common centroid (which is better)

[deepavs]

which is best common centroid pattern


A B B A
B A A B

or

A B A B
B A B A

I think in first case transistor A and B have common centroid in both x and y axis

where as 2nd case only in y direction centroid of A and B lies common

plz inform me whether my conclusion is wron or right and plz explain which is better

 

[shashikumar.22]

deepavs,

Yes you are right the first one is the best common centroid pattern .

 

[refugee]

which is best common centroid pattern


A B B A
B A A B

or

A B A B
B A B A

I think in first case transistor A and B have common centroid in both x and y axis

where as 2nd case only in y direction centroid of A and B lies common

plz inform me whether my conclusion is wron or right and plz explain which is better

deepavs, would you pls explain your choise more clearly?
because I don't understand your mean of transistor A and B have common centroid in both x and y axis

 

[ND700]

A B B A
B A A B

is better

 

[solidstate]

And for even better matching, use dummies left and right of both lines

 

[kapilsn]

I feel the second choice is better because the transistors are more evenly distributed.

 

[checkmate]

The first is better for lower diffusion capacitance and compactness, as it allows for diffusion sharing.

 

[deepavs]

if am not use diffusion share then which is good?

 

[Raviprasad_K]

2nd pattern in preferred.
For more discussions please refer the below topic:

 

[deepak242003]

pattern 1 is more preferable.. as far as gradient and offset voltage is considered..

if you are not saring difusion , considering STI stress and LOD effect,area and capacirtance will remain unchanged but your matching will be more easy in first option

.
let me know if you have any other view..

 

[ankitgarg0312]

first pattern in better as it will decrease mismatches due to cross/linear gradients...

 

[dmparmet]

Yes ABBA is the first choice but also try to keep the number of fingers on your components even to make dummy transistors identical in X & Y dimensions.

 

[kapilsn]

The four rules of common cetroid layout as given in art of analog layout are:
1)coincidence:centriods of the matched devices should coincide
2)Symmetry:The array should be symmetric around both the X and Y axes
3)Dispersion:The array should exhibit highest possible degree of dispersion
4)compactness

If we want symmetry than I think pattern 1 is better.But dispersion is better in pattern 2 .
If we want the source and drains of all the transistors onthe same side than pattern 2 is better.If that is not a constraint than patter 1 is better.We keep this constraint so that the doping styles of the drains or sources of all the transistors are same.

 

[gksivas]

Hi guys,

Both structures are having same thermal gradients, process variations (etching, oxide thickness etc)

I think 2nd pattern is good, because evenly distributed in that structure.Anyway in deep submicron we are not doing diffusion sharing, if we use sharing then 1st pattern is better, because area will reduce.

remaining the sharing, both structures are having same effects, but 2nd is better a bit more.

If it is in one row then 1st is better, means ABBA is betterthan ABAB to nullify the Thermal gradients

invite comments on this post.

 

[deepak242003]

Hi guys,

Both structures are having same thermal gradients, process variations (etching, oxide thickness etc)

I think 2nd pattern is good, because evenly distributed in that structure.Anyway in deep submicron we are not doing diffusion sharing, if we use sharing then 1st pattern is better, because area will reduce.

remaining the sharing, both structures are having same effects, but 2nd is better a bit more.

If it is in one row then 1st is better, means ABBA is betterthan ABAB to nullify the Thermal gradients

invite comments on this post.

It will further reduce your PLI variation effect .... so more desirable at USDM