[Magic-dev] Broken Nets

[Satpal Singh]

Dear Tim,

The problem of broken nets disappears if I extract the design using the 
same technology file (SCN3ME_SUBM.30.tech27) but use magic 6.5 instead of 7.1

The problem also disappeared when using magic 7.1, I used the option, 
"extract no capacitance".

The problem of broken nets was that on some designs, it appears to be 
random, on extraction and ext2spice, one net is broken into 2 nets, which 
leads to failure of simulation. In reality, there is only one net as can be 
verified by highlighting the net on the layout. Also using the "getnode" by 
selecting different portions of the net, the same node name is returned, 
however the "ext all" followed by ext2spice shows 2 nets.

It also seems that sometimes by shifting the position of the label with the 
same cell, or shifting part of the track in the layout, the problem goes 
away while it may appear at other areas.

I had also extracted the design using the standard scmos tech file with 
magic 7.1, and I still had the same nets shown as 2 nets, so the problem 
does not seem to be tech file specific.

I am not a programming expert, but I know enough to install Linux and 
magic, I have used magic 6.4 and 6.5 for 10 years now and have done several 
designs, so I do not think I am doing something terribly wrong (or may be I 
am and I was just lucky all along !! ).

Do you recommend that I dump magic 7.1 (which I had downloaded only 
recently) and move on to  higher versions, my choice of 7.1 was because it 
said its a stable version.

I would greatly appreciate your advise.

Regards,
Satpal Singh
Prime Circuits Inc. / GSFC




>X-POP3-Rcpt: singh at primecircuits.com
>Date: Tue, 15 Mar 2005 11:56:44 -0500
>From: "R. Timothy Edwards" <tim.edwards at multigig.com>
>To: singh at primecircuits.com
>Subject: RE: Broken nets
>Cc: magic-dev at csl.cornell.edu
>
>Dear Satpal,
>
> > I recently downloaded Magic 7.1 and am using the tech file
> > SCN3ME_SUBM.30.tech27 to design for AMI 0.5 um process. On extraction
> > (ext all) and running ext2spice to generate a netlist, I occasionally
> > (not always) find that a single net in the layout has been broken
> > into 2 different nets in the spice file, this leads to the failure of
> > simulation.
>
>There are a number of reasons this sort of thing can occur.  The
>reason that they seem "random" is that the extractor is not to
>blame---the extraction algorithm in magic is surprisingly robust---
>but there are a number of other subtle errors that can occur.
>
>One of them, which may have caused at least one of your broken
>simulations, is that labels can get copied from one cell into another,
>and end up on separate layer types (e.g., metal1 and metal2) that
>are not connected.  Extraction will short nets having the same
>(labeled) name.
>
>Usually it is a good idea to ensure that the circuit layout is correct
>by doing LVS (layout vs. schematic) as well as extraction+simulation.
>
>
>                                         Regards,
>                                         Tim