How to short small resistor for lvs purpose

Max,

Normally you need to have *.RESI in your CDL netlist to tell Dracula LVS to short resistors below a certain value. Do you have that?

I'm not aware of this having changed.

Regards,

Andrew.

thanks for the reply, however, my circuit was lvs clean when i was using dracula 4.8. Now since, i am using 4.9 is no longer short the resistor. i found no matter how the value of RESI is, it just does not work. here is my header cards:

*.BIPOLAR
*.RESI = 0.005
*.RESSIZE
*.CAPVAL
*.EQUATION
*.LDD
*.SCALE
*.DIOAREA
*.MEGA
.PARAM

something bugy in 4.9 dracula i think. BTW i am using IC5.41 and "lxRemoveDevice" property does not work for me.

i use this on the resistor instance as follow:

 lxRemoveDevice

Type: string

Name: lxRemoveDevice

Value: (SHORT(A B))

Thanx,

- Max

Max,

The problem is because you have *.RESSIZE in the netlist - it is therefore expecting the resistors to be defined by their size (dimensions) rather than by their value.

If you have this line omitted, or have *.RESVAL in the netlist, it will then short resistors less than 0.005 ohms (I tested this in Dracula  4.9.03-2010).

Note that using lxRemoveDevice to do this is only supported in IC61 onwards - it needs changes in the netlister to not netlist the component, and do the shorting at netlisting time.

Regards,

Andrew.

thanks for the quick response,  i already tried "*.RESVAL" once i use "*.RESVAL" the RESI is not being into accounted, therefore, no matter how much RESI value is, dracula short the resistor and i do not want this.

the CDF of my metal1 resistor shows L = 1nm, W = 400n, and shre/unitSq = 0.08ohm, so, the calculated value of the resistor = 200e-6ohm

as i said dracula version 4.8 has no problem with it.

for now to get around with this issue i am using "loglvs" with ver 4.8 and using "pdracula" with ver 4.9. this arrangement i get resistor shorted. i guess i looking for another permanent solutio, as you see i have to use *RESSIZE, since my resistor is defined as W&L.

- Max

 hi andrew, can you try *RESSIZE instead, if that works, i will link my dracula to ver 4.9.03-2010.

my resistor element is defined as w & l, so, i like to use *RESSIZE. btw do you know a way to short this resistor in netlist i am agian using 5.41?

- max

Max,

It's rather hard to see how *.RESI could work if you're using *.RESSIZE, given that the ignore threshold is specified as a resistance, and if the CDL netlist just has widths and lengths, it wouldn't know the resistance (the sheet resistance would not be known to LOGLVS).

In fact the documentation for *.RESI (ICinstDir/doc/dracularef/dracularef.pdf) says:

You must also specify the *.RESVAL command when using the *.RESI command. However, do not specify the *.RESSIZE command when using the *.RESI command as it may result in unpredictable behavior.

Maybe it worked in the older version due to the "unpredictable behavior" mentioned above - something that should not have been relied upon

Regards,

Andrew.

One other possibility would be to use a 0V source to short two nets in the CDL netlist. This is described in the LOGLVS chapter of the Dracula Reference manual mentioned above.

Regards,

Andrew.