The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Simulators >> Circuit Simulators >> time-domain noise analysis in switched capacitor https://designers-guide.org/forum/YaBB.pl?num=1185820774 Message started by hpham on Jul 30th, 2007, 11:39am |
Title: time-domain noise analysis in switched capacitor Post by hpham on Jul 30th, 2007, 11:39am Hi, I tried to use the time-domain noise feature of spectreRF for my switched capacitor amplifier design. I'm not sure how to select noise skip count and number of time poinst to have the accurate output noise density. I tried the sampled noise at only a single point (t=0) and vary maxsideband parameter and I already see the big difference for maxsideband = 20 & 50 case. I wonder for the case of amplifier, sampled noise at a single time point represent the accurate output noise density of a circuit OR i should choose more time points (at least two --> peak to peak voltage time points) ?? and how can I do that?? I've read the cadence document : Simulating switched-capacitor filters with SpectreRF (version 6c, 28 July 2006) best regards, ha |
Title: Re: time-domain noise analysis in switched capacit Post by horizon on Jul 31st, 2007, 6:01am Hi hpham, I suppose I am going to face the same problem as you as I am designing the same circuit. At the moment I still wonder which simulator between spectre and eldo is the best to do this. I read " simulating switched capacitor filters with spectre RF" and also "device noise simulation of sigma delta modulator". The spectre way of doing the simulation is not completly clear to me . Did you try to do the simulation with eldo and its noisetran analysis? I think I am going to test spectre RF on an ideal amplifier using ideal switch and ideal capacitor in order to check what is the output noise. have a nice day horizon |
Title: Re: time-domain noise analysis in switched capacit Post by sheldon on Aug 3rd, 2007, 9:10pm Ha, A comment on the setting of maxsidebands, what is the noise bandwidth of your circuit? You need to set maxsidebands so that all noise or at least the noise out to the noise bandwidth is included in the pnoise analysis. From your description, the circuit has a wide noise bandwidth so using 20 sidebands does not fold all the noise down. By increasing the maxsidebands to 50 more noise is folded down and the noise increases. Evaluate the noise bandwidth of the circuit and then set the number of maxsidebands. Best Regards, Sheldon |
Title: Re: time-domain noise analysis in switched capacit Post by hpham on Aug 7th, 2007, 10:18am Sheldon, Thanks for your reply. I though I set maxac freq big enough ( = 100M). More information about my setup : input freq = 12.5K, sampling clock = 200KHz, Bandwidth = 100KHz, amplifier's voltage gain = 72dB ( 3 stages in cascade, each = 24dB) . Input capacitor = 64pF, feedback capacitor = 4pF. When using maxsideband = 20, output noise density = 50.7nV2/Hz. msb = 50 , output noise = 87.7nV2/Hz. That's for the case of one single time point. I am thinking about choosing 2 time points to represent the peaks of output voltage, but i don't know how. Horizon, Thanks for your reply. I don't have access to Eldo. best regards, ha |
Title: Re: time-domain noise analysis in switched capacit Post by horizon on Aug 30th, 2007, 4:47am Hi everybody, At the moment I designed my circuit using spectre... but for the noise analysis , I used eldo (I am lucky guy, I have access to eldo and spectre) The results seems to be realistic. Nevertheless, It would be nice for me to use spectre also for the noise analysis. I don't understand how to have a comparable result with eldo. What I am doing with eldo: .TRAN 10p 100n .NOISETRAN FMIN=0 FMAX=5G NBRUN=20 MRUN My clock is working at 20Mhz, for the noisetran analysis, the first cycle is not measured. The result I measure is the rms noise when the data is sampled (let say I measure my noise at 100ns ) for example, I get a noise of 200uVrms Does somebody know how to do something comparable with spectre ? the spectre simulation I did gave different results. Somebody has a idea of the setting I should use ? Thank you very much have a nice day Horizon |
Title: Re: time-domain noise analysis in switched capacit Post by sheldon on Sep 6th, 2007, 6:41am Ha, I don't understand your comments about selecting two points at the peaks of the output voltage. Does your testbench have an input signal? Shouldn't the input be grounded when simulating noise? It is easier for me to think about this in terms of S/H behavior so I will describe the behavior that way. My understanding is that you need to evaluate two conditions: a) the sampling mode noise, measured at the instant the sampling when the S/H switches to hold mode b) the hold mode noise, measured at the time when the next stage samples the output For a S/H operating at 10MHz and a 50% duty cycle, I would select two time points, 50ns and 99ns. For a test circuit I built, the sample noise 1.25mVrms and the hold noise was 3.45mVrms so the total is 3.67mVrms from PNOISE. In my case the op-amp and sampling circuit are wideband so I needed to use a lot of sidebands, ~100. Hope this helps, Sheldon |
Title: Re: time-domain noise analysis in switched capacit Post by xiaodu2001 on Nov 13th, 2008, 7:15pm hpham wrote on Aug 7th, 2007, 10:18am:
may i ask u a question ? in the simulation u took above ,what is the parameter "stop" in pnoise analysis u choose? this is the problem bothering me so much these days.thanks a lot! |
Title: Re: time-domain noise analysis in switched capacitor Post by manodipan on Nov 14th, 2008, 2:27am Hi guys, I am also designing some switched capacitor filters and facing the same problem...i am following Ken's tutorial on SC filter simulation in Spectre RF...For time domain noise analysis what i do is,i make input dc voltage source zero and then no input source is selected ,noise type is selected as time domain....but it does'nt work at all??? also i am confused abt stop freq?? |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |