Ferkans — Interactive Telecom Tutor

Why Mean-Square Calculus?

In deterministic calculus, we differentiate and integrate functions. In stochastic calculus, we want to do the same with random processes — but what does it mean for a random process to be "continuous" or "differentiable"? Pointwise (sample-path) continuity is too strong for many purposes and too hard to verify. The point is that mean-square (m.s.) convergence gives us a weaker but far more useful notion: a process is m.s.-continuous if its autocorrelation is continuous, and m.s.-differentiable if the autocorrelation is smooth enough. These conditions are checkable directly from $r_{xx}(\tau)$ or $P_x(f)$ , without examining individual sample paths.

Mean-Square Convergence

A sequence of random variables $\{X_n\}$ converges in mean square (m.s.) to $X$ if $\lim_{n\to\infty} \mathbb{E}[|X_n - X|^2] = 0$ . We write $X_n \xrightarrow{\text{m.s.}} X$ or $\text{l.i.m.}_{n\to\infty} X_n = X$ .

Definition:
Mean-Square Continuity

A random process $\{X(t) : t \in \mathbb{R}\}$ with $\mathbb{E}[|X(t)|^2] < \infty$ for all $t$ is mean-square continuous at $t_0$ if $\text{l.i.m.}_{t \to t_0} X(t) = X(t_0),$ i.e., $\lim_{t \to t_0} \mathbb{E}[|X(t) - X(t_0)|^2] = 0$ . The process is m.s.-continuous if it is m.s.-continuous at every $t_0 \in \mathbb{R}$ .

For WSS processes, m.s.-continuity at one point implies m.s.-continuity everywhere, because the condition depends only on $r_{xx}(\tau)$ , not on the time origin.

,

Mean-Square Continuity

A process is m.s.-continuous at $t_0$ if $\mathbb{E}[|X(t) - X(t_0)|^2] \to 0$ as $t \to t_0$ . For WSS processes, this is equivalent to $r_{xx}(\tau)$ being continuous at $\tau = 0$ .

Related: Mean-Square Convergence

Theorem: Characterization of Mean-Square Continuity

Let $X(t)$ be a WSS process with autocorrelation $r_{xx}(\tau)$ . Then $X(t)$ is mean-square continuous if and only if $r_{xx}(\tau)$ is continuous at $\tau = 0$ .

Equivalently (by Wiener-Khinchin), $X(t)$ is m.s.-continuous if and only if $\int_{-\infty}^{\infty} P_x(f)\, df < \infty,$ i.e., the process has finite average power.

The m.s. difference $\mathbb{E}[|X(t+\tau) - X(t)|^2]$ depends only on the autocorrelation evaluated near the origin. If $r_{xx}$ jumps at the origin (as for white noise), the process is not m.s.-continuous. If $r_{xx}$ is smooth there, the process is.

Proof

Expand the mean-square difference

Compute $\mathbb{E}[|X(t+\tau) - X(t)|^2]$ by expanding the square: $\mathbb{E}[|X(t+\tau) - X(t)|^2] = \mathbb{E}[|X(t+\tau)|^2] - 2\,\text{Re}\,\mathbb{E}[X(t+\tau)X^*(t)] + \mathbb{E}[|X(t)|^2].$ For a WSS process, this equals $r_{xx}(0) - 2\,\text{Re}\,r_{xx}(\tau) + r_{xx}(0) = 2\,\text{Re}[r_{xx}(0) - r_{xx}(\tau)].$

Relate to continuity at the origin

The expression $2\,\text{Re}[r_{xx}(0) - r_{xx}(\tau)] \to 0$ as $\tau \to 0$ if and only if $r_{xx}(\tau) \to r_{xx}(0)$ , which is exactly continuity of $r_{xx}$ at $\tau = 0$ .

Spectral domain equivalence

By Wiener-Khinchin, $r_{xx}(\tau) = \int P_x(f) e^{j2\pi f\tau}\, df$ . Continuity of $r_{xx}$ at $\tau = 0$ is equivalent to $r_{xx}(0) = \int P_x(f)\, df < \infty$ and the dominated convergence theorem applies since $P_x(f) \geq 0$ is integrable.

,

Example: M.S. Continuity of the Ornstein-Uhlenbeck Process

The Ornstein-Uhlenbeck (OU) process has autocorrelation $r_{xx}(\tau) = \sigma^2 e^{-\alpha|\tau|}$ for $\alpha > 0$ . Is this process m.s.-continuous?

Solution

Check continuity at the origin

We have $r_{xx}(\tau) = \sigma^2 e^{-\alpha|\tau|}$ , so $\lim_{\tau \to 0} r_{xx}(\tau) = \sigma^2 = r_{xx}(0)$ . The autocorrelation is continuous at $\tau = 0$ .

Conclude

By TCharacterization of Mean-Square Continuity, the OU process is m.s.-continuous. Indeed, $P_x(f) = \frac{2\alpha\sigma^2}{\alpha^2 + (2\pi f)^2}$ is integrable, confirming finite average power.

Common Mistake: White Noise Is Not M.S.-Continuous

Mistake:

Assuming that "white noise $W(t)$ " is a well-defined, m.s.-continuous random process.

Correction:

Ideal white noise has $r_{xx}(\tau) = \frac{N_0}{2}\delta(\tau)$ , which is not continuous at $\tau = 0$ (it is a distribution, not a function). Therefore white noise is not m.s.-continuous — and in fact $\mathbb{E}[|W(t)|^2] = \infty$ . White noise is a generalized process (a random distribution), not a standard second-order process. In practice, we always work with bandlimited noise, which is m.s.-continuous.

Definition:
Mean-Square Derivative

The mean-square derivative of a process $X(t)$ is the process $X'(t)$ defined by $X'(t) = \text{l.i.m.}_{\Delta t \to 0} \frac{X(t + \Delta t) - X(t)}{\Delta t},$ provided this limit exists.

When $X'(t)$ exists, we also write $\frac{dX}{dt}$ in the m.s. sense.

,

Mean-Square Derivative

The m.s. derivative $X'(t)$ is the $L^2$ limit of the difference quotient $[X(t + \Delta t) - X(t)]/\Delta t$ as $\Delta t \to 0$ .

Theorem: Existence and Properties of the M.S. Derivative

Let $X(t)$ be a WSS process with autocorrelation $r_{xx}(\tau)$ and PSD $P_x(f)$ . The m.s. derivative $X'(t)$ exists if and only if $\frac{\partial^2}{\partial \tau^2} r_{xx}(\tau)\bigg|_{\tau=0} \text{ exists and is finite.}$

Equivalently, in the spectral domain, $X'(t)$ exists iff $\int_{-\infty}^{\infty} (2\pi f)^2 P_x(f)\, df < \infty.$

When $X'(t)$ exists, it is WSS with:

Mean: $\mathbb{E}[X'(t)] = 0$ (assuming zero-mean $X$ ).
Autocorrelation: $r_{X'X'}(\tau) = -r_{xx}''(\tau)$ .
PSD: $P_{X'}(f) = (2\pi f)^2 P_x(f)$ .
Cross-PSD: $P_{XX'}(f) = j2\pi f \, P_x(f)$ .

Differentiation in the time domain corresponds to multiplication by $j2\pi f$ in the frequency domain — exactly as in deterministic Fourier analysis. The existence condition says the PSD must decay fast enough at high frequencies that the "amplified" spectrum $(2\pi f)^2 P_x(f)$ remains integrable.

Proof

Form the difference quotient

Define $D_\Delta(t) = [X(t+\Delta t) - X(t)]/\Delta t$ . This is a WSS process (since $X$ is) with autocorrelation $r_{DD}(\tau) = \frac{r_{xx}(\tau + \Delta t) - 2r_{xx}(\tau) + r_{xx}(\tau - \Delta t)}{(\Delta t)^2}.$

Apply the m.s. convergence criterion

By the Cauchy criterion for m.s. convergence, $D_\Delta(t)$ converges as $\Delta t \to 0$ iff $\lim_{\Delta_1, \Delta_2 \to 0} \mathbb{E}[D_{\Delta_1}(t) D_{\Delta_2}^*(t)]$ exists. This limit equals $\lim_{\Delta_1, \Delta_2 \to 0} \frac{r_{xx}(\Delta_1 - \Delta_2) - r_{xx}(\Delta_1) - r_{xx}(-\Delta_2) + r_{xx}(0)}{(\Delta_1)(\Delta_2)},$ which is $-r_{xx}''(0)$ if this second derivative exists.

Spectral domain condition

By Wiener-Khinchin, $r_{xx}''(0) = \int_{-\infty}^{\infty} -(2\pi f)^2 P_x(f)\, df$ . This is finite iff $\int (2\pi f)^2 P_x(f)\, df < \infty$ .

PSD of the derivative

The PSD of $X'(t)$ is the Fourier transform of $r_{X'X'}(\tau) = -r_{xx}''(\tau)$ . Since differentiation in $\tau$ corresponds to multiplication by $j2\pi f$ in the spectral domain, two derivatives give $(j2\pi f)^2 = -(2\pi f)^2$ , so $P_{X'}(f) = (2\pi f)^2 P_x(f)$ .

,

Example: The OU Process Is Not M.S.-Differentiable

Show that the Ornstein-Uhlenbeck process with $r_{xx}(\tau) = \sigma^2 e^{-\alpha|\tau|}$ does not have a mean-square derivative.

Solution

Check the spectral condition

The PSD is $P_x(f) = \frac{2\alpha\sigma^2}{\alpha^2 + (2\pi f)^2}$ , which decays as $1/f^2$ for large $f$ . We need $\int_{-\infty}^{\infty} (2\pi f)^2 \cdot \frac{2\alpha\sigma^2}{\alpha^2 + (2\pi f)^2}\, df.$ For large $|f|$ , the integrand approaches $2\alpha\sigma^2$ , which is not integrable.

Equivalently, check the autocorrelation

$r_{xx}(\tau) = \sigma^2 e^{-\alpha|\tau|}$ has a kink (non-differentiable point) at $\tau = 0$ . The left and right derivatives of $e^{-\alpha|\tau|}$ at $\tau = 0$ are $-\alpha$ and $+\alpha$ respectively, so $r_{xx}'(\tau)$ has a jump discontinuity and $r_{xx}''(0)$ does not exist.

Physical interpretation

The OU process has too much high-frequency content (PSD decays only as $1/f^2$ ) for the derivative to exist. Loosely, the sample paths are continuous but "rough" — they look like Brownian motion at fine scales.

PSD of a Process and Its M.S. Derivative: $P_{X'}(f) = (2\\pi f)^2 \P_x(f)$

Compare the PSD of a WSS process $P_x(f)$ with the PSD of its mean-square derivative $(2\pi f)^2 P_x(f)$ . Observe how differentiation amplifies high-frequency content. The process type selector lets you see which processes are m.s.-differentiable (finite area under the derivative PSD) and which are not.

Parameters

Process type

W

(bandwidth parameter)2

Quick Check

A WSS process has PSD $P_x(f) = \frac{1}{1 + (2\pi f)^4}$ . Does the m.s. derivative exist?

Yes — the integral of $(2\pi f)^2 P_x(f)$ converges.

No — the PSD does not vanish at high frequencies.

Cannot determine without knowing $r_{xx}(\tau)$ .

Correction:

Yes — the integral of

(2\pi f)^2 P_x(f)

converges.

We need $\int (2\pi f)^2 / [1 + (2\pi f)^4]\, df < \infty$ . The integrand decays as $1/f^2$ for large $|f|$ , so the integral converges. The m.s. derivative exists.

Definition:
Mean-Square Integral

The mean-square integral of a process $X(t)$ over $[a, b]$ is $Y = \int_a^b X(t)\, dt \triangleq \text{l.i.m.}_{n\to\infty} \sum_{k=1}^{n} X(t_k)\,\Delta t_k,$ where $\{t_k\}$ is a partition of $[a, b]$ and $\Delta t_k = t_{k+1} - t_k$ .

We say the m.s. integral exists if this limit is the same for every sequence of partitions with mesh going to zero.

,

Mean-Square Integral

The m.s. integral $\int_a^b X(t)\,dt$ is the $L^2$ limit of Riemann sums of the process. It always exists when the process is m.s.-continuous on $[a,b]$ .

