Assuming the system is defined by the matrices {A,B,C,D} you have

x_n+1 = Ax_n + Bu_n

y_n = Cx_n + Du_n

You can modify this to have additional outputs, in this case the state vector itself, without changing the system dynamics:

x_n+1 = Ax_n + Bu_n

[ y_n ; x_n ] = [ C ; I ]x_n + [ D ; 0 ]u_n

This will result in a vector output from the block [ y_n ; x_n ] and you can separate this using a Selector block, sending the portion of the vector for y_n wherever you previously used that signal (e.g. feedback loop) and sending x_n to wherever you want to have the full state measurement.

Let's assume

• You've defined matrix variables A, B, C, and D, and a sample time Ts
• x_n has N states
• y_n has M measurements
• u_n has L inputs

Then instead of putting A, B, C, D, and Ts into the Discrete State Space block, you'd put A, B, [ C ; eye(N) ], and [ D ; zeros(M,L) ] as in the following: Block Parameters: Discrete State-Space

Note that my dummy example uses N = 4, M = 3, L = 2, and Ts = 1

Then, to get y_n add a selector block to get the elements 1:M like this: Selector for y

And similarly for x_n, you want elements (M+1):(M+N) like this: Selector for x

Since you have not modified the system matrices {A,B,C,D} you're free to continue using them for analysis or design tasks as is.