Taper the power radiated from different object x,y sources.  Increase radiated power toward the edge of the VCSEL emitting area. 

I usually have better results at the beginning by optimizing on spot size rather than wavefront.  While in sequential mode, don't check Afocal, and put a paraxial lens after your collimating element.  Then when you switch to NONSEQ, delete the NSC paraxial and just let the lens shine on the detector.

I am interested in returning to the purpose of the model.

I’ve used VCSELs in systems and the ones I’ve encountered seemed to have higher order modes up to TEM55 (see link for example).  Having (x1) or more VCSELs and using the word collimated has me concerned.

An extended source can never be collimated in the same manner as a TEM00 or infinitely small source.  So, what I’m trying to understand is your definition of collimated.  Do you mean:

1. Smallest total beam at infinity for the entire VCSEL array?
2. Smallest total beam at a given distance from the VSCEL array?
3. An array of collimated beams where each VCSEL is well collimated but not pointed in the same direction?
4. A method of combining all (x7) VCSELs such that they appear as one and it is collimated?

Case 1:  Typically this is achieved by imaging the entire array to infinity.  There is likely no other configuration that minimizes the total power in the smallest area.  This is not collimated as you cannot collimate an extended source.  Etendue will limit the size of the image of the source.

Case 2:  Same as Case 1, only image it to a finite distance.

Case 3:  This requires a good Fourier Transform lens and I would design such a lens in sequential mode, check the performance with point sources off-axis in the back focal plane and then import that to a non-sequential model.

Case 4:  I have done something similar to this using a lens array.  The array has (x1) lens for each VSCEL, projecting an image of the VCSEL to infinity.  Then I used a single lens to re-image that from infinite to a finite distance.

What Brian said.

Another approach to modeling VCSELs is to model just one, as an object with small but finite size. If it’s close to the first optical surface with power, use cos-cubed apodization, but if it’s far away then uniform is fine.

Then model the array of VCSELS by using multiple configurations to place each VCSEL at the correct location relative to the lens.

In general, sequential mode is best used to get the best imaging quality, but then move to non-seq to look at illumination profiles. In NS mode you won’t need multiple configurations to model multiple sources, either.

• Mark

Hello everybody,

thank you all for your advice!

Sorry for the delayed response – holidays 😊

Regarding Brians Questions:

• I want to achieve a total beam size/diameter, which should increase as little as possible over a distance range, e.g. from a distance of 0,3m to 10m away from the last surface of the lens.
  Of course, due to the finite size of the VCSEL-array there will always remain some degree of divergence, which can be estimated by theta=size / focal-length. Therefore the beam diameter will roughly have a beam diameter of the clear aperture at z=0 and will increase by tan(theta)*distance.
• On the other hand each VCSEL has a gaussian emission profile and will therefore have an effect on the divergence of the total beam. (of the VCSEL-array)
• During the optimization I’m a trying the find the object distance for which I can obtain a total beam with as little divergence as possible.

Based on the information above, what would be your recommendation ob how the achieve my goal?

Thank you very much,

Felix