I am adding a new TMA-specific Tips page,
courtesy of Carl Huster. Your contributions are welcome. (Nov 2, 1998).
PISCES allows you to regrid on several parameters, typically
doping. Actually, early manuals probably went overboard with the
doping regridding. It's more important to regrid the channel and
the depletion layers than the actual junction location (well, at
least for MOSFETs). Turns out that you can regrid on an arbitrary
box if you simply omit the regrid criteria (e.g. DOPING,
POTENTIAL). You usually need to specify SMOOTH=1 to get well
formed triangles, but there are cases where you are better off
without it. (7/98)
My favorite undocumented TCAD trick is the "!" in SUPREM4,
which allows you to execute a shell command from within SUPREM4.
Scott Irving (National) used this to call S3 in order to
do a sheet resistance extraction in S4. My typical application
is to remove files to make sure they aren't there. In S4,
you can save the structure at every time step during an
oxidation. This can be useful for very slow oxidations, such
as a stress-dependent LOCOS simulation. Also useful if the
program crashes during an oxidation. "dump=1" saves after
every time step; the default prefix for the saved structures is
"s", and the rest of the file name will be the elapsed time,
e.g., s000000.002. Before you execute this, you might want to
remove the structures from previous runs:
! remove s000*
diffus time=100 temp=900 wet dump=1
Turns out that SILVACO has extended this capability to other
programs within Deckbuild/VWF with SYSTEM. This is not well
documented, but I find it highly useful. (5/12/98).
I used to close PISCES log files by opening up new dummy
log files. Atlas allows "log off" to close a file. But beware
of a recent discovery of mine: while running deckbuild on a Sun,
and remotely executing Atlas on an HP, an extract failed because
the extract started before the log file was closed. (4/21/98).
Silvaco has lots of very nice example files that you can
load and run in deckbuild. Tip: just load the file, and rather
than running it, jump to the tonyplot statements and simply
execute them. Loading the examples also loads the results! (4/21/98).
It's possible to do "physical" extracts in Atlas, as long as
you first save the structure, and then specify the structure file.
By default, Atlas loads the last log file. Now, once you know
that, you can extract the inversion layer thickness, which is handy
for comparing to your grid spacing. Here is an example:
save outf=lin.sol -- no quotes here
extract init inf="lin.sol" -- quotes here
extract name="i50" x.val from curve(depth,impurity="Current Flow" \
material="Silicon" mat.occno=1 x.val=0) where y.val=0.5
extract name="i90" x.val from curve(depth,impurity="Current Flow" \
material="Silicon" mat.occno=1 x.val=0) where y.val=0.1
extract name="i99" x.val from curve(depth,impurity="Current Flow" \
material="Silicon" mat.occno=1 x.val=0) where y.val=0.01
The above extracts the depth where 50, 90 and finally 99% of the
current flows. The extraction is done at x=0, which was the center
of the channel in my example (4/21/98).