Jitter cacluation from phase noise??

Choosing the frequency range for integration is dependent on your application and what spec you're trying to meet - it depends on the bandwidth of the system. For a driven system (or when using PMjitter mode for autonomous) you should integrate beyond half the PSS fundamental because you would double count the noise (due to the ideal sampler included in the simulation).

In your other questions, question 1 - the number of harmonics in the PSS solution. When using shooting, this is not so important, because the PSS solution contains enough data to represent the solution for at least 40 harmonics by default (or 4 times the highest harmonic asked for if that is greater). Only if you want to include noise from high multiples of the PSS fundamental would you need to worry about this (and even then using the maxacfreq option on the options form could be an alternative way to handle this). With harmonic balance however, you have to include enough harmonics to represent the solution accurately, and also to include all the harmonics around which there will be significant noise contributions. It certainly does not (in either case) mean that fewer harmonics will  include more noise.

For the second question, it's the frequency range that you wish to analyze the noise over. The curves will be plotted from A to B where A and B are the limits of your sweep. The bigger the range, the more calculations the simulator has to do.

For sidebands, you need to set this high enough to include all the significant noise contributions. It will only include the noise contributions from sidebands around multiples of the PSS fundamental, up to the limit that you give. Any higher frequency noise contributors will be ignored (so they won't get folded in). Essentially the simulator needs to compute the noise at source at each of the sidebands, and then compute the transfer function from that frequency to the output frequency, and then sum up the noise powers at the output. So it has to have some idea over what range of sidebands to include the noise.

Maybe reading this paper Introduction to RF simulation and its application might help?

Regards,

Andrew.

You might also want to take a look at http://www.designers-guide.org/Forum/YaBB.pl?num=1224609785 .

Hello Andrew,

Thank you for your detailed reply.

It made me clarify the concept of PSS and PNOISE simulation, especially the meaning of number of harmonics, maximum sidebands, pnoise frequency sweep range.

Well, still I'm confused about phase noise integration range to extract jitter.

Assume one designs a free running VCO and this VCO drives ADC's sample & hold circuit.

Output frequency of VCO is 10MHz and ADC's maximum input frequency is 100kHz.

In my opinion, the maximum offset frequency(upper limit of phase noise integration) of VCO which affect ADC's input is 100kHz.

However, in ADI's application note MT-008(Converting Oscillator Phase Noise to Time Jitter), Walt said that phase noise integration has to be performed up to twice of sampling frequency. From Walt's note, maximum offset frequency is 20MHz.

The simulation results are quite different between above methods.

I'm not sure which is correct.

Can you give me some hints?

Thank you.

Hello Frank,

Thank you for your reply.

I read suggested posts and found your answer.(http://www.designers-guide.org/Forum/YaBB.pl?num=1092399689/22#22  reply #20)

In this post, you mean the frequency sweep range of pnoise simulation for driven circuit is 0Hz(or very low freq.)~Fout/2. Am I right?

Maybe, it'll be same in autonomous circuit(e.g. VCO).

If so, how can I choose phase noise integration range?

As you replied to the post "http://www.designers-guide.org/Forum/YaBB.pl?num=1092399689 reply #7", I found following comment in your suggested document.

"Determine c by choosing Δf well above the corner frequency ( fc) to avoid ambiguity and well below 'fo' to avoid the noise from other sources that occur at these frequencies."

Is that means phase noise integration range is from at least 'fc' to at most 'fo'?

From above comment, phase noise by flicker noise (1/f^3 band) does not affect as jitter. Am I right?

Can you explain with an example like my previous reply?

When output frequency of VCO is 10MHz and maximum input frequency of ADC is 100kHz, how should I decide phase noise integration range?

Following your suggested document, if in the simulation result, flicker corner frequency 'fc' were 100kHz, the phase noise integration range has to be 100kHz~10MHz. Am I right??

In pnoise jitter analysis, you always must integrate from 0 Hz to f0/2, where f0 is the fundamental frequency (or "beat frequency") of the pss analysis. This is also true for autonomous circuits. However, the long-term jitter of autonomous circuits is infinite. Only the k-cycle jitter Jc is meaningful for an autonomous circuit (see http://www.designers-guide.org/Forum/YaBB.pl?num=1224609785/9#9). Make sure to select PM jitter (not FM jitter) in your pnoise jitter simulation setup if you have an autonomous circuit.

The value of k that you should use for Jc depends on your application. It could for example be the maximum number of consecutive samples that you want to acquire with your ADC.

The frequencies fc and fo have nothing to do with integration. They simply specify the range that you should use to determine the parameter c of the equation.

I really appreciate your reply, Frank.

Your reply was been so helpful to understand the jitter measuring method.

In conclusion, when I want to know accumulation jitter of free running VCO for driving ADC, I should do following steps.

1. Pnoise analysis with frequency sweep from 0Hz to f0/2 should be performed.

2. Maximum sideband of pnoise analysis has to be large enough for accurate result.(at least 10)

3. Find fc to calculate c. The fc(delta f) is cut off frequency of phase noise(separation point between -30dB/dec and -20dB/dec) 

4. Caculate c and J. >> J=(c/fo)^0.5

However, I still have some questions about your above comments, Frank.

