Layout for differential pair with common centerriod configuration

[viren_s]

Hello Every one,

Can any one post the layout for differential pair , with common centerriod configuration. Or the complete layout with the detailed explanation of matching constraints and circuit. I have fair idea of matching and differential pairs. I am interested to make a script for it , to automate the layout for the differential pair. So for it i want at least one perfect example as reference. So if some one can contribute it for me , i will be thankful to him.

Regards,
viren

 

[viren_s]

Re: differential pair

i am attaching the pic of one pair. Please suggest me is this a good pair routed for all process from 90nm to 32 nm. or we need to change it in 45 or 32nm process.
I know importance of adding dummy poly .. so please igonre dummy poly for this.
Does it need improvisation.. i will post one more pic later today...[/img]

 

[JoannesPaulus]

differential pair

I am afraid this is not a good example for the layout of a differential pair. Here are some of the problems:
1. Un-matched parasitic capacitance at the inputs,
2. M2 going over the gate,
3. uneven poly distribution,
4. too many M1 lines between the pairs... you only need 2 lines.

I hope this helps!

 

[viren_s]

Re: differential pair

Dear Paulus,


Thanks for your time, I am modifying the pair and posting it . I hope rest is fine.

Pls find the attached pic of pair.


regards,
viren.

 

[JoannesPaulus]

Re: differential pair

Definitely an improvement! Attached is an example that might help you understanding point 2 and 3 in my previous post.

I hope it helps.

 

[thetas49]

Re: differential pair

You should also remember that proximity is an important factor in matching, you should endeavour to get the elements as physically close together as possible.
Cross coupled is not always the best solution for diff pairs, you should also consider interdigitation for your relatively small transistors.

Added after 3 minutes:

And if your inputs are really sensitive you should consider sticking diodes on the inputs to combat the "antenna" effect.