• Previous message: Antony Lyons: "subject"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Dear Albert,

   In the cifoutput section, "scalefactor 1" is an equivalent scale
to "scalefactor 10 nanometers".  That is, the default units are
"centi-microns" (1/100 micrometer units).  When the keyword "nanometer"
is added, however, all units in the CIF section are interpreted as
being in nanometers, not centimicrons.  The units for grow, shrink,
bloat, and squares statements are *always* either centimicrons
(default) or nanometers (with the "nanometers" keyword).  The
"scalefactor" determines the scaling between magic internal units
and physical units.

> Magic also complained that the layout doesn't fall on an integer
> grid while doing "calma write file.gds" command.

I suspect that the "scalefactor 1 nanometers" is an incorrect scalefactor,
as you probably don't want or need the magic internal grid on a 1
nanometer spacing.  However, this is what contributes to the GDS output
error.  This is a subtle problem about which I exchanged a few emails
with Jeff Sondeen recently.  Magic places contact cuts inside contact
rectangles automatically.  But because contact rectangles can overlap,
and because they can overlap between cells rotated 90 degrees with
respect to one another, the contact cuts have to be centered in the
rectangle, or it is possible for two overlapping cells to have an
offset in the contact cut, creating a non-square cut in the output
(which is illegal in every fabrication technology I know).  But,
centering is averaging, so if the contact rectangle is not the right
size, the cut can be placed on a half-grid spacing.  Generally
speaking, as long as a half-grid spacing is not smaller than the
smallest dimension accepted by the foundry, there is no problem with
this.  But as technologies get down to the 90nm scale and beyond,
half units sometimes end up at sub-nanometer spacing, which is usually
not acceptable (this limitation is usually imposed by the company that
makes the masks).

In this case, though, magic is not checking against the minimum
dimension (which it doesn't know how to do, by the way), but there
is a limitation in the way it writes GDS output that limits the scale
to unit nanometers.  Jeff has informed me how to get around this problem,
so I will be fixing it sometime fairly soon, and probably adding an
"angstroms" keyword for those forward-thinking nanotech (or at least
nanoscale) people out there.

For now, it is wise not to make the magic grid smaller than 2 nanometers
if you want correct GDS output.

					Regards,
					Tim

• Next message: R. Timothy Edwards: "mpla, mquilt, meg, and peg"
• Previous message: Antony Lyons: "subject"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]