In recent months, both Digium and Sangoma have released T1/E1 Interface cards for use with Asterisk Open Source PBX that provide hardware Echo Cancellation.

Echo can negatively impact QOS (Quality of Service) on a VOIP call and lead to an unsatisfactory user experience. There are two sources of echo on voice communication networks. The first, and most common cause of echo is impedance mismatches anywhere in the circuit-switched phone network. When telco cabling is spliced or terminated in connectors, or in the conversion of 4-wire phone circuits to 2 wires, a discontinuity occurs that causes an impedance mismatch on the phone circuit. The greater the extent of impedance mismatch, the more leakage of transmit audio on the receive side of the line when it terminates on a 2-wire phone circuit or device. The main cause of this impedance mismatch is the 2 to 4 wire hybrid that converts a 4-wire audio path to a local 2-wire loop. This tip-and-ring loop is the most-used type of telephone connection made to subscribers from your telephone company central office, and directly drives common analog telephone devices.

Acoustic coupling between the microphone and speaker of a telephone device leads to Acoustic echo, the second most common cause of echo on voice communication networks. Received caller audio can often “leak” to the microphone in cheaper speakerphones and hands-free cellular phones, because of sound pressure from a speaker or earpiece device

This inherent echo led to the development of Echo Cancellers, in order to improve the quality of voice communications. Echo cancellers are very complex digital signal processing devices, and the algorithms that drive them are produced by a handful of technically proficient companies. Naturally, as telecommunications evolves from the legacy PSTN network to the VOIP network, echo cancellation is still a very necessary component for ensuring QOS.


Digium offers hardware echo cancellation on their TE406P and TE411P telephony boards, as well as their new TDM2400 series full length analog PCI cards. Digium’s hardware echo cancellation provides 64ms across 32 channels; however, when it scales over 32 channels it is reduce to 16ms per channel across all channels.


Sangoma offers hardware echo cancellation on their A104D Quad T1 card, as well as on their as yet unreleased series of analog telephony boards. Unlike Digium where available ms of echo cancellation decreases as active channel density increases, Sangoma hardware echo cancellation provides a full 128ms of echo cancellation on all channels, regardless of density.

By providing hardware echo cancellation which extends to a full 128ms on all channels at full load, Sangoma provides a more robust compensation for echo which is more effective under extreme conditions. The echo tail length represents how long the effect of an echo extends after the time of the echo source. If a ping is transmitted at time zero, 16ms (16 ms = 128 taps at 8 samples per millisecond) of tail will deal with any echo from that ping that occurs within 16 ms. The echo itself dies away in about 8ms, so as long as the delay between the ping and the start of the echo is less than 8ms, then the echo canceller will work well. The problem arises when the signals get delayed. For instance traveling through one Telco switch adds about 5ms of two-way delay, so it is easy to see how a bit of extra switching could delay the echo right out of an echo canceller’s
range.


The above image illustrates the echo from a “ping” on a real line where there are
switching delays. In cases of extreme echo as pictured above, normal 128 tap echo cancellation would be ineffective at cancelling the echo present on the line.

Software echo cancellation works quite well for electrical echoes that are within the 16ms/128 tap range. Where hardware echo cancellation really has benefit is where the delays are longer. The Sangoma echo canceller is a carrier-grade device with 1024 taps (128ms) of tail to handle the most severe echo problems.

Discussion

2 Comments
Comment

Your email address will not be published.

  1. Hi,

    Would appreciate if you can send me the full length of the write up above. It was obviously cut since I remember seeing it before directly comparing Digium and Sangoma.

    I guess the original write up was a bit negative for Digium so that’w why they had to have the article modified?

    Thanks.

    Reply