Cadence Spectre error message for TSMC MM 0.18 PDK

[Analogworld]

spectre cadence 6.0 ic5033

Hi All,

Someone has posted this topic before, but no one has ever provided a solution to solve this issue yet. I am now having the same situation that Spectre can not find the transistor model in the correctly pointed tsmcrf library. I am sure that the require Spectre model file is setup correctly, and it works fine for the other devices (i.e. well or poly resistors) in the same file location. It just not work on the transistor models. Why?

I even looked into the pointed model file "rf018.scs" that provided by TSMC's cadence design kit. However, the transistor model named "nmos2v" (as shown in the error message), doesn't even exsit in the Spectre model file! And Cadence do not let me change the transistor's model name from the property menu.

Is there a problem with the TSMC's Cadence design kit? I doubt it though! Then how should I get it to work? Did anyone exprience this problem before? Any help would be greatly appreciated.

Spectre.out error message :
***********************************************************************************************
Command line:
/Cadence/IC5033/tools.lnx86/spectre/bin/spectre -env artist5.0.0 \
+escchars +log ../psf/spectre.out +inter=mpsc \
+mpssession=spectre0_3023_3 -format psfbin -raw ../psf \
input.scs
spectre pid = 3618

Loading /Cadence/IC5033/tools.lnx86/spectre/lib/cmi/3.0/libinfineon_sh.so ...
Loading /Cadence/IC5033/tools.lnx86/spectre/lib/cmi/3.0/libnortel_sh.so ...
Loading /Cadence/IC5033/tools.lnx86/spectre/lib/cmi/3.0/libphilips_sh.so ...
Loading /Cadence/IC5033/tools.lnx86/spectre/lib/cmi/3.0/libstmodels_sh.so ...
spectre (ver. 5.0.33.031705 -- 17 Mar 2005).
Includes RSA BSAFE(R) Cryptographic or Security Protocol Software from RSA
Security, Inc.

Simulating `input.scs'

Compiling ahdlcmi module library.
Finished compilation in 1 s (elapsed).
Installed compiled interface for bsource_1.
Installed compiled interface for bsource_2.
Installed compiled interface for bsource_3.
Installed compiled interface for bsource_5.
Installed compiled interface for bsource_41.
Installed compiled interface for bsource_42.
Installed compiled interface for bsource_56.
Installed compiled interface for bsource_68.
Installed compiled interface for bsource_69.
Installed compiled interface for bsource_83.
Installed compiled interface for bsource_85.
Installed compiled interface for bsource_90.
Installed compiled interface for bsource_101.

Error found by spectre during circuit read-in.
input.scs: M0 is an instance of an undefined model nmos2v.

spectre terminated prematurely due to fatal error
*************************************************************************************

 

[steer]

wrong model name in input.scs

It looks like something messed up in your CDF setups.
There should be no reference to nmos2v in your netlist. Change your CDF info to set the correct model names.

 

[vvchan]

modelname

I have exactly the same problem with this. I am trying to get a high resistive poly resistor. I can find it in the "input.scs" and when I call it into the schematic. The modelname is also matched with the name found in "input.scs" file. Does anyone experience it before?

 

[Analogworld]

Hi Steer,

Yes, I can use Edit component CDF in Cadence to change the cell name, but are we suppose to do that? What if we change the name to the other transistors that is not intend to be used? I just doubt that TSMC will have to let the users to do this manually. I think I do setup the files correctly. Any ideas? This is just driving me crazy!

 

[Teddy]

What I will say will probably used against me at th ecourt of law but anyway - TSMC does not create their PDKs. Those are done by Cadence and there ARE disconnects. I did not see ANy PDK working 100% yet.
My personal and private opinion is that if they would provide perfect pDK and something would go wrong , companies would use it against them and would hold them accounatable..... Again - it is just my private opinion and I can not provide any evidence.... well I'm trying to be pc...

I believe you MUST edit the PDK. I have seen before errors in pcells models etc.
If you got pdk from MOSIS then ask them. If you got it from tsmc - try to ask them for help.
but I bet you will end up modifying pdk to your needs or atleast to make it workable.
If device is calling wrong model name - ask tsmc for sure.

 

[Analogworld]

What I will say will probably used against me at th ecourt of law but anyway - TSMC does not create their PDKs. Those are done by Cadence and there ARE disconnects. I did not see ANy PDK working 100% yet.
My personal and private opinion is that if they would provide perfect pDK and something would go wrong , companies would use it against them and would hold them accounatable..... Again - it is just my private opinion and I can not provide any evidence.... well I'm trying to be pc...

I believe you MUST edit the PDK. I have seen before errors in pcells models etc.
If you got pdk from MOSIS then ask them. If you got it from tsmc - try to ask them for help.
but I bet you will end up modifying pdk to your needs or atleast to make it workable.
If device is calling wrong model name - ask tsmc for sure.

Thanks for the note Teddy. I got the PDK from MOSIS, I will probably email them to ask why. Is this is the way how we suppose to use the design kit? I just doubt that a big foundry like TSMC would have to let the designer to select/change their model manually.

 

[Teddy]

Why not - I have seen worse.