metal filling and it's impact on timing.

[qual_ti]

Hi,

Why metal filling is done in ASIC design flow ?

If metal fill done, how it impacts the timing and cross talk.

Please share information regarding this.

Thanks in advance
qual_ti

 

[yadavvlsi]

Metal fills are dummy fill of metal pieces to avoid minimum density problem. If density on chip is less than a specified value this can cause problems during Chemical-mechanical planarization which in turn effect the planarity of subsequent layers. This can cause dishing effect. As these are just dummy metal pieces, It has no impact on timing and cross-talk.

 

[qual_ti]

thanks for info.

How these meal fills impact on timing and cross talk.

thanks in advance
qual_ti

 

[yadavvlsi]

As far as I know metal fill has no impact on timing and cross-talk.

 

[qual_ti]

hi

see this paragraph...


" Using a signoff quality metal density which adheres to the rule deck of foundry for metal filling at every stage of routing ensures that we get a signoff quality metal fill and avoids any surprises down the flow. When metal fill is done in a design it could as well impact the timing of the design and have a direct impact on the crosstalk, hence the metal fill capability should be timing aware, so that we do not see any timing degradation due to metal fill. With the ICC-ICV In-Design links, designer can call signoff quality ICV based metal fill within IC Compiler during the stages of routing and post-route optimizations "


This is one of the SNUG paper.

Please explain this ,if i am wrong...

 

[matter]

Metal fill will add coupling capacitance to the actual routes and hence cause timing disturbances. Based on designs that I have worked on so far, this impact will be very less. Typically we will have enough signoff margin to cover those.

 

[cop02ia]

Hi,

Actually depends how tight your timing margins are against .
If you have high-enough margins, it should be sufficient.
Else, coupling capacitance between metal fill shapes & some critical nets can cause problems.

We do not live in a perfect world, so some designs may have very tight margins, so, it is better to check how sensitive critical nets in the design are towards metal fills.
Probably shielding these nets could be sufficient - or use filler metals near critical nets as shields by tying them to ground.

Of course, this argument is mostly theory to me as in practise, critical nets I've had were always sufficiently shielded on same metal layer and sometimes above & below as well.

Best regards,
I-FAB

 

[rca]

you could also connect the metal filling to ground or power to reduce the IR drop.
regarding the timing impact, the metal filler tool avoid to add to closer any metal filling to net, to avoid any coupling.

 

[matter]

Well About Metal Fill And impacting on timing As far As I know the metal fill really has no impact on timing as well.

Hi tooroopknotts,
Did you even read my post above or any of the other posts from cop02ia and qual_ti ? Why would you say that "As far As I know the metal fill really has no impact on timing as well". Posts like this can confuse anyone who is new to this field. Can you explain what is the reasoning behind your statement ?

It is pretty simple and straight forward to see that metal fill will cause timing impact. It has to. It might very tiny depending on how dense your existing routes are. But nevertheless it certainly does impact coupling cap, which will in turn impact timing.
As cop02ia and other mentioned, your overall slack numbers might not degrade(depending on your margin), but it does NOT mean that timing is not getting impacted. The metal filling tools will take all steps to make sure that the slack degradation is very minimal or nill, but once again it does NOT mean that timing is not getting impacted by metal fill. If timing does not gets impacted like you say, then why should EDA R&D people break their heads trying to come up with newer and more efficient algorithms to make sure that the slack/qor impact is kept at a minimum ?

Regards

 

[naribtech]

Yes , I agree with Matter statement.
In a Highly dense design this metal fill effect the timing.



Thanks,
Nari

 

[qual_ti]

Thanks for all who helped in this discussion.

My concern is , how the metal fills effect the coupling cap ?

Does the metal fills connected to either Power or ground. I thought these are floating .

How steady state state signal increase or decrease the coupling cap? or only switching nets can induce the coupling cap ?

Pls help me here .

Regards
qual_ti