Region of Convergence | DSP-101

Why Convergence Matters

In Lesson 4.1 we wrote down the Z-transform definition X(z) = Σ x[n] z⁻ⁿ without asking a critical question: for which values of z does this infinite sum actually converge? The answer is the Region of Convergence (ROC) — the set of complex numbers z for which the sum is finite.

The ROC is not a technicality to be glossed over. Two completely different sequences can share the same algebraic expression for X(z) but differ in their ROC, producing different inverse transforms. The ROC encodes whether a signal is causal, anti-causal, or two-sided, and it determines whether a given H(z) corresponds to a stable system. Understanding the ROC is therefore essential for correctly inverting Z-transforms and for stability analysis.

Absolute Convergence Condition

\sum_{n=-\infty}^{\infty} |x[n]|\,|z|^{-n} < \infty

The Z-transform X(z) converges wherever this sum is finite. Since |z⁻ⁿ| = r⁻ⁿ with r = |z|, convergence depends only on the magnitude of z, making the ROC an annular region in the complex plane.

Because the convergence condition depends only on r = |z|, the ROC always takes the form of an annulus (possibly degenerate): a set of circles centered at the origin. This geometric simplicity makes the ROC easy to draw on the z-plane and gives rise to three archetypal shapes, one for each type of sequence.

Right-Sided Sequences: ROC Outside a Circle

A right-sided sequence is zero for n < N for some finite N (it may start before n = 0 but eventually becomes zero to the left). The most common case is a causal sequence where x[n] = 0 for n < 0.

For such sequences the Z-transform sum is dominated by large-n behavior. The sum converges when z is large enough that the decay in |z⁻ⁿ| outpaces the growth in |x[n]|. Specifically, if x[n] grows no faster than r₀ⁿ, then X(z) converges for |z| > r₀ — the exterior of a circle of radius r₀.

Causal Exponential ROC

\mathcal{Z}\{a^n u[n]\} = \frac{z}{z-a}, \quad |z| > |a|

The causal exponential aⁿu[n] has ROC |z| > |a| — the exterior of the circle of radius |a|. The pole at z = a sits on the boundary of the ROC and is excluded from it.

Notice that the pole at z = a lies on the boundary of the ROC but is never inside it. This is a universal rule: the ROC never contains poles. Poles are the values of z where X(z) blows up — they mark the edge of where the sum can converge.

Left-Sided Sequences: ROC Inside a Circle

A left-sided sequence is zero for n > N for some finite N. The extreme case is an anti-causal sequence where x[n] = 0 for n > 0. For left-sided sequences the sum is dominated by large-negative-n behavior, and convergence requires z to be small enough — specifically |z| < r₀ for some r₀.

Anti-Causal (Causal Form)

\mathcal{Z}\{a^n u[n]\} = \frac{z}{z-a}, \quad |z|>|a|

Causal: aⁿu[n] → z/(z−a), ROC: |z| > |a|. The ROC is the exterior of a circle.

Anti-Causal Form

\mathcal{Z}\{-a^n u[-n-1]\} = \frac{z}{z-a}, \quad |z|<|a|

Anti-causal: −aⁿu[−n−1] → z/(z−a), ROC: |z| < |a|. Same expression X(z), but a different ROC and a different sequence.

This side-by-side comparison is the most important illustration of why the ROC matters. Both sequences produce exactly the same rational expression z/(z−a), yet they represent completely different time-domain signals — one causal and decaying (if |a| < 1), the other anti-causal and growing. Without specifying the ROC, the inverse Z-transform is ambiguous.

Two-Sided Sequences: Annular ROC

A two-sided sequence is nonzero for both n > 0 and n < 0. Its Z-transform must converge for both "directions" simultaneously, requiring that |z| be neither too large nor too small. The ROC is an annulus r₁ < |z| < r₂.

Two-Sided Example

x[n]=a^n u[n] - b^n u[-n-1] \;\Rightarrow\; X(z)=\frac{z}{z-a}+\frac{z}{z-b},\quad |b|<|z|<|a|

A two-sided sequence built from two exponentials has an annular ROC |b| < |z| < |a| (assuming |b| < |a|). If |b| ≥ |a|, the ROC is empty and the Z-transform does not exist.

An annular ROC is only possible if the inner radius is strictly less than the outer radius. When we decompose X(z) by partial fractions, each term contributes either an exterior or an interior ROC, and the ROC of the total is the intersection of all individual ROCs. If the intersection is empty, the sequence has no Z-transform.

The Five ROC Rules

Several universal rules govern the ROC regardless of the specific sequence. Mastering them lets you determine the ROC quickly from the pole-zero plot alone.

Rule 1

Annular Shape

The ROC is always an annulus r₁ < |z| < r₂ centered at the origin (with r₁ = 0 or r₂ = ∞ allowed as degenerate cases).

Rule 2

No Poles Inside

The ROC never contains any poles of X(z). Poles sit on the boundary or outside the ROC.

Rule 3

Right-Sided → Exterior

If x[n] is right-sided, the ROC is the exterior of a circle |z| > r₁, possibly including z = ∞ if x[n] is causal (N ≥ 0).

Rule 4

Left-Sided → Interior

If x[n] is left-sided, the ROC is the interior of a circle |z| < r₂, possibly including z = 0 if x[n] is anti-causal (N ≤ 0).

Rule 5

FIR → All z (Except 0/∞)

If x[n] has finite duration (FIR), the ROC is the entire z-plane except possibly z = 0 or z = ∞ (depending on sequence extent).

Rule 6

Rational X(z) → Boundary = Pole

If X(z) is rational, the boundary of the ROC passes through the outermost or innermost pole, never between poles.

ROC and System Stability

The connection between the ROC and system stability is one of the deepest results in Z-transform theory. Recall that an LTI system is BIBO stable if and only if its impulse response h[n] is absolutely summable: Σ|h[n]| < ∞. This is precisely the convergence condition for H(z) evaluated on the unit circle.

Stability Theorem

An LTI system with transfer function H(z) is BIBO stable if and only if the unit circle |z| = 1 is contained in the ROC of H(z).

For a causal system (right-sided h[n]), this means all poles of H(z) must lie strictly inside the unit circle.

Causal + Stable ⟺ All poles inside |z| = 1, ROC includes |z| = 1

This result unifies the frequency-domain and time-domain stability conditions. It also explains the geometric picture introduced in Lesson 4.1: the mapping z = e^sT carries the imaginary axis of the s-plane onto the unit circle of the z-plane, so “left half-plane = inside unit circle” directly corresponds to stable poles in both domains.

A practical consequence: when designing digital filters, we must ensure all poles of H(z) have magnitude less than 1. The ROC framework makes this requirement precise and testable — simply locate the poles on the z-plane and check whether they lie inside the unit circle.

Choosing the Right Inverse

The inverse Z-transform recovers x[n] from X(z) via a contour integral around a closed path inside the ROC. When X(z) is rational, partial-fraction decomposition is the practical method, and the ROC determines which inverse pair to use for each term.

Inverse Z-Transform (Contour Integral)

x[n] = \frac{1}{2\pi j}\oint_C X(z)\,z^{n-1}\,dz

The inverse Z-transform integrates X(z)z^(n−1) around a counterclockwise contour C inside the ROC. In practice, partial fractions and the table of known pairs replaces this integral.

The rule is simple: for a pole at z = a, if the ROC is |z| > |a|, the corresponding time-domain term is causal (aⁿu[n]); if the ROC is |z| < |a|, the term is anti-causal (−aⁿu[−n−1]). Specifying the ROC therefore pins down the unique inverse transform.

The next lesson introduces poles, zeros, and the pole-zero plot — the graphical representation of a transfer function on the z-plane that makes frequency response, stability, and ROC structure instantly readable.

Key Takeaways

The ROC is the set of z values for which the Z-transform sum converges absolutely; it always forms an annulus centered at the origin.
The ROC never contains any poles of X(z).
Right-sided (causal) sequences have ROC = exterior of a circle (|z| > r); left-sided have ROC = interior (|z| < r); two-sided have an annular ROC.
The same rational expression X(z) can correspond to different sequences depending on which ROC is specified — the ROC is essential for a unique inverse transform.
A causal LTI system is BIBO stable if and only if all poles lie strictly inside the unit circle (ROC includes the unit circle).
For partial-fraction inversion: pole with ROC |z| > |a| gives aⁿu[n]; ROC |z| < |a| gives −aⁿu[−n−1].