Designer’s Guide Community

EL

New Member

Offline

Posts: 4

Supply noise to phase noise at a Multi-Modulus divider output
Oct 31^st, 2013, 12:33pm

I have a question regarding a problem I am facing in a simulation.
I apologize in advance if the post is too long.

Background:
I am trying to simulate the supply-to-phase noise conversion in a digital frequency divider (multi modulus frequency divider). On the supply VDC I set a noise file (one column frequency vector the second column noise in V^2/Hz). The frequency divider is set to divide by 43. My target is to simulate how much of the noise applied on the supply is converted into phase noise at the divider output.

Simulation setup (pss)
The input clock to the divider is at 5GHz

The pss is set as follows:
-beat frequency:5G/43
-Number of harmonics:131
-errpreset: moderate (I tried conservative also)
-Stabilization time:80ns

My starting target was to find the transfer function from supply noise to output phase noise in the MMD divider. To have this I used a pss-pac analysis and look at the conversion from AM noise (around harmonic 0) on the supply to PM noise (around harmonic 1) at the divider output. My input noise profile multiplied by the Transfer function and scaled to the fundamental amplitude derived with pss-pac should give the phase noise at the divider output in dBc/Hz.

My question is:

I know from Ken paper that I should not use the pnoise for supply noise. Is this valid even if I switch to sources=modulated and I plot the PM component at the output?

If this is not the case, why I get a different result between the noise estimated with pss-pnoise, source modulated and PM plot respect to the method of using pss-pac analysis to get the TF from supply to output (plotting the AM around harmonic zero to PM around harmonic 1 and scaling it to the fundamental level)?

An extra question regards a warning which I really cannot understand in the pss-pac analysis
The way I set the pac is:
PSS beat=5G/43
relative harmonic=0 (because my noise is applied on the supply and therefore at baseband. Please correct me if I am wrong)
Span is 1k to 10M
In specialized analysis :
Input type (SSB/AM/PM)
Output modulated harmonic=1 (I want the PM noise around the 5G/43 component)
Input modulated harmonic=0 (because the supply is modulated at baseband)

The warning I get are those two:
The 'relharm' value specified for the first analysis of the modulated PAC/PXF simulation is not a positive value
The most worrisome though is the second one:
The 'relharm' value specified for the second analysis of the modulated PAC/PXF simulation is not a negative value. As a consequence, the calculation of the modulated PAC/PXF analysis will be wrong.

Now I cannot find any issue with my settings
but being the first time I use the specialized analysis I may be wrong.

If any body can suggest what is wrong with my simulation/setup or does have experience on simulating the supply noise to phase noise conversion in frequency dividers I would really appreciate his/her feedback.

thanks

EL

Back to top

IP Logged

tm123

Community Member

Offline

Posts: 67
Chicago, IL

Re: Supply noise to phase noise at a Multi-Modulus divider output
Reply #1 - Nov 1^st, 2013, 7:06am

Hi EL,

This is an interesting question. I think one important thing to realize is that the phase noise is only relevant at the zero crossing(s) of the output waveform, assuming the divider drives another edge-sensitive circuit such as a PFD using D FF's. Given that, using PSS-PAC to find the transfer function does not seem right to me. To find divider phase noise I use PNOISE type=timedomain to find the noise exactly at the output waveform zero crossing, then convert that to phase noise using the formula in Ken's paper. If you apply your supply noise vector, you should see some contribution due to that supply noise and perhaps you can infer the transfer function from that result. It is not the most direct way of finding the result you are looking for, maybe others will have different suggestions.

Hope this helps.

Tim

Back to top

IP Logged

EL New Member Offline Posts: 4	Re: Supply noise to phase noise at a Multi-Modulus divider output Reply #2 - Nov 1^st, 2013, 10:23am Hi Tim, i'll give a try to that and compare to maybe the results I have fro the other simulations testbenches. Maybe I can figure this out.
Back to top	IP Logged

EL

New Member

Offline

Posts: 4

Re: Supply noise to phase noise at a Multi-Modulus divider output
Reply #3 - Nov 1^st, 2013, 2:32pm

Hi Tim,

one more question regarding the strobed noise analysis. Is the spectral density at the output (Sfi(f)) in dBc or it needs to be scaled by the power at the fundamental frequency?

The question arises from the fact that I want to compare how the supply noise stands with respect to the MMD output phase noise which is typically expressed in dBc.

My take is that that power density needs to be scaled by the fundamental power but maybe given your experience on this method you can confirm if I am right or wrong.

thanks in advance

EL

Back to top

IP Logged

Frank Wiedmann Community Fellow Offline Posts: 678 Munich, Germany	Re: Supply noise to phase noise at a Multi-Modulus divider output Reply #4 - Nov 4^th, 2013, 12:44am I suggest that you use sampled pxf analysis (see http://www.designers-guide.org/Forum/YaBB.pl?num=1224609785/15).
Back to top	IP Logged

EL New Member Offline Posts: 4	Re: Supply noise to phase noise at a Multi-Modulus divider output Reply #5 - Nov 12^th, 2013, 2:47pm Thank you Frank
Back to top	IP Logged

tm123 Community Member Offline Posts: 67 Chicago, IL	Re: Supply noise to phase noise at a Multi-Modulus divider output Reply #6 - Nov 20^th, 2013, 6:41am EL, I believe the strobed noise analysis result is referenced to the fundamental.
Back to top	IP Logged