2.1. Structure and Performance Parameters
In an off-axis system, the decenters and tilts of the local coordinate system of a surface relative to the global coordinate system can be used to describe the positional properties of that surface. The local coordinate system is generally defined with a local coordinate system origin at the center of the surface. Here, the center of a surface is defined as the vertex of its best-fit conic surface. The coordinate system for a surface can be shifted in the
X,
Y, and
Z dimensions and tilted in the
YZ,
XZ, and
XY planes relative to the coordinate system of the previous surface [
32]. These shifts and tilts are illustrated in
Figure 1a. When the coordinate system of the previous surface is the global coordinate system, these shifts and tilts are the
X,
Y, and
Z decenters, and
α,
β, and
γ tilts are used to describe the position of the surface local coordinate system, in which the definitions of
α,
β, and
γ tilts are the same as those of tilt, pitching, and deflection in the mechanical field, which is conducive to the coordination of design and manufacture. Additionally, the coordinate system for the nonsymmetric systems is consistent with the definition of the optical surface coordinate system in CODEV, which is convenient for the further optimization of the starting point obtained by the proposed nonsymmetric design method.
We can therefore introduce the structure of the nonsymmetric system based on this type of description. The global coordinate system of the proposed optical system can be established, and the
Z axis of the global coordinate system coincides with the chief ray of the central field in the object space.
Figure 1b shows the global coordinate system of the nonsymmetric off-axis three-mirror freeform system and the local coordinate system of one surface.
In the defined coordinate system, a nonsymmetric system can be defined as an optical system in which the chief ray of the central field can be deflected out of the
YZ plane after reflection/refraction from the surfaces within the system. Each surface in this optical system can have
α,
β, and
γ tilts relative to the global coordinate system. In addition, except for the center of the primary mirror, which is located near the
YZ plane, the centers of the other surfaces are all located outside the
YZ plane. This means that each surface in the system can be decentered in the
X,
Y, and
Z dimensions relative to the global coordinate system. Like other optical systems, the light rays in a nonsymmetric system will eventually converge on the image plane to realize the required imaging. Since the chief ray of the central field in the image space is deflected outside the
YZ plane, the center of the image plane is also located outside the
YZ plane and the image plane can have three decenters and three tilts relative to the global coordinate system. In addition,
Figure 1b shows that the normal vector at the center of each surface in the system and the normal vector of the image plane are not located in the same plane, and the centers of the four surfaces are also not in the same plane.
In a coaxial optical system, the Z decenter can describe the position of the surfaces. The centers of the system’s surfaces can only be located on the Z axis, and these surfaces are not tilted. In an off-axis optical system with planar symmetry, the Y decenter, Z decenter, and α tilt can describe the positions of the surfaces. The centers of these surfaces are located in the YZ plane, and each surface has only one tilt. In a nonsymmetric system, however, the centers of the surfaces and the center of the image plane can be placed in any position and can also have any tilt relative to the coordinate system. When compared with the previous two system types, the structure of a nonsymmetric system is both more complex and more flexible.
2.3. Establish the Nonsymmetric Initial Plane System
The structure of the initial plane system is consistent with the initial solution calculated by the point-by-point method. This feature is the key to designing the nonsymmetric structure. Therefore, first, establish an initial plane system that is consistent with the expected system structure, but without optical power.
When designing the planar symmetry system, there were structures for reference or suitable construction methods. In the worst case, it can be constructed by patchwork. However, the spatial position of each surface of a nonsymmetric system has high degrees of flexibility, and its structural forms are also diverse. Therefore, it is difficult to piece together the initial plane structure with the imagination of the designer. Besides, there are few nonsymmetric system structures for reference. To ensure that the initial system calculated by the point-by-point method can obtain a convergence result, it is necessary to discuss the parameters that determine the structure of the nonsymmetric system.
According to the discussion in
Section 2.1, when the position parameters of each surface are determined, the structure of the system is determined, as shown in
Figure 1b. The vector
Li = (
xi,
yi,
zi,
αi,
βi) (
i = 1,2,3…
M, where
M is the number of surfaces including the image plane, for three mirrors system,
M = 4) can be used to describe the position parameters of the
ith surface. Next, determine the position vector of each surface in the system. In this work, a nonsymmetric initial plane system is constructed based on the following principles: (1) The offset of the FOV and aperture is not considered, (2) the chief ray of the central field passes through the origin of the global coordinate system, and its direction vector is (0,0,1), and (3) the chief ray of the central field passes through the center of each surface. (4) The position of the image plane has been provided under the actual requirements and the estimation of the system size, that is,
L4 is fixed. (5) The chief ray of the central field is incident perpendicular to the image plane.
