|
The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Simulators >> RF Simulators >> mixer minimum noise figure simulation https://designers-guide.org/forum/YaBB.pl?num=1208334972 Message started by Ian on Apr 16th, 2008, 1:36am |
|
Title: mixer minimum noise figure simulation Post by Ian on Apr 16th, 2008, 1:36am For LNA, it's common practice to use SP simulation to see the 'Gmin' and 'NFmin' and do the noise matching. Here I want to do the same thing for the mixer. I tried PSS + PSP for an acitve mixer, the simulated NF was very close to the result predicted by PSS+Pnoise at the bands dominated by thermal noise. However, the 'Gmin' there only gives a result of zero real part and zero imaginary part, and the 'NFmin' plot is empty. I don't know what's going on there. Any comments? Thanks. Ian |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 16th, 2008, 4:12am Ian wrote on Apr 16th, 2008, 1:36am:
NF results of PSS/Pnoise and PSS/PSP have to be completely same unless you do wrong setting in simulation. If your results are not same completely, your setting must be wrong, maybe for PSP. Quote:
I checked NFmin and Gmin in PSS/PSP for mixer using "spectre version 6.2.0 09/19/2007 14:35 (sfsol77) $". Both Gmin and NFmin give non zero value although I don't believe SpectreRF results at all. Since your NFmin is "NaN", your setting of PSP must ne wrong. PSP(QPSP) setting is relative difficult. I also confirmed Gmin and NFmin results of Agilent ADS(RFDE). In RF design, ADS(MDS, Series-IV) have long history and is most reliable. Spectre's Gmin is same as ADS. But NFmin of Spectre is apparently suspicious. You must not believe NFmin of Spectre. PSP and QPSP are not available in HB analysis. They are available only for shooting PSS/QPSS. As far as I know, PSP and QPSP are not used so much. Especially QPSP is not used and verified, I think. If they are not fixed, NF(dsb, ssb, ieee-ssb) of QPSP must be still wrong. But as far as my experience, NFmin and Gmin for mixer analysis are not so valuable even in ADS(RFDE). |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 16th, 2008, 8:08am Thanks, pancho_hideboo. I appreciate your comments and nice plot. I reported here some of my progress in the following. I simulated a conventional Gilbert-cell mixer, the results got from PSS/PSP and PSS/Pnoise settings were (almost) the same. Then I simulated a modified version of Gilbert mixer (using current-resue bleeding technique) using both methods, they coincided well in the flat band, but still not in the 1/f noise region. I adjusted the automatic sweep in the psp setting to linear sweep (i.e., 100pts), two curves match better than before (see the attached plot). Another interesting obeservation is that at the intended IF frequency (e.g., 100MHz in the attached plot), there's a spike in the NF curve generated from PSS/PSP. P.S.: my spectre version: $CDS: icfb.exe version 5.1.0 06/30/2005 00:12 (cicln03) $ |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 16th, 2008, 8:13am Try Pnoise and PSP with solver=std in option setting. Show me netlist portions regarding port or source(RF, LO, IF) statement and analysis statement. Ian wrote on Apr 16th, 2008, 8:08am:
This is DFII's version not Spectre's version. |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 16th, 2008, 8:52am pancho_hideboo wrote on Apr 16th, 2008, 8:13am:
--- Source Statement PORT_LO (_net0 0) port r=50 num=3 type=sine freq=flo ampl=vlo \ fundname="flo" PORT_RF (net081 0) port r=Rs_rf num=1 type=sine freq=frf ampl=vrf \ fundname="frf" PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc freq=flo ampl=vlo \ fundname="flo" --- Analysis Statement pss pss fund=100M harms=30 errpreset=moderate annotate=status psp psp sweeptype=relative start=-99.999M stop=30M lin=100 + portharmsvec=[-10 1] ports=[PORT_RF PORT_IF] donoise=yes + maxsideband=50 annotate=status freqaxis=out For the spectre version, funny enough, I couldn't find it out right now. |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 16th, 2008, 8:58am Ian wrote on Apr 16th, 2008, 8:52am:
Ian wrote on Apr 16th, 2008, 8:52am:
Teach me values of flo, frf. Also teach me options statement. BTW why do you set freq=flo ampll=vlo fundname="flo" in PORT_IF ? It is type=dc, though..... So it can be no effect. |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 16th, 2008, 9:08am pancho_hideboo wrote on Apr 16th, 2008, 8:58am:
