CONVERGENCE
Soft clients vs. IP phones
Independent testing gives soft clients the edge in terms of voice quality.
by Mike Hollier
One of the main attractions of IP telephony (IPT) to the enterprise is cost savings. Lower call costs promise operational savings for on-net voice calls and voice conferencing. In addition, the more flexible infrastructure can provide a wide range of new features–far beyond the capabilities of plain old telephony service (POTS)–to enhance productivity. This enrichment is a cornerstone of the unified communications argument currently receiving extensive attention from several leading vendors and system integrators.
Performance-management tools that consider only IP network performance may result in undetected and undiagnosed problems.
The industry typically describes two basic types of voice-over-IP (VoIP) end-user edge devices:
- IP-phones are dedicated hardware phones and often appear as familiar-looking handsets. They are typically specified as part of an IPT deployment from the vendors.
- Soft clients are software applications running on a PC that can utilize the network connectivity and high processing power of the PC.
IP phones include carefully designed acoustic transducers (a mouthpiece and earpiece) that can be selected to meet precise performance requirements. This helps to minimize variability across a deployment. Also, IP phones are often designed to look, act and sound like regular phones.
Many IP phones operate on a strict narrowband frequency response to mimic POTS. As a dedicated instrument, the IP phone does not have to share processing resources with unrelated applications. Soft clients typically utilize a wider range of less-stringently controlled peripherals, including various sound cards, headsets and handsets. This can lead to performance issues, including variation. Soft clients can benefit from PC desktop integration, readily incorporating additional information, such as address books.
In terms of cost, IP phones vary widely. A midrange IP phone from one leading vendor costs approximately $500; a high-end IP phone can cost in excess of $1,000. Soft clients are typically a hidden cost, as part of a service or desktop software suite, and rely on the provision of a PC to provide powerful signal processing.
Beyond the PC, soft clients require handset or headset peripherals, which can be relatively inexpensive. For example, a basic headset retails for $25, while a high-end noise-reducing headset may cost $100. Associated licensing and server costs should be factored into system costs.
When deploying IPT, two issues are paramount. First, products that fit within budgetary constraints should be selected. Also, be cognizant of adequate network management, including the rapid detection and identification of service-affecting issues that are perceptible to end-users.
The end-user’s quality of experience (QoE) for a voice call is determined by both the IP network performance and application-level factors, such as echo, noise, delay, speech levels and speech waveform distortion. Edge devices are critical to ensuring that application-level performance is satisfactory. Poorly designed, misconfigured or malfunctioning edge devices will result in unsatisfactory QoE and end-user complaints.
Performance-management tools that consider only IP network performance may result in undetected and undiagnosed problems. Solutions do exist, however, that measure and diagnose the full set of factors driving QoE telephony issues, including the application level. In a recent study, an independent testing laboratory made a detailed comparison of the performance of a soft client and IP phone from leading vendors. Objective measurements were made on network bandwidth utilization and delay.
Objective tools for voice quality measurement include ITU-T P.862 PESQ (perceptual evaluation of speech quality). A subjective test was preferred for this study because the sophisticated packet-loss concealment (PLC) strategy employed by the soft client falls outside the current scope for PESQ.
The subjective test followed well-established ITU-T guidelines and best practices, and utilized untrained listeners. The listeners provided their opinion on voice quality against ITU-T P.800 scales.
The test was conducted in a purpose-built subjective testing laboratory. The opinion score for multiple subjects was collected for each test condition and then averaged to provide a mean opinion score. The test cases used for the evaluation cover a range of real-world scenarios relevant to enterprise networks.
The report findings show that the soft client matched or outperformed the IP phone in terms of subjectively judged voice quality; however, the data was collected in a wideband performance context and the IP phone was narrowband-only. The subjective judgments are valid for users in a wideband environment. The experimental results cover performance in noise, with different bandwidths and a variety of IP impairments.
The independent test suggests that constraining the performance of voice systems to that of POTS when designing IPT solutions is not necessary. The increased fidelity of wideband solutions increases the perceived quality of speech and usability of the system.
Also, soft clients utilize part of the processing power of the PC to perform advanced PLC, providing high error resilience. The test measured higher voice-quality performance from the leading soft client than the leading-brand IP phone. In addition, the risk of performance variability from end-point peripherals, such as headsets and handsets, is generally greater with soft clients than with IP phones.
Delivering satisfactory end-user QoE in live networks requires in-service monitoring of both the IP network and application-level performance. Edge devices are a key component in delivering IPT QoE and can be a primary source of application-level performance issues. Network performance tools that focus only on the IP network cannot detect or diagnose application-level issues. Real-time, per-session measurements covering all these performance aspects are imperative.
Mike Hollier is the CTO for Psytechnics, Portsmouth, N.H.
For more information
(click here)