The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Does device mismatch important in circuit design?
https://designers-guide.org/forum/YaBB.pl?num=1385277513

Message started by tulip on Nov 23^rd, 2013, 11:18pm

Title: Does device mismatch important in circuit design?
Post by tulip on Nov 23^rd, 2013, 11:18pm

Hello everyone,
During my design of filter, I do not do monte carlo simulation, and the measurement results are ok.
Now I did monte carlo simulation by using statistical file provided by foundry. Because process variations can be corrected by tuning circuit of filter,for example, pll built by filter'gm and C. I only run monte carlo mismatch simulation. I found gm mismatch changed the filter frequency response much,mismatch of C is much smaller than mismatch of gm.
The measured response of filter is much better than worst case of monte carlo simulation, I think the reasons can be followings:
1. gm cell is carefully placed during layout, so device mismatch is much smaller.
2. monte carlo simulation simulate circuit with distritution provided by foundry, perhaps it is two pessimistic.

My question are:
1. Between the two reasons I bring forward above, which is of the major importance?
2. During the design of circuit, need the designer do much work to decrease mismatch effects? like scale the width and length of mosfet, this may tradeoff with other circuit performance, such as speed, area,etc.

Best regards :) :)

Title: Re: Does device mismatch important in circuit design?
Post by aaron_do on Nov 24^th, 2013, 5:07pm

Hi,

how many chips did you measure? Monte carlo is giving you a spread. You won't see the spread in measurement until you start measuring thousands of chips. Yes mismatch is important and you have to take care of it, but it depends on the application of your filter. If its not critical, then you can probably get away with a lot more variation.

Aaron

Title: Re: Does device mismatch important in circuit design?
Post by tulip on Nov 26^th, 2013, 2:02am

Thank you very much.
I just test 3 chips, perhaps this is not enough.

Best regards
tulip :) :)

Title: Re: Does device mismatch important in circuit design?
Post by jockeymonto on Dec 14^th, 2013, 3:42am

Hi tulip,

I have seen a lot of people ignoring MC simulations because its too time consuming for the entire systems. But they do always run MC on critical blocks and run atleast the 4 standard corners. If filter is your only block in the chip then you should take MC simulations. But you should know whether you are measuring the "process spread" or "device mismatch" thru MC. Incorporating process spread and device mismatch is likely to give you a lot of spread in your desired response. Lot of people only measure mismatch. Yes you are right careful layout can minimize device mismatch but in order to minimize the process spread you have to be careful in both circuit design and layout.
MC simulations show no of times you will get the desired response. Even if this no is 10 out of 100 runs, you have 10% chance of getting the wright result. So in this case you may find yourself lucky!