The Designer's Guide Community
Forum
Welcome, Guest. Please Login or Register. Please follow the Forum guidelines.
May 19th, 2019, 1:17pm
Pages: 1 2 3 
Send Topic Print
Question about phase noise simulation result (Read 36550 times)
YCY
Community Member
***
Offline



Posts: 32

Question about phase noise simulation result
Jun 28th, 2011, 3:49am
 
Hi,

I'm designing a CML frequency divider. The schematic is as shown:



When I use PSS+PNOISE to simulate the phase noise of this divider,
I find the phase noise measured before the selector (node fDIV3)
is larger than that after the selector (node fSEL).
The simulation result is:

I think this result is somewhat weird since the selector is
a voltage amplifier that should not improve the phase noise.  

So my question is:
Is this simulation result correct?
If yes, how can a voltage amplifier improve the phase noise of a signal?
If no, what's wrong with the simulation, do I have a wrong simulation setting?
My test condition is:
   input frequency fin=3.4 GHz
   modulus=8 (P0=P1=P2=0, P3=1)
   fSEL=fDIV3=425 MHz

PSS setup:
   shooting engine is used.
   fund=425 MHz
   No. of harmonics=16
   errpreset=conservative

PNOISE setup:
   sweeptype=relative
   relative harmonic=1
   maximum sidebands=16
   sweep frequency=10k~10M
   probe voltage: fDIV3 and fSEL

Could anybody please help me figure out this question?

Thanks in advance.

YCY
Back to top
 
 
View Profile   IP Logged
YCY
Community Member
***
Offline



Posts: 32

Re: Question about phase noise simulation result
Reply #1 - Jun 29th, 2011, 7:58am
 
Well, I found why.
When the noise type is set to source, Pnoise calculate the carrier-to-noise ratio rather than real phase noise or timing jitter.
Because the divider is a driven system and only the output threshold crossing is of concern.
One should use jitter to calculate its phase noise.
The detailed setup can be found by searching "Jitter Measurements Using SpectreRF" on Google.
Back to top
 
 
View Profile   IP Logged
neoflash
Community Fellow
*****
Offline

Mixed-Signal
Designer

Posts: 382

Re: Question about phase noise simulation result
Reply #2 - Jun 29th, 2011, 11:07pm
 
Interesting topic.

I was long for a discussion with SpectreRF expert face to face on their RF simulators' phase noise sim. However, I can always only reach to some AE. No real talk ever.

I was always expecting the pnoise in source mode with driven circuit to give also the correct phase noise plot. However, my self also encounter one weird phenomenon recently. I'm still working with Cadence AE hopelessly.

Do you have any solid source saying that source mode won't work for driven circuits? I ever seen an application note adopted source mode for VCO.
Back to top
 
 
View Profile   IP Logged
YCY
Community Member
***
Offline



Posts: 32

Re: Question about phase noise simulation result
Reply #3 - Jun 30th, 2011, 2:37am
 
Hi, neoflash,

I think I mislead you by saying "... divider is a driven system..."
Actually my point is that "only the threshold crossing is of concern."

Because I am simulating a frequency divider, only the timing jitter is important.
The source mode won't work here since the AM noise may dominate the plotted phase noise,
and lead to a useless information.
That's what happened in my previous post.

In fact, I am also curious why the PM-only phase noise in the modulated mode is not the same as Jee in jitter mode.
I used to consider the PM-only phase noise as the PSD of phase error, and shouldn't it be the same as 20log(|2*pi*f*Jee|)?

By the way, could you also share the experience you encounter, I am also interested in that.

Thanks

YCY
Back to top
 
 
View Profile   IP Logged
neoflash
Community Fellow
*****
Offline

Mixed-Signal
Designer

Posts: 382

Re: Question about phase noise simulation result
Reply #4 - Jun 30th, 2011, 10:22am
 
Guess what? We could be lucky if Ken can share some of his inputs on this. He and Andrew are the people who really know the insider info. LOL.

