A serious shielding questions

[mappycarol]

parallel shielding

I have several "quiet" signals need to be shielded by ground in a mix-signal chip with UMC 6 metal process.

For better result, a designer proposed that the signals should be shielded be 4 walls-- 2 vertical metal layers running below and above , 2 parallel running along. The thickness of the metal layers is about 0.3 um and the minimum width of signals is 0.2um.

I think this way is too costly, because it'll use up 3 metal layers routing together, plus the width is less than thickness. From parasitic capacitance point of view , parallel is dominant. Does this trade-off worth it?

What do you think?

And does anybody tell me the effect of 4-wall shieding and 2-wall shielding? [/b]

 

[jerryzhao]

The shielding by 4 lines like the coaxial cable, It can shield the singal best but large parasitic and area.
About the 2 lines parallel shielding, if the above and below no other signal cross, I think you can use 2 parallal lines shielding.
But some signals must cross this sensitive line.In this instance you'd better use coaxial cable shielding.
In order decrease the parasitic capacitor you can shield by 2 lines, only when some signals crossed by coaxial cable shielding.

 

[Humungus]

When your sensitive signal, say in metal 4, is going to cross a noisy line, say in metal 2, then beneath the crossing, make an isolated isle of metal 3 large enough. The isolated isle will decrease the coupling capacitance betwenn the metal 2 and metal 4 lines, so reducing cross talk.

 

[laglead]

It depends on where noise signal pass.Generally 2-side shielding is enough.