Configuring Echo Cancellation
Echo cancellation is configured at the voice port level. It is enabled by default, and its characteristics are configurable. Echo cancellation commands are as follows:
- echo-cancel enable—Enables cancellation of voice that is sent out through the interface and received back on the same interface. Sound that is received back in this manner is perceived by the listener as echo. Echo cancellation keeps a certain-sized sample of the outbound voice and calculates what that same signal looks like when it returns as an echo. Echo cancellation then attenuates the inbound signal by that amount to cancel the echo signal. If you disable echo cancellation, it will cause the remote side of a connection to hear echo. Because echo cancellation is an invasive process that can minimally degrade voice quality, you should disable this command if it is not needed. There is no echo path for a four-wire E&RM interface. The echo canceller should be disabled for this interface type.
echo-cancel coverage—Adjusts the coverage size of the echo canceller. This command enables cancellation of voice that is sent out through the interface and received back on the same interface within the configured amount of time. If the local loop (the distance from the interface to the connected equipment that is producing the echo) is longer, the configured value of this command should be extended.
If you configure a longer value for this command, it takes the echo canceller longer to converge. In this case, the user might hear a slight echo when the connection is initially set up. If the configured value for this command is too short, the user might hear some echo for the duration of the call because the echo canceller is not canceling the longer-delay echoes. There is no echo or echo cancellation on the network side (for example, the non-POTS side of the connection).
non-linear—The function enabled by the non-linear command is also known as residual echo suppression. This command effectively creates a half-duplex voice path. If voice is present on the inbound path, then there is no signal on the outbound path. This command is associated with the echo canceller operation. The echo-cancel enable command must be enabled for the non-linear command to take effect. Use the non-linear command to shut off any signal if near-end speech is not detected.
Enabling the non-linear command normally improves performance. However, some users encounter truncation of consonants at the ends of sentences when this command is enabled. This occurs when one person is speaking and the other person starts to speak before the first person finishes. Because the nonlinear cancellation allows speech in one direction only, it must switch directions on the fly. This might clip the end of the sentence spoken by the first person or the beginning of the sentence spoken by the second person.
ITU standard G.164 defines the performance of echo suppressors, which are the predecessors of echo cancellation technology. G.164 also defines the disabling of echo suppressors in the presence of 2100-Hz tones that precede low-bit-rate modems.
ITU standard G.165 defines echo cancellation and provides a number of objective tests that ensure a minimum level of performance. These tests check convergence speed of the echo canceller, stability of the echo canceller filter, performance of the nonlinear processor, and a limited amount of double-talk testing. The signal used to perform these tests is white noise. Additionally, G.165 defines the disabling of echo cancellers in the presence of 2100-Hz signals with periodic phase reversals in order to support echo-canceling modem technology (for example, V.34), which does not work if line echo cancellation is performed in the connection.
ITU standard G.168 allows more rigorous testing and satisfies more testing requirements. White noise is replaced with a pseudo-speech signal for the convergence tests. Most echo cancellation algorithms use a least mean square algorithm to adapt the echo cancellation filter. This algorithm works best with random signals and slows down with more correlated signals such as speech. Use of the pseudo-speech signal in testing provides a more realistic portrayal of the echo canceller's performance in real use.
If you speak into your telephone and hear your own voice a short time later, you are experiencing talker echo. As you learned earlier in the "Echo" section, talker echo is caused by the remote telephony circuitry's two-wire to four-wire hybrid circuit. Enabling echo-cancellation on your voice port will eliminate the problem. Depending on the return time of the echoed voice, you can further adjust using the echo-cancel coverage command. Table 3-6 compares echo cancellation standards.
Table 3-6. Comparing Echo Cancellation Standards
|
G.165 EC |
G.168 EC |
|
|
Tail Coverage |
Up to 32 ms |
Up to 64 ms |
|
Minimum Echo Return Loss (ERL) |
Greater than or equal to -6 dB |
Configurable to greater than or equal to -0 dB, -3 dB, or -6 dB |
|
Echo Suppression |
Up to 10 seconds |
Not required due to faster convergence |
|
Minimum Cisco IOS Software Release |
12.2(11)T, 12.2(8)T5, 12.2(12), and higher |
12.2(13)T, 12.2(8)YN, 12.2(15)T, 12.3(4)T, 12.3(4)XD, and higher |