Current Mirror Layout Orientation

[seamoss]

I have read in books that for better matching of current mirrors, it is better to have them laid out in same direction such that current always flows in the same direction for both the mirror devices. However, I am not sure how that would improve matching in mirror ? Would appreciate your thoughts..Thanks!

 

[dick_freebird]

Nothing beats identicalness when you are looking for matching.
So you'd prefer to just do it unless your real goal is being a
scholar.

It's not about improving - you know what ideal is, all you can
do is avoid detracting from it an any of the various ways
that you could.

 

[Dominik Przyborowski]

The recommendation to providing the same current direction in all current mirror transistors is related to asymetry between drain and source regions caused by non-perpendicular ion beam implantation.

If You want more informations check this thesis.

 

[seamoss]

thanks Dominik !

 

[Dominik Przyborowski]

One time I'm read one conference paper about layout of current mirrors and the conclusion was that the best matching is to place transistors in 8 wing star. If I find some time I will try to found this paper in my library at work ;-)

 

[ASIC_Norway]

Once a professor at my uni told us: Life is about matching! Anyway, my general rule for transistor layout, if it looks symmetrical and OK, then it is probably OK. Remember KISS principle, I have my doubts about unorthodox methods, even they show some improvements in performance, the trade-off does not really worth with respect to time and design effort. But, this depends whether you work in academia or in industry. Please, take a look at The art of analog layout by Hastings, if you havent done it yet.