I guess this is it: spectre (ver. 5.10.41_USR2.052705 -- 27 May 2005). Other info: 1. flo=1.1GHz and frf=1GHz. 2. Options Statement as follows: simulatorOptions options reltol=1e-3 vabstol=1e-6 iabstol=1e-12 temp=27 \ tnom=27 scalem=1.0 scale=1.0 gmin=1e-12 rforce=1 maxnotes=5 maxwarns=5 \ digits=5 cols=80 pivrel=1e-3 ckptclock=1800 \ sensfile="../psf/sens.output" checklimitdest=psf |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 16th, 2008, 9:42am Ian wrote on Apr 16th, 2008, 9:08am:
You are using fairly old Spectre. HB analysis is not available in your version. You might be required to do old style accuracy setting for SpectreRF. Also NFmin and Gmin of PSP in your version's Spectre might be completely broken, although they are still very suspicious even in relative new version I use. Show me analysis statement of Pnoise to compare with PSP setting. |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 16th, 2008, 7:48pm If I run PSP on SunOS 5.8 with @(#)$CDS: spectre version 6.2.0 09/19/2007 14:35 (sfsol77) $, PSP can't work and gives fatal error. But If I run PSP on Linux 2.4.21-32.ELsmp, PSP can work. Anyway PSP and QPSP have not been used and verified so much. First while you set harms=30 in PSS, you set maxsideband=50 in PSP. This could give uncorrect results. Set like following. simulatorOptions options reltol=1e-5 vabstol=3e-8 iabstol=1e-13 temp=27 \ tnom=27 scalem=1.0 scale=1.0 gmin=1e-12 rforce=1 maxnotes=5 maxwarns=5 \ digits=5 cols=80 pivrel=1e-3 sensfile="../psf/sens.output" \ checklimitdest=psf pss pss fund=100M harms=0 errpreset=moderate method=gear2only + relref=alllocal maxacfreq=6G annotate=status outputtype=freq + maxiters=15 pnoise pnoise start=1k stop=130M dec=11 maxsideband=50 + oprobe=PORT_IF iprobe=PORT_RF refsideband=-11 annotate=status psp psp start=1k stop=130M dec=11 portharmsvec=[-11 0] + ports=[PORT_RF PORT_IF] donoise=yes maxsideband=50 annotate=status + freqaxis=out Maybe there are wrong settings also in your Pnoise. Show me your Pnoise statement. If your setting is proper, you can see the following message in log file. **************************************************** Periodic Steady-State Analysis `pss': fund = 100 MHz **************************************************** *********************************************************** Periodic Noise Analysis `pnoise': freq = (1 kHz -> 130 MHz) *********************************************************** Input frequency range = (970 MHz -> 1.1 GHz). Using the operating-point information generated by PSS analysis `pss'. ******************************************************************** Periodic S-Parameter Analysis `psp': sweep freq = (1 kHz -> 130 MHz) ******************************************************************** Input port (PORT_RF) frequency range = (-1.1 GHz -> -970 MHz). Output port (PORT_IF) frequency range = (1 kHz -> 130 MHz). Using the operating-point information generated by PSS analysis `pss'. |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 17th, 2008, 12:07am pancho_hideboo wrote on Apr 16th, 2008, 7:48pm:
That's exactly the manual "SpectreRF Simulation Option User Guide (June 2004, Version 5.1.41)" tells me to do. Quote:
so you set PSS fundamental as IF frequency, instead of both LO and RF also recommended in the previously mentioned spectre manual. Then in the schematic, which port should be set as IF frequency, LO or RF? And I guess the other one has to be set as DC, otherwise it would be automatically show up in the PSS fundamentals. Quote:
Please be patient, right now my license server is down. |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 17th, 2008, 12:13am Ian wrote on Apr 17th, 2008, 12:07am:
It is not appropriate. There are many mistakes or wrong explanations in SpectreRF manual. But maybe it is not problem since maxstep is very small in PSS even though you set harms=30. Real problems exist in your setting of Pnoise and PSP. Show me log file. Quote:
You don't understand PSS. 100MHz is not IF frequency. This is fundamental frequency determined by flo and frf. I set harms=0 maxacfreq=6G in PSS. Port settings are following. These are yours. PORT_LO (_net0 0) port r=50 num=3 type=sine freq=flo ampl=vlo \ fundname="flo" PORT_RF (net081 0) port r=Rs_rf num=1 type=sine freq=frf ampl=vrf \ fundname="frf" PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc freq=flo ampl=vlo \ fundname="flo" Quote:
