Do You Hear What I Hear?�Part II: Defining Key Transmission Impairments
Due to its connectionless architecture, an IP network creates its own built-in obstacles to the smooth transmission of sequential data such as voice and video.
The previous tutorial in this series considered the very high standard that the existing Public Switched Telephone Network (PSTN) has set for reliable and high quality voice connections, and the likelihood that end users will not be satisfied with conversations that do meet those existing benchmarks. For example, we looked at the "Five Nines" or 99.999 percent reliability factor that is often quoted for PSTN central office switches. Most network managers and end users would agree that their present desktop computing systems do not approach the "two hours of downtime in forty years of operation" objective. To put it another way, when was the last time you had to reboot or restart your analog telephone?
But before we can compare the quality of the existing PSTN with a VoIP network connection, we must recall that these networks are fundamentally differentthe PSTN is circuit switched and connection-oriented, while VoIP networks are packet switched and connectionless. As a result, the signal transmission systems are also different, and physical impairments, which may not affect the PSTN, can produce significant degradations for a VoIP network connection. This tutorial will look at some of those sources of transmission impairments.
Many of the VoIP network Quality of Service (QoS) challenges come from the fact that we are asking a packet switched network to perform functionssuch as voice transmissionthat is was not originally designed to support. At the transmitter end of the packet switched network connection, the original data stream is divided into smaller units, or packets, such as the Internet Protocol (IP) datagrams, in a process called segmentation. At the receiver end, the reverse function is performed to put all of the packets back into a cohesive data stream in a process called reassembly. Both the segmentation and reassembly processes consume communication resources, both in terms of computing cycles, and in terms of network resources. Thus, asking the packet switched network to take a connection-oriented application (such as voice or video) and convert it into a series of packets for transmission, and then convert it back again, fundamentally goes against some of the original philosophies that were designed to support more traditional data applications, such as file transfer, email, and remote host access.
Thus, the packet network adds the following transmission impairments to the end-to-end network voice network connection:
- Packet Delay or Latency: is the difference in time between when the signal is transmitted, and when it is received. Delay is typically broken down into two components, a fixed delay and a variable delay. The fixed delays are found within the signal processing systems, such as the processing delays within the voice coders/decoders (codecs) that make the analog-to-digital signal conversions, and are also found within the physical transmission systems, such as the copper pairs. The variable delays come from queuing times at packet processing points, such as switches and routers, plus transmission variables, such as the path that a particular packet, or series of packets takes within the network. As those that have tried to carry out a telephone conversation over a satellite circuit with its intrinsic delays will attest, excessive delays (from any source) degrade the quality of the voice connection.
- Packet Jitter: measures the variation in arrival rates between individual packets. Since each packet can (theoretically) follow a unique path, it is possible that the time delay between successive packets can vary. In other words, packets number 1 and 2 might arrive 30 milliseconds apart, while packets 2 and 3 might arrive 40 milliseconds apart, because packet 3 took a different route from the first two. Excessive packet jitter adds complexity to the packet reassembly process, which must present the received voice signal to the end user as a continuous, smooth stream of information.
- Packet Loss: is a measure of the number of packets from the original data stream that do not find their way to the destination. Connectionless networks do not guarantee packet delivery, and the Packet Loss measures the severity of this condition. Voice packets are typically not retransmitted, as the information content is relatively small (generally around 20 milliseconds of voice), so the loss of a few voice packets is not severe. However, loss of many packets can degrade received signal quality.
Of the three impairments noted above, the packet latency issue is one most frequently discussed. Network designs typically include a "delay budget" and VoIP systems and subsystems are optimized with a view of minimizing that budget. We will dig deeper into the sources of latency, plus how the VoIP network can deal with this issue, in our next tutorial.
Copyright Acknowledgement: © 2005 DigiNet ® Corporation, All Rights Reserved
Mark A. Miller, P.E. is President of DigiNet ® Corporation, a Denver-based consulting engineering firm. He is the author of many books on networking technologies, including Voice over IP Technologies, and Internet Technologies Handbook, both published by John Wiley & Sons.