Designer’s Guide Community

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Sampled noise vs. unsampled noise
Feb 04^th, 2008, 2:05am

Hi Guys,

There is an example that demos the simulation of the switched-capacitor filter. It measure the unsampled noise and sampleds noise with pnoise analysis shown below.

// Measure noise
unsmpldNoise (out gnd) pnoise start=100 stop=25kHz maxsideband=200
smpldNoise (out gnd) pnoise start=0_Hz stop=0.5*CLOCK_FREQ maxsideband=200 \
noisetype=timedomain noisetimepoints=[0] numberofpoints=1

I don't understand the reason why the second pnoise analysis measures sampled noise while the first measures continuous noise. Does the pnoise analysis named smpldNoise calculate PSD at time 0 as defined by noisetimepoints=[0]?
Is the parameter numberofpoints that make smpldNoise calculate PSD at one time point only? That means sampled noise. right?

BTW, how does pnoise analysis calculate time domain noise?

Thanks
Yawei

Back to top

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: Sampled noise vs. unsampled noise
Reply #1 - Feb 4^th, 2008, 4:49am

ywguo wrote on Feb 4^th, 2008, 2:05am:

I don't understand the reason why the second pnoise analysis measures sampled noise while the first measures continuous noise. Does the pnoise analysis named smpldNoise calculate PSD at time 0 as defined by noisetimepoints=[0]?
Is the parameter numberofpoints that make smpldNoise calculate PSD at one time point only? That means sampled noise. right?

It's the noisetype=timedomain parameter that specifies strobed noise. If you are interested in strobed noise at specific points, I recommend setting numberofpoints=0 and specifying the points with the noisetimepoints parameter. Alternatively, you can use noisetype=pmjitter with the thresholdvalue and crossingdirection parameters to specify the timepoints. For details, type "spectre -h pnoise" (without the quotes) at the Unix prompt.

Back to top

IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: Sampled noise vs. unsampled noise
Reply #2 - Feb 4^th, 2008, 10:39pm

Hi Frank,

English is not my native language. Please forgive me if I ask you what strobed noise is. I am not very clear. Does strobed noise mean the transient noise at a specified time point?

If my guess about strobed noise is right, why do you recommend to set numberofpoints=0 and specifying the points with the noisetimepoints parameter? The definition for the above two parameters are shown below, which I get with the command "spectre -h pnoise".

26 noisetimepoints=[...]
Additional time points for time-domain noise analysis.
27 numberofpoints=5 Number of time points of interest in the period where to calculate time domain PSD. Simulator
divides the period evenly into N segments (N=numberofpoints) and calculates time domain PSD
on the starting time point of each segment. When < 0, the parameter is ignored.

I think that it is reasonable to set numberofpoints=1. If it's set to 0, does the simulator measure noise at any point?

Is there any books or literature explaining the principle, syntax, and usage of spectre? It is not so easy with the online help "spectre -help". :(

Thank you, Frank.
Yawei

Back to top

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: Sampled noise vs. unsampled noise
Reply #3 - Feb 4^th, 2008, 11:30pm

ywguo wrote on Feb 4^th, 2008, 10:39pm:

Hi Frank,

English is not my native language. Please forgive me if I ask you what strobed noise is. I am not very clear. Does strobed noise mean the transient noise at a specified time point?

Strobed noise, sampled noise and timedomain noise are different names for the same type of analysis. Historically, this analysis was first called timedomain noise. Now, ADE says "timedomain: strobed noise analysis" when you select Noise Type: timedomain". Ken calls it sampled noise in http://www.designers-guide.org/Analysis/sc-filters.pdf.

Quote:

If my guess about strobed noise is right, why do you recommend to set numberofpoints=0 and specifying the points with the noisetimepoints parameter? The definition for the above two parameters are shown below, which I get with the command "spectre -h pnoise".

26 noisetimepoints=[...]
Additional time points for time-domain noise analysis.
27 numberofpoints=5 Number of time points of interest in the period where to calculate time domain PSD. Simulator
divides the period evenly into N segments (N=numberofpoints) and calculates time domain PSD
on the starting time point of each segment. When < 0, the parameter is ignored.

I think that it is reasonable to set numberofpoints=1. If it's set to 0, does the simulator measure noise at any point?

It will measure the noise at exactly the points you specified with the noisetimepoints parameter (just try it). The numberofpoints parameter will calculate the noise at additional evenly-spaced points starting at the beginning of the period.

Quote:

Is there any books or literature explaining the principle, syntax, and usage of spectre? It is not so easy with the online help "spectre -help". :(

I don't know of any documentation other than the one from Cadence that comes with SpectreRF (for example http://sourcelink.cadence.com/docs/files/Release_Info/Docs/spectreRFvol2/spectre...) and Ken's paper mentioned above.

Back to top

IP Logged