The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Simulators >> RF Simulators >> Spectre pss & pac difficulties for SC filter. https://designers-guide.org/forum/YaBB.pl?num=1274020705 Message started by leon30 on May 16th, 2010, 7:38am |
Title: Spectre pss & pac difficulties for SC filter. Post by leon30 on May 16th, 2010, 7:38am Hello everybody, I'm designing a fully differential switched capacitor biquad filter with switching frequency of 1.33 MHz (actually two non-overlapping clocks). When analyzing the ideal circuit (constructed with ideal switches from analogLib and all others elements plus ideal fully differential opamp model constructed from me with vcvs) I should do the following steps(I've read many papers on the subject and I think I get the pss pretty clear, but the other (pac, pnoise) were explained for the oscillators and mixers mostly and I'm finding myself little lost there): 1) I put for input source of the circuit - one vdc with 0V dc voltage and pac magnitude of 100mv with pac phase 0, only for the one input, the other is set to ground; (Is this is correct ?) 2) I'm setting the pss analysis as follows: 2.1) In the "Engine" section I choose the shooting method; 2.2) In the "Fundamental tones" section there are the two clock signals with the same frequency of 1.33 MHz, so as "Beat frequency" I have also 1.33 MHz; 2.3) In the "Output harmonics" section for "Number of harmonics" I write 0, because I need the pss only for determining the periodic operating point of the circuit, and because I'm interested in the circuit behavior from the range of 0 Hz to 1.33 MHz. (First question should I entered 0 harmonics here, or I will need more in order to completely characterize an SC filter ?) 2.4) In the "Options..." section for "Integration method" I choose "gear2only" for better convergence; 2.5) When I run the analysis I check the SpectreRF log to see the "Conv. norm" value if it is less then 1 and if I have "pss: The steady-state solution was achieved in X iterations." everything is OK with this; 2.6) If not, I'm having convergence problems and I should change the initial condition (IC) parameter for all the capacitors in the circuit to 0, or to do step response of the circuit and see after witch moment of time it's become stable (let's call this moment ST) and then set the "tstab" parameter in the pss analysis form to a multiple of the clock period greater than ST. (Second question - are these procedures correct ? Am I doing good and sufficient pss analysis of SC filter ?) After that I'm staring to fill the form of the pac: 3) What should I choose for "Sweeptype" ? I need the behavior of the circuit from 0 Hz to the "Beat frequency" in this case 1.33 MHz, I mean that I need the AC transfer characteristic of the filter. 3.1) "Input frequency range" I think this is in strong relationship with the above parameter - I think I should choose for the above "absolute" and for that one - from 0.1 to 1.33 MHz, but I'm not sure ? 3.2) This "Sweep type" will be logarithmic with ten points per decade... 3.3) What is this parameter "Maximum sideband", what should I choose here ? What are the other analysis I need, except Pnoise ? Do I need "Quasi periodic analysis" or others ? Thanks in advance for your help. |
Title: Re: Spectre pss & pac difficulties for SC filter. Post by ahmadagha23 on May 17th, 2010, 7:52am hi Also I have the above problem. Why there is not any newer version of "simulating switched-capacitor filters with spectreRF" document for spectreRF2008? In the above doc the netlist defines only the period and maxacfreq; while in specreRF there are so more parameters required, for example "Fundamental tones" and beat frequency simultaneusly. And other parameters as you said. I think you should not set any "Fundamental tones" and only set beat freq to 1.33MHz. the "Maximum sideband" is described in sc-filter (section 2.1) make me aware if you achieved any more information |
Title: Re: Spectre pss & pac difficulties for SC filter. Post by sheldon on May 17th, 2010, 8:01am Leon, Hope this helps 1) Why don't you use the ideal_balun to create the input signal? Inputs are the differential input, one source VDC=0 and pac=1, pin d and the common-mode input, one source VDC=your choice, pin c connect pin p to the non-inverting input and pin n to the inverting input 2) You have one clock with two phases, give both sources the same frequency name 3) Probably need to define more output harmonics if you want to see anything, 20? 100? 4) Would not expect big convergence issues with this type of circuit. Certainly make sure that the tstab is long enough for the circuit to settle. At least you will need the common mode feedback circuits to stabilize. 5) You probably should select the frequency sweep based on your design's characteristics 6) Circuit with sampling have frequency translation, noise, etc. at one frequency are shifted to another frequency by the process of sampling. Maxsidebands tells the tool how many sidebands, to include when calculating the frequency translation 7) PNOISE will give you the noise PXF will give you power supply rejection You can use PSS + PAC to simulate distortion, IP3 Best Regards, Sheldon |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |