Spectre: weird warnings when using well-known design kits

S

spectrallypure

Guest
Hello all,

I always get strange warnings when simulating my designs using well-
known design kits. Here are some examples:

*Spectre 5.10.41.062906 with IBM PDK for technology bicmos7wl (latest
MOSIS release):
....
Notice from spectre during circuit read-in.
dpbemod: `is' has the unusually large value of 1 A.
dpbemod: `imax' has the unusually large value of 1e+99 A.
dpbcmod: `is' has the unusually large value of 1 A.
dpbcmod: `imax' has the unusually large value of 1e+99 A.
Notice from spectre during hierarchy flattening.
dpbsmod: `imax' has the unusually large value of 1e+15 A.
Further occurrences of this notice will be suppressed.
Notice from spectre in `nfet':`MN0', in `Terminacion_Simple':`I808',
during
hierarchy flattening.
I808.MN0.nch: `jsw' has the unusual value of 50 fA/m.
I808.MN0.nch: `n' has the unusual value of 1.02.
I808.MN0.nch: `ckappa' has the unusual value of 6.7867.
I808.MN0.ndio: `rs' has the unusually small value of 120 pOhm.
I808.MN0.cwgd: Terminals are connected together (to node
`net01595').
....

**Spectre 5.10.41.062906 with AustriaMicrosystems Hit-Kit v3.70 (tech
C34B5):
....
Notice from spectre in `modn', during circuit read-in.
modn.mosinsub: `jsw' has the unusual value of 130 pA/m.
Notice from spectre in `modnm', during circuit read-in.
modnm.mosinsub: `jsw' has the unusual value of 130 pA/m.
Notice from spectre in `modpm', during circuit read-in.
modpm.mosinsub: `jsw' has the unusual value of 610 pA/m.
Notice from spectre in `modp', during circuit read-in.
modp.mosinsub: `jsw' has the unusual value of 610 pA/m.
Notice from spectre in `modnh', during circuit read-in.
modnh.mosinsub: `jsw' has the unusual value of 130 pA/m.
Further occurrences of this notice will be suppressed.
....

Searching in the model files for both technologies I can see that the
models indeed define these parameters with those values. Then, why do
these warning messages appear? are they meaningful? how can I correct/
suppress them? Do they affect the simulation results? Can the
simulation time be somehow affected by the nature of these values (too
big or too tiny sometimes means overflow in calculations, right?).

The simulations continue to run despite these messages, and most of
the time (say 80%) the results seem to be congruent with what is
expected but sometimes they don't seem to converge at all. ...I really
wonder if what I am getting is the correct thing!

Thanks in advance for any help/ideas/comments.

Regards,

Jorge Luis.
 
On Sun, 18 Nov 2007 12:19:37 -0800 (PST), spectrallypure <jorgelagos@gmail.com>
wrote:

Hello all,

I always get strange warnings when simulating my designs using well-
known design kits. Here are some examples:

*Spectre 5.10.41.062906 with IBM PDK for technology bicmos7wl (latest
MOSIS release):
...
Notice from spectre during circuit read-in.
dpbemod: `is' has the unusually large value of 1 A.
dpbemod: `imax' has the unusually large value of 1e+99 A.
dpbcmod: `is' has the unusually large value of 1 A.
dpbcmod: `imax' has the unusually large value of 1e+99 A.
Notice from spectre during hierarchy flattening.
dpbsmod: `imax' has the unusually large value of 1e+15 A.
Further occurrences of this notice will be suppressed.
Notice from spectre in `nfet':`MN0', in `Terminacion_Simple':`I808',
during
hierarchy flattening.
I808.MN0.nch: `jsw' has the unusual value of 50 fA/m.
I808.MN0.nch: `n' has the unusual value of 1.02.
I808.MN0.nch: `ckappa' has the unusual value of 6.7867.
I808.MN0.ndio: `rs' has the unusually small value of 120 pOhm.
I808.MN0.cwgd: Terminals are connected together (to node
`net01595').
...

**Spectre 5.10.41.062906 with AustriaMicrosystems Hit-Kit v3.70 (tech
C34B5):
...
Notice from spectre in `modn', during circuit read-in.
modn.mosinsub: `jsw' has the unusual value of 130 pA/m.
Notice from spectre in `modnm', during circuit read-in.
modnm.mosinsub: `jsw' has the unusual value of 130 pA/m.
Notice from spectre in `modpm', during circuit read-in.
modpm.mosinsub: `jsw' has the unusual value of 610 pA/m.
Notice from spectre in `modp', during circuit read-in.
modp.mosinsub: `jsw' has the unusual value of 610 pA/m.
Notice from spectre in `modnh', during circuit read-in.
modnh.mosinsub: `jsw' has the unusual value of 130 pA/m.
Further occurrences of this notice will be suppressed.
...

Searching in the model files for both technologies I can see that the
models indeed define these parameters with those values. Then, why do
these warning messages appear? are they meaningful? how can I correct/
suppress them? Do they affect the simulation results? Can the
simulation time be somehow affected by the nature of these values (too
big or too tiny sometimes means overflow in calculations, right?).

The simulations continue to run despite these messages, and most of
the time (say 80%) the results seem to be congruent with what is
expected but sometimes they don't seem to converge at all. ...I really
wonder if what I am getting is the correct thing!

Thanks in advance for any help/ideas/comments.

Regards,

Jorge Luis.
Click to expand...
The model implementation has some checks built in to warn for excessively large
values - in case somebody has made a mistake in entering them.

BTW, you're using an old version of spectre - you should really use an MMSIM
version of spectre (e.g. MMSIM62) - although check what the hitkit suggests you
use.

You probably should talk to AMS about the warnings?

Regards,

Andrew.
--
Andrew Beckett
Senior Solution Architect
Cadence Design Systems, UK.
 
Thanks for your reply, Andrew. I understand the messages are due to
the range checks, and it is precisely because of this that I really
find it difficult to believe that both IBM and AMS are distributing
their kits with such trivial bugs (i.e.: since they set the both the
parameter values and the allowed ranges on when making the design
kits!) ...Honestly, I rather believe I'm missing some configuration or
parameter of the like...

I followed your suggestion and tried to use a newer version of
spectre, and the errors still appear. For instance, the IBM PDK for
their BiCMOS 7WL 0.18um technology says the following before
performing the simulations, this time with Spectre sub-version
6.0.2.290:

(...)
Command line:
/programs/cadence/MMSIM_6.0/tools.lnx86/spectre/bin/32bit/spectre
\
-env artist5.1.0 +escchars +log ../psf/spectre.out
+inter=mpsc \
+mpssession=spectre0_6126_1 -format sst2 -raw ../psf
+lqtimeout \
900 -maxw 5 -maxn 5 +param \
/programs/cadence/DESIGN_KITS/ams_v3.70/spectre/
ams_range.lmts \
input.scs
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libinfineon_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libnortel_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libphilips_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libsparam_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libstmodels_sh.so ...
spectre (ver. 6.0.2.290 -- 27 Jun 2006).
Includes RSA BSAFE(R) Cryptographic or Security Protocol Software from
RSA Security, Inc.

