the even harmonics should be smaller.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
ULPAnalog
Community Member
***
Offline



Posts: 97

Re: Reg peaking in noise spectrum in pnoise simulation with Spectre
Reply #4 - Jan 28th, 2019, 4:46pm
 
Thanks Ken, for clarifying. Indeed I would expect to see smaller peaks at even harmonics to account for the minor asymmetries but sometimes the peaks are larger than any other harmonics, odd included, which surprised me.

Best regards
Back to top
 
 
View Profile   IP Logged
Pages: 1
Send Topic Print
Copyright 2002-2019 Designer’s Guide Consulting, Inc. Designer’s Guide® is a registered trademark of Designer’s Guide Consulting, Inc. All rights reserved. Send comments or questions to editor@designers-guide.org. Consider submitting a paper or model.