The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Modeling >> Semiconductor Devices >> BSIM3 can give correct IM3 ? https://designers-guide.org/forum/YaBB.pl?num=1207474039 Message started by pancho_hideboo on Apr 6th, 2008, 2:27am |
Title: BSIM3 can give correct IM3 ? Post by pancho_hideboo on Apr 6th, 2008, 2:27am Hi. I've heard that most standard Spice MOS models such as Level 1, 2, 3, BSIM1, BSIM2, BSIM3, BSIM4, MOS9, or EKV 2.63 will not give correct results for IM3 simulation(Inter Modulation of Third Order). I've also heard that the only MOS models that give correct IM3 results are the PSP model from Philips/Penn State University, the MOS11 model from Philips, and the HiSIM model from Japan. Coincidentally, all three of these models are surface potential based models. IM3 given by standard Spice MOS models such as BSIM3 is very incorrect so that it can't be useful in actual circuit design ? More modern surface potential-based models like PSP, MOS11, and HiSIM will be absolutely needed ? I expect that standard Spice MOS models is still practical level regarding IM3. |
Title: Re: BSIM3 can give correct IM3 ? Post by didac on Apr 6th, 2008, 7:14am Hi, I've heard that this issue is caused by non-symmetry of the BSIM model, something related with non-zero second derivative of the current at vds=0. I just can only tell that the last time I made measurements on IM3 with circuits designed with BSIM3v3 the maximum error found between sims and measurement was around +/-0.5dB, so I don't really know were this issue comes as big player in design. |
Title: Re: BSIM3 can give correct IM3 ? Post by Geoffrey_Coram on Apr 7th, 2008, 5:07am Try plotting Cgd and Cgs as you sweep vds across zero. The values aren't equal at vds=0, so there's a jump as they swap places. A nice reference is "Validation of MOSFET model Source-Drain Symmetry" by McAndrew, IEEE Trans ED Sept 2006. The Compact Model Council had some nice plots showing IM3 plots from BSIM3/4 with the wrong slope (wasn't 3:1 output to input power), but I can't find it right now. |
Title: Re: BSIM3 can give correct IM3 ? Post by pancho_hideboo on Apr 7th, 2008, 5:23am Hi. I found some informations about this issue in pp.24-29 of the following. http://www.mos-ak.org/munich/papers/03_MOS-AK_Mattausch.pdf |
Title: Re: BSIM3 can give correct IM3 ? Post by didac on Apr 7th, 2008, 11:59am Hi, Looking at the CMC and around the net(and replicating the gummel test and the modified gummel test proposed in the reference provided by Geoffrey) I think that this issue becomes a player in design basically in passive mixers,switching circuits, but I'm not sure about standard Gilbert cell if the intermodulation is dominated by the transconductor. Can anybody confirm this? PS:another reference for the cause http://www.jazzsemi.com/docs/RFIC-Workshop.pdf |
Title: Re: BSIM3 can give correct IM3 ? Post by pancho_hideboo on Apr 7th, 2008, 9:35pm didac wrote on Apr 7th, 2008, 11:59am:
Passive mixer and Sampling mixer are very important building block. So I'm very anxious about validity of IM3 for passive mixer and sampling mixer. |
Title: Re: BSIM3 can give correct IM3 ? Post by Geoffrey_Coram on Apr 8th, 2008, 4:38am didac wrote on Apr 7th, 2008, 11:59am:
If you're not swinging across VDS=0, then it's not an issue. (There could be other issues; I've heard BSIM3 has a mathematical smoothing between strong and weak inversion, whereas the surface potential formulations have a physical description. So, you have to worry about how well the BSIM math fits the actual measurement.) |
Title: Re: BSIM3 can give correct IM3 ? Post by didac on Apr 8th, 2008, 4:43am Hi, Thanks for the tips, I'm playing with a passive mixer right now and with two types of transistors and I'm wondering if the finite on resistance would partially circunvent this issue(i.e. don't make a very good switch to avoid the singularity...), at the moment I'm seeing slopes of 2.5 sometimes 3 other times... |
Title: Re: BSIM3 can give correct IM3 ? Post by didac on Apr 8th, 2008, 7:16am Geoffrey_Coram wrote on Apr 8th, 2008, 4:38am:
In the reference that I found from jazz semiconductor it tells something about this region , which makes me wonder which model used in the papers that I read about "sweet spot" bias point(switch from moderate to strong inversion)-not taking into account the side effects associated with this biasing point and issues like feedback of the second harmonic in source degenerated devices-. |
Title: Re: BSIM3 can give correct IM3 ? Post by didac on May 4th, 2008, 2:34am Hi, If somebody is still interested in this issue I found yesterday a LNA with an IIP3 with a slope of 2 instead of 3... without entering in details about the topology I track down the problem to a switch in the signal path(I verified removing the switch and the slope it's 3-without changing nothing in the simulation configuration-), so I think that this issue extends to "any circuit with a switch in the signal path". |
Title: Re: BSIM3 can give correct IM3 ? Post by loose-electron on May 15th, 2008, 11:50am This is one of those things where you need to draw the difference between the limitations of the model, and the accuracy of the paremeters entered into the model. Looking at the device physics, we all know that they have to be equal for symmetric MOS devices. However, often models are an exercise in blind curve fitting, and checking if ring oscillators simulate and run at the same frequency. Also, 99% of model use is not for analog design, and the poor quality goes unchecked. -- Jerry Geoffrey_Coram wrote on Apr 7th, 2008, 5:07am:
|
Title: Re: BSIM3 can give correct IM3 ? Post by weber8722 on Feb 19th, 2013, 7:34am didac wrote on Apr 7th, 2008, 11:59am:
I made some experiments on my own with bsim3 models: A normal LNA and Gilbert mixer has not the big bad-IM-slope problem, only passive mixers have that problem in simulation. Bye Stephan |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |