How to get these two instance parameters, Mulu0 and Delvt0 in Spectre or Hspice?

[Alex Liao]

Hi,

I just noticed that in Spectre Manual (Virtuoso Simulator Components and Device Models Reference, V12.11 May 2013), there are two instance parameters.

One is:

19 mulu0=1 mobility multiplier. in BSIM3v3 Level-11 Model(bsim3v3) or 65 mulmu0=1.0 Mobility multiplier, alias of mulu0 in BSIM4 Level-14 Model (bsim4).

Another one is:

17 delvt0=0 V shift in zer0-bias threshold voltage vth0 in Bsim3v3. or in Bsim4.5 "A new instance parameter DELVTO that may be used to represent threshold voltage variation" And even the equation for delvto is given

However, these two parameters are only found in manual, not in model cards, or pint model parameters(psf) after simulation.
Do they really exist and if so how to retrirve them?

In Hspice 2010 vertion manual, these two are given in more detail. But I check hspice.out, still no appearance of these two as well as in model card. It looks both hspic and spectre documented them but I put a question mark on whether any simulator uses them. But if not use why document and why even giving equation?
Any reply is appreciated.

Regards,

Alex.

 

[erikl]

They can be used and - if so - they will be honoured by the simulator. But if they are not given in the foundry's model files, they simply have not been given attention by the model provider, perhaps they even have not been extracted from silicon, because this would need a lot of additional measurements to achieve good statistical relevance. These parameters are only necessary for local mismatch modeling, which in most cases is covered by MC mismatch parameters.

delvt0 is an important parameter for NBTI modeling, which usually isn't simulated by "normal" customers.

The only chance to get these parameters is to directly ask the model providers.

 

[Alex Liao]

Erikl, as a matter of fact, even these two my not be used in normal simulation I have found the shadow of them. After any simulation, in result browser, you could find element information in psf (resultant data). The reason why they are not appearing in model card or print model parameter is because they are not model parameter. They are instance parameter just like w,l, sa,sb. So they appear in psf and can be viewed **broken link removed**. The question here is those delvt0(V) = nan, mulu0 = nan. I cannot understand this. And the stimod is even stranger stimod = 2147483647.

Regards,
Alex

 

[erikl]

delvt0(V) = nan, mulu0 = nan. I cannot understand this. And the stimod is even stranger stimod = 2147483647.

Hi Alex, nan means not a number, and 2147483647 = 2e31 - 1 , the largest possible number for HSPICE on your computer (±32bit arithmetic). I think, the instance parameters aren't honoured, but I don't know, why.

 

[Alex Liao]

Hi Erikl, I just checked Hspice, in result browser there is no instance parameter (element information) filed whereas Spectre has. Those two parameters are generated after simulation, not as a input parameter. Those are mobility and Vth related and is a good verification on manually calculated mobility and vth. Otherwise the accuracy of mobility from a new model cannot be verified. Research may halt on this way and turn to others. Thank you so much on this thread, I would take those two as not used now for Hspice, but I would keep tracking my question on Cadence forum. If you have new idea on this, do not hesitate discussing with me.

Regards,
Alex

 

[dick_freebird]

Have you tried simply editing the model cards to add

+delvt0=0.0 mulu0=1.0

in the device params body, and see what happens? If it
doesn't barf, then try (say) setting

+delvt0=mydelvt0n mulu0=mymulu0n

instead, add those variables to the ADE list and see what
comes of varying them.

SPICE type simulators have always let you omit unused
params, that they are missing from a particular model
library doesn't really tell you anything. But a little bit of
experimenting might.

 

[Alex Liao]

Thank you dick_freebird.