Does Matching depend on process?

[penchal_gv]

Hi All,
Is Matching concept depend on process or not?
i,e. for example in general ABBA is better matching patten than ABAB to match A and B transistors.
But i seen in one pdk file they given ABAB is better than ABBA.
plz clarify me...

Thanks
penchal

 

[acbalbason]

Re: Matching

hi!

i don't think it is a matter of process. matching is matching. so any process would do, and your layout is better with matching. Ü

- al

 

[elbadry]

Re: Matching

I think that in 0.18 um process and beyond, it'll always be better to use an ABAB structure than the usual ABBA structure.

Probably, this is due to the LOD effect. In processes using STI, the STI OX has a stress effect which changes the effective mobility of the transistor. The LOD effect depends on the distance between the STI isolation and the channel.

Thus , for an ABAB structure both A and B transistors will have the same average distance from STI, hence they will be better matched. The ABBA structure,however, will make each one have a different avergae distance, hence worse matching.

 

[siva497]

Re: Matching

matching stuff is used for very critical pair of transistor where operation gets effected with very little voltage. as a layout designer u must take care that if some change in voltage comes while fabrication that should be for both transistors. so obviously ABAB is better that ABBA

 

[cretu]

Re: Matching

matching stuff is used for very critical pair of transistor where operation gets effected with very little voltage. as a layout designer u must take care that if some change in voltage comes while fabrication that should be for both transistors. so obviously ABAB is better that ABBA

The ABBA has the advantage of taking care of the x,y mismatches. Having a common point which is in the middle of BB, will insure that the average deviation of the 4 transistors is the same. first B has more than first A, second B less then second A in average they have the same deviation.

The common centroid is better for differential pairs due to the fact that takes into account also thermal gradients. This is valid down to 0.13um CMOS. I ahve donre it in silicon and it works.

 

[omsi]

Matching

Yeah cretu is right.

 

[waxtomato]

Matching

If gradient is major concern for mismatch, ABBA is better. If distance-related factor is major concern, ABAB is better. But in my option, ABBA is first choice.