The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> RF Design >> Discrepancy in Tran and PSS time domain outputs https://designers-guide.org/forum/YaBB.pl?num=1157393066 Message started by jobless on Sep 4th, 2006, 11:04am |
Title: Discrepancy in Tran and PSS time domain outputs Post by jobless on Sep 4th, 2006, 11:04am I have a circuit wherein two oscillators are coupled with frequency of one oscillator as 3 times that of the other. Upon carrying out Tran and PSS analysis, I get a mismatch in terms of fundamental frequency and amplitude. I have tried various tstab methods. For fundamental frequency, gear2 and Euler agree(with tran), while there is still no agreement on steady state oscillation amplitude. I am using a very tight 'maxstep' and have set 'errpreset' as 'conservative' with 'finitediff' and 'highorder' radio buttons ON. Which one should I trust, tran or pss? This disagreement start at a particular point as I vary the size of the cross-coupled MOSFETs in VCO structure. Also this point of disagreement seems to shift a little bit every time I sweep the transistor size in parametric analysis. I am really stuck and confused at this moment. Please help me out. |
Title: Re: Discrepancy in Tran and PSS time domain output Post by ACWWong on Sep 5th, 2006, 1:57am gear2only is a better algorithm to use for both tran and pss. Euler introduces more numerical damping. When you say "very tight" max step, what do you mean ? I would use a max step of no less than say 100 points of the oscillation freq... (often i would use only 50) .... going overboard with harmonics and too tight with tolerances might not be too clever (as it will slow you down/increas simulation time, and the models you are by no means exact, so will be working very hard to solve very very accurately non-exactly modelled devices.) Also when you say "Upon carrying out Tran and PSS analysis, I get a mismatch in terms of fundamental frequency...", how are you determining fundamental frequency in each case, and what magnitude is the error ? If you use tstab correctly with the same algorthm as used in tran then you do not expect any discernible differnce. For the tran i guess you are setting you dft correctly and fairly so that you would expect the same result in the first place? |
Title: Re: Discrepancy in Tran and PSS time domain output Post by jobless on Sep 5th, 2006, 11:37am Thanks a lot for your comments. I am not carrying out dft for evaluating fundamental frequency from Tran analysis waveform. I am just physically zooming on the waveform and measuring the time difference between two consecutive signal peaks. It is of course not accurate and precise, but the results mostly closely match with PSS. But in this case, I see a sudden jump in PSS fundamental frequency(300MHz) and amplitude(500mV). While in Tran analysis, frequency and amplitude both vary smoothly as I vary the transistor size. Also I am using the same tstab method for both PSS and Tran. Also what should be 'maxstep' for a signal frequency of 2 GHz? Regards -J |
Title: Re: Discrepancy in Tran and PSS time domain output Post by ACWWong on Sep 5th, 2006, 12:55pm So you are looking at the time waveform when doing the transient run and measuring the frequency with cursors after a certain amount of time where you think the ocsillation has sufficently settled (ie the tstab time you will use in the pss). I assume also you are expecting ~2Ghz on one output and ~666MHz on another output. Lets say you setup a max step of 10ps and use gear2only for both tran and pss runs (including tstab period). To make the simulations a bit quicker, let say you use errpreset moderate, and switch off high order (only really needed when looking at small signals close to what would otherwise be the numerical noise floor). Now on the pss run, I guess you are ploting/printing the pss fundamental frequency.... i have a couple of questions you might ask yourself to help solve this issue: 1) are you saving the initial transient to learn why you get the sudden unexpected jump in pss fundamental frequency ? 2) have you plotted the pss outputs voltage against time (using the main form) to see what the steady state waveform against time looks like on your 2 outputs ? This should closely match your transient simulation waveforms. 3) is your initial guess at the pss fundamental something close to 666MHz ? 4) you will only get a good idea of the amplitude at any given frequency by doing dft on transient waveform, or plotting "sprectrum" on the pss output form. Anyway hope these points help you resolve you issue. cheers aw |
Title: Re: Discrepancy in Tran and PSS time domain output Post by jobless on Sep 7th, 2006, 2:47am I precisely did what mentioned in your points. I follow you instructions and address issues in Points 1,2 and 3. I give pss fundamental as 700MHz. I plot time domain waveform at both the outputs from the pss time domain outputs. I put gear2only as the tstab method and 1ps as the maxstep. The pss time domain shows the amplitude as 630mV while the same from tran analysis is 1.3 Vs. I will also try out dft method. But tha, I feel, should not be drastically different from what you get by eyeballing. Please help me out. |
Title: Re: Discrepancy in Tran and PSS time domain output Post by ACWWong on Sep 7th, 2006, 3:26am Sounds very strange... maxstep of 1ps is a bit over the top. as i said 10ps should be easily small enough for good accuarcy... So you are no longer getting the frequency wrong now, just the amplitude ?? Are you sure you are not confusing single ended output against differential or peak against peak2peak ? If you are looking at both time waveforms of pss and tran after the same amount of time (ie after tstab for tran) and the waveforms are drastically different then I think your particular case needs addressing be Cadence themselves, who given the netlist, setup etc. can comment more appropriately for your exact situation. |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |