📚Signal Processing Unit 4 Review

Parseval's theorem connects two ways of measuring a signal's energy: in the time domain and in the frequency domain. The total energy you compute from the time-domain waveform will always equal the total energy you compute from its Fourier transform. This guarantee of energy conservation is what makes it safe to move between domains when designing filters, compressing data, or analyzing noise.

Fundamental relationship between time and frequency domains

Parseval's theorem states that the total energy of a signal equals the sum of the energies of its individual frequency components. In practical terms, no energy is created or destroyed by taking a Fourier transform. The energy just gets expressed differently: spread across time samples in one domain, and spread across frequency bins in the other.

This applies to both continuous-time and discrete-time signals, though the exact formulas differ slightly (covered below).

Analyzing energy distribution across frequencies

Because total energy is preserved, you can use the frequency-domain representation to study where a signal's energy lives. This is useful for:

Identifying dominant frequencies and their relative contributions to overall signal energy
Designing filters that target specific frequency bands (e.g., removing a 60 Hz hum while keeping speech frequencies intact)
Comparing signals by examining how their energy is distributed, even when their time-domain waveforms look very different

The squared magnitude of the Fourier transform, $|X(f)|^2$ , is called the energy spectral density. It tells you how much energy sits at each frequency.

Calculating signal energy

Continuous-time signals

For a continuous-time signal $x(t)$ with Fourier transform $X(f)$ , Parseval's theorem says:

$E_x = \int_{-\infty}^{\infty} |x(t)|^2 \, dt = \int_{-\infty}^{\infty} |X(f)|^2 \, df$

Both integrals run over the entire real line. The left side sums up squared amplitude over all time; the right side sums up squared amplitude over all frequencies. They give the same number.

If you use the angular frequency convention $X(\omega)$ instead of $X(f)$ , the frequency-domain integral picks up a factor of $\frac{1}{2\pi}$ :

$E_x = \int_{-\infty}^{\infty} |x(t)|^2 \, dt = \frac{1}{2\pi}\int_{-\infty}^{\infty} |X(\omega)|^2 \, d\omega$

Which form you use depends on which Fourier transform convention your course adopts. Just be consistent.

Discrete-time signals

For a discrete-time signal $x[n]$ of length $N$ with DFT $X[k]$ , the theorem becomes:

$E_x = \sum_{n=0}^{N-1} |x[n]|^2 = \frac{1}{N} \sum_{k=0}^{N-1} |X[k]|^2$

The left side sums squared magnitudes over all time samples. The right side sums squared magnitudes over all DFT bins, divided by $N$ .

The $\frac{1}{N}$ factor appears because the standard DFT definition accumulates a factor of $N$ in the forward transform. If your course uses a symmetric convention ( $\frac{1}{\sqrt{N}}$ on both the forward and inverse DFT), the $\frac{1}{N}$ disappears and both sides match directly.

Normalization and scaling factors

The most common source of errors with Parseval's theorem is getting the scaling wrong. Here's a quick reference:

Convention	Forward transform scaling	Parseval's frequency-side factor
$X(f)$ (ordinary frequency)	1	1
$X(\omega)$ (angular frequency)	1	$\frac{1}{2\pi}$
DFT (standard, no $\frac{1}{N}$ in forward)	1	$\frac{1}{N}$
DFT (symmetric)	$\frac{1}{\sqrt{N}}$	1

Always check which convention your textbook or problem set uses before plugging in numbers.

Energy conservation in Fourier transform

Redistribution of energy among frequencies

When you take a Fourier transform, you're decomposing a signal into sinusoidal components at different frequencies. Each component carries some portion of the total energy. The transform redistributes the energy across these components, but the total stays constant.

Think of it this way: a short, sharp pulse in time concentrates its energy in a narrow time window but spreads it broadly across frequencies. A long, steady sinusoid does the opposite. In both cases, adding up all the energy in either domain gives the same result.

Importance in signal analysis and interpretation

Energy conservation underpins many practical techniques:

Filtering: You can predict exactly how much energy a bandpass filter will remove by integrating $|X(f)|^2$ over the rejected frequency band.
Compression: Discarding low-energy frequency components (as in audio and image compression) lets you quantify the reconstruction error directly as the energy you threw away.
Noise analysis: If you know the noise power spectral density, Parseval's theorem lets you compute total noise energy and compare it against signal energy (SNR calculations).
Communications: Channel coding and equalization schemes rely on energy-preserving transforms to ensure transmitted signal power is accounted for correctly at the receiver.

The core takeaway: because the Fourier transform preserves energy, any analysis you do in the frequency domain is a faithful representation of what's happening in time. You're not losing or gaining anything by switching domains.

📚Signal Processing Unit 4 Review