Telephone networks can experience two types of echo: acoustic echo and electrical echo.

Acoustic echo frequently occurs with speakerphones, when the received voice on the speaker

excites the microphone and travels back to the speaker. Electrical echo occurs when there is an

electrical inconsistency in the telephony circuits. This electrical inconsistency is called

impedance mismatch.

If the lines have a good impedance match, the hybrid is considered balanced, with little or no

reflected energy. However, if the hybrid is inadequately balanced, and a portion of the transmit

voice is reflected back toward the receive side, echo results.

Some form of echo is always present. However, echo can become a problem under the

following conditions:

The magnitude or loudness of the echo is high.

The delay time between when you speak and when you hear your voice reflected is

significant.

The listener hears the speaker twice.

The two components of echo are loudness and delay. Reducing either component reduces

overall echo. When a user experiences delay, the conversation can get choppy, and the words of

the participants sometimes overlap.

There are two ways to solve an echo problem in your telephone network:

Echo suppression

Echo cancellation

The echo suppressor works by transmitting speech in the forward direction and prohibiting

audio in the return direction. The echo suppressor essentially breaks the return transmission

path. This solution works sufficiently for voice transmission. However, for full-duplex modem

connections, the action of the echo suppressor prevents communication. Therefore, when

modems handshake, the answering modem returns a tone of 2025 Hz to the calling modem,

which serves to disable the echo suppressors along the transmission path.

Echo suppression has shortcomings in addressing certain echo conflict situations. Echo

cancellation is a more sophisticated method of eliminating echo.

Rather than breaking or attenuating the return path (as in echo suppression), echo cancellation

uses a special circuit to build a mathematical model of the transmitted speech pattern and

subtract it from the return path.

You must configure the voice-port of your gateway:

echo-cancel enable and echo-cancel coverage are used to activate echo cancellation;

non-linear is used to activate echo suppression;

Check POTS/PSTN impedence too.

On LAN use only switch, not hub. Check the latency.