Should channel length be multiple of the minimum of the tech

Hi, all.

I have a question about deciding a channel length.

Say, I use 0.18um-tech to design analog circuits. Even though spectre doesn't complain about the channel length I use, is there a rule to follow to determine the channel length? In real design, can I use 0.2um or 0.21um?? Or, the channel length should be a multiple of the minimum channel length?

If non-multiple of the minimum channel length can be used, are there any pros and cons?


Want to hear from the real designers...

Re: Should channel length be multiple of the minimum of the

The channel length could be set to any value within the valid range, for example in TSMC 0.18um the valid range for the channel length is from 0.18um to 20um. There is no restriction upon the channel length to be multiple of the feature size.

Pros and Cons are:
1. Shorter channel length means lower oxide capacitance
2. Longer channel length means larger output resistance
3. Longer channel length means lower flicker noise

hello,
but i think there is some rule , for instance i cannot make 0.18111um there must be some minimum step , i think that using multiples of L is good practice.
"my opinion only and i am not a real designer"

Re: Should channel length be multiple of the minimum of the

Both previous posters are correct: there are no restrictions other than those noted above. However there are some practical considerations:

1) You must double check the grid size that the layout groups work with; increments are usually limited to 0.01 um.

2) You should also check the WAT structures used to derive the spice models, and stick as closely as possible to the lengths that they use. Otherwise, intermediate lengths are simulated with parameters that are extrapolated, and therefore less accurate. If you have a good modelling group, talk to them about it. Do your own checks as well: If you run a sim with a sweep of channel lengths, you often see a discontinuous curve as a result of "binning" the models at various lengths. Finally, be practical... if the minor inconsistencies won't affect your design, don't worry about it.

Re: Should channel length be multiple of the minimum of the

Thanks...guys....

It seems that a realizable channel length is more likely to be determined by its layout requirement, which is make sense to me..

Re: Should channel length be multiple of the minimum of the

Another concern from very short channel length is big leakage current, which could be very bad if your circuit scale is very big, you might not be able to meet standby current spec.