VOICE OVER IP AND MULTIMEDIA BASICS

Tags: #◇Computer Science #◇NSCOM01 #◇Conceptual #◇VoIP

---

◇Overview

Voice over IP (VoIP) is the transmission of voice and multimedia content over Internet Protocol (IP) networks. It replaces traditional circuit-switched telephone networks with packet-switched network technology. Providing a seamless conversational experience poses several networking challenges, as IP networks were originally designed for best-effort delivery of non-real-time data.

◇Key Definitions

• VoIP (Voice over IP): Technologies for delivering voice communications and multimedia sessions over IP networks.
• Delay Jitter: The variation in packet delay or latency over time.
• Playout Buffer: An application-level buffer that smooths out jitter by delaying the playout of packets to a fixed time.
• Talk Spurt / Silent Period: The alternating states of human speech.
• Interleaving: A loss concealment technique where audio chunks are divided and spread across multiple packets.

◇Theoretical Framework / Core Concepts

Delay Requirements

For a natural conversational flow, end-to-end delay must be strictly controlled:

• < 150 ms: Ideal for seamless conversation.
• > 400 ms: Highly noticeable, disrupting conversational flow and leading to "talk-over".

Delay Components

Total delay is the sum of network and application delays:
$Total\_Delay = d_{transmission} + d_{propagation} + d_{queueing} + d_{packetization} + d_{playout}$

Traffic Characteristics

1. Talk Spurts vs. Silent Periods: Speech is not continuous; VoIP systems often suppress transmission during silent periods to save bandwidth.
2. Data Rate: During talk spurts, audio is often encoded at 64 kbps.
3. Packetization: Audio is typically grouped into 20-msec chunks.
   • At 8 Kbytes/sec (64 kbps), each chunk = $160$ bytes.
   • These chunks get encapsulated (along with an application-layer header) into UDP (or TCP) segments every 20 ms.

Packet Loss & Delay Jitter

• Network Loss: Caused by router buffer overflows during network congestion.
• Delay Loss: A packet arriving later than its scheduled playout time is functionally "lost" to the application.
• Tolerance: VoIP can tolerate 1% to 10% packet loss depending on the codecs and error concealment used.

Playout Delay & Jitter Buffering

To handle delay jitter, the receiver utilizes a playout buffer, playing audio at exactly $t + q$, where $t$ is the generation timestamp and $q$ is the buffer delay.

• Large $q$: Decreases loss rate (allows more time for late packets) but worsens interactivity.
• Small $q$: Better interactivity but higher loss of late packets.

Loss Concealment: Interleaving

To mitigate packet loss without adding redundancy overhead, interleaving is used. A 20-ms chunk is split into four 5-ms units. These units are distributed across multiple consecutive packets.
If one packet is lost, the receiver still obtains 15 ms of every affected 20-ms chunk, slightly degrading audio quality rather than causing a complete gap. However, it increases end-to-end delay because the receiver must wait for all interleaved packets to reconstruct the audio.

◇Worked Examples / Calculations

Calculating Packet Size:
Given an audio encoding rate of 64 kbps and a packetization interval of 20 ms.

1. Convert kbps to bytes/sec: 64 kbps = 64,000 bits/sec = 8,000 bytes/sec.
2. Multiply by interval: 8,000 bytes/sec $\times$ 0.020 sec = 160 bytes per chunk.

◇Summary & Key Takeaways

• End-to-end delay must be kept below 150 ms for optimal VoIP performance.
• Network congestion leads to packet loss and jitter.
• Playout buffers smooth out jitter but introduce additional application delay.
• Interleaving conceals packet loss by spreading chunks across multiple packets, trading off increased latency for better resilience.

◇References & Navigation

• Next Topic: ◇Real-Time Protocols
• Related Notes: ◇Applications