Frequently Asked Question
How to deal with echo and buffer empty alarms.
Last Updated 6 years ago
uffer empty and echo are caused by latency issues. In a pure TDM network, transmission latency is measured in micro-seconds (or even nano-seconds) where as in a packet network, we typically measure latency in milli-seconds.
Many voice communication systems that run over E1 or T1 have “echo cancellation” built-in. This may be automatic or it may require manual configuration. There are settings that can be optimized to provide both stable transmissions and to minimize 'echo'.
When deploying IPM in a network, the two most important tuning parameters are the Jitter Buffer and Packet Size settings.
On a fast network that has very low latency, It is possible to optimize the IPM with small packet sizes and smaller jitter buffer. In the table attached, if we set a packet size of 178 Bytes and set the Jitter Buffer to 0 (11ms), the actual buffer will be 1.3ms per the table. This is the absolute minimum jitter buffer size that can be realized with the IPM. If the network consistently has less than 1ms latency (and or if the CESoETH traffic can be groomed with high priority) then, this minimal setting could be used and the echo may no longer exist. For this example, the settings on both IPM need to match and would be:
pktsize x 2
jitterdb x 0
Where “x” is the channel number in the IPM.
When doing this ‘tuning’, we suggest that if this minimum setting results in buffer empty alarms, try only increasing the jitter buffer but leave the packet size the smallest. So, per the table, if we set the buffer to 1 (23ms), the actual buffer with 178 byte packets would be 2.7ms per the table.
The IPM has a built-in ping command. Ping will give a latency figure for “round-trip”. The jitter buffer only needs to cover half of that, i.e., the one-way latency plus some overhead.
Many voice communication systems that run over E1 or T1 have “echo cancellation” built-in. This may be automatic or it may require manual configuration. There are settings that can be optimized to provide both stable transmissions and to minimize 'echo'.
When deploying IPM in a network, the two most important tuning parameters are the Jitter Buffer and Packet Size settings.
- Packet Size: It is always best to use small packets over larger packets in a full duplex network. They are smaller and can be processed quicker. The loss of a small packet (which is then filled with random data) has less impact on the voice quality than on the loss of a larger packet. If the connecting network is half-duplex, such as in a WiFi or Microwave (wireless) connection, then it is better to use large packets.
- Jitter Buffer: As CESoETH (Circuit Emulation Services over Ethernet) packets are received at the IPM, a buffer is employed so that any latency changes (jitter) in the network can be compensated for, allowing the packets to be constantly processed. If the buffer becomes empty, due to longer latency in the network or because of traffic congestion, the IPM alarm will display JBUFEMP (jitter buffer empty).
- Buffer and packet size relationship: The jitter buffer settings (0~7) value are related to the packet size. The attached table shows this relationship.
On a fast network that has very low latency, It is possible to optimize the IPM with small packet sizes and smaller jitter buffer. In the table attached, if we set a packet size of 178 Bytes and set the Jitter Buffer to 0 (11ms), the actual buffer will be 1.3ms per the table. This is the absolute minimum jitter buffer size that can be realized with the IPM. If the network consistently has less than 1ms latency (and or if the CESoETH traffic can be groomed with high priority) then, this minimal setting could be used and the echo may no longer exist. For this example, the settings on both IPM need to match and would be:
pktsize x 2
jitterdb x 0
Where “x” is the channel number in the IPM.
When doing this ‘tuning’, we suggest that if this minimum setting results in buffer empty alarms, try only increasing the jitter buffer but leave the packet size the smallest. So, per the table, if we set the buffer to 1 (23ms), the actual buffer with 178 byte packets would be 2.7ms per the table.
The IPM has a built-in ping command. Ping will give a latency figure for “round-trip”. The jitter buffer only needs to cover half of that, i.e., the one-way latency plus some overhead.