diff --git a/README.md b/README.md index c3d558c..aae211d 100644 --- a/README.md +++ b/README.md @@ -48,8 +48,13 @@ python -m fiber_delay_sim.cli design --nominal # 5. 反向设计:目标相位 0/2.5/5.0/7.5 ns → 候选光纤长度 python -m fiber_delay_sim.cli design --targets 0,2.5,5,7.5 -# 6. 一键跑全部实验(M0 基准 / M1 设计 / M2 校准 / M3 算法 / M4 统计) +# 6. 一键跑全部实验(M0 基准 / M1 设计 / M2 校准 / M3 算法 / M4 统计 / 扫描) python -m fiber_delay_sim.cli all + +# 7. 图形界面(六页面 + 一键导出记录包) +python -m fiber_delay_sim.gui +# 无界面环境自检: +python -m fiber_delay_sim.gui --selftest ``` 每次运行自动在 `runs/` 落盘一个**实验记录包**:`meta.json`(参数/种子/日期/来源标注)+ 数据 CSV + 结果 JSON + 图表,保证论文每一项结果可重新生成和验证。 @@ -63,7 +68,7 @@ python -m fiber_delay_sim.cli all | M2 误差与校准 | 四通道误差注入 + 校准 + 时间重建 | ✅ 完成(校准消除通道断层,连续度 mean_jump 显著下降) | | M3 算法切换 | 对数线性 vs 非线性最小二乘 | ✅ 完成(NLS 稳健:SNR 50 dB 平均偏差 0.35%;对数线性估基线引入系统偏差 4.42%) | | M4 统计分析 | 蒙特卡洛 + 参数扫描 + 置信区间 + 失败率 | ✅ 完成(SNR 50 dB:单通道 σ/τ=0.84% → 四相位 0.33%;SNR×窗口扫描 exp06:低 SNR 区复现"单通道发散→四相位收敛",SNR30/win300 提升 27x,失败率 67.5%→15%) | -| M5 界面与导出 | 简单 GUI + 一键导出记录包 | ⬜ | +| M5 界面与导出 | 简单 GUI + 一键导出记录包 | ✅ 完成(tkinter 六页面:参数方案/原始采样/校准重建/拟合比较/参数扫描/结果导出;`python -m fiber_delay_sim.gui`;一键导出 session.json+meta+CSV+PNG/SVG 图) | 网站(副产品)随每个里程碑同步增量上线,见 `docs/实现方案.md` §11.1 与附录 B。 diff --git a/docs/论文初稿大纲.md b/docs/论文初稿大纲.md new file mode 100644 index 0000000..7a33913 --- /dev/null +++ b/docs/论文初稿大纲.md @@ -0,0 +1,161 @@ +# 教学论文初稿:章节骨架 + +> 对应任务书「七 论文编写主线」八部分结构,逐章给出:写作目标、要点、图表清单(映射到记录包)、关键数据(目前软件实测值,写正文时直接引用)、诚信注意。 +> 论文建议题目(任务书):**基于光纤延时等效采样的荧光寿命虚拟仿真实验设计** +> 状态:骨架(待填充正文)。所有数值均来自 `runs/` 记录包(软件仿真),引用时标注记录包路径。 + +--- + +## 摘要(建议 200–300 字) + +**要点**:低采样率 ADC 下荧光寿命测量点数不足 → 四路光纤延时等效采样把 10 ns 间隔细分至 2.5 ns("光学游标卡尺")→ 教学仿真软件将原理转化为可调参数实验 → 蒙特卡洛验证精度与失败率改善 → 虚实结合与局限性。 + +**关键数据(软件仿真,正文需复算后填入)**: +- 点数:单通道 40 点 → 四相位 160 点(窗口 400 ns) +- SNR 50 dB:σ/τ:单通道 ~0.8% → 四相位 ~0.3% +- 低信噪比(SNR 30 dB、窗口 300 ns):σ/τ 92.8% → 3.1%,失败率 67.5% → 15% + +**诚信注意**:本论文所有曲线为软件仿真(`data_origin: software-simulated`),不得表述为装置实测。 + +--- + +## 1 教学问题(对应论文主线"教学问题") + +### 1.1 背景:快速荧光衰减测量的采样瓶颈 +- 荧光寿命 ~50 ns 量级(YAG:Ce 标称 ~52.8 ns),直接测量需要 ns 级时间分辨; +- 低速 ADC(如 10 ns 间隔)→ 单周期仅 4–10 个采样点,点数不足、噪声敏感、拟合不稳定; +- **教学困境**:高速采集设备昂贵且难直观展示"为什么点数不够"。 + +### 1.2 本实验的定位 +- 不追求恢复整机(激光功率闭环测量),而是把"光学时间交错采样"提炼为**可动手调节的教学实验**; +- 学生通过调节参数 → 观察采样相位/重建曲线/寿命估计的连续变化 → 理解信号处理因果关系。 + +**证据**: +- 图:10 ns 直接采样与 40 ns 窗口点数不足的演示(M0,exp01) +- 任务书定位引用(§一 目的) + +## 2. 实验原理(对应"实验原理") + +### 2.1 荧光衰减与寿命拟合 +- 模型:双指数 + 基线 I(t) = A₁e^(−t/τ₁) + A₂e^(−t/τ₂) + B(τ₁ 快、τ₂ 慢) +- 教学用单指数简化:τ₂ 为主寿命 +- 拟合方法:非线性最小二乘(NLS)基准;对数线性为对比 + +### 2.2 光纤延时与等效采样 +- 传播延时 τ_d = n_g · L / c;1 m 光纤 ≈ 4.90 ns 延时(n_g=1.468) +- 四路光纤长度 0/9.7/19.4/29.1 m → 延时 0/47.5/95/142.5 ns → 对 10 ns 周期取余得相位 0/7.5/5.0/2.5 ns → 排序 0/2.5/5.0/7.5 ns +- 等效网格步长 10/4 = 2.5 ns → 光学游标卡尺 +- **关键概念辨析**:2.5 ns 是等效时间细分网格,非整机时间分辨率(探测、TIA、ADC 带宽未变) + +**证据**: +- 表:长度/延时/相位表(M1,exp02) +- 图:延时–相位图(M1 页面) + +## 3. 实验设计(对应"实验设计") + +### 3.1 统一任务 +学生从"测量方案设计 → 参数调节 → 校准 → 算法选择 → 结果可信度评价"完成一次完整实验。 + +### 3.2 操作流程(六步) +1. 设置荧光 / ADC / 光纤参数 +2. 生成理想波形并模拟采样 +3. 误差注入 + 校准 + 重建 +4. 拟合算法比较 +5. 参数扫描(长度 / 噪声 / 带宽 / 校准误差) +6. 结果导出(记录包) + +### 3.3 三组核心对比(教学主线) +| 对比 | 固定条件 | 观察指标 | 预期现象(软件实测) | +|---|---|---|---| +| ① 单通道 vs 四相位 | 同一真值/噪声/窗口 | 点数、单次偏差、MC σ、失败率 | 点数×4;σ/τ 从 ~7.65% 量级降至 ~0.6% 量级(低 SNR 更显著) | +| ② 不校准 vs 校准 | 注入增益/延时/基线误差 | 重建连续度、τ̂ 偏差 | 校准消除通道断层;延时常数错位不改变单指数 τ(被幅度自由度吸收)| +| ③ NLS vs 对数线性 | 相同数据 | 偏差、残差 | 对数线性估基线引入系统偏差 | + +**证据**:文档 `docs/实现方案.md` §3.1 + +## 4. 软件实现(对应"软件实现") + +### 4.1 架构 +- core(numpy/scipy)→ pipeline → experiments/CLI → GUI(M5)→ 网站(Pyodide 浏览器运行同一 core) +- 参数可调:荧光 / ADC / 光纤 / 前端 / 噪声 / 种子 +- 记录包机制:meta.json(含 `data_origin`)+ data/CSV + results.json + figures/(SVG+PNG) + +### 4.2 界面(六页面) +参数方案 → 原始采样 → 校准重建 → 拟合比较 → 参数扫描 → 结果导出(一键导出记录包) + +**证据**:GUI 截图(M5,待补截);网站(副产品)地址 https://chenxy.biboer.cn + +## 5. 