The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Period Jitter Skill in Cadence Calculator
https://designers-guide.org/forum/YaBB.pl?num=1262939836

Message started by Mayank on Jan 8^th, 2010, 12:37am

Title: Period Jitter Skill in Cadence Calculator
Post by Mayank on Jan 8^th, 2010, 12:37am

Hi all,
I am a bit confused about what formula does the cadence calculator uses in it's inbuilt skill function for period jitter.

Acc. to me,
If periods(1:1000) denotes 1000 samples of Time Periods of a jittery waveform measured over 1000 cycles, then

jitter_zeromean = periods - mean(periods);
Single_period_jitter = sqrt ( mean ( jitter_zeromean .^2 ) );

But, this formula shows different results than cadence's inbuilt period jitter skill.

Can somebody pls explain the formula used by this skill OR if something'swrong in my formula to calculate period jitter ??.

Also, for adjacent cycle period jitter, what should be the bin size in this function ??

--
mayank.

Title: Re: Period Jitter Skill in Cadence Calculator
Post by Andrew Beckett on Jan 18^th, 2010, 7:03am

Hi Mayank,

I've had a look in the code, and I see an error in the calculation. I'm assuming you're running something like:

Code:

period_jitter(v("op" ?result "tran-tran") "rising" ?mode "auto" ?binSize 0 ?xName "time" ?outputType "sd" )

In other words, outputType as "sd" and binSize as 0 - this seems to be the problematic case. There's a sign wrong in the calculation of the mean, and so it ends up adding the mean rather than subtracting it from each period.

I did some searching, and CCR 731797 is covering the same issue, and is fixed in IC614 ISR1 (due out very shortly). Having quickly reviewed the code for the upcoming IC614 USR1, I see this sign has been corrected.

Regards,

Andrew.

Title: Re: Period Jitter Skill in Cadence Calculator
Post by Mayank on Jan 18^th, 2010, 9:54am

Hello Andrew,
I tried matching my results from matlab post-processing to cadence calculator's period jitter skill...

i found out that period jitter skill in cadence calculator uses the same formula
Quote:

jitter_zeromean = periods - mean(periods);

but the mean is a moving window average whose window size is equal to binsize...

i.e. if T1, T2, T3,... denote consecutive periods of the waveform, & random variable X denotes period jitter,
then X1 = T2 - (T1+T2)/2; X2 = T3 - (T2+T3)/2; X3 = T4 - (T4+T3)/2 and so on....

Am i correct ??

--
Mayank.

Title: Re: Period Jitter Skill in Cadence Calculator
Post by Andrew Beckett on Jan 18^th, 2010, 10:03am

Hi Mayank,

That's correct. If the binSize is greater than 0, the answer is as you say.

It does say this in the documentation by the way:

Quote:

Bin Size is the width of the moving average window, The deviation of value at the paricular point from the average of this window is the jitter.
If binsize=0, all frequencies are used to calculate the average.
If binsize=N, the last N frequencies are used to calculate the average.

Note, it is being updated to say "periods" rather than "frequencies" as it's a bit confusing otherwise.

The problem is that with binSize of 0, the answer is wrong, because of the mistake in the calculation of the overall mean period.

Regards,

Andrew.

Title: Re: Period Jitter Skill in Cadence Calculator
Post by Mayank on Jan 18^th, 2010, 8:50pm

Thank you Sir,
That explains it all... :-)
& yeah 1 more question.....If we do PLL Macro Model Simulation, And use Direct Plot Form's Plot PLL Noise PSD, It generates a PLL Phase Noise plot, but it also indicates a value of center frequency & Jitter alongside.
What Jitter metric is this ??
----> Single Period Jitter calculated from time domain waveform of PLL o/p which goes to freq. meter (with bin size 2) ??
----> Or is it calculated from frequency domain phase noise plot that PLL PSD option generates ??
If it is calculated from frequency domain, which conversion formula does it use ?? -- Zanchi / Demir / Frank Wiedmann's / Allan Variance / or other ??

thanx,
Mayank.

Title: Re: Period Jitter Skill in Cadence Calculator
Post by Andrew Beckett on Jan 19^th, 2010, 6:25am

Mayank,

I'm afraid I don't have the time right now to go through the flow and do the experiments and checks to see precisely what it does, but to a similar question a few days ago, R&D said:

Quote:

At present, the PLL Noise-Aware flow provides only cycle-cycle and phase jitter

Not sure if that's really precise enough for what you're asking. If I have time in the next few days, I'll try it out and look to see exactly what it is doing (I think my quoted statement above doesn't tell you much more than it says in the PLL.pdf in the <mmsimDir>/tools/spectre/examples/SpectreRF_workshop dir

Regards,

Andrew.