Files
chenproject/fiber_delay_sim/tests/test_core.py
2026-10-12 14:26:51 +08:00

143 lines
5.6 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

"""M0/M1 验收测试(docs/实现方案.md §11–§12)。"""
import numpy as np
import pytest
from fiber_delay_sim import pipeline
from fiber_delay_sim.core import fiber, fitting, signal
# ---------- fiber.py:延时与相位 ----------
def test_delay_per_meter():
"""1 m 石英光纤 ≈ 4.90 ns(报告 2.4.1:c/n_g = 20.42 cm/ns)。"""
d = fiber.delay_from_length(1.0)
assert 4.89 < d < 4.92
def test_nominal_lengths_match_design_delays():
"""名义长度按 n_g 反算的延时应与设计延时一致(±0.01 ns)。"""
for L, d in zip(fiber.NOMINAL_LENGTHS_M, fiber.NOMINAL_DELAYS_NS):
assert abs(fiber.delay_from_length(L) - d) < 0.01
def test_nominal_phases():
"""名义四组:相位 {0, 7.5, 5.0, 2.5},排序后 {0, 2.5, 5.0, 7.5},严格均匀。"""
fwd = fiber.forward_design(delays_ns=fiber.NOMINAL_DELAYS_NS)
assert np.allclose(fwd["phases_ns"], [0.0, 7.5, 5.0, 2.5], atol=1e-12)
assert np.allclose(fwd["sorted_phases_ns"], [0.0, 2.5, 5.0, 7.5], atol=1e-12)
assert fwd["uniformity"]["max_gap_dev_ns"] < 1e-9
assert fwd["order_index"] == [0, 3, 2, 1] # 相位 0→CH1, 2.5→CH4, 5→CH3, 7.5→CH2
def test_inverse_design_reproduces_original():
"""反向设计 0/2.5/5/7.5 ns 应复现原项目 0/142.5/95/47.5 ns 与名义长度。"""
inv = fiber.inverse_design([0.0, 2.5, 5.0, 7.5])
assert inv["is_uniform_grid"]
rec = inv["recommended"]
delays = sorted(e["delay_ns"] for e in rec)
assert np.allclose(delays, [0.0, 47.5, 95.0, 142.5], atol=1e-9)
by_phase = {e["phase_ns"]: e["length_m"] for e in rec}
assert abs(by_phase[0.0] - 0.0) < 1e-9
assert abs(by_phase[7.5] - 9.70) < 0.01
assert abs(by_phase[5.0] - 19.40) < 0.01
assert abs(by_phase[2.5] - 29.10) < 0.01
def test_inverse_forward_roundtrip():
"""反向设计 → 长度 → 正向复算:相位应还原目标(M1 验收)。"""
targets = [0.0, 2.5, 5.0, 7.5]
inv = fiber.inverse_design(targets)
lengths = [e["length_m"] for e in inv["recommended"]]
fwd = fiber.forward_design(lengths_m=lengths)
assert np.allclose(fwd["sorted_phases_ns"], sorted(targets), atol=1e-9)
def test_nonuniform_targets_flagged():
"""非均匀目标相位:应明确提示网格不均匀(教学点)。"""
inv = fiber.inverse_design([0.0, 1.0, 4.0, 7.5])
assert not inv["is_uniform_grid"]
assert inv["recommended"] is None
# ---------- signal.py:预设 ----------
def test_signal_presets_values():
t = np.array([0.0])
y = signal.generate(t, signal.PRESETS["monte_carlo"])
assert abs(y[0] - 1.0) < 1e-12 # 0.16 + 0.84
y2 = signal.generate(t, signal.PRESETS["single"])
assert abs(y2[0] - 1.0) < 1e-12
def test_signal_single_exp_decay():
t = np.array([0.0, 52.8])
y = signal.generate(t, signal.PRESETS["single"])
assert abs(y[1] - np.exp(-1.0)) < 1e-12
# ---------- M0:最小程序验收 ----------
def test_m0_ideal_recovery():
"""M0 验收①:理想无噪条件恢复设定寿命(相对偏差 < 1e-6)。"""
rec = pipeline.run_m0({"preset": "single", "tau": 52.8,
"snr_db": None, "bits": None, "window_ns": 400.0}, seed=1)
assert rec["n_samples"] == 40
assert rec["rel_error"] < 1e-6
assert rec["fit"]["rmse"] < 1e-9
def test_m0_reproducibility():
"""M0 验收②(复现性红线):同种子两次运行逐位一致。"""
a = pipeline.run_m0({"preset": "single", "tau": 52.8, "snr_db": 50, "bits": 12}, seed=7)
b = pipeline.run_m0({"preset": "single", "tau": 52.8, "snr_db": 50, "bits": 12}, seed=7)
assert np.array_equal(a["y_samples"], b["y_samples"])
assert a["tau_hat"] == b["tau_hat"]
def test_m0_noise_reasonable():
"""M0 验收③:加噪后多次运行分布合理(无发散、无明显偏差)。"""
reps = pipeline.run_m0_repeats(
{"preset": "single", "tau": 52.8, "snr_db": 50, "bits": 12, "window_ns": 400.0},
repeats=30, seed=100)
assert 0 < reps["rel_std"] < 0.05 # 有随机性但不发散
assert abs(reps["bias"]) < 0.02 # 无明显系统偏差
def test_m0_short_window_fewer_points():
"""采样点数随窗口变化:40ns 窗口只有 4~5 点(对应原项目 0.76τ 教学案例)。"""
rec = pipeline.run_m0({"preset": "single", "tau": 52.8, "snr_db": None,
"bits": None, "window_ns": 40.0}, seed=3)
assert rec["n_samples"] == 4
# ---------- 拟合算法 ----------
def test_log_linear_ideal_known_baseline():
"""对数线性(已知基线 B=0)在理想单指数下应精确还原寿命——与 NLS 数学等价。"""
t = np.arange(0.0, 400.0, 10.0)
y = 1.0 * np.exp(-t / 52.8)
r = fitting.fit_log_linear(t, y, b=0.0)
assert not r["baseline_estimated"]
assert abs(r["tau"] - 52.8) / 52.8 < 1e-6
assert r["dropped_points"] == 0
def test_log_linear_estimated_baseline_bias():
"""教学点:对 B=0 纯信号,默认尾段估基线引入明显偏差(基线估计误差)。"""
t = np.arange(0.0, 400.0, 10.0)
y = 1.0 * np.exp(-t / 52.8)
r = fitting.fit_log_linear(t, y) # 默认尾段估基线
assert r["baseline_estimated"]
rel = abs(r["tau"] - 52.8) / 52.8
assert 0.05 < rel < 0.5 # 偏差显著但有限(约 20%)
def test_double_exp_nls_ideal_recovery():
"""双指数 NLS 基准:理想数据恢复双寿命(对应原项目"标准拟合"验证)。"""
t = np.arange(0.0, 400.0, 4.0)
y = signal.generate(t, signal.PRESETS["monte_carlo"])
r = fitting.fit_double_exp_nls(t, y)
assert abs(r["tau2"] - 52.80) / 52.80 < 1e-4
assert abs(r["tau1"] - 0.60) / 0.60 < 1e-3