Simulating `input.scs' on LETSBEGREAT2 at 12:18:55 AM, Sat Nov 24,
2007.
(...)
Notice from spectre in `nfet':`I808.TN0', in `nfetest':`I808.TN0', in
`Terminacion_Simple':`I808', during hierarchy flattening.
I808.TN0.nfetest.nch: `n' has the unusual value of 1.02.
I808.TN0.nfetest.nch: `jsw' has the unusual value of 50 fA/m.
I808.TN0.nfetest.nch: `ckappa' has the unusual value of 6.7867.
I808.TN0.nfetest.ndio: `rs' has the unusually small value of 120
pOhm.
Notice from spectre in `nfet':`I888.I115.I0.TN1', in
`nfetest':`I888.I115.I0.TN1', in `_sub53':`I888.I115.I0', in
`Seccion_SISO1_Inicial_3':`I888.I115', in `SISO1':`I888', during
hierarchy flattening.
I888.I115.I0.TN1.nfetest.nch: `n' has the unusual value of 1.02.
I888.I115.I0.TN1.nfetest.nch: `jsw' has the unusual value of 50 fA/
m.
Further occurrences of this notice will be suppressed.
I888.I115.I0.TN1.nfetest.ndio: `rs' has the unusually small value
of 120 pOhm.
Notice from spectre in `pfet':`I888.I115.I0.TP1', in
`pfetest':`I888.I115.I0.TP1', in `_sub53':`I888.I115.I0', in
`Seccion_SISO1_Inicial_3':`I888.I115', in `SISO1':`I888', during
hierarchy flattening.
I888.I115.I0.TP1.pfetest.pch: `noib' has the unusually large value
of 1.38e+06.
I888.I115.I0.TP1.pfetest.pdio: `rs' has the unusually small value
of 120 pOhm.
Notice from spectre in `pfet':`I888.I115.I4.I1.TP8', in
`pfetest':`I888.I115.I4.I1.TP8', in `_sub69':`I888.I115.I4.I1', in
`_sub70':`I888.I115.I4', in `Seccion_SISO1_Inicial_3':`I888.I115', in
`SISO1':`I888', during hierarchy flattening.
I888.I115.I4.I1.TP8.pfetest.pch: `noib' has the unusually large
value of 1.38e+06.
I888.I115.I4.I1.TP8.pfetest.pdio: `rs' has the unusually small
value of 120 pOhm.
Further occurrences of this notice will be suppressed.

(Circuit inventory: nodes 188932, equations 272799, iprobe 80, bsim3v3
41852, capacitor 125556, diode 83704, isource 992, resistor 167408,
vsource 3)
(...)

I really haven't succeeded to find any reference whatsoever in the
design kit documentation about these warnings. I am not entering ANY
model value by hand, just using the model files as provided by MOSIS
(I already checked on their site and I have the latest release of the
kit--no patches or the like available). This is why I really find it
difficult to believe that I'm the only one getting these errors!!!

Thanks once again for any help/ideas/comments.

Regards,

Jorge Luis.
 
I believe that they might intentionally use offrange values if that allow to better fit the model to
real device behavior. If that's the case, there should probably be some mention of it in the kit
documentation, but it's not always very complete..

Anyway your best bet is to ask your kit provider.

Stéphane

spectrallypure wrote:
Thanks for your reply, Andrew. I understand the messages are due to
the range checks, and it is precisely because of this that I really
find it difficult to believe that both IBM and AMS are distributing
their kits with such trivial bugs (i.e.: since they set the both the
parameter values and the allowed ranges on when making the design
kits!) ...Honestly, I rather believe I'm missing some configuration or
parameter of the like...

I followed your suggestion and tried to use a newer version of
spectre, and the errors still appear. For instance, the IBM PDK for
their BiCMOS 7WL 0.18um technology says the following before
performing the simulations, this time with Spectre sub-version
6.0.2.290:

(...)
Command line:
/programs/cadence/MMSIM_6.0/tools.lnx86/spectre/bin/32bit/spectre
\
-env artist5.1.0 +escchars +log ../psf/spectre.out
+inter=mpsc \
+mpssession=spectre0_6126_1 -format sst2 -raw ../psf
+lqtimeout \
900 -maxw 5 -maxn 5 +param \
/programs/cadence/DESIGN_KITS/ams_v3.70/spectre/
ams_range.lmts \
input.scs
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libinfineon_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libnortel_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libphilips_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libsparam_sh.so ...
Loading /programs/cadence/MMSIM_6.0/tools.lnx86/cmi/lib/4.0/
libstmodels_sh.so ...
spectre (ver. 6.0.2.290 -- 27 Jun 2006).
Includes RSA BSAFE(R) Cryptographic or Security Protocol Software from
RSA Security, Inc.

