Compare the two routes in the worked x(2t) example. A one-second delay before time scaling produces an onset at 0.5 seconds; delaying the output gives an onset at 1 second. At 0.75 seconds the outputs differ, disproving time invariance. Compare gain, squaring and clock weighting to see why linearity and time invariance are independent properties. A matching probe alone never proves invariance.
At t = 1.5 s, the unit step and flipped exponential overlap from τ = 0 to 1.5 s, giving output about 0.777. Switch to the worked 2-second pulse at t = 3 s and see that input has stopped while the output remains about 0.318. Change time and pulse width to connect the causal overlap, the peak at switch-off and the release tail.
Narrow in time, wide in frequency; delay lives in phase
$X(f)=\tau\,\mathrm{sinc}(f\tau)e^{-j2\pi ft_0}$
Compare the worked 1 ms pulse with a 0.1 ms pulse on the same displayed spectrum. Its first null moves from 1 kHz to 10 kHz while its area falls by ten. Shift the pulse centre and see the complex Fourier coefficient rotate without changing magnitude. Probe a null and a negative sinc lobe to see why magnitude alone does not describe the signal.
The idea, with a picture. Sampling multiplies the signal by an
impulse train; in frequency, the spectrum replicates at multiples of
$f_{s}$. The Nyquist criterion: a signal band-limited to
$f_{max}$ が完全に復元できるのは $f_{s} > 2f_{max}$ の場合である。それ以下では、
複製が重なり、エイリアスによって高周波数が折り重なって
導出過程。 サンプラー出力
Sampled at 6 kHz, a 4 kHz tone is indistinguishable from 2 kHz
The worked 4 kHz cosine sampled at 6 kHz shares every sample with a 2 kHz cosine. Change phase to see why a negative-frequency fold must reverse cosine phase. At the Nyquist boundary, a sine can vanish at every sample. The audio example uses a 25 kHz tone at 44.1 kHz, which aliases to 19.1 kHz; raising the rate to 96 kHz avoids folding that chosen tone.
要約。$H(s) = Y/X$;部分分数分解は変換を指数関数に戻し、
exponentials; poles set the modes, the ROC the causality, the left
half-plane the stability, and $s = j\omega$ the frequency response.
探索
Move the zeros; watch the tone-control curve change
Start with the worked tone control, whose zeros are at −1 ± j9.95 and whose fixed poles are at −1 and −10. Its gain near 10 rad/s is about 0.14, while its DC gain is 10. Move the zeros farther from the imaginary axis to soften the dip, or onto it for an exact null. The causal system stays stable because the poles and ROC stay fixed. The distance products connect the s-plane directly to the gain curve.
概念と図解。 畳み込みの離散版:
$y[n] = \sum_{k} x[k]\,h[n-k]$. 差分方程式で記述されるシステム
方程式**、例えば $y[n] = 0.5\,y[n-1] + x[n]$ — 加算器
homogeneous mode is $a^{n}$ where $a$ is the root of the characteristic equation; iterating one step at a time is the direct solution.
特性方程式の1ステップごとの反復は、直接
解法である。
$y[n]=a\,y[n-1]+x[n],\quad |a|<1\text{ for causal BIBO stability}$
Recover the worked step sequence 1, 1.5, 1.75, 1.875 from the stated zero initial state. Compare the alternating approach for a = −0.8. At a = −1, a step trace stays bounded, but an alternating input of the same bounded amplitude produces growing output. The strict root condition decides stability, rather than the appearance of one short trace.
** 要約。** $a^{n} \leftrightarrow 1/(1-az^{-1})$、ROCは外側;
因果系において、単位円内の極 = 安定性;$z = e^{sT}$ は減衰を
radius and frequency to angle — and the DC gain $1/(1-a)$ closes the
反復によってループを閉じる。
探索
The ROC decides which sequence and which stability claim
$H(z)=\frac{1}{1-az^{-1}},\quad\text{stability requires the unit circle in the ROC}$
Start with the causal pole at 0.5 and a probe on the unit circle, where the geometric sum converges to 2. Keep the same rational expression but choose anti-causal support; the ROC moves inside the pole circle and the unit circle is excluded. A pole at 1.2 can instead give a stable anti-causal sequence. Probe the ROC boundary away from the pole to see why a finite algebraic value does not prove series convergence.