• Previous message: Graham Petley: "Updated vsclib and vxlib libraries available"
• Maybe in reply to: Graham Petley: "Problem with choice of names for nfet anf pfet?"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Dear Graham,

> According to the documentation, the name of the fet can be a
> user defined string with meaning to the simulation program
> (like Spice). But I find that the name must be either nfet or
> nmos (or pfet or pmos), otherwise the diffusion area and
> perimeter values that appear in the Spice deck after exttospice
> are wrong.

The default names can be overridden by passing information to
ext2spice with the "-j" option (see the man page for ext2spice).

However, note that in version 7.2 (which revision, I'd have to
research), the TCL-based version uses the command "exttospice"
instead of the standalone program "ext2spice".  I coupled the
"exttospice" command more tightly to the program itself, so that
it picks up information from the technology from the internal
database, not from the .ext file (although it checks against the
information in the .ext for basic sanity checking).  This allows
exttospice to pick up all the defined FET names from the tech
file, instead of generating only the two defaults.  Note that
because this can ONLY be done when the exttospice program has
complete access to the magic database, it ONLY works in the TCL
version!

There is also a problem with ext2spice in that it assumes a
specific ordering of the "resistclass" lines in the technology
file.  Otherwise, all the resistance classes of source/drain
materials will get screwed up.  Again, this is corrected in
the TCL "exttospice" command.  Note that there is *no way* to
get around this particular error other than to *not touch*
the "resistclass" lines in the tech file.

As a final note, magic version 7.2 implements the "device mosfet"
method to replace the "fet" lines of the original tech file
format.  This method gets rid of a lot of problems stemming from
the "fet" format, including device names, and the ability to
correctly extract annular (ring) devices.  Also, in order to
avoid having to specify every type of device as some sort of
mutant FET, the "device" method also includes "device bjt",
"device capacitor", and "device resistor".  If you are interested
in how these work, please let me know.  I have not documented
them yet, except for a few references in the magic-7.2 web pages.
The "device capacitor" and "device resistor" methods are nice, in
that they allow you to get a "C" or "R" record in the SPICE deck
along with a capacitance or resistance record (or length and width,
if "device resistor" defines a model type).  The short explanation
of the syntax for "device" was mentioned in an email I sent to Charles
Krinke and CC'd to magic-dev.  To reiterate:

----------------------------------------------------------------------
        device mosfet <name> <gate_types> <source,drain_types> \
                        <substrate_types>|None \
                        <default_substrate_name> \
                        [perim_cap] [area_cap]
        device bjt <name> <base_types> <emitter_types> \
			<collector_types>
        device resistor <name> <resistor_types> <terminal_types>
        device capacitor <name> <top_types> <bottom_types> \
			[<perim_cap>] <area_cap>
        device subcircuit <name> <gate_types> [<terminal_types> \
                        [<substrate_types>]]
----------------------------------------------------------------------

I hope that answers your questions!

					Regards,
					Tim

• Next message: Gordon (Jake) Jacobs: "Postscript section for SCN5M process"
• Previous message: Graham Petley: "Updated vsclib and vxlib libraries available"
• Maybe in reply to: Graham Petley: "Problem with choice of names for nfet anf pfet?"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]