Simulating `input.scs' on LETSBEGREAT2 at 12:18:55 AM, Sat Nov 24,
2007.
(...)
Notice from spectre in `nfet':`I808.TN0', in `nfetest':`I808.TN0', in
`Terminacion_Simple':`I808', during hierarchy flattening.
I808.TN0.nfetest.nch: `n' has the unusual value of 1.02.
I808.TN0.nfetest.nch: `jsw' has the unusual value of 50 fA/m.
I808.TN0.nfetest.nch: `ckappa' has the unusual value of 6.7867.
I808.TN0.nfetest.ndio: `rs' has the unusually small value of 120
pOhm.
Notice from spectre in `nfet':`I888.I115.I0.TN1', in
`nfetest':`I888.I115.I0.TN1', in `_sub53':`I888.I115.I0', in
`Seccion_SISO1_Inicial_3':`I888.I115', in `SISO1':`I888', during
hierarchy flattening.
I888.I115.I0.TN1.nfetest.nch: `n' has the unusual value of 1.02.
I888.I115.I0.TN1.nfetest.nch: `jsw' has the unusual value of 50 fA/
m.
Further occurrences of this notice will be suppressed.
I888.I115.I0.TN1.nfetest.ndio: `rs' has the unusually small value
of 120 pOhm.
Notice from spectre in `pfet':`I888.I115.I0.TP1', in
`pfetest':`I888.I115.I0.TP1', in `_sub53':`I888.I115.I0', in
`Seccion_SISO1_Inicial_3':`I888.I115', in `SISO1':`I888', during
hierarchy flattening.
I888.I115.I0.TP1.pfetest.pch: `noib' has the unusually large value
of 1.38e+06.
I888.I115.I0.TP1.pfetest.pdio: `rs' has the unusually small value
of 120 pOhm.
Notice from spectre in `pfet':`I888.I115.I4.I1.TP8', in
`pfetest':`I888.I115.I4.I1.TP8', in `_sub69':`I888.I115.I4.I1', in
`_sub70':`I888.I115.I4', in `Seccion_SISO1_Inicial_3':`I888.I115', in
`SISO1':`I888', during hierarchy flattening.
I888.I115.I4.I1.TP8.pfetest.pch: `noib' has the unusually large
value of 1.38e+06.
I888.I115.I4.I1.TP8.pfetest.pdio: `rs' has the unusually small
value of 120 pOhm.
Further occurrences of this notice will be suppressed.

(Circuit inventory: nodes 188932, equations 272799, iprobe 80, bsim3v3
41852, capacitor 125556, diode 83704, isource 992, resistor 167408,
vsource 3)
(...)

I really haven't succeeded to find any reference whatsoever in the
design kit documentation about these warnings. I am not entering ANY
model value by hand, just using the model files as provided by MOSIS
(I already checked on their site and I have the latest release of the
kit--no patches or the like available). This is why I really find it
difficult to believe that I'm the only one getting these errors!!!

Thanks once again for any help/ideas/comments.

Regards,

Jorge Luis.
Click to expand...
 
Hi again! I just wanted to comment, for the sake of completeness, that
I finally closed the case on these two issues:

-With the IBM PDK, Andrew you were right; changing to MMSIM
effectively eliminates the warnings (sorry for my previous post, I was
using a corrupted library!)
-Wtih the AMS HitKit, you guys were also right: I consulted with the
provider and turns out its a known bug that occurs when using spectre.
They told me it might be solved in newer releases of the kit.

Thanks a lot to everyone for your valuable help!

Regards,

Jorge Luis.
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top