The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> ultrasim vs spectre
https://designers-guide.org/forum/YaBB.pl?num=1176167388
Message started by arsenal_he on Apr 9th, 2007, 6:09pm

Title: ultrasim vs spectre
Post by arsenal_he on Apr 9th, 2007, 6:09pm
hi,

i am runing some simple simulation using ultrasim and spectre.
For a transistor with fixed biases, spectre shows id=260uA while ultrasim only gives out 50uA with the same model and corner and highest accuracy setting.
And the spice model shows that spectre simulation gives the correct results, then what can i do such that ultrasim runs fine？

thanks
arsenal

Title: Re: ultrasim vs spectre
Post by sheldon on Apr 10th, 2007, 6:05am
Arsenal,

  You need to provide more information about your circuit and any
warnings in the log file. In particular, was the dc operating point
calculation correct?

                                                      Best Regards,

                                                         Art Schaldenbrand

Title: Re: ultrasim vs spectre
Post by arsenal_he on Apr 10th, 2007, 11:19am
hi art,

thank you for your reply.
I didn't find any warnings but the vth of ultrasim simulation is 0.658v while that of spectre is only 0.420v.
Then why is so big difference of vth?

thanks
arsenal

sheldon wrote on Apr 10th, 2007, 6:05am:
Arsenal,

  You need to provide more information about your circuit and any
warnings in the log file. In particular, was the dc operating point
calculation correct?

                                                      Best Regards,

                                                         Art Schaldenbrand

Title: Re: ultrasim vs spectre
Post by Geoffrey_Coram on Apr 10th, 2007, 12:11pm
Do your models include WPE or STI stress effects (parameters SA, SB, SC)?

Title: Re: ultrasim vs spectre
Post by arsenal_he on Apr 10th, 2007, 1:31pm
hi Geoffre,

I found some warnings complaining lack of sa1 so ultrasim use sa instead. BUt I did'nt find any sa1 or sa in the model file. And I have checked both the netlist of ultrasim and spectre, both of them have the parameters of SCA, SCB, SCC

thanks
arsenal

Geoffrey_Coram wrote on Apr 10th, 2007, 12:11pm:
Do your models include WPE or STI stress effects (parameters SA, SB, SC)?

Title: Re: ultrasim vs spectre
Post by Geoffrey_Coram on Apr 12th, 2007, 4:39am
SA/SA1 are instance parameters, meaning that the instance should have values that indicate how that particular device is laid out; AD, AS, PD, PS are also instance parameters.  The model file describes common parameters.  If ultrasim is using the wrong values for the stress parameters, then it will incorrectly compute the shift in threshold voltage due to stress.

If it's a small test netlist, you could delete all the sa/sb/sc parameters and see if you get the same threshold voltage.  Then we can work on seeing if we can fix the parameter names.  Eg, "sa1" is not an official BSIM parameter, it is part of a foundry-specific STI stress model; there are values SA1 through SA9 for up to 9 gate segments.  It may be that ultrasim doesn't support these unofficial parameters.

Title: Re: ultrasim vs spectre
Post by arsenal_he on Apr 12th, 2007, 11:35am
hi Geoffrey,

thank you very much.
I found that it was due to the calculation of SCA,SCB and SCC, which should change as to W/L, but my those parameters in my pdk are fixed! Still i don't know where to get a new pdk for tsmc65gp

thanks
arsenal

Geoffrey_Coram wrote on Apr 12th, 2007, 4:39am:
SA/SA1 are instance parameters, meaning that the instance should have values that indicate how that particular device is laid out; AD, AS, PD, PS are also instance parameters.  The model file describes common parameters.  If ultrasim is using the wrong values for the stress parameters, then it will incorrectly compute the shift in threshold voltage due to stress.

If it's a small test netlist, you could delete all the sa/sb/sc parameters and see if you get the same threshold voltage.  Then we can work on seeing if we can fix the parameter names.  Eg, "sa1" is not an official BSIM parameter, it is part of a foundry-specific STI stress model; there are values SA1 through SA9 for up to 9 gate segments.  It may be that ultrasim doesn't support these unofficial parameters.

Title: Re: ultrasim vs spectre
Post by Geoffrey_Coram on Apr 13th, 2007, 6:45am
You may need to get a layout-extractor that supports WPE (SCA,SCB,SCC).

Title: Re: ultrasim vs spectre
Post by arsenal_he on Apr 13th, 2007, 3:35pm

ya, im using qrc to do the RCX, and it seems SCA,SCB and SCC affect vth too much! i did an experiment and the pre-layout vth is about 470mv while the post-layout is 690mv!!! How can i minimize the discrepancy in layout?

thanks

Geoffrey_Coram wrote on Apr 13th, 2007, 6:45am:
You may need to get a layout-extractor that supports WPE (SCA,SCB,SCC).

Title: Re: ultrasim vs spectre
Post by sheldon on Apr 13th, 2007, 10:22pm
Arsenal,

  Can you modify the transistor layouts? These parameters are a
function of layout. BTW, did a quick search and found an interesting
paper that might give you insight into what needs to be changed,

      http://www.ieee-cicc.org/06-8-6.pdf

                                                     Best Regards,

                                                          Sheldon

Title: Re: ultrasim vs spectre
Post by arsenal_he on Apr 17th, 2007, 3:47pm
Sheldon,
Thank you very much.

arsenal

sheldon wrote on Apr 13th, 2007, 10:22pm:
Arsenal,

  Can you modify the transistor layouts? These parameters are a
function of layout. BTW, did a quick search and found an interesting
paper that might give you insight into what needs to be changed,

      http://www.ieee-cicc.org/06-8-6.pdf

                                                     Best Regards,

                                                          Sheldon

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.