Related: Mean-Square Derivative

Theorem: Existence of the Mean-Square Integral

If $X(t)$ is m.s.-continuous on $[a, b]$ , then the m.s. integral $Y = \int_a^b X(t)\, dt$ exists. Moreover:

Mean: $\mathbb{E}[Y] = \int_a^b \mathbb{E}[X(t)]\, dt$ .
Second moment: $\mathbb{E}[|Y|^2] = \int_a^b \int_a^b r_{xx}(t_1 - t_2)\, dt_1\, dt_2$ .
Variance: $\text{Var}(Y) = \int_a^b \int_a^b C_X(t_1 - t_2)\, dt_1\, dt_2$ , where $C_X(\tau) = r_{xx}(\tau) - |\mu_X|^2$ is the autocovariance.

M.s.-continuity is sufficient for the m.s. integral to exist — we do not need differentiability. Integration is a smoothing operation: even "rough" processes like the OU process can be integrated in the m.s. sense.

Proof

Cauchy criterion for Riemann sums

Let $S_n = \sum_{k=1}^n X(t_k)\Delta t_k$ be a Riemann sum. We need $\mathbb{E}[|S_n - S_m|^2] \to 0$ as $n, m \to \infty$ . Expanding: $\mathbb{E}[|S_n - S_m|^2] = \sum_{k,l} \Delta t_k \Delta t_l \cdot \big[r_{xx}(t_k - t_l) - r_{xx}(t_k' - t_l')\big] + \cdots$ where the primed indices come from partition $m$ .

Use continuity of the autocorrelation

Since $r_{xx}(\tau)$ is continuous (by m.s.-continuity of $X$ ), it is uniformly continuous on the compact set $[-(b-a), b-a]$ . Therefore as both partitions are refined, the double Riemann sums converge to $\int\int r_{xx}(t_1 - t_2)\, dt_1\, dt_2$ , and the Cauchy criterion is satisfied.

Compute the moments

The mean follows from linearity of expectation and the limit: $\mathbb{E}[Y] = \mathbb{E}[\text{l.i.m.}\, S_n] = \lim \mathbb{E}[S_n] = \int_a^b \mu_X(t)\, dt$ . The second moment uses $\mathbb{E}[|Y|^2] = \lim \mathbb{E}[|S_n|^2] = \int\int r_{xx}(t_1 - t_2)\, dt_1\, dt_2$ .

,

Example: Mean-Square Integration of the OU Process

Let $X(t)$ be a zero-mean WSS process with $r_{xx}(\tau) = \sigma^2 e^{-\alpha|\tau|}$ . Compute $\mathbb{E}[|Y|^2]$ where $Y = \int_0^T X(t)\, dt$ .

Solution

Apply the double integral formula

$\mathbb{E}[|Y|^2] = \int_0^T \int_0^T \sigma^2 e^{-\alpha|t_1 - t_2|}\, dt_1\, dt_2.$ $By symmetry, this equals$ 2\sigma^2 \int_0^T \int_0^{t_1} e^{-\alpha(t_1 - t_2)}, dt_2, dt_1$.

Evaluate the inner integral

$\int_0^{t_1} e^{-\alpha(t_1 - t_2)}\, dt_2 = \frac{1}{\alpha}\bigl(1 - e^{-\alpha t_1}\bigr).$ $

Evaluate the outer integral

$\mathbb{E}[|Y|^2] = \frac{2\sigma^2}{\alpha}\int_0^T \bigl(1 - e^{-\alpha t_1}\bigr)\, dt_1 = \frac{2\sigma^2}{\alpha}\left[T - \frac{1}{\alpha}(1 - e^{-\alpha T})\right].$ $For large$ T $, this grows linearly:$ \mathbb{E}[|Y|^2] \approx 2\sigma^2 T / \alpha$. This makes sense: integrating the process accumulates power linearly in the observation interval.

Common Mistake: Do Not Interchange Expectation and Calculus Without Justification

Mistake:

Writing $\mathbb{E}[X'(t)] = \frac{d}{dt}\mathbb{E}[X(t)]$ or $\mathbb{E}[\int X(t)\,dt] = \int \mathbb{E}[X(t)]\,dt$ without checking conditions.

