|
|
Magic Mailing List |
|
From: Robert Penny (rob AT network DOT ucsd.edu) Date: Fri Mar 30 2001 - 12:34:12 EST
On Fri, 30 Mar 2001, Philippe O. Pouliquen wrote: > It is true that historically, Magic has prohibited the use of CIF when > lambda is not an even integer number of centimicrons. I believe that > this may be due to an early definition of CIF prohibiting features > smaller than 1 centimicron (coupled with Magic contact placement), or > perhaps an early version of the DS statement didn't have a > denominator. Maybe somebody with a copy of Mead & Conway can check on > this. I'm a novice at this, but as luck would have it I do have a copy of Mead & Conway next to my desk. Quoting from "Introduction to VLSI systems", Mead & Conway 2nd printing 1980. *************************************************** Measurements ------------ The intermediate form uses a right-handed coordinate system shown in Fig. 4.18, with x increasing to the right and y increasing upward. (Directions and distances are always represented in terms of the front surface of the finished chip, not in terms of the various sizes and mirrorings of the intermediate artifacts.) The units of distance measurements are hundredths of a micron (um); there is no limit on the size of a number. [Programs reading numbers from CIF files should check carefully that the number does not overflow the number of bits in the internal representation used, and should specify their own limits, if any.] *************************************************** Now I'm speculating here (like I say, I'm no expert) but the even multiple problem may arise from that fact that CIF specifies a center and size for a box, whereas Magic specifies corners. If Magic didn't use the even centimicron multiple then you'd generate CIF centers that are sub-centimicron in size. Is this where the limitation came from? -Rob Penny (rob AT ucsd DOT edu)
|
|
|
|