(a) "Only the k-cycle jitter Jc is meaningful for an autonomous circuit"

>> Cycle jitter means the timing difference between ideal clock and actual output, Cycle-to-cycle jitter means the timing difference between previous cycle and next cycle. Then VCO is autonomous circuit and has no ideal clock as reference. Why the k-cycle jitter Jc is meaningful for an autonomous circuit? From the definition of Jc and Jcc, the meaningful jitter has to be Jcc.  

(b) "The value of k that you should use for Jc depends on your application. It could for example be the maximum number of consecutive samples that you want to acquire with your ADC."

>> If all internal blocks of ADC(e.g. clockgen, S/H..) are drived by VCO, k the number of cycle is 1, isn't it?

(c) "The frequencies fc and fo have nothing to do with integration. They simply specify the range that you should use to determine the parameter c of the equation."

>> If the frequency fc is 100kHz, selecting delta f 1MHz or 10MHz or any other frequency lower than fo does not affect the result?

Following is another question.

What I really want to know is the method to calculate jitter by spectreRF.

For this, I orginized a simulation flow.

>> pnoise simulation >> direct plot >> main form >> select Jcc >> define number of cycle >> define integration limits from fc to fo/2 >> plot >> Jcc will be calculated.

Here, the phase noise under fc(Flicker noise area) is not contributed to accumulating jitter and selecting lower limitation of integration range has to do not affect the result.

Am I right?

I appreciate your help. Your replies were really helpful.

Unfortunately, I still have some questions. Can you give me more answers?

1. From your suggested documents, meaningful jitter for autonomous circuit is Jc.

However in 10page "Jitter measurements using SpectreRF application note" by Cadence, Jc is timing difference between ideal reference  and output.

There is no reference OSC in case of VCO. From definition the meaningful jitter of VCO is Jss, isn't it?

2. And you said that one should make jitter property as PM jitter not FM jitter for autonomous circuit.

However I can't do it. It is selected automatically. How can I change the jitter property in analysis setup.

3. If all blocks in ADC are drived by VCO, how can I choose k the number of cycle? If I choose k higher than 1, the accumulated jitter increase infinitely.

4. What I really want to know is the method to calculate jitter by spectreRF.
Therefore I orginized a flow as below.
>> Direct plot >> Main form >> pnoise jitter >> jc or jcc >> define integration range >> plot

However there is a problem, you said the frequencies fc and fo do not affect the integration.

But the calculated jitter value is varied with the integration range.

Is there no correct method to calculate jitter by spectreRF?

I appreciate your help. Your replies were really helpful.

Unfortunately, I still have some questions. Can you give me more answers?

1. From your suggested documents, meaningful jitter for autonomous circuit is Jc.

However in 10page "Jitter measurements using SpectreRF application note" by Cadence, Jc is timing difference between ideal reference  and output.

There is no reference OSC in case of VCO. From definition the meaningful jitter of VCO is Jss, isn't it?

2. And you said that one should make jitter property as PM jitter not FM jitter for autonomous circuit.

However I can't do it. It is selected automatically. How can I change the jitter property in analysis setup.

3. If all blocks in ADC are drived by VCO, how can I choose k the number of cycle? If I choose k higher than 1, the accumulated jitter increase infinitely.

4. What I really want to know is the method to calculate jitter by spectreRF.
Therefore I orginized a flow as below.
>> Direct plot >> Main form >> pnoise jitter >> jc or jcc >> define integration range >> plot

However there is a problem, you said the frequencies fc and fo do not affect the integration.

But the calculated jitter value is varied with the integration range.

Is there no correct method to calculate jitter by spectreRF?

Because posting did not work for him at the time (his posts from January 30 appeared here only yesterday), Chadong sent me his questions via the contact form on my website. For the record, here are the answers I gave him:

1. From your suggested documents, meaningful jitter for autonomous circuit is Jc. However in 10page "Jitter measurements using SpectreRF application note" by Cadence, Jc is timing difference between ideal reference and output. There is no reference OSC in case of VCO. From definition the meaningful jitter of VCO is Jss, isn't it?

Jc tells you how much the transitions vary from their ideal positions, so I think that Jc is relevant in most applications. Perhaps in your application Jcc is more relevant, I don't know.

2. And you said that one should make jitter property as PM jitter not FM jitter for autonomous circuit. However I can't do it. It is selected automatically. How can I change the jitter property in analysis setup.

This feature is only available in more recent versions, see http://www.designers-guide.org/Forum/YaBB.pl?num=1158924743/4#4.

3. If all blocks in ADC are drived by VCO, how can I choose k the number of cycle? If I choose k higher than 1, the accumulated jitter increase infinitely.

I have already answered this question. It depends on your application and could for example be the maximum number of consecutive samples that you want to acquire.

4. What I really want to know is the method to calculate jitter by spectreRF.Therefore I orginized a flow as below.>> Direct plot >> Main form >> pnoise jitter >> jc or jcc >> define integration range >> plot. However there is a problem, you said the frequencies fc and fo do not affect the integration. But the calculated jitter value is varied too much with the integration range. Is there no correct method to calculate jitter by spectreRF?

I did not say the frequencies fc and fo do not affect the integration. I said that the frequencies fc and fo mentioned in the paper by Ken Kundert do not specify an integration range. As I already mentioned, the integration range in pnoise jitter analysis should always be from 0 to half of the fundamental frequency of the pss analysis.

I just found above pile of my questions which approved lately.

I really appreciate Frank's reply again.

Thank you.