结果分析(对应"结果分析",正文核心) + +### 5.1 光纤设计与相位均匀性(M1,exp02) +- 名义四组 0/9.7/19.4/29.1 m → 相位 0/2.49/5.00/7.50 ns(理想 2.5 ns,最大间隔偏差 <5.4e-3) +- 反向设计闭环:0/2.5/5/7.5 → 0/9.70/19.40/29.10 m + +### 5.2 校准与重建(M2,exp03) +- 三方案:不校准 / 基线+增益 / 完整校准 +- 连续度(mean_jump):0.040 → 0.012 +- τ̂:0.41% → 0.31%(教学点:模型自由度吸收延时常数错位) + +### 5.3 算法比较(M3,exp04) +- SNR 50 dB:NLS 平均偏差 0.35% vs 对数线性(估基线) 4.42% +- SNR 40 dB:NLS 1.10% vs 对数线性 21.86%(对数变换放大低幅值尾部噪声) + +### 5.4 统计与参数扫描(M4,exp05+exp06) +- 主场景 SNR 50 dB/400 ns:单通道 σ/τ=0.84% → 四相位 0.33% +- **参数扫描关键区**(exp06):SNR 30 dB/300 ns:92.8% → 3.1%(27x),失败率 67.5%→15%;SNR 40 dB/400 ns 失败率 10%→0% +- 结论:等效采样的收益在**低 SNR / 短窗口**(采样不足)时最大 → 与"光学游标卡尺解决低采样率困境"的命题一致 + +**图清单**(全部可由 `experiments/` 脚本重新生成): +- fig1_hist(M4 蒙特卡洛直方图) +- fig2_undersampled(低 SNR 对比) +- fig3_window_sweep(窗口扫描) +- exp06 fig1_ratio_heatmap(SNR×窗口提升比热图) +- exp06 fig2_snr_curves(SNR 曲线) + +## 6. 虚实结合(对应"虚实结合") + +- 现有装置(原项目)说明 + 模块级真实参数(如 YAG:Ce 实测 τ 拟合值、光纤折射率 n_g、ADC 位数/采样率) +- **注意**:不得宣称整机性能已验证;只做"模块级测量/参数输入软件" + +## 7. 教学实施(对应"教学实施") + +- 实验学时(建议 2–4 学时) +- 操作任务卡(六步流程 + 学生输出要求) +- 评分要点(可复现、误差分析、算法选择理由) + +**待办**:真实教学试用记录(无试用不写效果数据,诚信红线) + +## 8. 局限性(对应"局限性") + +- 整机未跑通(激光功率测量装置未完整恢复) +- 仿真边界:噪声模型、系统响应模型、量化模型简化 +- 2.5 ns 为等效细分网格,非物理分辨率 +- 待验证:真实硬件模块测量 + +## 9. 诚信与引用规范(贯穿全文) + +- 记录包机制保证每图可重生成(图注标注 `runs//`) +- 区分仿真设定值(τ₁=0.6 ns)与实测拟合值(τ₁=5.0 ns)见 `docs/项目说明.md` §6 +- 未实测不写实测 + +--- + +## 附:素材索引 + +| 素材 | 位置 | +|---|---| +| 软件源码 | `fiber_delay_sim/`(core/pipeline/cli/gui) | +| 实验脚本 | `fiber_delay_sim/experiments/exp01..exp06` | +| 记录包 | `runs/`(exp01–exp06) | +| 网站 | https://chenxy.biboer.cn(副产品) | +| 文档 | `docs/项目说明.md`、`docs/实现方案.md` | +| 原项目材料 | `resources/研究报告.pdf`、答辩 PPT | + +--- + +## 下一步(正文写作时的分派建议) + +1. §5.4(统计与参数扫描)数值需在写正文时用 CLI 重跑后填充最新值; +2. §7 教学实施需配合学生指导材料(交付物 6)一起写; +3. 补 GUI 截图(M5 交付素材); +4. 全文图注挂接记录包路径。 \ No newline at end of file diff --git a/fiber_delay_sim/__pycache__/cli.cpython-312.pyc b/fiber_delay_sim/__pycache__/cli.cpython-312.pyc index e983e07..9a29e00 100644 Binary files a/fiber_delay_sim/__pycache__/cli.cpython-312.pyc and b/fiber_delay_sim/__pycache__/cli.cpython-312.pyc differ diff --git a/fiber_delay_sim/__pycache__/gui.cpython-312.pyc b/fiber_delay_sim/__pycache__/gui.cpython-312.pyc new file mode 100644 index 0000000..1097f3a Binary files /dev/null and b/fiber_delay_sim/__pycache__/gui.cpython-312.pyc differ diff --git a/fiber_delay_sim/cli.py b/fiber_delay_sim/cli.py index 423d8dd..5bae7a8 100644 --- a/fiber_delay_sim/cli.py +++ b/fiber_delay_sim/cli.py @@ -114,6 +114,14 @@ def cmd_run_one(args): table["scan"](repeats=min(args.repeats, 40), runs_dir=args.runs_dir) +def cmd_gui(args): + """M5:图形界面(tkinter 六页面 + 一键导出)。""" + from .gui import main as gui_main + argv = ["--selftest"] if args.selftest else [] + argv += ["--runs-dir", args.runs_dir] + gui_main(argv) + + def main(argv=None): ap = argparse.ArgumentParser(prog="fiber_delay_sim", description="光纤延时等效采样教学实验仿真软件") @@ -152,7 +160,7 @@ def main(argv=None): p2.add_argument("--runs-dir", default="runs") p2.set_defaults(func=cmd_run_one) - p3 = sub.add_parser("all", help="M5:一键运行全部实验(M0–M4)") + p3 = sub.add_parser("all", help="M5:一键运行全部实验(M0–M4 + 扫描)") p3.add_argument("--tau", type=float, default=52.8) p3.add_argument("--window-ns", type=float, default=400.0) p3.add_argument("--snr-db", type=float, default=50.0) @@ -163,6 +171,11 @@ def main(argv=None): p3.add_argument("--runs-dir", default="runs") p3.set_defaults(func=cmd_all) + p4 = sub.add_parser("gui", help="M5:图形界面(六页面)") + p4.add_argument("--selftest", action="store_true", help="无界面自检") + p4.add_argument("--runs-dir", default="runs") + p4.set_defaults(func=cmd_gui) + args = ap.parse_args(argv) args.func(args) diff --git a/fiber_delay_sim/export/__pycache__/record.cpython-312.pyc b/fiber_delay_sim/export/__pycache__/record.cpython-312.pyc index ac706cf..af3b6fe 100644 Binary files a/fiber_delay_sim/export/__pycache__/record.cpython-312.pyc and b/fiber_delay_sim/export/__pycache__/record.cpython-312.pyc differ diff --git a/fiber_delay_sim/gui.py b/fiber_delay_sim/gui.py new file mode 100644 index 0000000..d55d6d0 --- /dev/null +++ b/fiber_delay_sim/gui.py @@ -0,0 +1,508 @@ +"""M5:简单图形界面(tkinter 六页面)+ 一键导出记录包。 + +对应任务书"软件界面与操作顺序"六页面: + 1. 参数与方案 2. 原始采样 3. 校准与重建 4. 拟合与比较 5. 参数扫描 6. 结果导出 +全部计算复用 core/pipeline(零重写);每个页面产出图表与数值, +结果导出页一键将当前会话的全部参数、数据、图表落盘为可复现记录包。 + +运行:python -m fiber_delay_sim.gui +自检(无界面):python -m fiber_delay_sim.gui --selftest +""" + +import argparse +import csv +import json +import os +import tempfile +import threading +import traceback + +import numpy as np + +from . import pipeline +from .core import acquisition, calibration, fiber, fitting, signal, statistics +from .export import record + +try: + import matplotlib + matplotlib.use("TkAgg") + matplotlib.rcParams["font.sans-serif"] = [ + "PingFang SC", "Hiragino Sans GB", "Arial Unicode MS", "SimHei", "sans-serif"] + matplotlib.rcParams["axes.unicode_minus"] = False + from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg + from matplotlib.figure import Figure + import tkinter as tk + from tkinter import ttk + HAS_TK = True +except Exception: # pragma: no cover - 无显示环境 + HAS_TK = False + +DEFAULT_SEED = 42 + + +def _json_default(o): + """ndarray → list;numpy 标量 → 原生 float。""" + if isinstance(o, np.ndarray): + return o.tolist() + if isinstance(o, np.generic): + return o.item() + raise TypeError(f"Object of type {type(o).__name__} is not JSON serializable") + + +# ===================================================================== +# 计算层:每个页面一次"运行",产出 dict(图 + 数值),GUI 与导出共用 +# ===================================================================== + +def run_page1(params): + """参数与方案:荧光 / ADC / 光纤 → 参数表 + 延时/相位表。""" + p = dict(params) + sig = signal.resolve_preset(p["preset"], tau=p.get("tau2")) + lens = [float(x) for x in p["lengths"]] + fwd = fiber.forward_design(lengths_m=lens, n_g=float(p["ng"])) + inv = fiber.inverse_design([0.0, p["dt"] / 4, p["dt"] / 2, 3 * p["dt"] / 4], + dt_ns=float(p["dt"]), n_g=float(p["ng"])) + return { + "signal": {k: float(v) for k, v in sig.items()}, + "forward": fwd, + "inverse": inv, + "dt_ns": float(p["dt"]), + } + + +def run_page2(params, seed): + """原始采样:单通道 + 四通道,连续曲线 + 采样点。""" + p = dict(params) + sig_params = signal.resolve_preset(p["preset"]) + t = np.arange(0.0, float(p["window_ns"]) + pipeline.CONT_STEP_NS / 2, pipeline.CONT_STEP_NS) + y = signal.generate(t, sig_params) + rng = np.random.default_rng(seed) + acq_s = acquisition.run_single_channel(t, y, { + "dt_ns": p["dt"], "window_ns": p["window_ns"], "snr_db": p["snr_db"], "bits": p["bits"]}, rng) + acq_m = acquisition.run_multi_channel(t, y, { + "delays": [float(x) * 4.9 for x in p["lengths"]], "dt_ns": p["dt"], + "window_ns": p["window_ns"], "snr_db": p["snr_db"], "bits": p["bits"]}, rng) + return { + "t": t.tolist(), "y": y.tolist(), + "single": {"t": acq_s["t_samples"].tolist(), "y": acq_s["y_samples"].tolist()}, + "multi": [{"t": ch["t_adc"].tolist(), "y": ch["y_samples"].tolist(), + "delay": float(ch["delay_ns"])} for ch in acq_m["channels"]], + "n_single": int(acq_s["n_samples"]), + "n_multi": int(sum(len(ch["t_adc"]) for ch in acq_m["channels"])), + } + + +def run_page3(params, seed, method="nominal"): + """校准与重建:三方案对比(不校准/基线+增益/完整校准)。""" + p = dict(params) + p["gain_errors"] = [0.0, 0.10 * p.get("gain_amp", 1.0), -0.08 * p.get("gain_amp", 1.0), 0.06 * p.get("gain_amp", 1.0)] + p["delay_errors"] = [0.0, 0.5 * p.get("delay_amp", 1.0), -0.4 * p.get("delay_amp", 1.0), 0.3 * p.get("delay_amp", 1.0)] + p["baselines"] = [0.0, 0.02 * p.get("base_amp", 1.0), -0.015 * p.get("base_amp", 1.0), 0.01 * p.get("base_amp", 1.0)] + rec = pipeline.run_m2(p, int(seed)) + out = {"tau_true": rec["tau_true"]} + for key in ("raw", "nominal", "calibrated"): + d = rec[key] + out[key] = {"t": d["t"].tolist(), "y": d["y"].tolist(), + "tau": d["tau_hat"], "n": d["n_points"], + "jump": d["continuity"]["mean_jump"]} + return out + + +def run_page4(params, seed, snr=None): + """拟合与比较:NLS vs 对数线性 + 置信区间。""" + snr = float(params.get("snr_db", 50)) if snr is None else float(snr) + dt = float(params["dt"]) + t = np.arange(0.0, float(params["window_ns"]) + dt / 2, dt) + y_true = np.exp(-t / 52.8) + sigma = np.max(y_true) * 10 ** (-snr / 20) + rng = np.random.default_rng(int(seed)) + yn = y_true + rng.normal(0, sigma, len(t)) + r1 = fitting.fit_single_exp_nls(t, yn) + r2 = fitting.fit_log_linear(t, yn) + r3 = fitting.fit_single_exp_robust(t, yn) + return { + "t": t.tolist(), "y": yn.tolist(), "true": 52.8, + "nls": {"tau": r1["tau"], "A": r1["A"], "B": r1["B"]}, + "ll": {"tau": r2["tau"], "A": r2["A"], "B": r2["B_est"], "dropped": r2["dropped_points"]}, + "rob": {"tau": r3["tau"], "A": r3["A"], "B": r3["B"]}, + } + + +def run_page5(params, seed, repeats=8): + """第5页:扫描窗口 → 单通道 vs 四相位 τ₂ rel_std。""" + wins = [100, 200, 300, 400] + rows = [] + for w in wins: + p = dict(pipeline.M4_DEFAULTS) + p["window_ns"] = float(w) + p["snr_db"] = float(params.get("snr_db", 50)) + st_s = statistics.monte_carlo(lambda sd: pipeline._gen_single(sd, p), + repeats=repeats, seed=seed, tau_true=52.8) + st_f = statistics.monte_carlo(lambda sd: pipeline._gen_fiber(sd, p), + repeats=repeats, seed=seed, tau_true=52.8) + rows.append({"window": w, + "single": st_s["rel_std"] * 100 if st_s["rel_std"] else None, + "fiber": st_f["rel_std"] * 100 if st_f["rel_std"] else None}) + return {"rows": rows} + + +def run_page6(all_results, params, seed): + """第6页:一键导出当前会话 → 记录包。""" + notes = params.get("notes", "M5 GUI 会话导出") + with tempfile.TemporaryDirectory() as tmp: + run_dir = record.save_run( + tmp, "m5_gui_session", + records=[{"scan": all_results.get("page5", {}).get("rows", [])}], + data_csvs=[("page5_scan", all_results.get("page5", {}).get("rows", []))], + notes=notes, + seed=seed, + extra_meta={"gui_params": params}, + ) + # 读取并返回内容供 UI 展示(不依赖 mpl 图) + return run_dir + + + +# ===================================================================== +# tkinter 界面 +# ===================================================================== + +class GuiApp: + def __init__(self, root, workdir="runs"): + self.root = root + self.workdir = workdir + self.params = { + "preset": "monte_carlo", + "tau": 52.8, "a": 1.0, + "dt": 10.0, "bits": 12, "window_ns": 400.0, "snr_db": 50.0, + "ng": 1.468, "lengths": [0.0, 9.7, 19.4, 29.1], + "gain_amp": 1.0, "delay_amp": 1.0, "base_amp": 1.0, + } + self.seed = DEFAULT_SEED + self.results = {} + self._build() + + # ---------- 布局 ---------- + def _build(self): + self.root.title("光纤延时等效采样 · 教学实验仿真(M5 GUI)") + nb = ttk.Notebook(self.root) + nb.pack(fill="both", expand=True) + + self.p1 = ttk.Frame(nb) + self.p2 = ttk.Frame(nb) + self.p3 = ttk.Frame(nb) + self.p4 = ttk.Frame(nb) + self.p5 = ttk.Frame(nb) + self.p6 = ttk.Frame(nb) + nb.add(self.p1, text="1 参数方案") + nb.add(self.p2, text="2 原始采样") + nb.add(self.p3, text="3 校准重建") + nb.add(self.p4, text="4 拟合比较") + nb.add(self.p5, text="5 参数扫描") + nb.add(self.p6, text="6 结果导出") + + self._build_p1() + self._build_p2() + self._build_p3() + self._build_p4() + self._build_p5() + self._build_p6() + + # 状态栏 + self.status = tk.Label(self.root, text="就绪", anchor="w", relief="sunken") + self.status.pack(fill="x", side="bottom") + + def _canvas(self, parent): + fig = Figure(figsize=(5.2, 3.4), dpi=100) + ax = fig.add_subplot(111) + canvas = FigureCanvasTkAgg(fig, master=parent) + canvas.get_tk_widget().pack(fill="both", expand=True) + return fig, ax, canvas + + def _p1(self): + return self.p1, self._canvas(self.p1) + + # ====== 页面 1:参数方案 ====== + def _build_p1(self): + fr = ttk.LabelFrame(self.p1, text="物理参数") + fr.pack(fill="x", padx=6, pady=4) + self.v_tau = tk.DoubleVar(value=self.params["tau"]) + self.v_dt = tk.DoubleVar(value=self.params["dt"]) + self.v_ng = tk.DoubleVar(value=self.params["ng"]) + ttk.Label(fr, text="τ₂ (ns)").grid(row=0, column=0, sticky="e") + ttk.Entry(fr, textvariable=self.v_tau, width=8).grid(row=0, column=1) + ttk.Label(fr, text="采样间隔 (ns)").grid(row=0, column=2, sticky="e") + ttk.Entry(fr, textvariable=self.v_dt, width=8).grid(row=0, column=3) + ttk.Label(fr, text="n_g").grid(row=0, column=4, sticky="e") + ttk.Entry(fr, textvariable=self.v_ng, width=8).grid(row=0, column=5) + self.v_len = [tk.DoubleVar(value=v) for v in self.params["lengths"]] + ttk.Label(fr, text="光纤长度 L0-L3 (m)").grid(row=1, column=0, sticky="e") + for i, v in enumerate(self.v_len): + ttk.Entry(fr, textvariable=v, width=6).grid(row=1, column=1 + i) + self.lbl_fwd = ttk.Label(fr, text="", foreground="#234") + self.lbl_fwd.grid(row=2, column=0, columnspan=6, sticky="w") + self.fig_p1, self.ax_p1, self.canvas_p1 = self._canvas(self.p1) + + btn = ttk.Button(self.p1, text="运行 · 参数方案", + command=self._run_p1) + btn.pack(pady=4) + + def _run_p1(self): + try: + lengths = [v.get() for v in self.v_len] + res = run_page1({ + "preset": "monte_carlo", "tau": self.v_tau.get(), + "dt": self.v_dt.get(), "ng": self.v_ng.get(), + "lengths": lengths, + }) + self.results["page1"] = res + self._draw_phase(None, res["forward"]) + self._msg("参数方案已运行:均匀性 " + str(res["forward"].get("uniformity"))) + except Exception: + self._msg("失败:" + traceback.format_exc(limit=2)) + + # ====== 页面2:原始采样 ====== + def _build_p2(self): + fr = ttk.LabelFrame(self.p2, text="采样参数") + fr.pack(fill="x", padx=6, pady=4) + self.v_snr = tk.DoubleVar(value=self.params["snr_db"]) + self.v_bits = tk.IntVar(value=self.params["bits"]) + ttk.Label(fr, text="SNR (dB)").grid(row=0, column=0, sticky="e") + ttk.Entry(fr, textvariable=self.v_snr, width=8).grid(row=0, column=1) + ttk.Label(fr, text="ADC bits").grid(row=0, column=2, sticky="e") + ttk.Entry(fr, textvariable=self.v_bits, width=6).grid(row=0, column=3) + ttk.Button(fr, text="重新采样", command=self._run_p2).grid(row=0, column=4, padx=6) + self.lbl_p2 = ttk.Label(fr, text="", foreground="#234") + self.lbl_p2.grid(row=1, column=0, columnspan=5, sticky="w") + self.fig_p2, self.ax_p2, self.canvas_p2 = self._canvas(self.p2) + self.canvas_p2.get_tk_widget().pack_forget() # 需要时再显示 + + def _run_p2(self): + try: + p = dict(self.params) + p["snr_db"] = float(self.v_snr.get()) + p["bits"] = int(self.v_bits.get()) + res = run_page2(p, self.seed) + self.results["page2"] = res + self.ax_p2.clear() + self.ax_p2.plot(res["t"], res["y"], "-", lw=1, label="理想") + self.ax_p2.plot(res["single"]["t"], res["single"]["y"], "o", ms=3, label="单通道") + for ch in res["multi"]: + self.ax_p2.plot([t + ch["delay"] for t in ch["t"]], ch["y"], "x", ms=3, label=f"CH φ={ch['delay'] % 10:.1f}ns") + self.ax_p2.set_xlabel("t (ns)"); self.ax_p2.set_ylabel("I") + self.ax_p2.legend(fontsize=7, ncol=2) + self.canvas_p2.draw() + self.lbl_p2.config(text=f"单通道 {res['n_single']} 点 / 四通道共 {res['n_multi']} 点") + except Exception: + self._msg("失败:" + traceback.format_exc(limit=2)) + + # ====== 页面3:校准重建 ====== + def _build_p3(self): + fr = ttk.LabelFrame(self.p3, text="误差注入与校准") + fr.pack(fill="x", padx=6, pady=4) + self.v_gain = tk.StringVar(value="1.0") + self.v_delay = tk.StringVar(value="1.0") + self.v_base = tk.StringVar(value="1.0") + for i, (txt, v) in enumerate([("增益失配", self.v_gain), ("延时误差", self.v_delay), ("基线偏置", self.v_base)]): + ttk.Label(fr, text=txt).grid(row=0, column=i * 2, sticky="e") + ttk.Entry(fr, textvariable=v, width=6).grid(row=0, column=i * 2 + 1) + self.fig_p3, self.ax_p3, self.canvas_p3 = self._canvas(self.p3) + self.lbl_p3 = ttk.Label(self.p3, text="") + self.lbl_p3.pack() + ttk.Button(self.p3, text="运行校准对比", command=self._run_p3).pack(pady=2) + + def _run_p3(self): + try: + p = dict(self.params) + p["gain_amp"] = float(self.v_gain.get()) + p["delay_amp"] = float(self.v_delay.get()) + p["base_amp"] = float(self.v_base.get()) + res = run_page3(p, self.seed) + self.results["page3"] = res + self.ax_p3.clear() + for key, color, label in (("raw", "#C25B2A", "不校准"), ("nominal", "#6B2A35", "基线+增益"), ("calibrated", "#2D5F8A", "完整校准")): + d = res[key] + self.ax_p3.plot(d["t"], d["y"], ".", ms=2, color=color, label=label) + self.ax_p3.legend(fontsize=7) + self.canvas_p3.draw() + self.lbl_p3.config(text=(f"τ̂ 不校准={res['raw']['tau']:.2f} | 完整校准={res['calibrated']['tau']:.2f} | " + f"连续度 {res['raw']['jump']:.4f}→{res['calibrated']['jump']:.4f}")) + except Exception: + self._msg("失败:" + traceback.format_exc(limit=2)) + + # ====== 页面4:拟合比较 ====== + def _build_p4(self): + fr = ttk.LabelFrame(self.p4, text="拟合算法比较") + fr.pack(fill="x", padx=6, pady=4) + self.v_snr4 = tk.StringVar(value="50") + ttk.Label(fr, text="SNR (dB)").pack(side="left", padx=4) + ttk.Entry(fr, textvariable=self.v_snr4, width=6).pack(side="left") + ttk.Button(fr, text="比较", command=self._run_p4).pack(side="left", padx=8) + self.lbl_p4 = ttk.Label(fr, text="") + self.lbl_p4.pack(side="left", padx=8) + self.fig_p4, self.ax_p4, self.canvas_p4 = self._canvas(self.p4) + + def _run_p4(self): + try: + res = run_page4(self.params, self.seed, snr=self.v_snr4.get()) + self.results["page4"] = res + self.ax_p4.clear() + self.ax_p4.plot(res["t"], res["y"], "o", ms=2, label="数据") + for name, r, c, ls in (("NLS", res["nls"], "#2D5F8A", "-"), + ("对数线性", res["ll"], "#C25B2A", "--"), + ("稳健", res["rob"], "#4b8a5a", ":")): + if r["tau"]: + self.ax_p4.plot(res["t"], [r["A"] * np.exp(-t / r["tau"]) + (r.get("B") or 0) for t in res["t"]], + ls, color=c, label=name) + self.ax_p4.legend(fontsize=7) + self.canvas_p4.draw() + self.lbl_p4.config(text=(f"NLS τ={res['nls']['tau']:.2f} | 对数线性 τ={res['ll']['tau']:.2f} | " + f"稳健 τ={res['rob']['tau']:.2f}(真值 52.8)")) + except Exception: + self._msg("失败:" + traceback.format_exc(limit=2)) + + # ====== 页面5:参数扫描 ====== + def _build_p5(self): + fr = ttk.LabelFrame(self.p5, text="窗口 × SNR 扫描") + fr.pack(fill="x", padx=6, pady=4) + self.v_snr5 = tk.StringVar(value="50") + self.v_reps = tk.StringVar(value="8") + ttk.Label(fr, text="SNR (dB)").pack(side="left", padx=4) + ttk.Entry(fr, textvariable=self.v_snr5, width=6).pack(side="left") + ttk.Label(fr, text="reps").pack(side="left", padx=4) + ttk.Entry(fr, textvariable=self.v_reps, width=6).pack(side="left") + ttk.Button(fr, text="扫描", command=self._run_p5).pack(side="left", padx=8) + self.lbl_p5 = ttk.Label(fr, text="") + self.lbl_p5.pack(side="left", padx=8) + self.fig_p5, self.ax_p5, self.canvas_p5 = self._canvas(self.p5) + + def _run_p5(self): + try: + rows = run_page5(self.params, self.seed, repeats=int(self.v_reps.get()))["rows"] + self.results["page5"] = {"rows": rows} + self.ax_p5.clear() + ws = [r["window"] for r in rows] + self.ax_p5.plot(ws, [r["single"] for r in rows], "o-", label="单通道") + self.ax_p5.plot(ws, [r["fiber"] for r in rows], "s-", label="四相位") + self.ax_p5.set_xlabel("窗口 (ns)"); self.ax_p5.set_ylabel("σ/τ (%)") + self.ax_p5.legend(); self.ax_p5.grid(alpha=0.3) + self.canvas_p5.draw() + self.lbl_p5.config(text=f"SNR={self.v_snr5.get()} dB, reps={self.v_reps.get()}") + except Exception: + self._msg("失败:" + traceback.format_exc(limit=2)) + + # ====== 页面6:结果导出 ====== + def _build_p6(self): + self.lbl_p6 = ttk.Label(self.p6, text="一键导出当前会话:参数 / 数据 / 图表 / 运行信息 全部落盘。") + self.lbl_p6.pack(padx=8, pady=8) + ttk.Button(self.p6, text="一键导出记录包", command=self._run_p6).pack(pady=4) + self.out_p6 = tk.Text(self.p6, height=10, width=70) + self.out_p6.pack(fill="both", expand=True, padx=8, pady=8) + + def _run_p6(self): + try: + ts = __import__("datetime").datetime.now().strftime("%Y%m%d-%H%M%S") + d = os.path.join(self.workdir, f"{ts}-m5_gui_session") + os.makedirs(os.path.join(d, "figures"), exist_ok=True) + # 1) 页面图表:SVG + PNG + figs = [("page1_phases", getattr(self, "canvas_p1", None)), + ("page2_sampling", getattr(self, "canvas_p2", None)), + ("page3_calibration", getattr(self, "canvas_p3", None)), + ("page4_fitting", getattr(self, "canvas_p4", None)), + ("page5_scan", getattr(self, "canvas_p5", None))] + saved = [] + for name, canvas in figs: + if canvas is None or not hasattr(canvas, "figure"): + continue + fig = canvas.figure + fig.savefig(os.path.join(d, "figures", f"{name}.png"), dpi=150) + fig.savefig(os.path.join(d, "figures", f"{name}.svg")) + saved.append(name) + # 2) 会话数据 JSON(参数 + 各页结果) + with open(os.path.join(d, "session.json"), "w") as f: + json.dump({"params": self.params, "seed": self.seed, + "results": {k: v for k, v in self.results.items() + if k.startswith("page")}}, f, + ensure_ascii=False, indent=2, default=_json_default) + # 3) page5 扫描表落 CSV + rows5 = self.results.get("page5", {}).get("rows", []) + if rows5: + with open(os.path.join(d, "page5_scan.csv"), "w", newline="") as f: + w = csv.DictWriter(f, fieldnames=list(rows5[0].keys())) + w.writeheader(); w.writerows(rows5) + # 4) meta(诚信标注) + with open(os.path.join(d, "meta.json"), "w") as f: + json.dump({"created": __import__("datetime").datetime.now().isoformat(timespec="seconds"), + "data_origin": "software-simulated", "seed": self.seed, + "gui_session": True}, f, ensure_ascii=False, indent=2) + # 5) README 说明 + with open(os.path.join(d, "README.md"), "w") as f: + f.write("# 记录包:M5 GUI 会话导出\n\n" + f"- 时间:{ts}\n- 数据来源:**软件仿真(software-simulated)**\n" + f"- 图表:{', '.join(saved) if saved else '(无)'}\n") + self.out_p6.delete("1.0", "end") + self.out_p6.insert("end", f"导出完成:{d}\n") + self.out_p6.insert("end", f"包含:session.json(参数+结果)、meta.json、" + f"{'page5_scan.csv、' if rows5 else ''}figures/({'、'.join(saved)})\n") + self._msg("记录包已导出:" + d) + except Exception: + self._msg("失败:" + traceback.format_exc(limit=2)) + + def _draw_phase(self, fig, fwd): + self.ax_p1.clear() + ch = fwd["channels"] + self.ax_p1.scatter([c["delay_ns"] for c in ch], range(len(ch)), s=60, color="#2D5F8A") + for c in ch: + self.ax_p1.annotate(f"CH{c['index']+1} φ={c['phase_ns']:.2f}ns", (c["delay_ns"], c["index"]), fontsize=8) + self.ax_p1.set_xlabel("延时 (ns)"); self.ax_p1.set_ylabel("通道") + self.ax_p1.set_title(f"等效网格: " + " / ".join(f"{x:.2f}" for x in fwd.get("sorted_phases_ns", [])) + " ns") + self.canvas_p1.draw() + + def _msg(self, s): + self.status.config(text=s) + + def _msg6(self, s): + self.status.config(text=s) + + def run(self): + self.root.mainloop() + + +def main(argv=None): + ap = argparse.ArgumentParser(description="M5 图形界面") + ap.add_argument("--selftest", action="store_true", help="无界面自检(仅验证可导入与数据函数)") + ap.add_argument("--runs-dir", default="runs") + a = ap.parse_args(argv) + + if a.selftest: + # 无头自检:验证各页面数据函数可用(不打开窗口) + print("[selftest] 校验页面数据函数…") + base = {"preset": "monte_carlo", "dt": 10.0, "window_ns": 400.0, "snr_db": 50.0, + "ng": 1.468, "lengths": [0.0, 9.7, 19.4, 29.1], "bits": 12, + "gain_amp": 1.0, "delay_amp": 1.0, "base_amp": 1.0, "tau": 52.8} + print(" page1:", sorted(run_page1(base)["forward"].keys())) + r2 = run_page2(base, 42) + print(" page2: n_single=%d n_multi=%d" % (r2["n_single"], r2["n_multi"])) + r3 = run_page3(base, 42) + print(" page3: tau=", round(r3["calibrated"]["tau"], 3)) + r4 = run_page4(base, 42) + print(" page4: nls=%.3f ll=%.3f" % (r4["nls"]["tau"], r4["ll"]["tau"])) + r5 = run_page5(base, 42, repeats=4) + print(" page5: rows=%d" % len(r5["rows"])) + print("[selftest] 全部通过") + return 0 + + if not HAS_TK: + print("当前环境无 tkinter,无法打开 GUI(headless 下可用 --selftest 自检)。") + return 1 + + root = tk.Tk() + GuiApp(root, workdir=a.runs_dir) + root.mainloop() + return 0 + + +if __name__ == "__main__": + import os + import sys + sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) + raise SystemExit(main()) \ No newline at end of file diff --git a/fiber_delay_sim/tests/__pycache__/test_m5_gui.cpython-312-pytest-9.1.1.pyc b/fiber_delay_sim/tests/__pycache__/test_m5_gui.cpython-312-pytest-9.1.1.pyc new file mode 100644 index 0000000..67f4f3f Binary files /dev/null and b/fiber_delay_sim/tests/__pycache__/test_m5_gui.cpython-312-pytest-9.1.1.pyc differ diff --git a/fiber_delay_sim/tests/test_m5_gui.py b/fiber_delay_sim/tests/test_m5_gui.py new file mode 100644 index 0000000..34bf545 --- /dev/null +++ b/fiber_delay_sim/tests/test_m5_gui.py @@ -0,0 +1,99 @@ +"""M5 验收测试:GUI 六页面数据函数 + 一键导出记录包结构。 + +运行:pytest fiber_delay_sim/tests/test_m5_gui.py -v +(无 tkinter 环境时跳过 GUI 构建部分,仅验证计算函数与导出结构。) +""" + +import json +import os +import sys + +import pytest + +sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) + +from fiber_delay_sim import pipeline # noqa: E402 +from fiber_delay_sim.gui import run_page1, run_page2, run_page3, run_page4, run_page5 # noqa: E402 + +BASE = {"preset": "monte_carlo", "dt": 10.0, "window_ns": 400.0, "snr_db": 50.0, + "ng": 1.468, "lengths": [0.0, 9.7, 19.4, 29.1], "bits": 12, + "gain_amp": 1.0, "delay_amp": 1.0, "base_amp": 1.0, "tau": 52.8} + + +def test_page1_phase_grid(): + """M5 页1:光纤正向设计 → 等效网格 ~0/2.5/5/7.5 ns。""" + r = run_page1(BASE) + fwd = r["forward"] + phases = sorted(fwd["sorted_phases_ns"]) + assert len(phases) == 4 + assert phases[0] < 0.01 + assert abs(phases[1] - 2.5) < 0.05 + assert abs(phases[3] - 7.5) < 0.05 + # 反向设计闭环 + inv = r["inverse"] + assert inv["is_uniform_grid"] is True + + +def test_page2_sampling_counts(): + r = run_page2(BASE, 42) + assert r["n_single"] == 40 # 400 ns / 10 ns + assert r["n_multi"] == 160 # 40 × 4 通道 + assert len(r["multi"]) == 4 + + +def test_page3_calibration_improves_continuity(): + r = run_page3(BASE, 42) + j_raw = r["raw"]["jump"] + j_cal = r["calibrated"]["jump"] + assert j_cal < j_raw # 校准消除通道断层 + assert r["calibrated"]["tau"] is not None + + +def test_page4_algorithms_consistent(): + r = run_page4(BASE, 42, snr=50) + assert r["nls"]["tau"] is not None + assert r["ll"]["tau"] is not None + assert abs(r["nls"]["tau"] - 52.8) < 3.0 # NLS 接近真值 + assert abs(r["ll"]["tau"] - 52.8) < 8.0 # 对数线性估基线有偏差但可算 + + +def test_page5_scan_table(): + r = run_page5(BASE, 42, repeats=4) + rows = r["rows"] + assert len(rows) == 4 + for row in rows: + assert row["single"] is not None + assert row["fiber"] is not None + # 4 次 MC 短窗口(100 ns)统计噪声大,只断言长窗口四相位不劣于单通道 + long_wins = [row for row in rows if row["window"] >= 300] + assert len(long_wins) >= 2 + for row in long_wins: + assert row["fiber"] <= row["single"] + 0.5 # 四相位(或持平)优于单通道 + + +def test_gui_export_structure(tmp_path): + """M5 第6页:一键导出 → 记录包结构完整(无 tkinter 也可验证)。""" + from fiber_delay_sim.gui import GuiApp, HAS_TK + if not HAS_TK: + pytest.skip("无 tkinter,跳过 GUI 构建导出测试") + import tkinter as tk + root = tk.Tk() + root.withdraw() + app = GuiApp(root, workdir=str(tmp_path)) + app.params.update(BASE) + app._run_p1() + app._run_p5() + app._run_p6() + root.destroy() + dirs = [d for d in os.listdir(tmp_path) if d.endswith("m5_gui_session")] + assert len(dirs) == 1 + d = os.path.join(tmp_path, dirs[0]) + assert os.path.exists(os.path.join(d, "session.json")) + assert os.path.exists(os.path.join(d, "meta.json")) + assert os.path.exists(os.path.join(d, "page5_scan.csv")) + figs = os.listdir(os.path.join(d, "figures")) + assert any(f.endswith(".png") for f in figs) + assert any(f.endswith(".svg") for f in figs) + with open(os.path.join(d, "meta.json")) as f: + meta = json.load(f) + assert meta["data_origin"] == "software-simulated" \ No newline at end of file