[SOLVED] Are Dual Contacts mandatory?

[swordfish01]

[solved] Are Dual Contacts mandatory?

Hi,

I'm doing the layout for some circuits for the H35B4 Technology from AMS. Normally I use multiple contacts when layout transistors (e.g. for drain, source etc.), but now I have an inverter that is pretty small (0.8 over 0.35 and 0.4 over 0.35). Should I layout the thing with also two contacts per terminal? Until now I never had the problem, since the transistors where not that small...
We learned to use multiple vias when switching metal layers, there it is not a problem, here it starts to look really awkward, since the contact have a minimum size of 0.4...

Regards,
Andreas

 

[jgk2004]

I would strongly recommend doing atleast 2 contacts.. I know its hard/a pain when your in min size transistors but its a good habit to always have atleast two!
Jgk

 

[swordfish01]

Hi,

Thanks for your quick answer. I made most of those transistors with two contacts already, but I was wondering whether I could save the area/work to do it...
So be it, let there be contacts ...

Regards,
Andreas

 

[erikl]

I'm doing the layout for some circuits for the H35B4 Technology from AMS. Normally I use multiple contacts ...

... here it starts to look really awkward, since the contact have a minimum size of 0.4...

AFAIR double contacts are not enforced for 0.35µm processes (I think this changed at process sizes ≤ 0.18µm). However: would you risk a yield loss due to a loss of function of this transistor, for such a small win of area?

 

[swordfish01]

@erikl: No surely I would not want to have a malfunctioning transistor, but since there is no according design rule I wanted to ask.

 

[dick_freebird]

This is purely a yield calculation. Personally I don't ever,
but I do double up vias as these are a much lower yielding
chain in my processes. Never seen a contact chain fail.

What is your (say) -3sigma contact yield, how many
single contact devices are in your circuit, and how much
yield does that say you'll eat? If it's sub-%, fuggedaboudit.
10%, start driving the single devices out of the design.

 

[erikl]

Never seen a contact chain fail.

Designing analog - with usually enough space in our long transistors - we'd probably never see when one of the multiple contacts fail.

What is your (say) -3sigma contact yield, how many
single contact devices are in your circuit, and how much
yield does that say you'll eat? If it's sub-%, fuggedaboudit.
10%, start driving the single devices out of the design.

That's exactly what I wanted to tell! ;-)