Is it a good practice to design with minimum length?

[suhas_shiv]

Designing with minimum L

Is it a good practice to design with minimum length? For example, with 0.25um technology, is it better to go with 0.25um or say 1um as length? I have designed a LVDS receiver and it seems to work better with 0.25um rather than 1um. Even the common mode range(0 - 2.4V) is unaffected (maybe since I am using a symmetrical OTA kind of structure with both nmos and pmos input transistors). I know there will be short channel effects with minimum length. What do you think are the other pros and cons? Any inputs on this would be appreciated.

Also when you say common mode voltage, it could either be a DC, AC or a combination of both I suppose. Please correct me if I am wrong.

Thanks.

 

[gevy]

Re: Designing with minimum L

Usually minimum L is used in digital circuits.
Longer L is used in analog circuits for best matching characteristics.

 

[johnsmith101]

Re: Designing with minimum L

Typically, longer L's are used when good matching is required. If your circuit is sensitive to parasitic capacitance, you may need to use minimum L for the process.

 

[sunking]

Designing with minimum L

the longer L can help matching and noise.

 

[pthoppay]

Re: Designing with minimum L

Apart from the technical point. There might be difference in the spice models that are used for minimum length device. What I mean is, usually the models are not accurate for minimum length device, so there might be difference in performance between simulated and the final product. Hence people prefer to have long length devices.

This argument usually applies when one goes for less than 90nm channel length. Since 250 um, is a kind of matured technology, there shouldnt be a probelm.

Prakash.

 

[jimjim2k]

Re: Designing with minimum L

the longer L can help matching and noise.

Longer L is suitable for near ideal current sources too.


tnx