• Previous message: R. Timothy Edwards: "CIF and nanometers"
• In reply to: R. Timothy Edwards: "CIF and nanometers"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

R. Timothy Edwards writes:
 > Dear Magic Hackers (esp. Jeff Sondeen and Philippe Pouliquen):
 > 

 >    There were some email posts to the magic-dev group a while back
 > about the implementation of the "nanometers" keyword in the
 > cifoutput section of the magic techfile.  As I recall, Philippe
 > argued that the 1 centimicron limitation was not a limitation of
 > the CIF spec, per se, because the DS command can always be
 > specified with a scale divisor such that "DS 0 1 10" would
 > effectively put all measurements in nanometers, assuming all prior
 > scaling is 1:1.  If so, then to deal with tech feature sizes like
 > 0.13 um, the cifoutput section should really allow a "scale
 > expander" in addition to the "scale reducer", so the CIF output
 > file can make use of this scale expander instead of having a scale
 > expander fixed at 2.  "scale" values which are odd numbers in the
 > cifoutput section would automatically cause both the scale and the
 > expander to be doubled (so box centers still fall on integer
 > boundaries).  "nanometers" would effectively be a shorthand for a
 > scale expander of 10.  "calmaonly" would be unnecessary, and would
 > do nothing.
 > 

 >    The real question, then, is whether any CIF readers (especially MOSIS)
 > make an implicit assumption about minimum CIF values being integer
 > centimicrons, or whether they follow the spec to the letter and allow

my understanding (that Philippe pointed out) is that the units of the
cif file don't imply anything about the minimized sized object in the
cif file.  just the units it's expressed in, and that mosis honors
this.  however, remember when mosis reads SCALABLE RULES (cif or gds),
they snap all vertices to the nearest 1/2 lambda, so we're really
talking about what mosis does when it reads vendor design rules, which
i believe is nothing except what is required as 'mask preparation
operations' specified by the wafer vendor.  

 > any scaling.  I know that MOSIS rounds off, such as the TSMC 0.25 um
 > process becomes a 0.24 um process in the CIF output sent to MOSIS, and
 > I presume that MOSIS scales it by a factor of 25/24 before passing it

nope. i don't know where you're getting these .25/.24 numbers.  i
think mosis does nothing special with layout given in vendor rules.

for layout in scalable rules, the way Mosis works is that there's
aways enough slack in the scalable rules (with respect to the required
manufacturer's mask bloats/shrinks, not necessarily the manufacturers'
own drawing rules), so that Mosis can bloat/shrink the layout into
legal manufacturer's rules.

 > on to the vendor.  How would MOSIS handle a CIF file that used a scale
 > expander to exactly represent lambda=0.125?  Would the CIF file parser
 > hiccup and die, or would it actually take the input as written?  (Jeff,
 > can you answer this question?)
 > 

i hope this is answered.

 >    If MOSIS can handle sub-centimicron measurements in CIF, I intend
 > to rewrite the CIF section to take a scale expander from the tech file
 > and generate output accordingly, as outlined above.  Suggestions,
 > please?
 > 

great idea!!!

/jeff

 > 						Regards,
 > 						Tim

• Next message: R. Timothy Edwards: "process scaling"
• Previous message: R. Timothy Edwards: "CIF and nanometers"
• In reply to: R. Timothy Edwards: "CIF and nanometers"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]