A Digital Echo Canceller is a four-port device that supports bidirectional voice traffic between the two ends of a connection. The two ends of the connection are referred to as the far end and the near end. Voice signals from the far end are carried to the near end on the receive path. Voice signals from the near end are carried to the far end on the send path. Note that the signal pathways are labeled from the point of view of the near end. A typical configuration is shown in the figure below.
The picture above indicates conversation between two distant persons. The signal traverses the receive path into the echo canceller's Rin port, which monitors the signal and forwards it along through its Rout port.
The signal continues unaltered to the hybrid, a device that interfaces the PSTN's four-wire circuit to the two-wire subscriber line. This interface is not perfect, resulting in a reflection of a portion of the signal energy being reflected back into the four-wire circuit. This reflection manifests itself as echo, a delayed, attenuated and dispersed version of the original speech signal, which traverses the send path back to the original speaker. In some cases where, the echo is significantly delayed, it can be very annoying to the far end speaker, to the point of interfering with the conversation.
The function of the echo canceller is to detect and remove the echo. It does this by estimating the transfer function of the hybrid that produces the echo, which it accomplishes by comparing the echo to the original signal. The process of generating the estimate is normally iterative in nature -more the signal is seen, more refined is the estimate.
The estimated transfer function is referred to as the echo path model. The process of refining the echo path model as more and more signal is seen is referred to as adaptation. When adaptation has proceeded to the point that a reasonably good estimate has been calculated, the echo canceller is said to have converged. The speed with which the echo canceller converges, and how accurate the echo path model is, are crucial measures of the echo canceller's performance in a network.
Once a reasonable echo path model has been calculated, the echo canceller can predict the echo that will result from the far end signal that it is monitoring. It can thus remove that echo by subtracting it from the signal on the send path (that is, by subtracting the estimated echo from the actual echo).
- Allows real-time and offline processing.
- Interfaces directly with A-Law or m-Law encoded signals.
- G.168-2000 compliant, tests 2A, 2B, 2C, 3A, 3B, 3C, 4, 5, 6, 7.
- 16, 32, 64 or 128 ms tail length; programmable tail offset.
- Comfort noise generator with adaptation to background noise level.
- Echo path visibility.
- Continuous reporting of echo path delay, ERL, and dispersion.
- Event logging.