Correction:

The interchange of expectation with m.s. differentiation requires $X'(t)$ to exist in the m.s. sense; the interchange with m.s. integration requires m.s.-continuity. These are provable results (TExistence and Properties of the M.S. Derivative and TExistence of the Mean-Square Integral), not automatic identities. In particular, pathwise differentiation may not commute with expectation even when m.s. differentiation does.

Common Mistake: Mean-Square vs. Sample-Path Properties

Mistake:

Concluding that sample paths of an m.s.-differentiable process are differentiable.

Correction:

M.s.-differentiability is a property of the ensemble, not of individual realizations. The Ornstein-Uhlenbeck process is m.s.-continuous but its sample paths (Brownian-motion-like) are nowhere differentiable. Conversely, m.s.-differentiability does not guarantee sample-path differentiability either, though for Gaussian processes with sufficiently smooth autocorrelation, one can often prove sample-path smoothness via the Kolmogorov continuity theorem.

Definition:
Higher-Order Mean-Square Derivatives

The $n$ -th mean-square derivative $X^{(n)}(t)$ is defined recursively: $X^{(n)}(t) = [X^{(n-1)}]'(t)$ . It exists if and only if $\int_{-\infty}^{\infty} (2\pi f)^{2n} P_x(f)\, df < \infty.$ When it exists, the PSD is $P_{X^{(n)}}(f) = (2\pi f)^{2n} P_x(f)$ .

Smoothness-Bandwidth Tradeoff

There is a fundamental tradeoff: the faster $P_x(f)$ decays at high frequencies, the smoother the process is in the m.s. sense. If $P_x(f) = O(|f|^{-2k-2})$ as $|f| \to \infty$ , then $X(t)$ is $k$ -times m.s.-differentiable. A bandlimited process ( $P_x(f) = 0$ for $|f| > W$ ) is infinitely m.s.-differentiable. This is the stochastic analogue of the Paley-Wiener theorem.

Summary: Mean-Square Operations on WSS Processes

Operation	Existence Condition	Spectral Condition	Result PSD
Continuity	$r_{xx}(\tau)$ continuous at $\tau = 0$	$\int P_x(f)\, df < \infty$	N/A
Derivative $X''(t)$	$r_{xx}''(\tau)\big\|_{\tau=0}$ exists	$\int (2\pi f)^2 P_x(f)\, df < \infty$	$(2\pi f)^2 P_x(f)$
$n$ -th derivative	$r_{xx}^{(2n)}(0)$ exists	$\int (2\pi f)^{2n} P_x(f)\, df < \infty$	$(2\pi f)^{2n} P_x(f)$
Integral $\int_a^b X(t)\,dt$	M.s.-continuity on $[a,b]$	$\int P_x(f)\, df < \infty$	N/A (result is a scalar RV)

Historical Note: The Origins of Mean-Square Calculus

1940s-1950s

The rigorous development of stochastic calculus in the mean-square sense was formalized by Joseph Doob in the 1940s and 1950s, building on earlier work by Norbert Wiener and Andrey Kolmogorov. The key insight — that convergence in $L^2(\Omega)$ provides a complete normed space in which limits, derivatives, and integrals can be defined — transformed the study of random processes from heuristic reasoning into rigorous mathematics. The spectral characterization of m.s. smoothness (relating PSD decay to differentiability order) was developed by Harald Cramér in his foundational 1940 paper on spectral representations.

Historical Note: Wiener's Definition of the Stochastic Integral

1920s-1930s

Norbert Wiener introduced the concept of integrating with respect to Brownian motion in his 1923 paper, laying the groundwork for what would become Itô calculus. The mean-square integral we study here — integrating a random process against ordinary (Lebesgue) measure — is simpler and older. But Wiener's work showed that even this "simple" integral required care: standard Riemann-Stieltjes integration fails for Brownian motion paths, and one needs the $L^2$ framework to make sense of stochastic integrals in full generality.

Quick Check

If $W(t)$ is white noise with PSD $N_0/2$ , does $\int_0^T W(t)\,dt$ exist as an m.s. integral?

No — white noise is not m.s.-continuous, so the integral does not exist.

Yes — the integral always exists for finite-power processes.

Yes — integration is a smoothing operation.

Correction:

No — white noise is not m.s.-continuous, so the integral does not exist.

