Interdigitisation is preferred over Common Centroid for Current mirror Layout. Why?

[ajay181]

Hi Friends,

I have read in most articles/books that for current mirrors Interdigitisation is better.
My query is : Are there any drawbacks of using Common centroid for Current mirrors ? [assuming the routing complexity is taken care of]

Same way what is the drawback if I go with Interdigitised pattern for Diff pairs ?

Have heard lot on gradients in X, Y and all. That is known, but besides that are there any electrical/functional reason behind this?

Thanks
Ajay

 

[Dominik Przyborowski]

No it isn't. In interdigitisation You have only smaller capacitances and area of current mirror, but the layout is affected by gradient effects and also different current direction on each transistor.

 

[dick_freebird]

Interdigitation can produce the same balance as common
centroid. Perhaps finer grained, or not, depending.

Downsides may include more close-in parasitic coupling
if your interdigitation causes more parallel routes in a
bigger wireball due to distributing (say) an 8-stripe
mirror device across 4 m=2 locations instead of two
m=4 (common centroid can be had as low as two
sub-locations).

 

[Dominik Przyborowski]

Interdigitation can produce the same balance as common
centroid. Perhaps finer grained, or not, depending.

No, it's no produce the same balance.

Of course everything depends to accuracy requirements, but in interdigitation your transistors are in different directions what's cause different drain to source profiles due to non perpendicular ion implantation of drain/source region. I don't want to writing deliberations about matching and physical situations in current mirrors, so I can only propose to read Drennan thesis about mismatch - it's quite old (1999) but its better than Pelgrom consideration and discus some things bounded with topic of this thread.

It's easy to find at google:
https://www.google.de/url?sa=t&rct=...ZfNgWtVMVOY0cWA&bvm=bv.59568121,d.Yms&cad=rja