In telephony, the phenomenon known as “echo” happens when a person hears what he has just said a few milliseconds after speaking. This can be experienced by either one or both parties. This issue seems to affect VoIP users more often than not, and the effect can range from irritating to unacceptable.

This echo is essentially feedback of one’s own voice coming from the telephone network. This feedback usually helps the person on the telephone know that the network is picking up his voice. However, if there is latency from anywhere between the terminal and the local exchange, this feedback will come in some time after speaking, and this could cause confusion.

There are several causes of echo in an Asterisk installation. First to check is your PSTN card, particularly whether its loadzone is correctly set. For instance, if your PSTN card is set to FCC (standard for US), but you’re in a UK line, then you are likely to experience harsh echo. You will therefore have to switch to UK mode.

Second, the connection from your gateway to the local exchange, or the connection of your called party to his own local exchange might be improperly balanced. This way, some of the signals transmitted are reflected back, therefore causing echo. This can be caused, among other things, by wet or damaged telephone cables, use of untwisted telephone wire on either end, or bridge taps. Another reason may be the analog handset used on the local line. Also, the use of speakerphones by the called party, could be a reason for echo (and this is further amplified by latency).

Solutions

There are basically three ways of dealing with echo. First is by eliminating the echo at the source, particularly by ensuring that all hybrids (connections from your VoIP gateway to the local exchange) are balanced. This is only theoretical, though. In practice, even balanced hybrids are likely to experience some echo.

Echo suppression is a solution used in earlier intercontinental calls. This method turns off transmission whenever a person stops talking. This is not useful, though, when both parties are talking simultaneously.

Lastly, echo cancellation uses a mathematical approach to removing echo from the signal. This essentially analyzes the signal that is transmitted when a person speaks, and subtracts the exact negative from any incoming signal if it arrives after a pre-set number of milliseconds. Therefore, echo is removed.