Gateway VOIP

The popularity of the Internet and advancements in voice coding techniques have convinced many vendors that the time is right. It's too soon to tell if they'll succeed, but not too soon to start kicking the tires. We wanted to get some of these voice-over-IP (VOIP) gateway products into the lab and give them a workout, so we invited the dozen or so vendors who are shipping products to participate in a side-by-side evaluation.

Four vendors supplied their VOIP gateway products for testing: long-established Lucent Technologies and Micom (now a division of Nortel), as well as relative newcomers Nuera Communications and Selsius Systems . While their products differ considerably in shape, size, price per port and almost every other criterion you can imagine , they all satisfied our voice quality and price/performance expectations.
All four products can reasonably fulfill what we expect to be their most common enterprise-network application: augmenting and in some cases even displacing existing voice tie-lines, trunks and carrier services between sites. In our test configuration, shown in Figure 1, all four products successfully passed "normal" voice phone calls over an IP-routed data connection. In short, these products work, they are affordable and they provide value. Table 2 summarizes and compares our weighted scoring for each product tested.

Quality Counts

We gave product performance a relatively heavy weighting of 30 percent in our overall evaluation, and most of the performance metrics we applied relate directly to voice quality. Although the rest of the metrics relate to efficiency how much data-network bandwidth each product uses and requires as a minimum, per voice conversation there seems to be a definite correlation between efficiency and quality.
We stressed the quality issue because it is the most common concern people have about voice-over-IP gateways: Is the quality acceptable for business use? The answer is: Yes, most of the time, under normal network conditions. Of course, this depends to a considerable degree on your definition of "acceptable."

The VOIP gateways we tested turned in consolidated scores between 5.0 and 8.1. A rating of 10 would represent the best possible PSTN voice phone connection originated on an analog loop, converted once to a 64 kbps, pulse code modulated (PCM) digital DS0 channel, sent a short distance via T1 and then converted back to the destination's analog loop. Most dial up long distance calls would probably score from 7 to 8 on this scale. There is also a point on this scale at which the voice quality would be considered minimally acceptable for normal business voice communications. Our admittedly subjective judgment places this point somewhere around 4.0.

In contrast to the scores for ideal conditions, the voice quality ratings of all the products fell to between 4.0 and 7.0 when burst errors were applied on the routed T1 link connecting the gateways. In this test scenario, we introduced errors into about 10 percent of passing data packets to simulate the not-uncommon manifestations of electrical noise on telco T1 circuits.