Even if license server is down, you can see netlist and log file in unix tree unless you delete them. Do you need two large signal truely ? See http://www.designers-guide.org/Forum/YaBB.pl?num=1184111670 |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 17th, 2008, 12:54am My PSS+Pnoise settings: pss pss fund=1.1G harms=0 errpreset=moderate method=gear2only + relref=alllocal maxacfreq=6G annotate=status outputtype=freq + maxiters=15 pnoise pnoise start=1K stop=130M dec=11 maxsideband=50 + oprobe=PORT_IF iprobe=PORT_RF refsideband=-1 annotate=status In PSS+Pnoise, fund=1.1GHz, however, in the PSS+PSP simulation, you are right. fund=100MHz. That's why I got confusion earlier. Another comment is that I used the newly pnoise settings, the result (use auto-sweep) still deviates from the one got from psp simulation at 1/f noise region, as in my previous attached plot. Quote:
I understand for the pnoise simulation, rf port set as dc. For psp simulation, if so, then fund=1.1GHz (no more intended 100MHz when frf=1GHz). I doubt the psp settings you gave to me earlier, particularly 'portharmsvec=[-11 0]. Now the equivalent frf became -12.1G -- -11.97G. I even tried just set flo=100MHz, and use your psp settings exactly, the resulting NF is way too big, around 30dB larger. |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 17th, 2008, 1:00am Ian wrote on Apr 17th, 2008, 12:54am:
The above is different from following. Also do you understand meaning of harms=0 maxacfreq=6G in PSS ? Show me original netlist of PSS/Pnoise and PSS/PSP where you observe difference of results between them. Show me original log file of PSS/Pnoise and PSS/PSP. What do you want to do ? Do you want to do PSS/Pnoise and PSS/PSP under two large signal drive ? And you want to know why PSS/Pnoise and PSS/PSP results don't match ? Ian wrote on Apr 16th, 2008, 8:52am:
|
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 17th, 2008, 2:00am Quote:
Well, of coz, I want to have matched results between PSS/Pnoise and PSS/PSP. That's why I adapted my settings to yours to try to get it. To remind you, the intended flo=1.1GHz and frf=1GHz. Teach me what's 'maxacfreq=6G'? -- Original Log Message for PSS+Pnoise: **************************************************** Periodic Steady-State Analysis `pss': fund = 1.1 GHz **************************************************** *********************************************************** Periodic Noise Analysis `pnoise': freq = (1 kHz -> 130 MHz) *********************************************************** Input frequency range = (970 MHz -> 1.1 GHz). Using the operating-point information generated by PSS analysis `pss'. -- Original Log Message for PSS+PSP: **************************************************** Periodic Steady-State Analysis `pss': fund = 100 MHz **************************************************** ******************************************************************** Periodic S-Parameter Analysis `psp': sweep freq = (1 kHz -> 130 MHz) ******************************************************************** Input port (PORT_RF) frequency range = (-1.1 GHz -> -970 MHz). Output port (PORT_IF) frequency range = (1 kHz -> 130 MHz). Using the operating-point information generated by PSS analysis `pss'. Original Netlist Options for PSS+Pnoise: pss pss fund=1.1G harms=0 errpreset=moderate annotate=status pnoise pnoise start=1K stop=130M dec=20 maxsideband=50 + oprobe=PORT_IF iprobe=PORT_RF refsideband=-1 annotate=status PORT_LO (_net0 0) port r=50 num=3 type=sine freq=flo ampl=vlo \ fundname="flo" PORT_RF (net081 0) port r=Rs_rf num=1 type=dc PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc Original Netlist Options for PSS+PSP: pss pss fund=100M harms=30 errpreset=moderate annotate=status psp psp sweeptype=relative start=-99.999M stop=30M lin=100 + portharmsvec=[-10 1] ports=[PORT_RF PORT_IF] donoise=yes + maxsideband=50 annotate=status freqaxis=out PORT_LO (_net0 0) port r=50 num=3 type=sine freq=flo ampl=vlo \ fundname="flo" PORT_RF (net081 0) port r=Rs_rf num=1 type=sine freq=frf ampl=vrf \ fundname="frf" PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 17th, 2008, 2:19am Quote:
