Common centroid VS moving apart

[pavelni]

common centroid layout of analog transistors

Hello all!
I've got the next problem:
I'm designign the LNA for RF (about 1GHz). I have one issue with designing of layout of the differential pair.
There is a usual way of design - using of common-centroid techique for differential pair. But instead of common centroid I want to move each transistor from each other and do not mix parts of them.
Could you please tell me if there any advantages and disadvantages between this two techniques?
How can I evaluate influence of the transistors on each other in both cases?
Could you please advise me some books about this?
Thanks in advance![/img]

 

[JoannesPaulus]

common cetroid

This layout will make the LNA suffer a huge input offset due to mismatch between the two transistors in the differential pair. This is the main reason why the differential pair layout is usually a common-centroid layout.

 

[pavelni]

differential pair common centroid layout

But I saw the working scheme of LNA using this kind of layout. That's why I've got this question.
How can I evaluate the mismathch between two transistor in the layout of this kind. Usual Spectre modeling doesn't take into account geometrical disposition of transistors in layout.

 

[JoannesPaulus]

common-centroid

In your design manuals there are plots of transistor matching for both common-centroid layout and far-away transistors. You can figure out how much bigger you will need to design this differential pair if they are far apart. For that size you won't be able to make them work in the GHz range...

 

[pavelni]

common centroid differential pair

Unfortunately, I don't have design manuals where I can figure out this plots
Besides Spectre simulation shows that both common-centroid and far-away layout provide GHz range. Transistor length is 0.35 um.
Is there any equation that shows the distance between two transistors in layout?

 

[JoannesPaulus]

common centroid

If you do not change the size of your devices to meet your matching requirements for input offset, spectre does not know how far apart these transistors are (as you also said in one of your previous posts).

There are no equations that show how the distance between two transistors affects matching but only empirical data, reported in the design manual for that particular technology. If you do not have it, you should get one!

 

[pavelni]

Could you please answer the following question:
Is there any influence of the transistors on each other when they are far away?
And if yes - what is the influence? Maybe substrate current or something else?
Thanks in advance!

 

[JoannesPaulus]

We are not talking about any "influence" of transistors but just physical effects due to the fabrication of the devices. If they are far apart they will experience different environments and therefore their electrical characteristics will be different (thus creating offset as a result). Moreover, if there is anything going on in the substrate, you want to have the differential pair transistor to experience the same effects so that they will be suppressed by the common mode rejection.

 

[naveen.suthar]

We are not talking about any "influence" of transistors but just physical effects due to the fabrication of the devices. If they are far apart they will experience different environments and therefore their electrical characteristics will be different (thus creating offset as a result). Moreover, if there is anything going on in the substrate, you want to have the differential pair transistor to experience the same effects so that they will be suppressed by the common mode rejection.

Normally Common centroid not preffered for High speed application cos it introduces lot of parasitic cap on the input and output both. that slow down the speed.

Solution for that it can be solved by breaking the active and placing them in interdigitized matching with higher no. of fingers and for noise isolation you can keep single /double guarding as per area available .