The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Measurements >> Phase Noise and Jitter Measurements >> Jitter problem
https://designers-guide.org/forum/YaBB.pl?num=1214222160

Message started by acezhy on Jun 23rd, 2008, 4:55am

Title: Jitter problem
Post by acezhy on Jun 23rd, 2008, 4:55am

Hi all,

For CML buffer working at rather low frequency, and considering two cascaded buffer, the jitter of the preceding block is independent of the succeeding block. But for the same cascaded CML buffer operating at a high frequency (the signal becomes a continuous one and no high/low level achievable), does the same rule valid?

Recently, I simulated the cascaded CML buffer with pnoise jitter method operating at high frequency, and by changing the parameter of the last stage, and checking the noise contribution table, I found more noise power from the preceding component becomes larger. It seems that the jitter of the preceding stage is amplified? Anyone could help explaining this case?

Thank you all!

Title: Re: Jitter problem
Post by Stefan on Jun 23rd, 2008, 6:52am

That might be, because amplitude noise becomes more important due to the not-so-steep slope...
Check your circuit by isolating amplitude/phase noise contribution, maybe this gives a hint to you ...

Title: Re: Jitter problem
Post by acezhy on Jun 23rd, 2008, 7:28am

Stefan,

Thanks for your suggestion!  ;)
But what I have done is the time domain noise, in which the AM noise should not be considered. And the power supply is ideal.
And I still fail to figure out how the succeeding stage could affect the previous stage's noise contribution. I thought the subsequenct stage could only affect the transfer function of the previous stage noise sources instead of its value. Where am I wrong?

Title: Re: Jitter problem
Post by thechopper on Jun 23rd, 2008, 6:00pm

Hi acezh,

Could you please be more specific as to what parameters you changed on your second stage? May be you are dealing with a loading effect on the gain of your first stage which at the end of the day might affect the noise contribution from it.

Regards
Tosei

Title: Re: Jitter problem
Post by acezhy on Jun 24th, 2008, 5:26am

Hi Tosei,

Let's see a more special case. The first stage is a CML buffer, the second stage is a CML2CMOS converter and the third stage is a CMOS buffer. When I change the size of the CMOS buffer I found the noise power from the first stage increase obviously. Does this phenomenon reasonable?

Thank you!

Title: Re: Jitter problem
Post by thechopper on Jun 24th, 2008, 7:14pm

Hi

At first sight it does not seem right to me. However, from you description of the high frequency behavior of your circuit I guess you might be dealing with some kind of noise integration, this noise being generated in the first stage. As you increase the size of your CMOS buffer you might get more capacitance for integrating such noise given you are driving the circuit a very high frequency. Therefore the effective noise contribution from the firs stage might be more noticeable as you either: go faster or integrate more such noise.
This is just a guess since I´m not sure if I´m picturing the circuit correctly.

Hope this helps
Regards
Tosei

Title: Re: Jitter problem
Post by acezhy on Jun 25th, 2008, 4:59am

Hi Tosei,

Thanks for your advice.
Another problem, what does the noise power given by 'noise summary' stands for? Has the noise source power been shaped by transfer function (such as noise integration) or not?
For example, if I do a pnoise (time domain) simulation, I found I got a smaller jitter value while the total noise power given by 'noise summary' table larger.

Thanks again!

Title: Re: Jitter problem
Post by thechopper on Jun 25th, 2008, 8:09pm

You can ask for spot or integrated noise power, and input or output referred. In any case the transfer function for the noise is evaluated and you could even see how the noise is shaped by such function if you plot it.

Hope this answers what you were asking
Tosei

Title: Re: Jitter problem
Post by acezhy on Jun 26th, 2008, 6:19am

Hi Tosei,

Sorry, my statement is not clear enough.

What I asked is that whether the component noise power shown in 'noise summary' table has been shaped by noise transfer function of each component to the output or has not been shaped by the transfer function?
Because sometimes I got a lower jitter but have a larger noise power in noise summary table. I don't think this reasonable.

By the way, you said that the transfer function can be ploted. How can I plot this? By pac simulation? Or has a more easier way?

I am very appreciated for your answers!

Title: Re: Jitter problem
Post by thechopper on Jun 26th, 2008, 7:09pm

In this moment I do not recall exactly whether the noise summary lists the input referred or output referred noise. If it lists the output referred noise the it is definitely shaped by the transfer function.
One easy way to check whether it is input or output referred is by taking the first - say - 10 biggest contributors and sum them quadratically. If that calculation gives a number close to the total output referred noise, which is provided at the end of the list, the it is output referred. Otherwise it should be input referred.

Using the Direct plot menu you can plot the output noise, which was already shaped by the transfer function.

Finally, it might not be crazy to me to eventually get lower jitter while larger noise power from the noise summary. While the noise summary is a result of a small signal analysis, the jitter is a time domain phenomenon. As the slope of your input signal is larger (for the same noise power generated in your circuit) you will get smaller jitter. The jitter is inversely proportional to the input signal slope and obviously proportional to the sigma of the noise.

Hope this helps
Tosei

Title: Re: Jitter problem
Post by acezhy on Jun 26th, 2008, 9:46pm


HdrChopper wrote on Jun 25th, 2008, 8:09pm:
In any case the transfer function for the noise is evaluated and you could even see how the noise is shaped by such function if you plot it.



Hi Tosei,

Thank you very much. You've done me a big favor!

Could you please tell how to plot the transfer function?




Title: Re: Jitter problem
Post by thechopper on Jun 27th, 2008, 7:40pm

Hi,

May be I was not clear enough: with the noise analysis you cannot plot the transfer function although if you plot input referred noise and output referred noise, you can get an idea of how the transfer function looks like.

If you want the transfer function you should run a pxf analysis.

I´m glad I could help
Tosei

Title: Re: Jitter problem
Post by acezhy on Jun 28th, 2008, 3:51am

Hi,

Thanks for your patient answer! It helps me so much!

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.