According to the principles, we can calculate the direction vector of the chief ray of the central field after being reflected by the tertiary mirror and determine the position vector of the primary mirror, which is L1 = (0,0,z1,α1,β1). Besides, the relative position of the image plane and the primary mirror determines the volume of the system, so the z1 coordinate of the primary mirror can be obtained according to the requirements of the system volume and the position of the image plane. At this time, α1 and β1 are still unknown, so we assume that they are at a certain angle. Next, the direction vector of the chief ray of the central field after being reflected by the primary mirror can be calculated according to the law of reflection. If the distance d1 from the primary mirror to the secondary mirror is provided, the center of the secondary mirror (x2,y2,z2) can be obtained by a simple geometric relationship. In the same way, if the distance d3 from the tertiary mirror to the image plane is provided, the center of the tertiary mirror (x3,y3,z3) can be obtained. Next, according to principle (3), the direction vector of the outgoing ray of the chief ray of the central field on the secondary mirror can be calculated, which is also the incident ray of the chief ray of the central field on the tertiary mirror. Finally, the normal vectors of the secondary mirror and the tertiary mirror can be calculated based on the reflection law, also obtaining α2, β2 and α3, β3. From the above discussion, it can be found that L2 and L3 are uniquely determined by α1, β1, d1, and d3, so the structure of the system is uniquely determined by α1, β1, d1, and d3. Then, we can provide a series of α1, β1, d1, and d3, and select a group from them to establish the nonsymmetric initial plane system. The selection criteria are: the rotation angle of each surface is close, the surface distance is appropriate, and the size in the XYZ dimensions is close.
2.4. Obtaining the Starting Point for Optimization
Before the calculations can be performed, the feature light rays as well as the starting and target points of the feature light rays must be defined. The feature light rays are the sample rays used for the calculation and are sampled from multiple field angles and using different pupil coordinates. When designing a system with rotational symmetry, it is only necessary to sample the FOV along a radial direction. For a planar symmetric system, only half of the system’s full FOV needs to be sampled. However, the nonsymmetric system proposed in this manuscript does not have such symmetry and thus needs to be sampled across the full FOV, which means that designing such systems is more challenging.
Figure 2 shows the sampling fields for these three system types.
Then, we can choose polar grids or rectangular grids to uniformly sample the rays of different pupil positions. When the ray beam is incident on the surface obliquely, although the feature light rays are uniform on the pupil, the distribution of the intersection points of the feature light rays on the surface is usually uneven, and the unevenness of the feature light rays from different fields is also inconsistent, as shown in
Figure 3.
The feature data points calculated by using uneven feature light rays cannot obtain the best fitting result. In a nonsymmetric system, it is even worse. One solution is to increase the number of feature rays, but as the number of rays increases, the number of calculations will increase. In this work, we propose an effective solution. There can be a different number of sampled rays for different fields and different surfaces. Taking polar grids as an example, for a certain surface, for a certain field, a series of polar angles can be evenly divided on the pupil. Different numbers of rays can be sampled at different polar angles, and the number of rays is determined by the footprint map on the surface. Through real ray tracing, the coordinates of the intersection of the chief ray and the edge ray on the surface can be obtained, and then the distance,
d(
α), between the intersection of the chief ray and the intersection of the edge ray can be calculated, where
α is a polar angle. If the distance corresponding to the sampling polar angle is short, the number of sampled rays is small, and vice versa. The number,
N(
α), of sampled rays in sample polar angle
α is,
where
N is the number of sampled rays corresponding to the polar angle with the shortest distance.
Next, we must determine the starting and target points of the feature light rays, where the starting points can be the intersections of the feature rays with any virtual surface located before the first surface of the system (this virtual surface can then be the
YZ plane of the global coordinate system) and the target points are the ideal image points. The image heights on the image plane are then obtained according to the given object–image relationship, i.e., the coordinates of the ideal image points are obtained in the local coordinate system of the image plane. However, the point-by-point method was performed within the global coordinate system, while the image plane of the nonsymmetric system is both decentered and tilted relative to the global coordinate system. Therefore, the coordinates of the image points in the local coordinate system of the image plane must be converted into the corresponding coordinates in the global coordinate system. When the image plane position is determined, the coordinates of the ideal image points in the global coordinate system can then be calculated based on the specified object–image relationship obtained by Equations (2) and (3):
Here, T is the rotation transformation matrix from the local coordinate system to the global coordinate system, and xI, yI, and zI and αI, βI, and γI are the decenters and tilts of the image plane relative to the global coordinate system, respectively. f is the focal length of the system, ωn,x and ωn,y are the nth field angles in the X and Y dimensions, respectively, and Hn represents the coordinates of the ideal image points of the nth feature field angles in the global coordinate system.
Next, we can perform the point-by-point calculation process [
9,
10,
29] to construct the freeform surfaces. To better understand this, here, we briefly introduce the data calculation process. During the construction process, a Nearest-Ray algorithm [
29] is used to determine the order of calculating data points and their coordinates. Then, we calculate the normal vectors of these data points based on Fermat’s principle and Snell’s law. These data points can then be used to perform the surface fitting. Surface fitting refers to the fitting of the surface shape and acquiring the position of the surface. The general form of a mathematical expression for a freeform surface involves the addition of freeform terms on the basis of a conic surface, as shown in Equation (4):
where
c is the curvature of the surface,
k is the conic coefficient,
Ai is the coefficient of the
ith freeform term, and
gi(
x,
y) is the freeform surface term described using polynomials such as Zernike polynomials or
XY polynomials. In the work introduced here, we use the
XY polynomials with the highest order of six. Therefore, the freeform surface terms can be provided by Equation (5):
The freeform surface of the planar symmetric system only needs to be described using the even-order terms of x in the XY polynomial, but the surface of the nonsymmetric system must be described using all the terms in the first N orders. According to Equations (4) and (5), when N = 6, the number of surface parameters used to describe the planar symmetric freeform surface is 15, while the number required for the nonsymmetric freeform surface is 27. This indicates that the surface of the nonsymmetric system has greater flexibility and is more conducive to the realization of high-performance and special applications. Greater freedom also means greater design difficulty, and designing such a system is a challenge.