The m.s. integral requires m.s.-continuity (TExistence of the Mean-Square Integral). White noise has $r_{xx}(\tau) = (N_0/2)\delta(\tau)$ , which is discontinuous at $\tau = 0$ , so it is not m.s.-continuous. The integral $\int_0^T W(t)\,dt$ does not exist as an m.s. Riemann integral. (It can be defined in a generalized sense via Itô integration, yielding a Wiener process.)

Key Takeaway

Mean-square calculus provides a rigorous framework for differentiating and integrating random processes. The key results are elegantly simple: m.s.-continuity requires only that $r_{xx}(\tau)$ be continuous at zero; m.s.-differentiability requires $\int (2\pi f)^2 P_x(f)\, df < \infty$ ; and m.s.-integration exists whenever the process is m.s.-continuous. Each m.s. derivative multiplies the PSD by $(2\pi f)^2$ , amplifying high frequencies — a fact that directly determines which processes are "smooth enough" for practical signal processing.

BIBO Stability

A system is bounded-input bounded-output (BIBO) stable if every bounded input produces a bounded output. For LTI systems, BIBO stability is equivalent to absolute integrability of the impulse response: $\int |h(t)|\, dt < \infty$ .

Related: Mean-Square Continuity

Autocovariance Function

The autocovariance of a WSS process is $C_X(\tau) = r_{xx}(\tau) - |\mu_X|^2$ , where $\mu_X = \mathbb{E}[X(t)]$ is the constant mean. For zero-mean processes, $C_X(\tau) = r_{xx}(\tau)$ .

Related: Mean-Square Continuity

Quick Check

A WSS process has PSD $P_x(f) = e^{-f^2}$ . How many times is it m.s.-differentiable?

Infinitely many times.

Twice.

Once.

Correction:

Infinitely many times.

The Gaussian PSD decays faster than any polynomial: $\int (2\pi f)^{2n} e^{-f^2}\, df < \infty$ for every $n$ . Therefore $X^{(n)}(t)$ exists for all $n$ .

Continuity, Differentiation, and Integration in Mean Square

Why Mean-Square Calculus?

Mean-Square Convergence

Definition: Mean-Square Continuity

Mean-Square Continuity

Theorem: Characterization of Mean-Square Continuity

Expand the mean-square difference

Relate to continuity at the origin

Spectral domain equivalence

Example: M.S. Continuity of the Ornstein-Uhlenbeck Process

Check continuity at the origin

Conclude

Common Mistake: White Noise Is Not M.S.-Continuous

Definition: Mean-Square Derivative

Mean-Square Derivative

Theorem: Existence and Properties of the M.S. Derivative

Form the difference quotient

Apply the m.s. convergence criterion

Spectral domain condition

PSD of the derivative

Example: The OU Process Is Not M.S.-Differentiable

Check the spectral condition

Equivalently, check the autocorrelation

Physical interpretation

PSD of a Process and Its M.S. Derivative: PX′(f)=(2pif)2¶x(f)P_{X'}(f) = (2\\pi f)^2 \P_x(f)PX′​(f)=(2pif)2¶x​(f)

Parameters

Quick Check

Definition: Mean-Square Integral

Mean-Square Integral

Theorem: Existence of the Mean-Square Integral

Cauchy criterion for Riemann sums

Use continuity of the autocorrelation

Compute the moments

Example: Mean-Square Integration of the OU Process

Apply the double integral formula

Evaluate the inner integral

Evaluate the outer integral

Common Mistake: Do Not Interchange Expectation and Calculus Without Justification

Common Mistake: Mean-Square vs. Sample-Path Properties

Definition: Higher-Order Mean-Square Derivatives

Smoothness-Bandwidth Tradeoff

Summary: Mean-Square Operations on WSS Processes

Historical Note: The Origins of Mean-Square Calculus

Historical Note: Wiener's Definition of the Stochastic Integral

Quick Check

Key Takeaway

BIBO Stability

Autocovariance Function

Quick Check

Definition:
Mean-Square Continuity

Definition:
Mean-Square Derivative

PSD of a Process and Its M.S. Derivative: $P_{X'}(f) = (2\\pi f)^2 \P_x(f)$

Definition:
Mean-Square Integral

Definition:
Higher-Order Mean-Square Derivatives