I am simulating PSS+Pnoise on my LC oscillator for jc and jcc jitter measurement. From the phase noise plot I got positive value at low frequency offset, which I think makes the jitter calculation incorrect. I searched and read through some old threads with the same topic. It seems this issue is connected to the MMSIM version, but nobody clarified it. I am likely working with IC61+MMSIM60, although I need to check with the server manager. So is it the case that if I upgrade MMSIM, this issue "positive phase noise" can be solved? Or is there anything else I am missing? Thanks in advance.
This is discussed in Cadence Online Support article 1832229.
Here is the complete link:
Starting around MMSIM 7.2, there is a parameter called "lorentzian". This parameter determines if the Lorentzian plot is used in the oscillator noise analysis. If you want to plot phase noise without the "non physical" results, select lorentzian=yes in the pnoise options.
You really need to be using MMSIM12.1.1 or 13.1 latest ISRs. You're using a simulator version that is 5+ years old and no longer supported. Please have your system administrator download the latest ISR from http://downloads.cadence.com .
Sr. Staff Support AE, Global Customer Support
Cadence Design Systems, Inc.
Note that it's not really wrong - it's a matter of how the curve is interpreted. However, it is important to understand that close to the carrier the small signal assumptions are no longer valid, and there is a frequency (sometimes referred to as "line width" and sometimes as "corner frequency") where that assumption stops being valid. A new parameter on the pnoise (and hbnoise) analysis was introduced in MMSIM711 ISR21 and MMSIM72 ISR3 called "lorentzian" . The default for this is "cornerfreqonly" but it can be set to "yes" in which case it computes the traditional Lorentzian shaped phase noise curve - the default just outputs the corner frequency.
Note I don't believe this impacts the jitter calculations (at least not significantly) when running the pnoise in "jitter" mode.
In reply to Tawna:
Thanks a lot for your reply. I am not the server manager who has the registration account, so I cannot read the content of that article. Could you please paste it here?
I asked my server manager to upgrade MMSIM. He told me now the version is 11.1. But when I check the Pnoise analysis setup window, somehow there is still no "lorentzian" option. How do I check the MMSIM version in my console or Virtuoso? Thanks in advance.
In reply to Andrew Beckett:
Thanks a lot for your explanation. So for the jitter calculation (jc / jcc), I should find the point of the corner frequency (fc) which is generally -3dB from the phase noise plot. Disregard the data on the low frequency less than fc, and set the frequency integration range from fc to fo/2 for the jitter calculation. This will give me a relatively accurate result, am I right? Please correct me if my understanding is wrong. Thanks.
In reply to bjbit:
Are you trying to do the integration yourself, or are you using the "jitter" noisetype on the pnoise form? That's the recommended way of doing this - the integration is harder than you might imagine (particularly ensuring that the frequency range makes sense, and also that the fact that you don't want it to do linear interpolation on a linear frequency range).
I am using "jitter" noisetype and then setting the integration range in jc / jcc plot of "directly plot from...".
OK, that should be fine then. I'd be surprised if it was very sensitive to the start frequency of the integration anyway. Maybe you could experiment and see what happens?
Here I attach my phase noise plot and the jc measurements with different start frequencies of the integration. Actually the measurements of single cycle (k=1) are similar around 47fs. But when I measure the long-term jitter, you can see from the attachment that the variation is large. 135Hz is the -3dB point. My beat frequency is 800MHz, so k=400M means 0.5 sec. Any thoughts about this issue? Thanks.
Andrew and I have conferred on your issue. Given that this is going to require some more detailed investigation and checking with supported releases, it would be best if you contact Customer Support for assistance.
Also, the Support AE will be able to email you the contents of the Article that I mentioned.
(we don't typically paste the contents of Articles from Cadence Online Support into the Cadence Community).
Thanks a lot for the reply. The server manager is trying to figure out if he installed MMSIM11.1 correctly. We will contact customer support if he think it is needed.
By the way, the reason that "lorentzian" doesn't show up on the pnoise options form is that you're using too old an IC version. I can tell from the graph appearance that you're definitely using IC614 or earlier. You didn't say the precise version but the fact that you were also using MMSIM60 suggests to me that it was a very old version.
You really should update to something newer. You could just type in lorentzian=yes in the additional parameters field on the pnoise options form, but it's highly likely that there are other fixes that could be important too - so my recommendation would be to update to IC616 (latest hotfix) first - and ideally MMSIM131 - before contacting customer support if you still have an issue.
You are right. It is IC614 I am using. My server manager showed me the content of the Cadence setup file, and it does show MMSIM version is 11.1, although he is not 100% sure it works correctly. So probably it is the old version of IC which fails to be compatible to the new "lorentzian" option? I will let him know your recommendation and try to update to a newer version. Thanks a lot for your information and time.
Sorry I bring this topic back. I finally got IC615+MMSIM11.1 which is the latest version my server manager can get from his current license. Now I have the "lorentzian" option to use, and got the phase noise and jitter plot as shown in the attachment. The red curve, related to the left Y axis, is plotted from "directly plot" -> "pnoise" -> "phase noise", which flattens the "non physical" data in low offset frequency. The green curve, related to the right Y axis, is plotted from "directly plot" -> "pnoise jitter" -> "phase noise", and it seems the "lorentzian" method is disabled here. Then I get the same issue on the jitter calculation (jc). For only one cycle, the jitter results from different start integration frequency is not much. But with a long term, the jitter calculation gives very different results.
Please let me know if I need to submit a support request. It takes time since I have to go through my server manager for the account. Thanks a lot for your time.
Please contact customer support. Sorry if that's more work for you - I'm pretty tied up this week.