How to use the Matrix data in the Camera Node?

In the Camera Node, View Tab,the first parameter Projection has a option:Custom Projection Matrix.
What is meanning? And how to use the Matrix data?
I find some information in wiki,but can’t understand.

===================================
Projection /projection - A pop-up menu lets you choose from Perspective and Orthographic projection types. A third option Perpective to Ortho Blend enables the Projection Blend parameter below which can be used to blend between perspectives. A new 4th option Custom Projection Matrix allows you to specify a custom 4x4 projection matrix using a CHOP or a DAT.

Matrix CHOP/DAT /projmatrixop - When Custom Projection Matrix is selected, this parameters should be filled in with either a CHOP or a DAT with a custom 4x4 projection matrix. If a CHOP is used the first sample of the first 16 channels of the CHOP are used to create a 4x4 matrix. The channels can be thought as being read row-by-row or column-by-column. If a DAT is given a 4x4 table should be used. The matrix convention used is column major, which means vectors/points are multiplied on the right of the matrix.

The camera Comp has a built-in python getter for its projection matrix

derivative.ca/wiki088/index.php? … ss#Methods

projection(x, y) → Matrix

Try turning this matrix into a 4x4 DAT or 16-channel CHOP and see if you get the same results as using the same camera without a custom projection matrix.

If you want to see a projection matrix in action open up the arcball camera from the palette. It uses mouse interactions on a panel to update the projection matrix for the camera.