The variability or statistical dispersion of latency measurements across a sequence of packets or events; commonly quantified by metrics such as standard deviation, interquartile range, or maximum-minimum spread over an interval.

Principle

Jitter arises from non-deterministic queuing, scheduling, reordering, retransmission behavior, and variable processing times; it reflects instability in timing rather than absolute delay magnitude.

Demonstration

During a VoIP call, measured packet arrival intervals show varying delays: some packets arrive 20 ms after send, others 70 ms; the spread and frequency of these deviations constitute jitter that affects audio playout and may require buffering.

Misapplication

Interpreting a single high-latency sample as 'jitter' without analyzing distribution, or assuming low average latency implies low jitter; confusing jitter with packet loss or bandwidth limitations.

Consequence

Quantifying and controlling jitter (via buffering, traffic shaping, priority scheduling) stabilizes playback and control loops, improving perceived quality in streaming and real-time systems, though mitigation can increase end-to-end latency.

Reversal

Treating jitter as the same objective as latency reduction — minimizing jitter sometimes requires added buffering that increases median latency, the inverse trade-off for smoothness versus responsiveness.

Boundary

Applies to variability in timing across repeated events; excludes deterministic periodic variability that is part of a protocol design and excludes rate-based throughput measures unless they induce timing variation.

Semantic Tension

Jitter competes conceptually with latency and throughput in performance discussions; stakeholders may prioritize low median latency while others demand low jitter for consistent experience, creating trade-off debates.

Synthesis

Jitter is the timing instability of repeated transmissions, measured as the statistical dispersion of latency; it originates in scheduling and queueing dynamics, degrades real-time performance, and is controlled by techniques that may trade off increased average latency for reduced variability.