wave op• 数据种类:cimage → cimage
• 调用:import optics; optics.angular_spectrum_propagate(field, wavelength_um=0.55, distance_um=100.0, pixel_pitch_um=1.0)(或 opsoptics.get("angular_spectrum_propagate"))
复场的标量自由空间传播,精确解(角谱法)。
> 以下的详细说明为原文 —— 摘要与标题已翻译。
`U(z) = IFFT{ FFT{U(0)} * exp(i*2*pi*z*sqrt(1/lambda^2 - fx^2 - fy^2)) }`
in the `exp(-i*omega*t)` convention, so a positive *distance_um*
propagates forward. Components beyond the propagating cone
(`fx^2 + fy^2 > 1/lambda^2`) are attenuated by
`exp(-2*pi*|z|*sqrt(fx^2 + fy^2 - 1/lambda^2))`, which is the physical
evanescent decay — not zeroed, so `distance_um = 0` is an *exact*
identity and the transfer function is continuous through it.
Unlike Fresnel propagation this makes no paraxial approximation: it is the
exact solution of the Helmholtz equation for a band-limited field, valid
from a fraction of a wavelength outward.
Returns a complex128 array with the same shape as *field*.
Ground truth it reproduces (measured): `distance_um = 0` returns the field
bit-identically (it short-circuits the transform pair); propagating `+z`
then `-z` returns the original to a relative L2 error of 4.3e-16 to
5.3e-16 for a band-limited field (measured on three: 64x64 random at
+/-50 um, a 64x64 Gaussian at +/-250 um, a 128x128 random at +/-500 um);
total power is conserved to between 0 and 3.5e-16 relative on the same
three. A field *with*
evanescent content does not round-trip — those components are gone by
construction, in both directions, because that is what physically happens.
*field* is a field in the space domain, not a spectrum: do not hand it
the fftshifted output of :func:complexops.cx_fft. Real input is promoted
to complex, which loses nothing.
Aliasing: the discrete transfer function is periodic, so a field that
diffracts past the array edge wraps around. The practical guard is the
usual one — pad the field so the propagated support stays inside, and keep
`pixel_pitch_um below lambda/(2*NA)`. No warning can detect this
reliably from the array alone, so none is invented.
Raises `ValueError`: *field* is not 2-D, smaller than 2x2, larger than
:data:MAX_FIELD_ELEMENTS, masked, or non-finite; non-positive or
non-finite *wavelength_um* / *pixel_pitch_um*; non-finite *distance_um*.
optics 的每个算子都先校验输入再计算(不让任何东西无声通过):
• 单位写进参数名 —— _mm / _um / _deg / _mrad。把 mm 和 µm 弄混不会崩溃,而是给出「看着合理却是错的答案」,所以用命名来防。这里绝不从数值大小去猜单位。
• **字符串一律 ValueError** —— float('50') 会成功,于是未解析的配置值会被当成长度混进来(实测:thin_lens('50', '200') 曾返回看着合理的 66.667 mm)。bool 也按 True == 1 的隐式提升拒绝。
• **complex / masked array 一律 ValueError(仅接受实数槽位;拒绝无声丢弃虚部或剥掉掩码)。所有输入中的 NaN/Inf 一律 ValueError**。
• 逐项点名拒绝除零及其近亲:焦距 0、曲率半径 0、折射率 <= 0、全不透明光阑(全为 0,归一化变成 0/0)、总和 <= 0 的 PSF、S0 = 0 的 Stokes 矢量、物体位于前焦点(像在无穷远)。
• 只有两个算子会返回非有限值,而且都写进了契约:depth_of_field 在超焦距以外返回 far_mm = inf(这正是超焦距的定义),gaussian_beam 在束腰处返回 wavefront_radius_mm = inf(平面波前的曲率半径)。两者都同时返回一个有限的搭档(far_is_infinite / curvature_per_mm)。**除此之外的无声 NaN/Inf 都在内部检出并 ValueError** ——「float64 溢出了」和「答案是无穷大」是两种不同的主张,不能拿后者的脸去交付前者。
• 尺寸上限:生成网格受 optics.MAX_GRID(4096)限制,传入的场/PSF/光阑受 optics.MAX_FIELD_ELEMENTS(2^24),ABCD 元件序列受 optics.MAX_SYSTEM_ELEMENTS(1024),Zernike 受 MAX_ZERNIKE_TERMS(512)/ MAX_ZERNIKE_ORDER(40)/ MAX_ZERNIKE_BASIS(2^25)。以 fail-closed 堵住「小参数引发巨大内部分配」的路径(实测:n_max=40 × 4096² 需要 108 GB)。
• 物理上不可能的状态同样拒绝:偏振度 > 1 的 Stokes 矢量、负透过率、负强度、n-|m| 为奇数等非法 Zernike 指标。
• 示例数据目录(下载 URL / 许可证) —— 2-D 用 skimage.data(BSD/公有领域)加合成图,3-D 给出真实数据源(Stanford/PDS 等)的下载 URL。
• 算子来历与参考文献 —— 该算子族所依据的研究/方法出处。
• 算法的正典(作者・年份)与用途见上面的族使用指南。
• optics_imaging — py -3.11 examples/optics_imaging.py
cimage 作为输入)wave)airy_pattern · fraunhofer_pattern · gaussian_beam
*Provenance: optics.py — OPTICS 算子登记表。本条目由 tools/opdocs.py md 自动生成(请勿手工编辑)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.