Convergence? Try voice over frame

A frame relay frame may contain multiple subframes from one conversation, single frames from multiple logical channels, or both. A subframe can be any digital stream - voice or data.

Consolidation of traffic on a single Data Link Connector Identifier - a frame relay virtual circuit number - eliminates the need for additional DLCIs. This simplifies the job of the backbone: the carrier configures only one frame relay virtual circuit, and the number of frames to process is cut dramatically. That's important because most switches reach their packet/second limit first, before bits/second top out.

On the downside, building larger frames from multiple subframes of one voice stream increases accumulation time and round-trip delay, potentially impairing voice quality. An alternative strategy is to send frames at short intervals, filling them with any subframes that are ready. This approach minimizes jitter at the source.

To transport voice, it must be digitally encoded. Smaller frame relay access devices (FRAD) and routers typically offer analog interfaces to phones and PBXs. The analog-to-digital conversion is done internally, by a coder/decoder. A CODEC first converts voice to pulse code modulation (PCM), the standard 64K bit/sec voice format of the public switched telephone network.

The most common frame relay service is on a 56Kbit/sec access line. That's why voice FRADs usually compress the PCM bit stream. The compression algorithm, which runs on a digital signal processor (DSP) chip, largely determines the perceived quality of voice transmissions. Devices implementing FRF.11 call out the CELP (code-excited linear prediction) method defined by ITU recommendation G.729, which produces good voice quality.

Each voice path produces a separate stream of subframes. Each subframe is marked with its own channel identifier to indicate whether it is: part of a voice utterance; the last packet before the sender stops transmitting (to minimize bandwidth while the speaker pauses); fax; or signaling information.

Signaling is done in two ways: In dual-tone multifrequency (DTMF) transfers, the compression processor detects tone pairs and converts them to codes that indicate which button is pushed and when the tone starts and stops. With the second method, dial pulses, frequent sampling of a two-wire analog input detects switch closings/openings that define dial pulses. This signal mode captures a very precise picture of what the sender does. It also conveys simple on-hook/off-hook status.

During a signaling event, such as dialing, the dialing end originates signaling subframes continuously, replacing voice with DTMF or dial-pulse packets. To protect against loss of frames, each interval of signaling information is sent in three consecutive subframes with overlapping content. Sending only voice or signaling (not both) prevents misinterpretation.

Fax transmission occurs in the same subframe format. Rather than apply voice coding to the fax modem noise, a voice FRAD typically swaps its DSP software from voice compression to fax modem. The data going into the packets is the original data generated by the fax scanner before passing through the sending fax's modem. With 9600 bit/sec fax machines operating at full speed over frame relay, just over 10K bit/sec of bandwidth is used.

Data subframes contain segments of whatever protocol the terminal equipment sends to the FRAD. Subframe headers indicate the start, continuation and end of a data block. When mixing voice and data, the size of data subframes can be tuned to minimize jitter in voice packets when they have to wait behind data frames. Most FRF.11 voice implementations also prioritize voice over data.

Feedback

Tell us your thoughts on this article or the issues raised in it. We'll cc: the author and editors on all comments.

Comments:

Name:
E-mail address:

Can we post your comments in an online forum on the topic?
Yes No

What did you think of this article?
Very useful Somewhat useful Not at all useful

Would you want to see:
More articles on this topic
Fewer articles on this topic

Thank you! When you click Submit, you'll be taken back to this article.