I haven’t opened your model, but from the system sketch you provided it appears that if you simply place a *single* dual-axis tip/tilt scanning mirror in the front focal plane of your last lens, the beam will be telecentric in your image space focal plane.  Is your plan to use a dual-axis mirror for beam steering, or is the other flat mirror in your setup also supposed to steer the beam?

A lens that is image-space telecentric has an exit pupil at infinity, which in turn means the aperture stop is located in the front focal plane  -- that’s the underlying basic principle.  The Edmund Optics website has a nice discussion here Telecentric Design Topics :

For a beam-steering and telecentric-focusing application, this means the steering function should be implemented in the front focal plane.  Here is a very simple example using the even asphere sample lens:

If the steering is done in a different plane, then the image space is no longer telecentric:

I’ve used this principle to construct a fiber-optic switch (ref: Fiber Bundle Switch (see Sec. 4) ):

If a single dual-axis mirror cannot be placed in the front focal plane, then two orthogonal mirrors can be placed in different locations, but they should both have their pivot points located in the focal plane.  Here is an example from the fiber switch:

Hope this helps…

Jeff

To design a simple singlet lens that is telecentric in image space should be relatively straightforward.  You can place the stop in the approximate front focal plane, use the Merit Function wizard to construct a default merit function (say RMS spot size), then add a few operands to constrain the effective focal length and enforce telecentricity.  Here is a very simple example using three field angles:

But a singlet lens with spherical surfaces will definitely have aberrations that might limit performance in your application (not sure what your requirements are).  For example, for larger field angles, field curvature may become a problem.  In that case a more complicated design with a field flattener can be tried (e.g., as described by Bentley and Olson, SPIE Field Guide to Lens Design, p. 19):

I’ve never designed optics for a terahertz application, so I have no idea what sort of material and/or fabrication constraints exist.  However,  I would think that standard visible lens design concepts can still be utilized, assuming the resulting components are feasible for fabrication.

Hi @Jeff.Wilde, 

Could you please help me designing it for my use case or point me to a good starting point.

I am looking to design a retroreflection system, which is lot like the above question proposed.

The difference being that 

1. Source/receiver being a single mode fibre @780nm (MFD: 5.3um)
2. System is a double pass system with the image plane in above question replaced by a mirror and beam couples back to the source fibre.
3. Instead of scanning mirror, I am using a transmission grating which is tuneable. And deflection half angle is max 1 degree. (The angle seems to be small but the precision of retroreflection required to couple back makes it tricky for me.) 
4. The beam diameter at grating is about 200um.

 Regards,

Saurabh