Generally, the freeform surface shape is a function of the surface sag, which varies with the coordinates in the surface local coordinate system, such as those provided by Equations (4) and (5), and this function should thus be fitted into the local coordinate system. However, the calculation process for the data points is performed in the global coordinate system. Therefore, it is first necessary to convert the coordinates and the normal vectors of the data points under the global coordinate system into the local coordinate system, i.e., to determine the relative positional relationship between the two coordinate systems.
First, the data points must be fitted on a basic conic surface, where the radius and the coordinates of the basic sphere center and the conic coefficient can be fitted effectively using the least-squares method. We can then determine the position of the surface. In
Section 2.1, a description of the decenter and tilt is presented. This description essentially describes conversion from the global coordinate system of the optical system through shifting and rotation to the local coordinate system of the optical surface. The displacement of the shift and the angle of rotation are the exact decenter and tilt values of the local coordinate system relative to the global coordinate system that must be calculated here. The line that connects the intersection
D of the chief ray of the central field on the surface to the center
C of the basic sphere of the conic surface can be defined as the
Z’ axis of the local coordinate system, and the intersection of the ray passing from
C to
D with the basic sphere can be defined as the origin
O of the local coordinate system, as illustrated in
Figure 4.
In
Figure 4, the
XYZ coordinate system is the global coordinate system for the entire system and the
XʹYʹZʹ coordinate system is the local coordinate system for the surface. In this work, we use decenters in three dimensions and the two tilts
α and
β to describe the positioning of the surface local coordinate system relative to the global coordinate system.
After the global coordinate system is shifted and rotated, the origin
G then coincides with the local coordinate system origin
O, and the
Z axis direction coincides with the local coordinate system’s
Z’ axis direction. The decenters in three dimensions and the two tilts of the local coordinate system, which are
xd,
yd,
zd, and the
α and
β tilts, respectively, can be calculated according to this conversion relationship. Suppose the coordinates of the center
C and the origin
O in the global coordinate system are provided as (
xc,
yc,
zc) and (
xo,
yo,
zo). The decenters and tilts are then given as follows:
After the decenters and tilts of the local coordinate system are calculated, the conversion of the data point coordinates and the normal vectors from the global coordinate system to the local coordinate system can be completed using the shift and rotation relationships between these coordinate systems. Suppose the coordinates and the normal vectors of a data point in the global coordinate system are (
x,
y,
z) and (
l,
m,
n), the corresponding data point coordinates and normal vectors in the local coordinate system are then (
x’,
y’,
z’) and (
l’,
m’,
n’), which are given by the following equations:
where
T−1 is the inverse matrix of the rotation transformation matrix
T. In the local coordinate system, the freeform surfaces can be constructed using any kind of fitting method, e.g., [
33], to fit the freeform surface expressed by Equation (4). Thus, a good initial solution is obtained.
Compared with the previous point-by-point design method, the design method of the nonsymmetric system proposed in this manuscript has many different features. First, the previous point-by-point design method is used for the design of optical systems with symmetry, and it cannot be used for the design of systems completely without symmetry. Second, a method for constructing an unobscured nonsymmetric plane system is proposed in this manuscript. Limited by the designer’s experience and knowledge, since the structure of nonsymmetric systems is very flexible, it is difficult to build an initial plane system that can be converged during the process of iterations. In this manuscript, a method to build the plane structure according to the system’s geometry and size requirements is presented. This method can conveniently build a plane system that meets the requirements and will be converged during the process of iterations. Third, different from the sampling method of feature rays in the previous point-by-point method, a sampling method that can ensure the uniformity of feature rays in nonsymmetric systems is adopted in this manuscript. In a nonsymmetric system, the surface position is flexible, and the surface shape has no symmetry. If the previous feature ray sampling method is used, it will lead to the uneven distribution of feature data points. However, the use of uneven distributed feature data points will lead to a decrease in the surface fitting accuracy, or even failure of the fitting. Generally, one solution is to increase the number of feature rays, but as the number of rays increases, the amount of calculation will increase, and may also cause the calculation to fail. According to the feature ray sampling method in this manuscript, the feature data points on each surface can be uniformly distributed without increasing the number of rays. In this way, a better starting point can finally be obtained after construction and iterations. Fourth, a method for nonsymmetric surface fitting which simultaneously considers coordinates and normal vectors is proposed in this manuscript.