10.2.3   Optical Components

As shown in the schematic, the deformable mirror and lenslet array of the
wavefront sensor are strategically positioned conjugate to the eye’s pupil.
There are two motivations for this, both of which stem from the fact (or
assumption) that the primary sources of aberrations in the system are the
cornea and crystalline lens. First, current correctors provide only phase correction.
Corneal and lenticular aberrations manifest themselves as essentially
pure phase (i.e., a distorted wavefront) at the eye’s pupil. At other planes,
propagation of the ocular aberrations generates intensity fluctuations (scintillation)
in which the original phase-only wavefront becomes encoded as both
phase and intensity. Scintillation is most significant at the image plane. Since
phase-only correctors cannot compensate directly for scintillation, their
maximum effectiveness will occur at a plane in the system where scintillation
is negligible, which for the eye will occur at (or near) its pupil.

Second, just as the ocular aberrations intrinsically vary with field angle, so
do the (compensatory) aberrations that are purposely introduced by the AO
system. This variation limits optimal correction to only one field point. To
maximize the field of view (i.e., isoplanatic patch size) about this point, field
variation in the total aberration pattern must be minimized, and this occurs
when the mirror (assuming only one corrector element) is positioned at (or
near) the pupil. At this position, field sizes at least as large as two degrees
have been observed with no loss in image quality.

The same rationale for positioning the mirror at the eye’s pupil also applies
to minimizing the intrinsic aberrations of the ophthalmoscope’s optical
system. Specifically the magnitude and physical origin of the optical system’s
aberrations impact how effectively these aberrations can be corrected by the
AO system and the field size over which the retinal image remains sharp.

Relaying the light from the eye through the optical system can be done
with lenses or curved mirrors. Lenses have an advantage in that they can be
used on-axis and have tolerable off-axis aberrations. Both the original
Rochester AO ophthalmoscope (RAOI) and the Indiana AO ophthalmoscope
(IAO) rely entirely on lenses and planar mirrors [3, 4]. The lenses were
precision achromats that minimized spherical aberration and coma when
operating at infinite conjugate ratios and yielded diffraction-limited performance
at the design wavelength. A problem with lenses is that their surface
reflections may project back onto the wavefront sensor and science CCD
camera and mask the signal from the retina. Such reflections are often diffi-
cult to remove as a properly aligned lens (i.e., one that maximizes optical
performance) generates reflections that follow the optical axis of the system.
Antireflective coatings covering the range of wavelengths in use noticeably
help but are rarely sufficient as the effective retinal reflection (<0.1% of the
light entering the eye) is significantly dimmer. An effective design step is to
simply minimize the number of optical surfaces in those channels of the
system that produce unwanted reflections (e.g., channels containing the subject’s
eye). Note the system in Figure 10.1 contains no optics between the BS
and the eye (except for trial lenses, which must be tilted to avoid back reflections).
If the BS is a cube beamsplitter, its primary internal reflections are
discarded by rotating the cube. A pellicle is a clean solution, as it creates no
disturbing reflections, although image quality may be slightly compromised.
An inevitable reflection originates at the anterior corneal surface and is
avoided in a couple of ways, which are described in later subsections.

Curved mirrors can be used instead of lenses. Mirrors have some advantages
over lenses, including the facts that they have no chromatic aberration,
have no back reflections, and provide the flexibility to make the optical system
more compact (by folding the beam). The trade-off of mirrors include their
expense and restriction to off-axis imaging, which gives rise to unwanted
aberrations, in particular astigmatism. Rochester’s second-generation flood-
illuminated AO ophthalmoscope (RAOII) relies on long focal length mirrors
to image the pupil onto a 97-channel Xinξtics deformable mirror. To overcome
off-axis aberrations, they use off-axis parabolas, which provide aberration-
free imaging at one point in the field of view. The choice of spherical
mirrors would have been more economical but may have given rise to unwanted
aberrations in the system. It is generally worthwhile to model the performance
of an AO instrument using commercial ray tracing software (e.g.,
Zemax, OSLO, etc.). Proper modeling permits optimization of the system,
establishes upper bounds on system performance, and helps avoid unwanted
surprises prior to the purchase of costly equipment.