• Next message: R. Timothy Edwards: "more on contacts. . ."
• Previous message: Stefanos Sidiropoulos: "Re: CVS, web pages, ftp sites, mirrors, etc."
• In reply to: Stefanos Sidiropoulos: "Re: CVS, web pages, ftp sites, mirrors, etc."
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

> 
> > maybe i've misunderstood you.  what i'm talking about is that, if you
> 
> You indeed misunderstand me...
> 
> 
> > look at the attached .mag file with a non-silicided process techfile,
> > like SCNA.80.tech27, the wells labelled 'pw' and 'nw' will *not* be
> > connected to gnd and vdd, respectively, since there will be n+/p+
> > junctions between the transistor diffusions and the ohmic diffusions
> > (well-contacts).  however, if you use a silicided process techfile,
> > like SCN3M_SUBM.30.tech27, you will see that the wells are connected
> > to their respective supplies, since the silicide removes those n+/p+
> > junctions, reflecting the fabricated part.  it may be a "weak"
> > connection but it doesn't have to carry much current (hopefully) to
> > keep the wells biased.  however, how can i make this clearer, it's NOT
> > weak!
> > 
> 
> Now take a look on the attached magic file (which is the same as yours
> with just a slight change). I am not saying that a designer would do that
> willingly but the point of extraction/LVS/simulation is to catch
> such mistakes. However, your SCN3M_SUBM.30.tech27 will happilly say 
> that the source of the PMOS is connected nicely to vdd. You can argue
> that this particular cell might work but a similar connection can carry 
> a LOT of current and definitely is VERY WEAK. Hope that is clear now.
> 

yes, it's clear now, the silicide connection is considered "soft" (a
"softconnect" in dracula lvs terminology), so your counter example is
a problem. it would be nice if magic knew about "soft" versus "hard"
connections, and could enforce hard connections except to
well/subtrate, or at least "flag" soft connections where they occur.
based on your input i'll separate out the connection from ohmic to
diffusion in the "connect" section, so it's easy to comment in or out.

i'm not sure why you think it's "VERY WEAK", though.  i'm quite sure
HP's electromigration limits for silicide in diffusion (in
amps/cm/hour) are comparable to that of aluminum metal (but granted
that's for "same flavor" diffusion and i don't know how the junction
changes that, if it does).

thanks for your input,

/jeff

• Next message: R. Timothy Edwards: "more on contacts. . ."
• Previous message: Stefanos Sidiropoulos: "Re: CVS, web pages, ftp sites, mirrors, etc."
• In reply to: Stefanos Sidiropoulos: "Re: CVS, web pages, ftp sites, mirrors, etc."
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]