The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Spectre-VerilogA and Calculator_psd function
https://designers-guide.org/forum/YaBB.pl?num=1136912606

Message started by ppg on Jan 10^th, 2006, 9:12am

Title: Spectre-VerilogA and Calculator_psd function
Post by ppg on Jan 10^th, 2006, 9:12am

I am running Fract-N PLL simulations at Behavioural level as per LISTING18 in Ken Kundert paper "Predicting the phase noise and jitter of PLL-based frequency synthesizers". A "Sigma_Delta.va" block is included controlling the VCO_with_divider.va block. However I do not have access to Matlab inside my unix environment. So I am unable to display PLL spurs / quantization noise using the proposed MatLab script Listing 17. Any hints for :
1. Why is the psd function in Cadence ADE / Wavescan calculator not used instead of Listing 17?
2. How one should best use such psd function with Spectre results? For instance what Spectre settings (like : strobe period, reltol, ... ) do provide suited data for such psd function in order to improve its noise floor and resolution?

So far I indeed experienced bad noise floor and resolution using the Cadence ADE / Wavescan calculator psd function.

Thank You for any recommendation herewith.

Title: Re: Spectre-VerilogA and Calculator_psd function
Post by svensl on Jan 13^th, 2006, 3:50am

ppg,

This link might help answer some of your questions regarding spectral characterization.

http://www.designers-guide.org/Forum/?board=ms_design;action=display;num=1118555245

Regards,
sven

Title: Re: Spectre-VerilogA and Calculator_psd function
Post by ppg on Jan 13^th, 2006, 12:21pm

Hi svensl,

Thank You for providing me direction to the referred link. My present issue is indeed similar to the one presented. Still in my case the FFt / dft calculator function can not be used, instead the psd calculator function must be used because one does not know the exact periodicity of fract-N PLL signals. Similarly to the link info, I was already pursuing as follows.

The output signal for which the psd is observed is the VCO phase deviation (vs. the target 0 ppm value freq0) :
phdev = 2*`M_PI*idt(freq-freq0, 0.0).
I am looking at the fract-N PLL phase noise = psd(phdev) :
- from 1 KHz offset frequency => tran Stop Time = (1024*4 + 1024*1024)*2*5ns = 10526.72us. The first chunk of periods passes over transients which dies out already after 20us.
- up to 100 MHz offset frequency => strobeperiod=5n.

Phase noise is given by psd Calculator function :
dB10(psd(VT("/phdev"),40.96us,10526.72us,1024*2*1024,
?windowName "Hanning",?smooth 1,?windowSize 1024*2*256,
?detrending "Linear",?cohGain 1))
I am similarly watching at the better known Sigma Delta modulator psd output.

Results obtained so far does not get any decent noise floors even when one pushes the "Simulation - Options - Analog" down to say : reltol=5e-6, vabstol=1e-8 and iabstol=1e-14(?). Pressing reltol seems to be the most important(?). Spectre simulation (errpreset = conservative) with all VerilogA blocks runs then for 10m 23.2s >:(. Spectrums stay quiet blurred and are not as nice as in our MatLab results (available only on PC) :
- Sigma Delta modulator output psd at 1 KHz offset in MatLab = -300 dBc in Spectre VerilogA (and psd Calculator function) = -130 dBc,
- Phase noise at 10 KHz offset (ideal Charge-Pump Phase Detector) in MatLab = -140 dBc in Spectre VerilogA (and psd Calculator function) = -90 dBc.
(Worst the results do not seems to be nicely reproducible).

Our fract-N PLL model allows for a multiplicity of Phase Detector models : from Linear (as a simple substractor) to more sophisticated Charge Pump type including all kind of imperfections easily implemented in Spectre VerilogA. The intend is (was?) to study stepwise the impact of all such imperfections on the fract-N PLL phase noise. So far it looks to be unfeasible or at least we are not able to get useable results from there our cry for help! :'(

Thank You again for providing me some direction.

Any further guidelines our recommendations herewith are welcome.

Best Regards,
ppg

Title: Re: Spectre-VerilogA and Calculator_psd function
Post by ppg on Jan 16^th, 2006, 8:43am

In previous e-mail I forgot to mention that the fract-N PLL simulations were only targetting the study of the Sigma Delta Quantization noise on the PLL phase noise performance. Thus the reported Spectre Veriloga transient analysis were ran without any added noise for Vco or Reference or Charge-pump, ... etc; only the real Sigma Delta behaviour was included. From there the expectation (with ideal charge-pump phase detector) of low-phase noise at low offset frequency.

ppg