关于「反馈式降噪的水床:伯德灵敏度积分的数值账本」,《耳内降噪、堵耳效应与听力努力》给出的定量结论是什么?

结论:第三章数值验证伯德灵敏度积分:示意反馈环路安静一侧与吵闹一侧各为 32.9 dB·kHz。

数据口径:书中的耳道尺寸、噪声谱、环路增益与认知参数均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何产品或聆听方式对听力、耳鸣或健康有任何益处。 第一章在合成噪声上运行最小均方与滤波 x 最小均方算法,说明教科书步长界 2/(L·P) 对高斯输入偏松:16 抽头时精确界是 0.1111,而次级通路模型的相位误差越过 90 度时算法发散。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 32.9

K = 10, d = 2 samples (41.7 us); max |closed-loop root| = 0.6206
f (Hz)|S| dB
50-20.71
100-20.37
200-19.23
500-15.04
1000-9.92
2000-3.97
3000-0.25
40002.34
50003.99
60004.74
80004.32
120001.70
20000-1.23
integral of ln|S|: quiet side -0.49621, loud side 0.49621
net 0.00e+00 (Bode: zero)
as area in dB x kHz: quiet -32.9, loud 32.9
查看完整验算程序
# Chapter 3. p31: The waterbed, computed: Bode's integral for feedback ANC.
# Model (illustrative): sampling 48 kHz; open loop of the error mic
# feedback canceller L(z) = K z^-d (1 - a) / (1 - a z^-1): a loop gain K,
# a one-pole roll-off at 300 Hz (a = exp(-2 pi 300 / fs)) and d samples
# of delay standing for driver, converters and acoustics. S = 1/(1 + L)
# is the ratio of noise with control to noise without. With L stable and
# strictly proper, the discrete Bode integral says
#   integral over 0..pi of ln|S(e^jw)| dw = 0  (sum of log closed-loop
#   instabilities, zero here). Closed-loop roots are checked directly.
import cmath
import math

FS, FP = 48000.0, 300.0
A = math.exp(-2 * math.pi * FP / FS)


def loop(f, k, d):
    z1 = cmath.exp(-2j * math.pi * f / FS)
    return k * z1 ** d * (1 - A) / (1 - A * z1)


def sens_db(f, k, d):
    return 20 * math.log10(abs(1 / (1 + loop(f, k, d))))


def roots(k, d):
    # z^d - a z^(d-1) + k (1 - a) = 0, Durand-Kerner iteration
    c = [1.0, -A] + [0.0] * (d - 1)
    c[d] += k * (1 - A)
    z = [0.9 * cmath.exp(2j * math.pi * (i + 0.25) / d) for i in range(d)]
    for _ in range(500):
        new = []
        for i, zi in enumerate(z):
            p = 0j
            for co in c:
                p = p * zi + co
            den = 1 + 0j
            for j, zj in enumerate(z):
                if j != i:
                    den *= zi - zj
            new.append(zi - p / den)
        z = new
    return z


def integral(k, d, n=40000):
    h = math.pi / n
    vals = []
    for i in range(n):
        w = (i + 0.5) * h
        f = w * FS / (2 * math.pi)
        vals.append(math.log(abs(1 / (1 + loop(f, k, d)))))
    neg = math.fsum(v for v in vals if v < 0) * h
    pos = math.fsum(v for v in vals if v > 0) * h
    return neg, pos


if __name__ == "__main__":
    k, d = 10.0, 2
    rmax = max(abs(r) for r in roots(k, d))
    print("K = {:.0f}, d = {} samples ({:.1f} us); max |closed-loop root| "
          "= {:.4f}".format(k, d, d / FS * 1e6, rmax))
    print("   f (Hz)   |S| dB")
    for f in (50, 100, 200, 500, 1000, 2000, 3000, 4000, 5000, 6000,
              8000, 12000, 20000):
        print("{:>9}  {:>7.2f}".format(f, sens_db(f, k, d)))
    neg, pos = integral(k, d)
    print("integral of ln|S|: quiet side {:.5f}, loud side {:.5f}".format(
        neg, pos))
    print("net {:.2e} (Bode: zero)".format(neg + pos))
    scale = 20 / math.log(10) * FS / (2 * math.pi) / 1000
    print("as area in dB x kHz: quiet {:.1f}, loud {:.1f}".format(
        neg * scale, pos * scale))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《耳内主动降噪最小均方自适应滤波、堵耳效应低频骨导增益与信噪比认知恢复》第 3 章「反馈式降噪的水床:伯德灵敏度积分的数值账本」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-164 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026164,
  author    = {Fan, Yafu},
  title     = {LMS Adaptive Filtering in In-Ear Active Noise Cancellation, Occlusion-Effect Bone-Conduction Gain and SNR-Driven Cognitive Recovery},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论