Telos Zephyr Xstream User Manual page 135

probably the most widely used system. In our view, this technology is acceptable for mono voice
where high fidelity is not necessary. It is good also for cueing and intercom channels.
We have included G.722 in ZephyrXstream because:
•
•
This method was invented in the late 70s and adopted as a standard in 1984 by the CCITT, the
Consultive Committee for International Telephony and Telegraphy, a division of the United
Nations. The technique used is Sub‐Band ADPCM (Adaptive Delta Pulse Code Modulation),
which achieves data reduction by transmitting only the difference between successive samples.
G.722 does this in two audio frequency sub‐bands: 50‐4kHz and 4kHz‐7kHz.
t
DEEP TECH NOTE!
Only two bits are allocated per sample for audio frequencies above 4 kHz – sufficient for
conveying the sibilance in voice signals, but not very good for intricate musical sounds. Also,
the "predictor model" used to determine the step size in the adaptive function is designed
only for speech. This is why music transmitted via G.722 has a distinct 'fuzzy' quality.
G.722 has a frequency response extending to 7.5 kHz with fairly poor fidelity. Unless there is no
alternative, it should not be used for music.
G.722 uses a procedure called "statistical recovery timing" or "statistical framing" to lock the
decoder to the data stream. (We use the procedures specified in ANSI standard T1.306‐1989.)
This process usually happens instantaneously, but can take up to 30 seconds.
Other strange effects may be observed. Tones and noises may be present before locking occurs,
and some continuous audio tones may cause momentary unlocking. Please note this is inherent
in G.722's statistical framing and is not an implementation problem with Zephyr Xstream.
w
IMPORTANT!
The locking can be sensitive to audio present on the G.722 path, as it relies on properties of
the audio itself. Some audio material and tones can prevent lock from ever happening. Low
level noise is the most reliable signal for locking, and undistorted voice is usually OK. The
most common problems are with sine tones and distorted voice or music signals, in which
case turning off the audio signal — or lowering it 12dB or so — for a few seconds will
generally let the system relock. In very rare cases, it may be necessary to disconnect and
redial.
In cases such as this, the AUTO Receive mode will not be capable of determining the correct
settings and you may need to set the Receive setting to G.722.
USER'S MANUAL
It had been the most popular coding method early on, so there are
many of these codecs in use. Because it is a standard, codecs from
various manufacturers have a good probability of being able to
interwork with one‐another. (We've tested with many units and
have found no problems so far.)
G.722 has the lowest delay of all popular coding methods.
Section 6: AUDIO CODING REFERENCE 123