fund=1.1G, maxsideband=50 So maximum frequency of noise folding is fund*maxsideband=55GHz Quote:
fund=100M, maxsideband=50 So maximum frequency of noise folding is fund*maxsideband=5GHz In addition to difference of 1tone drive and 2tone drive, maximum noise folding frequency is different in your setting for Pnoise and PSP. If you want to simulate NF of mixer, you have to choose 1tone(only LO) drive or multitone drive(LO and RF). See http://www.designers-guide.org/Forum/YaBB.pl?num=1184111670 Here assume flo=1100MHz, frf_interference=1000MHz and fif=1kHz~130MHz. frf_desire will be flo-fif or flo+fif. Then you chose frf_desire=flo-fif, LSB Mixer. If you choose 1tone drive where there is no RF interference, PSS/Pnoise/PSP setting has to be like following. PORT_LO (_net0 0) port r=50 num=3 type=sine freq=flo ampl=vlo fundname="FLO" PORT_RF (net081 0) port r=Rs_rf num=1 type=dc PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc pss pss fund=1.1G harms=0 maxacfreq=5.5G errpreset=moderate annotate=status outputtype=freq pnoise pnoise start=1k stop=130M dec=11 maxsideband=5 + iprobe=PORT_RF oprobe=PORT_IF refsideband=-1 annotate=status psp psp start=1k stop=130M dec=11 maxsideband=5 + ports=[PORT_RF PORT_IF] portharmsvec=[-1 0] donoise=yes annotate=status + freqaxis=out If you choose 2tone drive where LO and RF(interference) are large, PSS/Pnoise/PSP setting has to be like following. PORT_LO (_net0 0) port r=50 num=3 type=sine freq=flo ampl=vlo fundname="FLO" PORT_RF (net081 0) port r=Rs_rf num=1 type=sine freq=frf_interference ampl=vrf fundname="FRF" PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc pss pss fund=100M harms=0 maxacfreq=5.5G errpreset=moderate annotate=status outputtype=freq pnoise pnoise start=1k stop=130M dec=11 maxsideband=55 + iprobe=PORT_RF oprobe=PORT_IF refsideband=-11 annotate=status psp psp start=1k stop=130M dec=11 maxsideband=55 + ports=[PORT_RF PORT_IF] portharmsvec=[-11 0] donoise=yes annotate=status + freqaxis=out In relative old SpectreRF, "errpreset=moderate" in pss setting is not helpful. Instead of this, tighten accuracy parameters. (1) Add the followings to pss setting method=gear2only relref=alllocal maxiters=15 (2) Set accuracy parameter in options statement like following. reltol=1e-5 vabstol=3e-8 iabstol=1e-13 options Current SpectreRF don't need both (1) and (2). "harms=0" in pss setting mean we don't save large signal harmonics. "maxacfreq=5.5G" in pss setting mean maximum frequency for small signal analysis such as pac, pxf, pnoise and psp. Here I chose it as 5.5GHz=5*flo. maxacfreq has to be larger than 5.5GHz. So maxacfreq=6G is also OK. In shooting PSS, "maxstep" is very important parameter. Generally it is determined by "fund" and "harms". But if your fundamental frequency is very small, this is very inefficient in memoy usage. So you had better use "maxacfreq". And in small signal analysis such as NF analysis, we don't need to observe large signal harmonics. So I set "harms=0". |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 17th, 2008, 7:24am I followed your lead, and I've got perfectly matched results from either pss+pnoise or pss+psp, no matter one tone or two tones. The accuracy is within 0.005 dB. pancho_hideboo, many many thanks. Frankly speaking, at the very beginning, I didn't expect someone could help me in such great detail, but you surely are! I appreciate it very much. |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 17th, 2008, 8:11am Ian wrote on Apr 17th, 2008, 7:24am:
Maybe your RF interference ampl=vrf is very small so it doesn't affect NF characteristics. Many people often say results of PAC, PXF, Pnoise and PSP don't match. But they must be always matched. Just your settings are wrong or you don't understand equations used in Direct Plot Form. :P After many many corrections. we can be back to your first question. Now you do correct setting for PSP, do you come to be able to get NFmin and Gmin as meaning value ? I will not use SpectreRF. Still I don't believe SpectreRF results at all. 8-) http://www.designers-guide.org/Forum/YaBB.pl?num=1205223090 |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 21st, 2008, 2:17am pancho_hideboo wrote on Apr 17th, 2008, 8:11am:
