|
Is
voice quality really that important?
Yes,
from our discussions with a number of carriers and end customers,
voice quality is the #1 post-installation operational and customer
satisfaction issue.
"From
our experience, voice quality is the most sensitive issue
for business customers" Greg Schreiber, VP of Product Development
- Vaspian (carrier customer, U.S.A.)
It
seems that Murphy's Law comes to surface with many new VoIP installations.
After many hours of testing and successful operations, as soon
as the new customer's CEO picks up the phone, a technician elsewhere
in the organization will start a huge file transfer over the broadband
link, causing network congestion. With poor quality gateway equipment,
the CEO's conversation will be plagued with broken and choppy
speech.
Customers
have told us that poor voice quality is definitely one of the
items that will cause a "change in vendor".
|
What
affects voice quality?
The
performance of the compression algorithm, reliability of
the Voice Activity Detector, the number of lost packets
on the network and a number of other conditions affect the
perceived quality of voice in a VoIP telephone conversation.
Excellent voice quality is a balancing act that is highly
dependent on the data network's bandwidth and reliability.
In
some cases, there is an interaction between the different
algorithms that isn't obvious. For example: in commercial
broadband applications where the data network link bandwidth
is limited, many times the customer will find that using
a Low Bit Rate coder (like G.723.1) produces a better overall
voice quality experience than using uncompressed G.711.
Why? Because the Low Bit Rate coder compresses the speech,
thereby causing less congestion on the network, resulting
in fewer dropped packets, yielding smoother, more consistent
speech.
|
 |
How
is AudioCodes different?
After
many years of developing speech compression algorithms, AudioCodes
has developed a number of standard and proprietary techniques
to produce very high quality speech. Just a sampling of what makes
AudioCodes voice quality tops in the industry:
-
Dynamic Jitter Buffer — dealing with network jitter is
always a challenge, and it is especially difficult on shared
broadband links. AudioCodes has developed a proprietary algorithm
to manage the depth of the jitter buffering, maintaining a
smooth stream of speech without unnecessary latency.
-
Echo Cancellation — starting with standard G.168 echo
algorithms, AudioCodes has enhanced its technology with detection
and acquisition algorithms, minimizing or eliminating annoying
initial echo conditions
-
Packet Loss Concealment — packets get lost — it's
a fact of real life — and AudioCodes has developed a
number of algorithms to smooth the speech over the holes left
by any missing packets.
How
is voice quality tested?
Voice
quality is tested on a range of performance attributes. The core
of the testing is Listening Speech Quality Scoring using Mean
Opinion Scoring (MOS). Modern MOS scoring is performed using a
set of algorithms to measure what a bank of humans used to perform,
scoring the quality of speech from 1 to 5. To perform this testing,
gateways or other devices are put on a test bench that sends a
range of pre-recorded human speech clips into the system; then
the resulting output speech is analyzed, comparing it against
the original. The tests are run to test a number of voice criteria
and in a number of degraded network conditions.
What
testing has AudioCodes done recently?
AudioCodes
participated in the 3rd Annual European Telecommunications Standards
Institute (ETSI) Speech Quality Test Event held in the UK and
US. This year's ETSI testing pitted 10 different vendors against
each other, testing the performance of supplied test gateways.
AudioCodes
submitted a Mediant™ 2000 digital gateway for the testing.
The Mediant 2000 shares core enabling technology with a wide range
of AudioCodes gateway and media server platforms, testing the
core of a large number of AudioCodes products.
The
results of the testing are shown by a color-coded "Gateway Pie"
chart, with the results of the AudioCodes Mediant 2000 shown below.
Each segment of the pie represents one battery of tests. The size
of the pie slice represents how well AudioCodes Mediant 2000 performed
in each of the battery of tests. The color of the pie slices indicates
whether the results were below standard (red), met the standard
(yellow) or exceeded the standard (green). As you can see, there
are no test results below standard, and a number of tests exceeded
the standard significantly.
How
did other vendors do in the testing?
ETSI
performs their testing anonymously, releasing all the results
to each vendor, but without indicating the identity of the competing
vendors. Each vendor is then told which result is theirs and therefore
the individual vendor can decide whether to publish their own
results independently.
The
results of all of the vendors are shown in the following set of
"Gateway Pie" diagrams:
Summary:
AudioCodes
has worked long and hard to distinguish itself with excellent
voice quality and with the ETSI test results. After we released
our results officially, we can now confirm what customers and
colleagues have been telling us for some time. To quote from AudioCodes'
press release of January 24, 2005:
"Voice
quality is essential for VoIP to continue expanding and being
deployed in the most demanding networks. Achieving the highest
levels of voice quality requires technical expertise, commitment
to quality, and also a critical mass of field deployments in many
diverse conditions, networks and regions," said Jeremy Duke, President
and CEO of Synergy Research. "In the ETSI test, AudioCodes proved
it had the skill, quality, commitment and VoIP field-proven experience
to achieve the leading voice quality offering. AudioCodes customers
and the industry will benefit from this achievement."
|
