Echo Cancellation and VoIP

There are a number of impediments that can affect the overall quality of a voice over IP (VoIP) call, including echo, latency, delay, and others. Each of these can degrade the perceived quality of a voice call to various degrees ranging from being slightly annoying to downright unacceptable.

This problem is not unique to VoIP, however through years of trial and error, and simple experience, the voice quality issues facing the PSTN (define - news -alerts) have been addressed through network optimization and better network management. As IP networks converge with the PSTN, these voice quality degradations need to be dealt with.

So, what is echo anyway? Dictionary.com offers this standard definition: Repetition of a sound by reflection of sound waves from a surface. In telecommunications parlance, and as defined by Tehranis IP Telephony dictionary, echo is ...a type of transmission impairment in which a signal is reflected back to the originating source.

Simply put, echo is the sound of the speaking partys voice returning to their ear via the handset or headset speaker. Lest we think all echo is bad, it is common practice to allow a little echo (with delay in the range of 25 ms) so that the speaker can hear their own voice. Studies have shown that this is a reassuring sound. But aside from some minor allowances, echo can wreak havoc on a VoIP conversation by degrading the conversation to unacceptable levels.



According to the VoIP Wiki at voip-info.org, there are three potential sources of echo in a standard system. From the Callers perspective, these are:

Within the callers telephone; a certain amount of the signal from the microphone is fed straight back to the earpiece.

At the hybrid at the callees end. An improperly balanced hybrid wont correctly filter out the entire transmitted signal, and will reflect some of it back down the other half of the trunk. (A Hybrid is a device that combines transmit and receive signals from two pairs of lines to one single pair).

At the handset at the callees end. If the callee isnt holding the handset against his head, or if the handset is poorly designed, its possible for the microphone to pick up the sounds coming from the earpiece, and reflect the audio back down the line.

Likewise, according to the Wiki, there are three ways of dealing with echo:

Elimination at the source. This would mean going out and making sure all the hybrids in a network are tuned and properly balanced.

Echo suppression. A simple voice activated switch which turns off transmission from the speaker to the listener whenever the speaker is silent. Negative side effects of this approach include voice clipping and choppiness, which may result in voice degradation, the likes of which the solution was meant to avoid in the first place.

Echo cancellation. A mathematical approach to subtract exactly the right portion of the transmitted signal from the return signal to eliminate the echo.

This last approach is often the most feasible and practical.

Echo Cancellation
Echo cancellation is the process of extracting an original transmitted signal from the received signal that contains one or more delayed signals (copies of the original signal). Echoes may be created in a baseband or broadband signal. When echoes occur on an audio baseband signal, it is usually through acoustic feedback, where some of the audio signal transferring from a speaker into a microphone. When echoes occur on a broadband signal, it is usually the result of the same signal (such as a radio signal) that travels on different paths to reach its destination. In either case, echoed signals cause distortion and may be removed by performing via advanced signal analysis and filtering.

In order to properly determine the signal echo, it is necessary to measure the echo. Among the characteristics that need to be gauged is something referred to as echo return loss or ERL. ERL is the amount of attenuation or loss of a signal that is returned to the originating speaker. Essentially, this is how much lower the echo audio sounds compared to what was originally sent. ERL is critical to effectively canceling echo, because often an echo canceller cannot deal with echo that has not been attenuated to at least some degree.

Companies like Sangoma are actively studying the various impediments that can degrade voice quality. Through the use of dedicated processing power and finely tuned telephony cards, and the efficient management of system resources, Sangoma is able to design systems that deliver high-quality reliable voice. IT

Greg Galitzine is editorial director of Internet Telephony magazine and the newly launched IMS Magazine.