Hi Nadav,

This is an interesting condition. It looks like what is happening here is that the Diffraction Grating has a non-linear response to changes in angle for each ray, using the formula n2sin(theta2) - n1sin(theta1) = M*lambda/d. This means that unlike a folding flat mirror, changing the input angle by some angle theta does not just rotate the beam by 2theta but rather by some factor that is itself dependent on theta. If the incoming beam were parallel collimated this would just cause a rotation by a different amount than 2theta, but the beam is converging, so each part of the beam is changed by a different factor. You will have to re-focus the beam for each possible rotation angle of the grating, so the merit function and choice of variables will have to be adjusted accordingly.