How we did it

All voice-over-IP calls we placed and tested during this review had exactly the same impairments applied to them. Our "Internet" profile, based on ongoing research of Internet quality-of-service conditions, applied - in each direction - 100 millisec of latency, 20 millisec of jitter, plus 1% of packet loss, applied in "bursts" of eight packets at a time. Packet loss tends to be unpredictably intermittent over the real Internet, but 1% does represent actual conditions that many encounter from time to time. And loss does tend to occur in bursts, typically the result of a momentarily congested router interface.

Our test bed provided a bidirectional T-1-capacity connection to the telecommuter site, such as DSL access would provide. However, in a worst-case scenario, with Nortel's VPN VoIP connection and concurrent videoconferencing, total bandwidth consumption barely hit 200K bit/sec in each direction.

The tester was a party in every phone conversation rated for this review, typically in the role of teleworker but also often at the headquarters end, checking to see if the same voice quality was experienced at both ends of the VoIP connection. Every connection was rated as excellent, good, fair, or poor, based on latency, voice clarity, and background noise. Compared to traditional Mean-Opinion-Score ratings, this would loosely equate to: excellent, MOS 4.0 or higher; good, 3.50 to 3.99; fair, 3.00 to 3.49; and poor, less than 3.0.

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

All voice-over-IP calls we placed and tested during this review had exactly the same impairments applied to them. Our "Internet" profile, based on ongoing research of Internet quality-of-service conditions, applied - in each direction - 100 millisec of latency, 20 millisec of jitter, plus 1% of packet loss, applied in "bursts" of eight packets at a time. Packet loss tends to be unpredictably intermittent over the real Internet, but 1% does represent actual conditions that many encounter from time to time. And loss does tend to occur in bursts, typically the result of a momentarily congested router interface.

Our test bed provided a bidirectional T-1-capacity connection to the telecommuter site, such as DSL access would provide. However, in a worst-case scenario, with Nortel's VPN VoIP connection and concurrent videoconferencing, total bandwidth consumption barely hit 200K bit/sec in each direction.

The tester was a party in every phone conversation rated for this review, typically in the role of teleworker but also often at the headquarters end, checking to see if the same voice quality was experienced at both ends of the VoIP connection. Every connection was rated as excellent, good, fair, or poor, based on latency, voice clarity, and background noise. Compared to traditional Mean-Opinion-Score ratings, this would loosely equate to: excellent, MOS 4.0 or higher; good, 3.50 to 3.99; fair, 3.00 to 3.49; and poor, less than 3.0.

We loaded and ran all the soft phones on the same, high-performance Compaq/HP laptop, running Windows XP, to ensure an even playing field. All the vendors except Mitel offered a soft phone as part of their teleworker offerings. The soft phone setup in all cases also used a fairly inexpensive Plantronics headset, which plugged into the audio in/out jacks (sound card) of the laptop. This is known to incur more latency, compared to more expensive, USB-based headsets, but it is the way we consistently tested all the soft phone contestants. Several of the vendors lamented that they strongly encourage their soft phone customers to use better, USB-based headsets for improved voice quality.

Read more about voip & convergence in Network World's VoIP & Convergence section.