I often use bi-telecentric lenses at work, which means both object-side and image-side chief rays are parallel. However, as I don't really have an optics background, I'm a bit baffled at how this might work. I have an OK understanding of how a single thick lens behaves under paraxial approximation, but telecentricity just seems a bit magical.
Is it even possible to design a telecentric lens system with gaussian optics only? What would it look like? I'm primarily interested in object-side telecentricity, but bi-telecentricity would be ideal, of course.