Hi Joseph

My colleague Michael Cheng has created a macro that calculates the polarization. It can be found on the Code Exchange (please login to the be able to it):

It works in sequential mode (at least I was able to reproduce the results in sequential mode).

I have attached the file. I think you will be able to find the difference with the macro.

Let us know if not and we can have a look.

Hi Sandrine and Michael,

Thank you for your prompt response and the detailed macro that tracks the Jones matrix through sequential surfaces.  This is quite useful, however I do have a follow-up question.  I translated this code to a MATLAB script wherein I use my own 3D Snell's implementation and Fresnel coefficient calculation.  In order to get agreement with the Zemax result, I noticed several pecularities which I am hoping you might be able to clarify:

1. In the ZOS field editor, the X-Angle and Y-Angle values do not appear to directly correspond to a source with Y-tilt and X-tilt angles, respectively.  For example, if I define the X- and Y-Angles as 10deg and -40deg, respectively, the first transition matrix is

========== Surface 0 (STANDARD) ============


Material: AIR

Rotation matrix (transition):


0.752057766 0.000000000 0.659097198


0.000000000 1.000000000 0.000000000


-0.659097198 0.000000000 0.752057766

Using the convention described in to extract tilt angles from the rotation matrix, we obtain XTilt=ZTilt=0 and YTilt=41.23=sqrt(XAngle^2 + YAngle^2).   

However, if I were to compute ROTab(:,:,1) = Rz(ZTilt)Ry(YTilt)Rx(XTilt) with the field angles converted to tilts (XTilt=-40deg, YTilt=10), I would get for the first transition rotation matrix

ROTab(:,:,1) =

0.984807753012208 -0.111618897048950 0.133022221559489


0 0.766044443118978 0.642787609686539


-0.173648177666930 -0.633022221559489 0.754406506735489

This method leads to different results for all the vectors and ultimately very different Efield and intensity results.  In order to achieve the same results, I therefore redefine the tilt angles in the following way

YTilt = -40, XTilt=10


YTilt' = sqrt(XTilt^2 + YTilt^2),  XTilt'=0


ROTab(:,:,1) = Rz(ZTilt)Ry(YTilt')Rx(XTilt')

and I recover the first transition rotation matrix reported by the macro.  Is it therefore incorrect to equate the X- and Y-field angles with a source rotated by the same amount about the Y- and X-axes, respectively?  My concern in redefining the tilts is that some information may be lost regarding direction of the ray.  For example the incident ray k-vector for the first case is

0.659097197742238


0


0.752057766350650





whereas in the second case, it is 

0.133022221559489


0.642787609686539


0.754406506735489

I suppose what I am ultimately wondering is how does ZOS convert the specified field angles into tilt angles?

Note:  My actual application is a non-sequential diffractive system with NO rotational symmetry, which is why I am concerned about maintaining signs of the field (tilt) angles.

Hi Jospeh

The rotation matrix and tilts are described here: Rotation Matrix and Tilt About X/Y/Z in OpticStudio. I'm going to have a look at your description but you may go faster looking at the article.

Sandrine

So for example I have defined a sequential system:

The GCRS (Global Coordinate Reference Surface) is set to 1. The Rotation Matrix in the Prescription Data is:

Then in non-sequential I have defined:

I can again read the rotation matrix in the prescription data:

So that is identical to the sequential mode.