How we did it
 
|
|||
 | |||
|
|
|
|||
| |||
|
In this review we did not evaluate the platforms on which these soft phones were running. The rationale is that the user won't know or care about the platform-only the experience with the soft phone itself.
We conducted interactive voice-quality ratings, which are the average of at least six ratings made by at least three experienced Miercom voice-over-IP lab testers. In rotation, testers conducted real-time, two-way conversations and special tests, over separate voice-over-IP connections, between two PCs running the soft phone under test, one installed on the Compaq Armada M700 Laptop PC using a Plantronics LS1 analog headset, and the other installed on the Compaq Deskpro desktop PC using the Plantronics DSP 400 Universal Serial Bus headset. All machines supported 1-GHz Pentium III processors and 256K bytes of RAM. We then rated conversational quality of each connection based on the effect of latency; clarity and whether the voice quality was the same in both directions.
All vendors participating in this test were required to provide and technically support the IP-PBX or server-based voice-over-IP equipment on which the soft phone operated during the tests. (Test-bed equipment used in these tests) Voice encoding happens on the PCs, not on the back-end systems, so the media is already set up as voice-over-IP traffic when it gets there. In assessing soft phones, we set up our test plans to evaluate the user experience on the soft phone - that is, how does the voice quality sound to the user, based on our interactive tests.
To measure latency, we looped a voice path to begin and end at a PC running Syntrillium Software's Cooledit 2000, a tool that allows recording, editing, mixing and translation of audio signals. We configured this loop by plugging a lead from a PC running Cooledit 2000 directly into the microphone input of the transmitting phone. We then used a line splitter to bring the sound file back to the Cooledit program from the transmitting and receiving phones on separate channels. We played a .wav file and Cooledit recorded the sound as it came from the two soft phones. The difference between the timestamps on the recordings yielded our latency metric. We ran the latency tests three times and computed the average of those runs to represent the final score.
We used an Agilent Internet Advisor protocol analyzer to capture packets to verify coder/decoders, analyze the boot process and complete calls between soft phones.
RELATED LINKS
