Discus and Help: Extraction Parameter for EKV 2.6 Model

sonnt

Dear all,
I'm doing Extraction Parameter for EKV v2.6 MOSFET Model using Levenberg–Marquardt algorithm (LMA). LMA:
In this algorithm, I can extract and optimize:



I can optimize all parameters of EKV model for 1 devices; but when I want to optimize for many devices, I don't know how to combine the Jacobian Matrix of each device and how to combine the residual error of each device.
Example:
Jacobian matrix: J = J1 + J2 + J3 + ....       (Ji: Jacobian Matrix of Device i)
Residual error: ri = fi - f(xi,p)      : i--> device i


For many devices, I don't know how to process, combine these Residual errors.
Please help me or give me a idea if you had done it.
Thank you so much.