Unfortunately, the 'NFmin' plot is still 'NaN' and Gmin sits at the center of the smith chart. Abosultely meaningless. Lately, I upgraded my spectre to higher version, i.e., Version 6.2.0.420 -- 11 Jun 2007, which makes no difference on these results. Ian |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 21st, 2008, 2:51am Ian wrote on Apr 21st, 2008, 2:17am:
Do you realy do set PSP correctly ? In original netlist, you set like following where it is natural even if you get "NaN" as NFmin. psp psp sweeptype=relative start=-99.999M stop=30M lin=100 + portharmsvec=[-10 1] ports=[PORT_RF PORT_IF] donoise=yes + maxsideband=50 annotate=status freqaxis=out |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 21st, 2008, 6:41am pancho_hideboo wrote on Apr 21st, 2008, 2:51am:
Well, this time, I believe, 100% good in settings ;) PS: pss pss fund=100M harms=0 errpreset=moderate maxacfreq=5.5G + annotate=status outputtype=freq maxiters=15 psp psp start=1K stop=130M portharmsvec=[-11 0] ports=[PORT_RF+ PORT_IF] donoise=yes + maxsideband=55 annotate=status freqaxis=out PORT_RF (vrf 0) port r=50 num=1 type=sine freq=1G ampl=vrf fundname="frf" PORT_LO (vlo 0) port r=50 num=3 type=sine freq=1.1G ampl=vlo fundname="flo" PORT_IF (Vloadp Vloadn) port r=50 num=2 type=dc |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 21st, 2008, 7:09am Ian wrote on Apr 21st, 2008, 6:41am:
It seems OK except for typo. Which method do you use to plot NFmin, Direct Plot Form or Result Browser ? Try Result Browser. If both can't work, you should report to Cadence. SpectreRF have not been used deeply in RF design community, so there could be many bugs. I don't believe SpectreRF results at all, so I will not use SpectreRF for RF design. |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Apr 23rd, 2008, 12:12am Comparing PSP's results with Agilent RFDE(ADSsim) and GoldenGate, I noticed "freqaxis" has to be "absin" when I want to evaluate S-parameters including Gmin(=Sopt) in PSP of SpectreRF. Setting of "freqaxis" should affect only independent variable of output results. But it can affect S-parameter value. Maybe this is also bug of SpectreRF. Quote:
In PSP of SpectreRF, unless "freqaxis" is "absin", S-parameters are wrongly calculated like following. S11=b1(-RF_freq)/a1(-RF_freq) S21=b2(IF_freq)/a1(-RF_freq) S12=b1(-RF_freq)/a2(IF_freq) S22=b2(IF_freq)/a2(IF_freq) where RF_freq is positive value. So S11 is wrongly calculated as complex_conjugate of true S11. S21 and S12 are also wrongly calculated. |
|
Title: Re: mixer minimum noise figure simulation Post by Ian on Apr 25th, 2008, 2:56am pancho_hideboo wrote on Apr 23rd, 2008, 12:12am:
That explains! I did observe the S11 from PSP is the conjugate of the S11 got from SP. |
|
Title: Re: mixer minimum noise figure simulation Post by new_guy on Mar 7th, 2010, 10:47pm Hi Pancho, I followed this post with great intent. Would like to find out if I running spectre simulation for a subharmonic mixer, what would the reference side band should be for spectre PSS PNOISE analysis? Thanks Regards New_guy |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Mar 8th, 2010, 1:46am new_guy wrote on Mar 7th, 2010, 10:47pm:
http://en.wikipedia.org/wiki/Subharmonic_mixer BTW, have you been able to understand the following ? http://www.designers-guide.org/Forum/YaBB.pl?num=1193135745/6#6 |
|
Title: Re: mixer minimum noise figure simulation Post by new_guy on Mar 9th, 2010, 6:32pm Hi pancho, Thanks a lot. However even if the setup is single tone (large signal is LO and RF set to DC), the reference sideband might not be -2 or +2. -2 or +2 works only if LO and RF are multiples, eg RF = 2.4GHz and LO=1.2GHz. If RF =2.4GHz and LO=1.0 GHz, reference sideband would not be -2 or +2, correct? Thanks |
|
Title: Re: mixer minimum noise figure simulation Post by pancho_hideboo on Mar 10th, 2010, 2:00am new_guy wrote on Mar 9th, 2010, 6:32pm:
You are completely misunderstanding. Again have you been able to understand the following ? http://www.designers-guide.org/Forum/YaBB.pl?num=1193135745/6#6 new_guy wrote on Mar 9th, 2010, 6:32pm:
No one except for you thinks it is correct. I assume IF frequency where you observe noise is 0.4GHz. RF=IF+2*LO, Here RF=2.4GHz, LO=1.0GHz, IF=0.4GHz. In this case, reference sideband is "+2". Surely read message of "spectre -h pnoise" as unix command. |
|
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |