how to calculate jitter from Cadence simulation results?

[beabroad]

Now i designed a PLL, and want to calculate the output jitter from Cadence simulation results.
i want to plot the histogram, and calculate the rms jitter.
how can i do that? thank you.

 

[uncle_urfi]

Two methods:

------------------------------------ Detailed explanation

Sample: Use the calculator in Analog Artist to sample the waveform.

On the calculator there is a printvs function. Use this to print out ascii data in a file.

Use the histo() function on matlab to plot the histogram.

Each of these tools have a manual that you can read for more info.
---------------------------------------


Get a tool like the 'Calculator functions' from advICo Microelectronics from Germany that will do this for you automatically.

www.advico.de/german/products/download/Calculator_Tools_Manual.pdf

 

[amarnath]

Two methods:

Sample the waveform and use printvs to print the results to a text file and read that file in Matlab and use the histo() function on it. Use ascii text and experiment.


Get a tools like the 'Calculator functions' from advICo Microelectronics from Germany that will do this for you automatically.

www.advico.de/german/products/download/Calculator_Tools_Manual.pdf

hi can u explain the first method in greater detail.

regards
amarnath

 

[amarnath]

hi can u explain the first method in greater detail.

regards
amarnath

 

[mr_chip]

How do you simulate PLL for jitter?

 

[amarnath]

i did finally calculate jitter.u can do it by following the procedure given in chapter 5 of cadence documentation(oscillators).i did get realistic values ,but how can i plot a histogram.i would like to see a histogram or a plot which is usually seen in most IEEE papers.the cadence tool only gives me the value of the cycle to cycle and long term jitter etc.

regards
amarnath

 

[amarnath]

i did finally calculate jitter.u can do it by following the procedure given in chapter 5 of cadence documentation(oscillators).i did get realistic values ,but how can i plot a histogram.i would like to see a histogram or a plot which is usually seen in most IEEE papers.the cadence tool only gives me the value of the cycle to cycle and long term jitter etc.


please help....any help will be appreciated
regards
amarnath

 

[moisiad]

Another way to get jitter is by the eye diagram. Simply run a transient analysis, and after the PLL is locked apply the eye diagram function (form the calculator in cadence). The wide area in the eye diagram is the peak to peak jitter

 

[amarnath]

hi mosaid how do u calculate the fequency to be entered in the eye diagram calculations,do u just take the average frequency,please elaborate further.i believe u need to take average frequency after the pll has locked.

 

[moisiad]

Hi amarnath

In the start and stop time of the eye diagram function give the time period where the PLL is locked.

In order to define the period do the following: There is a function called "frequency". Apply this to the transient signal (again only in the locked state) and you get the frequency that your PLL is locked. Transform frequency into period and set it now into the eye diagram function.

I have noticed that in order to get correct results you need to define the period with high accuracy (as given by the calculator) and not rounded values.

Regards

 

[neoflash]

HI, beabroad:

I am curious that why you simulate jitter in cadence.

Since transient can not have noise in device normally, the jitter you see in the waveform will be meaningless.

Why you care about that?

 

[flushrat]

In spectre you can use veriloga models to measure jitter.
You can write by yourself or use models from bmslib

 

[beabroad]

HI, beabroad:

I am curious that why you simulate jitter in cadence.

Since transient can not have noise in device normally, the jitter you see in the waveform will be meaningless.

Why you care about that?

i believe there is added noise in transient simulation. otherwise, the eye diagram will be perfect.

 

[xieyan2001]

Usually we use equipment to detect jitter.

It is because the simulation result is different from real result.
Jitter is unpredictable.

 

[amarnath]

i have finally come to the conclusion that jitter cannot be calculated using a simple tran analysis.the jitter results we get are wrong cause we are taking the average of the pll output frequency evenbefore it locks.i believe those of us who have used this method have done a tran analysis and then taken average of the PLL output frequency using the calculator,here is where the mistake lies since the tool will calculate average from time 0 to transient time end.so this will include the time when the pll has still not locked,so naturally we will get good amount of jitter and if everything after this goes wrong ,we finally interpret the results as if they are right.for anybody who has to predict jitter i would advice u to read the chapter on simulating oscillators in the cadence doc.i was able to get proper values which tallly closely with that given in literature using this method.

regards
amarnath

 

[kiat]

how to check the version of my calculator in cadence? cos my calculator does not seem to have the eye diagram function...

 

[flushrat]

Obviously it is wrong if the freq or jitter are calculated from time 0s because of including unlock state. Fortunally, we can calculate transient freq and jitter. So I dont think tran anaysis is wrong for calculating jitter. Another problem is the library doesnt provide noise model for tran anaysis, so we need to add noise.
A mothod is shown as:
"Predicting the Phase Noise and Jitter of PLL-Based Frequency Synthesizers"
can be downloaded from www.designers-guide.org

 

[amarnath]

Obviously it is wrong if the freq or jitter are calculated from time 0s because of including unlock state. Fortunally, we can calculate transient freq and jitter. So I dont think tran anaysis is wrong for calculating jitter. Another problem is the library doesnt provide noise model for tran anaysis, so we need to add noise.
A mothod is shown as:
"Predicting the Phase Noise and Jitter of PLL-Based Frequency Synthesizers"
can be downloaded from www.designers-guide.org

i did not say i calculated jitter from time 0us.i entered the jitter from the time of lock to the end of transient simulation time,here when u the tool asks u to enter the average time period u click the output signal and then calculate the average using the calculator,i said doing this is wrong since this value entered decides the closing of the eye in the eye diagram.so just use the clip command in the calculator and clip the signal from the transient lock to transient end time.then take the average.when i did this for my pll,i got almost zero jitter.but when i calculated jitter using the method given in the spectre manual ,i did get a very correct value for jitter.

regards
amarnath