Hi, Guys,

As well known, the mismatch of current sources determines the linearity of DAC. Here I am confused by the coefficient in the current mismatch formula.

In Geert A. M. Van der Plas, "A 14-bit Intrisic Accuracy Q² Random Walk CMOS DAC," the formula is (1) in the following attachment. In Marcel J. M. Pelgrom, et. al, “Matching Properties of MOS Transistors,” the formula is (2) in the following attachment. The former has a coefficient 1/2, while the latter has a coefficient 1.

Which one is correct?


Thanks
Yawei

They are both correct, but they are for different things.  Pelgrom's formula (2) is for the mismatch between a pair of devices.... but pairs don't make sense in a dac -- you have to consider each contribution individually.  To get around this (1) normalizes the mismatch contribution to a single device.  That way you can calculate the mismatch contribution of each device and add them up.  

Note that if you take the contributions of two devices using (1), you get (2).  

Bottom line is that you want to use (1).

rg

Hi, Robert and Raduga,

Thank you very much.


Raduga, I read the Pelgrom's paper again. The equation is for the mismatch of a pair of MOS devices there. So I agree with Robert.


Yawei

Hi, Raduga,

I think that equation (6) in Pelgrom's paper is for a pair of devices composed of 4 cross-coupled idential MOS transisotr with demension ration W/L. That means area of each device is 2×WL . Because 4 cross-coupled MOS are not placed in one row, equation (6) is a little different from equation (5). The random variation is affected by both x and y directions in equation (6) while only by x direction in equation (5).


Best regards,
Yawei

Hi,

RobertZ wrote on May 25^th, 2007, 12:26am:


Would you please introduce how you set up the simulation? Do you simulate with 2 MOS transistors? Does your model for MOS transistor have mismatch parameter? Where do you get the constants Avt, Abeta from?


Thanks
Yawei