I agree that the noise output in sources mode will be dominated by the AM noise. Currently, I'm getting stuck in this because for some reason the jitter mode doesn't run. And the source mode only gave me blended AM/PM noise.
Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2161
Silicon Valley
Re: Question about phase noise simulation result
Reply #5 - Jun 30th, 2011, 11:04am
 
The so-called 'sources' mode produces SSB noise. If you plot then plot the 'phase noise' you are actually plotting the SSB noise power normalized to the carrier (the give away is in the units, which are dBc/Hz). I know these terms are either meaningless ('sources') or misleading ('phase noise') and I repeatedly encouraged the Cadence simulation environment developers to change them, but was unsuccessful. I don't know where the term 'sources' comes from, but the term 'phase noise' was used because back in the early days of RF simulation we were primarily focused on computing oscillator phase noise, and in that situation normalized noise is effective proxy for phase noise because in oscillators they are equal for most frequencies of interest.

For information about the difference between SSB noise, AM/PM noise, and jitter, see Noise in mixers, oscillators, samplers, and logic: an introduction to cyclostationary noise on www.designers-guide.org/Theory.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
neoflash
Community Fellow
*****
Offline

Mixed-Signal
Designer

Posts: 382

Re: Question about phase noise simulation result
Reply #6 - Jun 30th, 2011, 12:15pm
 
Ken,

In deed thank you for offering your inputs.

One long existing question is that why Cadence seems to use "source" mode for their VCO pnoise simulation in one of their application note. I took it as the recommended way but more knowledge seems to convince me this is not a good idea.

Regards,
Neo

Ken Kundert wrote on Jun 30th, 2011, 11:04am:
The so-called 'sources' mode produces SSB noise. If you plot then plot the 'phase noise' you are actually plotting the SSB noise power normalized to the carrier (the give away is in the units, which are dBc/Hz). I know these terms are either meaningless ('sources') or misleading ('phase noise') and I repeatedly encouraged the Cadence simulation environment developers to change them, but was unsuccessful. I don't know where the term 'sources' comes from, but the term 'phase noise' was used because back in the early days of RF simulation we were primarily focused on computing oscillator phase noise, and in that situation normalized noise is effective proxy for phase noise because in oscillators they are equal for most frequencies of interest.

For information about the difference between SSB noise, AM/PM noise, and jitter, see Noise in mixers, oscillators, samplers, and logic: an introduction to cyclostationary noise on www.designers-guide.org/Theory.

-Ken

Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2161
Silicon Valley
Re: Question about phase noise simulation result
Reply #7 - Jun 30th, 2011, 4:49pm
 
Neo,
   Using 'source' mode for VCO phase noise simulation make sense for a VCO even though that mode computes normalized SSB noise rather than phase noise make sense because all oscillators produce almost exclusively phase noise in the frequencies of interest. Thus, using 'source' mode will give the same result as the more expensive am/pm mode.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
neoflash
Community Fellow
*****
Offline

Mixed-Signal
Designer

Posts: 382

Re: Question about phase noise simulation result
Reply #8 - Jun 30th, 2011, 5:34pm
 
Ken:

Thanks for the reply.

In EldoRF, it always split noise into PM noise and AM noise. In VCO sim, I still see high AM noise in the beat tone region.

Also, I don't understand why VCO has low AM at the 1st harmonic band while clock buffer and dividers have a much higher. I think both will be up-converted from DC to carrier band.

Regards,
Neo

Ken Kundert wrote on Jun 30th, 2011, 4:49pm:
Neo,
   Using 'source' mode for VCO phase noise simulation make sense for a VCO even though that mode computes normalized SSB noise rather than phase noise make sense because all oscillators produce almost exclusively phase noise in the frequencies of interest. Thus, using 'source' mode will give the same result as the more expensive am/pm mode.

-Ken

Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2161
Silicon Valley
Re: Question about phase noise simulation result
Reply #9 - Jun 30th, 2011, 10:27pm
 
Neo,
   I don't know what the 'beat tone' region is.

If you want to know why the noise produced by oscillators is predominantly phase noise, see Noise in mixers, oscillators, samplers, and logic: an introduction to cyclostationary noise on www.designers-guide.org/Theory.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
neoflash
Community Fellow
*****
Offline

Mixed-Signal
Designer

Posts: 382

Re: Question about phase noise simulation result
Reply #10 - Jul 1st, 2011, 12:16am
 
Ken,

I read your slides and they are wonderful.

I understand that many oscillators will saturate their outputs in the buffering stages trailing the oscillator core. However, there are quite a few cases that these limiters still introduce AM noise.

Again, appreciate your inputs and slides.

Regards,
Neo

Ken Kundert wrote on Jun 30th, 2011, 10:27pm:
Neo,
   I don't know what the 'beat tone' region is.

If you want to know why the noise produced by oscillators is predominantly phase noise, see Noise in mixers, oscillators, samplers, and logic: an introduction to cyclostationary noise on www.designers-guide.org/Theory.

-Ken

Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2161
Silicon Valley
Re: Question about phase noise simulation result
Reply #11 - Jul 1st, 2011, 2:41pm
 
In an oscillator, amplitude noise is suppressed by the amplitude limiting mechanism of the oscillator, whereas the phase noise is not suppressed. In fact, the phase noise accumulates. That is why close-in phase noise in oscillators always dominates over the amplitude noise.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
neoflash
Community Fellow
*****
Offline

Mixed-Signal
Designer

Posts: 382

Re: Question about phase noise simulation result
Reply #12 - Jul 1st, 2011, 3:18pm
 
What if the oscillator output is buffered with a chain of CMOS inverters supplied with a noisy regulator?

Is that buffer chain going to add AM noise?

- Neo

Ken Kundert wrote on Jul 1st, 2011, 2:41pm:
In an oscillator, amplitude noise is suppressed by the amplitude limiting mechanism of the oscillator, whereas the phase noise is not suppressed. In fact, the phase noise accumulates. That is why close-in phase noise in oscillators always dominates over the amplitude noise.

-Ken

Back to top
 
 
View Profile   IP Logged
YCY
Community Member
***
Offline



Posts: 32

Re: Question about phase noise simulation result
Reply #13 - Jul 1st, 2011, 6:42pm
 
Hi, Ken

Thanks for your suggestion. I've read your paper and slides. They are really helpful.
Yet I still have some questions:

1. Can we use the "modulated" mode and plot PM noise to analyze
the jitter or phase noise of a driven circuit, like frequency divider?
I don't really understand the difference between "jitter" and
''modulated" mode when only PM noise is considered.

2. In the slides page 21 "Removing Cyclostationarity", you say the
"noise at any frequency f1 is uncorrelated with noise at any other
frequency f2 as long as both f1 and f2 are within the passband".
I don't understand why. Since after the mixer the noises at different frequencies have been correlated,
how can they be separated by a bandpass filter?
(I borrow your picture in the paper to express my thoughts.
The redline stands for the passband of the bandpass filter.)


Could you please help me figure out these questions?
Your help is really appreciated.

Regards,

YCY
Back to top
 
 
View Profile   IP Logged
Ken Kundert
Global Moderator
*****
Offline



Posts: 2161
Silicon Valley
Re: Question about phase noise simulation result
Reply #14 - Jul 2nd, 2011, 8:14am
 
Neo,
    First consider the case where you have an oscillator followed by a linear amplifier which contributes white noise to the output. The noise from the amplifier will be added to the noise of the oscillator. In this case, since the noise from the oscillator is large but drops at 20dB per decade of offset frequency, there will be an offset frequency beyond which the noise will be dominated by that of the amplifier. At those frequencies the noise will be unmodulated, meaning that there will be equal amounts AM and PM noise.

Now consider the circuit you described: an oscillator followed by a chain of buffers. Presumably these buffers are acting as limiting amplifiers. Like oscillators, it is the nature of limiting amplifiers to suppress AM noise. Furthermore, since you are passing the signal through limiting amplifiers, you presumably are driving circuits that are boolean in nature (switching mixers or logic). In this case any incidental AM noise would be ignored by the subsequent circuitry. So in the case you describe, you can generally ignore the AM component of the noise.

-Ken
Back to top
 
 
View Profile WWW   IP Logged
Pages: 1 2 3 
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.