The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Phase noise analysis of autonomous ring oscillator
https://designers-guide.org/forum/YaBB.pl?num=1263835149

Message started by iep on Jan 18^th, 2010, 9:19am

Title: Phase noise analysis of autonomous ring oscillator
Post by iep on Jan 18^th, 2010, 9:19am

I've been asked to design a simple low freq (100MHz) low phase noise ring oscillator. I've not had to work in osc design before so I'm just getting to grips with the tools.

Most importantly, I would like to be able to use the PSS and PNOISE feaures of the ADE tool to establish the phase noise of my design. I have the analyis running but I've got a couple of questions that perhaps someone can help me with:

I have the PSS and Pnoise windows set up as below:

PSS:

Engine = Harmonic Balance

Beat Frequency = 100MHz (frequency of osc)

Oversampling Factor = 10 (not really sure what to put here)

Number of Harmonics = 10

Accuracy Defdaults = Liberal

tstab = 10u

Oscillator Mode selected

Oscillator node = clock_output

Reference node = gnd

Osc initial condition = default

Pnoise:

PSS Beat Freq = 100MHz

Sweeptype = relative

Relative Harmonic = 0 (I think this represents the fundamental)

Start = 1

Stop = 1G

Sweep Type = Automatic

Maximum sideband = 10

Output = Voltage

Positive Output Node = clock_output

Negative Output Node = gnd

Noise Type = modulated

The resulting output from Results -> Direct Plot -> Mainform

Analysis = pnoise

Function = Phase Noise

is attached.

So, questions:

1. Does this plot look reasonable?

2. Obviously, there is some error in the analysis since the noise extends up to 37dBc. I understand this relates to the Lorentian process. However, is there an easy way to get a more accurate result around 0Hz from carrier or, are there any guidlines in how best to interpret what I have?

Cheers,

iep

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by loose-electron on Jan 18^th, 2010, 10:02am

phase noise of ring oscillators can be difficult to properly analyze - you need to include power/ground noise and cross coupling iussues bvefore anything meaningful comes out.

Doing a simple PSS and phase noise analysis is not going to show you all the second order effects, which are the things that are going to bite you.

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by ACWWong on Jan 18^th, 2010, 12:09pm

Hi Iep,

Just a few comments:

iep wrote on Jan 18^th, 2010, 9:19am:

PSS:

Engine = Harmonic Balance

Since its a ring oscillator (square waves) you would probaby get better results with shooting rather than HB. Using HB is OK, but you would probably need more harmonics... you should re-run the simulation with increased harmonics to see whether the result changes

iep wrote on Jan 18^th, 2010, 9:19am:

Oversampling Factor = 10 (not really sure what to put here)

It is not a good idea to randomly assign values to variables when you do not understand what they do. Consult the SpectreRF manual docs in your installation path if in doubt. In this case the "oversampling factor" is a way to get HB to more accuarate accommodate sharp transistions (like square waves)... anyway i would still recommend shooting.

iep wrote on Jan 18^th, 2010, 9:19am:

Relative Harmonic = 0 (I think this represents the fundamental)

the fundamental (100MHz) is harmonic 1. so your pnoise analysis is relative to DC rather than the oscillation frequency.

I think you should consider going through some SpectreRF examples (which are included in the installation path), to better learn how to simulat what you want.

Best Regards,

aw

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by iep on Jan 19^th, 2010, 2:56am

Thanks for the responses.

Loose-electron, I understand that the phase noise figure I will get from this analysis is not representative of the oscillator on chip. However, I am currently interested in optimising the oscillator's own noise performace (from internal noise sources) before including external factors.

AW, thanks for detailed response. I had tried the HB analysis after finding that the Shooting analysis gave the same issue as I reported in my original post (that the phase noise close in to the carrier sits above 0dBc). I can see that I would need a greater number of harmonics for a truly accrate response but the figure of 10 was suggested within the SpectrRF manual as a good starting point.

Thanks for the comment on the 0 harmonic. I had tried a value of 1 and found that my phase noise close in to carrier was above 0dBc (I had go the epoint where i was grabbing at straws). Some other bits of software I have used refer to the fundamental as the 0th harmonic so I thought I'd give it a shot.

I have now followed the recommended process from the SpectreRF manual to the letter (Periodic Steady State and Phase Noise with PSS and Pnoise) and have tried setting the 'augmented' box in the pnoise menu to 'yes' and 'fast'. Unfortunately I am still left with my original issue.

Can anyone suggest how I can set up the tool to give a phase noise value close in to carrier that is not above 0dBc. Or, alternatively offer any advice on how to interpret the data I currently have (latest plot attached).

Many thanks.

iep

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by pancho_hideboo on Jan 19^th, 2010, 3:19am

See the followings.
http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395
http://www.designers-guide.org/Forum/YaBB.pl?num=1260521995
http://www.designers-guide.org/Forum/YaBB.pl?num=1258339986/42#42

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by iep on Jan 19^th, 2010, 8:05am

Hi Pancho,
Thanks for the links I have reviewed (and re-reviewed these) many times but am still struggling to come up with an approach that allows me to confidently interpret my own results.

However, I think my understanding is improving. Would I be correct to say:

1. The nature of the pnoise tool within Spectre RF is such that it cannot accurately predict phase noise perfromance below the corner frequency of the phase noise ? This is simply because the Lorentzian spectrum models are inaccurate in the presence of flicker noise?

2. Using augmented mode in HB mode will cause the spectrum of the noise to flattened out below the flicker noise limit. In shooting mode it is not flattened out and will still exceed 0dBc as it approaches 0Hz offset from the carrier. In which case, what is the real benefit of augmented mode? Does it improve accuracy above the flciker noise offest from the carrier?

3. There seems to be a broad consensus that the augmented mode does not really improve accuracy below the flicker noise limit of the phase noise?

4. The results I am seeing at frequencies greater than the flicker noise corner are probaly accurate?

5. Other posts have reffered to the Linewidth, does the linewidth simply refer to the frequency offset from the carrier below which flicker noise dominates?

Unfortunately, I am particularly interestd in the phase noise performance close in to the carrier. It apperas these results will be very difficult to establish.

Thanks,

ip

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by pancho_hideboo on Jan 19^th, 2010, 8:29am

iep wrote on Jan 19^th, 2010, 8:05am:

1. The nature of the pnoise tool within Spectre RF is such that it cannot accurately predict phase noise perfromance below the corner frequency of the phase noise ?

You may think so.

iep wrote on Jan 19^th, 2010, 8:05am:

This is simply because the Lorentzian spectrum models are inaccurate in the presence of flicker noise?

No.
It is simply due to limitation of small signal noise analysis in frequency domain.
See http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395/26#26

iep wrote on Jan 19^th, 2010, 8:05am:

2. Using augmented mode in HB mode will cause the spectrum of the noise to flattened out below the flicker noise limit.
In shooting mode it is not flattened out and will still exceed 0dBc as it approaches 0Hz offset from the carrier.
In which case, what is the real benefit of augmented mode?
Does it improve accuracy above the flciker noise offest from the carrier?

Again see http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395/28#28

iep wrote on Jan 19^th, 2010, 8:05am:

3. There seems to be a broad consensus that the augmented mode does not really improve accuracy below the flicker noise limit of the phase noise?

Ideally augmented mode should be helpful for improving accuracy of phase noise for small offset frequency region except for non-physical value region.

Even in Agilent ADSsim, phase noise analysis using HB had to be improved.
Again see http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395/20#20
This improvement for Agilent ADSsim has been completed in 2001 or 2002.

But I think there are still many many bugs in augmented pnoise of Cadence Spectre.

iep wrote on Jan 19^th, 2010, 8:05am:

4. The results I am seeing at frequencies greater than the flicker noise corner are probaly accurate?

It is not accurate result generally.
Again see http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395/26#26

iep wrote on Jan 19^th, 2010, 8:05am:

5. Other posts have reffered to the Linewidth, does the linewidth simply refer to the frequency offset from the carrier below which flicker noise dominates?

I don't know how Cadence Spectre defines "linewidth".

I guess it forces an adaptation of phase noise as "the Lorentzian spectrum form" regardless of flicker noise existence.
Then it defines "linewidth" so that it has an integrated area of one.

But I'm not sure.

See http://www.cadence.com/community/forums/T/12466.aspx
Still there is no answer.

iep wrote on Jan 19^th, 2010, 8:05am:

Unfortunately, I am particularly interestd in the phase noise performance close in to the carrier.
It apperas these results will be very difficult to establish.

Use full-spectrum oscillator device noise analysis(oscnoise) of BDA's Analog FastSPICE RF.

This analysis is fairly different from conventional phase noise analysis based on small signal noise analysis in frequency domain which is major in many vendor's simulators.

Again see http://www.designers-guide.org/Forum/YaBB.pl?num=1258339986/42#42

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by Andrew Beckett on Jan 20^th, 2010, 1:26am

pancho_hideboo wrote on Jan 19^th, 2010, 8:29am:

But I think there are still many many bugs in augmented pnoise of Cadence Spectre.

Yawn. Here we go again. :'(

There was a problem with augmented (mainly in shooting) which were resolved in MMSIM71 (may have been MMSIM711). I have no evidence to say that there is an outstanding problem with it.

The curious results in the thread you reference http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395/0 may well be due to the bug which was fixed - looking back through the thread, it seems that Mayank was not using MMSIM71. I've offered to look at it, but Mayank needs to get clearance to send me the testcase (I think that was the last status).

Hardly "many many".

Useful to see the other approaches to low-offset phase noise being discussed though.

Andrew.

Title: Re: Phase noise analysis of autonomous ring oscillator
Post by Mayank on Jan 20^th, 2010, 4:15am

Hi,

@ iep : You can keep a watch on this thread, http://www.designers-guide.org/Forum/YaBB.pl?num=1050465395/30#31

@ everyone : This thread actually turns out to be a pretty cool one....thanx to Pancho & Andrew...

@ Andrew : Yeah, i was on MMSIM70 when i posted those result.

--
Mayank.