Home > Articles > Networking > Voice/IP Communications

This chapter is from the book

Voice Connections

A voice connection is an end-to-end permanent virtual circuit (PVC) that originates and terminates on MGX VISM endpoints. The connection receives PCM voice samples and signaling and converts them into ATM cells using AAL2 (voice traffic and CAS signaling) or AAL5 (CCS signaling) and transports the cells to the remote endpoint. All voice connections are bidirectional, meaning that traffic flows in both directions.

A single AAL2 voice connection can carry multiple voice calls. This is called AAL2 multiplexing. The channel identifier (CID) differentiates voice traffic streams on the connection.

You can add two types of voice connections in an MGX network: feeder and local. Feeder connections go from the VISM to the PXM trunk and are transported through the ATM backbone network (BPX or other) to the destination MGX switch. Local connections go from one endpoint on an MGX switch to another endpoint on the same switch. Local connections can be between endpoints on the same or different VISM cards. Feeder and local connections perform the same functions.


All VISM connections terminate on endpoints. An endpoint is a channel (DS0 or timeslot) on a T1 or E1 line. An endpoint is defined by its endpoint number and the T1 or E1 line and channel number. You must create VISM endpoints before you can terminate connections on them.

Channel Identifiers

A CID identifies a specific voice traffic stream. You use it when multiplexing multiple voice calls across a single ATM connection. On the MGX switch, the CID also links the connection to the endpoint. When you create a CID, you identify the compression type and other voice processing characteristics, such as echo cancellation and dual-tone multifrequency (DTMF) transport. Figure 22-1 shows the VISM with endpoints, connection, and CIDs.

Figure 22-1 Channel Identifiers

Feeder Example

A voice feeder connection has three connection segments:

  • A master connection between the VISM endpoint and PXM1 trunk on one MGX switch.

  • A routing connection through the ATM backbone network. The routing connection should be rt-VBR for voice traffic and nrt-VBR for signaling traffic.

  • A master connection between the VISM and PXM1 trunk on the other MGX switch.

Figure 22-2 shows a voice feeder connection in an MGX network.

Local Example

An ATM local connection has two connection segments:

  • A slave connection on one VISM. The slave connection must be added first.

  • A master connection on the other VISM pointing to the slave connection.

Figure 22-2 Feeder Example

How PCM Samples Are Converted into ATM Cells

How are ATM cells created on Voice over AAL2 (VoAAL2) services on the VISM?

Figure 22-3 shows a high-level view of the traffic flow of voice traffic on the VISM.

Figure 22-3 Traffic Flow

Pulse Code Modulation (PCM) voice samples come in from the line and are processed by echo cancellation digital signal processors (DSPs) to eliminate any echo (if echo cancellation is enabled). Next, the PCM samples are processed by the compression DSPs. This is where the 8-bit PCM samples are converted into a variety of different samples (length and frequency), depending on the compression method in use. The compressed samples go to the segmentation and reassembly (SAR) processor to be loaded into ATM cells for transport to the cell bus and across the network.

AAL2 Segmentation

ITU-T Recommendation I.363.2 specifies the basic AAL2 structure for VoAAL2. Figure 22-4 shows three voice samples (from the same or different voice calls) and how they are made into an ATM cell.

Figure 22-4 Voice Sample Conversion to ATM Cell

Table 22-1 describes each stage of the AAL2 segmentation process.

Table 22-1 AAL2 Segmentation Process



Voice samples

The fixed-length voice samples come from the codec. In this example, each voice sample is 10 bytes long, but the length depends on the compression method used.

CPS (Common Part Sublayer) packet

A 3-byte CPS header is added to each voice sample. Because the voice samples could come from different traffic streams (channels or calls), the CPS header differentiates the channels with the CID.


The CPS-PDU includes a 1-byte start field at the beginning of the CPS packets. Padding is added to the data to make the CPS-PDU exactly 48 bytes long.

ATM cell

The 5-byte ATM header is added to the CPS-PDU (ATM cell payload), and the cell is then ready to transport through the network.

LI (Length Indicator)

The number of cells with a payload length that does not match the LI in the CPS packet header.

AAL2 Coding

Here are the coding or compression methods supported on the VISM:

  • G.711u—PCM with µ-law coding

  • G.711a—PCM with A-law coding

  • G.726—ADPCM

  • G.729—CS-CELP with 10-, 20-, and 30-millisecond (ms) cell times

The coding type you use affects the voice sample size and the sample and cell frequency. Ultimately, this determines how much bandwidth the voice traffic utilizes. Figure 22-5 shows each coding type and the ATM cell created from the voice samples, assuming that multiplexing (multiple voice streams on the same ATM connection) is not in use.

Figure 22-5 AAL2 Coding Types

  • + Share This
  • 🔖 Save To Your Account