You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I would like to know if someone has a bit hands on experience on using MOSRA commands in HSPICE for doing the Aging analysis particularly for NBTI. I am trying to simulate an inverter (for the purpose of circuit/transistor level aging) with mosra commands with 32nm MG High-K PTM. My goal is to compute the delta VTH vs stress time (e.g t=1 year to 5 years).
The problem I am facing right now is that the VTH values are not changing (means it remains constant) through out the stress time of 5 years. I believe that I am doing a little mistake in adjusting Vgs/Vds/ Vbs values or may be in transient analysis in mosra commands. Following is an example netlist of inverter aging for NBTI:
After reviewing the White Paper it may presented at too high of a level level, so I would just email the local Synopsys AE for support. Another contact at Synopsys that could help is Hany Elhak.
Thanks for your quick reply. In-fact the help which I could get is only from few papers of Bogdan Tudor (ex-Synopsys employee- I do not know his current email ID). The rest of the limited documentation of MOSRA is available in Hspice manual. But the problem is that there is no such example with proper netlist from which I could see the aging affect. I would really appreciate if you could provide me of any contact (email) such as "Hany Elhak" to my following email ID.
I recommend that instead of following professors' advice
you pull (say) midpoint and endpoint models from NBTI
stressed samples. All that model param twiddling needs
to be grounded in reality or it's pointless, physics of
failure has to precede any prediction of what the device
will do if you want to get from theory to individual cases.
NBTI is also quite dependent on the whole use cycle, in
some cases more than half of the max drift can anneal
back (there are reversible and irreversible components).
This means that a "worst case" test could tax you unduly
in design, but can also mean that your problem is not just
single ended shifts but service-variable shifts, mismatch
developing over time in analog blocks and so on (not an
inverter problem, per se, but perhaps a late-in-life timing
skew problem arising from different paths developing
different delays due to duty cycle / rep rate differences).
@usmanstar2 - Your netlist seems OK. You have to make sure that your models are capable of reliability modeling for NBTI. Since I don't know your models then I'm going to do some speculation. NBTI only manifest itself mainly at high temperatures and although one could expect changes at any temp, I cannot see from your netlist the value you are using, neither do I know on whether your models have a temp threshold below which NBTI is ignored. Besides, NBTI has a recovery effect and again, this is in the models and I'm not sure what's in there. HSPICE definitely has a step by step example for you to run NBTI through MOSRA. They even provide an example with user URI (Unified Reliability Interface). Again, there is not enough in your description to debug. You could have posted a logFile to see any warnings about your run. You can contact me: riad dot kaced at intel.
I would recommend using the Designer's guide forum for such questions: The Designer's Guide Community Forum - Index.
Riad.
Thank you dick_freebird and Riad for your valuable comments. My simulations is running fine as I have followed properly Mosra manual examples.
However, I have a short question regarding the difference between PMOS "Threshold Voltage (Nominal/absoulte) of PDK/PTM" Vs "Threshold Voltage (fresh/pre-stress or normally get in mosra outfile at time=0)? I thought it should be same as nominal Vth but the values are different (as I am considering NBTI PMOS aging affect).
You don't say how different the "nominal" and the "fresh"
VT values are. Perhaps for purposes of reliability analysis
they picked a "fresh" value that's the worst case starting
position (since NBTI tends to be unidirectional, |VT|-increasing,
I would pick the high end of the foundry |VT| control range
for NBTI sims and let it degrade from there). But I have no
idea what your foundry data shows, or how they (or you)
might choose to worst-case things.
Hi usmanstar2,
I am not sure if you had your problem solved. But if your only goal is determine a Vth shift temporally, then why not just use the reaction-diffusion model outside the simulator and use this del-vth you obtain through the model into your nestlist (delvt0) and account for nbti. Like one of the previous mentioned comments, NBTI is largely dependent on the duty-cycle (ratio between time period during which the device is under stress and under recovery). You can find the models for RD theory right inside the PTM website from which you obtained you hspice model cards. If you need some help with the simulation, let me know.
Dear all
I'm simulating a simple circuit with "MOSRA Hspice", every thing is OK, but there is two problem:
1- Transistors age even there is no load on them (VCC and GND are dicsonnected)
2- No recovery process occurs when PMOS transistors are in relaxation state
I think I have missed some settings.
I'll be appreciated if any one could help.
Hi usmanstar2,
I am not sure if you had your problem solved. But if your only goal is determine a Vth shift temporally, then why not just use the reaction-diffusion model outside the simulator and use this del-vth you obtain through the model into your nestlist (delvt0) and account for nbti. Like one of the previous mentioned comments, NBTI is largely dependent on the duty-cycle (ratio between time period during which the device is under stress and under recovery). You can find the models for RD theory right inside the PTM website from which you obtained you hspice model cards. If you need some help with the simulation, let me know.
Thank you for your valuable suggestion. The only thing I am concerened that there are several authors who propose RD and Trapping/detrapping models. Kindly can you refer to a precise paper(s) which is giving very authentic/accurate model of BTI for RD and Trapping/Detrapping?
Hi,
I want to the HSPISE MOSRA simulation for NBTI and HCI. but I don't know how to set the parameters of this simulation (for example the values for TITTD, TITTC, and etc) for different technologies, e.g. 90nm and 45nm. when I do the simulation with the default values, for higher temperatures delta vth is lower than the lower temperatures, which is not actually true. would you please help me how I can define these values? Regards, Farzaneh Nakhaee
Hi,
I want to the HSPISE MOSRA simulation for NBTI and HCI. but I don't know how to set the parameters of this simulation (for example the values for TITTD, TITTC, and etc) for different technologies, e.g. 90nm and 45nm. when I do the simulation with the default values, for higher temperatures delta vth is lower than the lower temperatures, which is not actually true. would you please help me how I can define these values? Regards, Farzaneh Nakhaee
Hi,
unfortunately no. in addition to the problems posted above, I had a new problem in which different versions of hspice gave me different values for delta vth. I examine the same procedure on hspice 2014, 2013, 2012, and 2008 and different results has been obtained. have any one experienced the same problem?
best regards,
farzaneh nakhaee
