|
|
Magic Mailing List |
|
From: Philippe O. Pouliquen (philippe AT alpha DOT ece.jhu.edu) Date: Fri Mar 30 2001 - 16:06:34 EST
Robert Penny <rob@network.ucsd.edu> and Andrew Lines <lines AT avlsi DOT com> both mentioned the fact that CIF rectangles are specified by their center coordinates and size. I had indeed forgotten about that. I think the contact issue is more severe, in the sense that other layout programs, which expect all layer rectangles to be on a lambda grid, balk at reading Magic's CIF. (For those of you not familiar with the contact issue, if you draw a 5x5 contact in magic, the actual contact cut (2x2) will be on a half-lambda boundary.) In addition, Andrew Lines reported: > Yes, the problem is that CIF rectangles are center specified, so the > center must line up on a centimicron grid. So if you have odd sized > rectangles and an odd lambda, you're in trouble. Naturally, CIF > write gives no warnings, so you just get a mask set with lots of > tiny gaps. This is the crux of the matter: CIF *does not* restrict items to lie on the centimicron grid. If you have *actual experience* of a correctly written CIF file, which when fabricated had gaps in its layers, who fabbed it? My experience is that things get fabbed just fine. In fact, I've used Magic to write CIF for a process for which lambda was an odd number of centimicrons (TMSC 0.35 IIRC). The process I went through was to write CIF with lambda=0.5 microns, and then change all the DS statements in the CIF to divide by 2. Certainly, this does not preclude Magic from producing incorrect CIF. Robert Penny <rob AT network DOT ucsd.edu> wrote: > I'm a novice at this, but as luck would have it I do have a copy of > Mead & Conway next to my desk. Don't let anybody borrow it! :-) Since you have the book, why don't you check to see if there is any restrictions on the CIF DS statement (i.e. that the resulting scale must be an integer) and settle this once-and-for-all. Philippe Pouliquen The Johns Hopkins University
|
|
|
|