Voice Compression and Silence Suppression

Buy Shoes
 Voice Compression and Silence Suppression  





We will look at several schemes for performing voice compression over communications links. Voice compression over new networks, such as asynchronous transfer mode (ATM), is compared with older time division multiplexing (TDM) links, such as conventional T1. Although some limited compression has been possible on a TDM link, these schemes were rigid and restricted traffic types. Silence suppression is a method of saving bandwidth.


New telecommunications service platforms such as ATM networks test, moment by moment, whether or not active speech is occurring. Traffic is only generated during active speech. Although TDM could suppress silence, no bandwidth saving occurred because the free bandwidth could not be reassigned.

During a voice conversation, silence represents more than 50% of all voice traffic, so the importance of preventing this wasteful bandwidth consumption is evident. ATM can deliver high-quality voice using much less bandwidth. Active silence suppression schemes deliver a full 2:1 compression over the voice encoding techniques offered on TDM platforms.

This return of bandwidth to the network allows companies to reduce operating costs or add new services to their network. The full spectrum of multimedia services can be mixed at the multimedia workstation to and throughout the network. This important advance will make it possible to merge all traffic types onto one common network architecture.


Voice compression technology is outstanding in its high-connection capacity, but several complications are usually encountered when implementing compression within a network. The remainder of this chapter describes current compression schemes and some that are soon to be available. They are evaluated in terms of:

  Their quality and ability to suppress bandwidth consumption during periods of silence.
  The use of faxes and high-speed modems.

The question is not whether voice compression should be used, but how much to use and how much it will cost or save the company.

The economic advantages are great, and voice quality is maintained. Currently, a T1 trunk between the U.K. and the U.S. can cost almost $70,000 a month. Which is more economical—24, 96, or even 192 voice connections on that link, or 96 voice ports and a 768K-bps LAN connection, which would also be possible?


With the advent of pulse code modulation (PCM) little more than a decade ago, voice telecommunications made a major technological advance. Today, PCM voice quality is the standard by which all other compression techniques are tested.

PCM can support voice, data, fax, and proprietary forms of interdevice communication (i.e., over a signaling channel).

How PCM Works

With PCM, the delay imposed by sample creation is low and is governed by the 8-KHz sample rate (one sample every 125 microseconds). It functions well, but telecommunications users may wonder how the PCM scheme was designed. Why 8-bit samples? Why an 8-KHz sampling? Why “mu” law and not “A” law encoding?

Analog voice communication by telephone was designed as a bandpass filter, allowing a frequency range from 300 Hz to 3,000 Hz. A basic law of 8-bit PCM states that the sample rate must be at least two times the highest frequency allowed—this means the minimum allowable sample rate is 6 KHz.

More Capacity without More Wires.  The engineers may have chosen 8 KHz to ensure the delivery of excellent quality voice. Therefore 8,000 times per second, an 8-bit word is created that measures the voltage of the current analog signal, with this conversion taking place in a PBX or telco central office. The measurements are sent to the distant end of the connection, where the PBX or central office switch reconstructs the voice for delivery to the called party.

The reason for creating PCM had nothing to do with digital technology or computers. It had only to do with wire—specifically, outside plant wire that connects the central office to anything outside the central office building. It took up to eight wires to support each interoffice tie line voice conversation. Copper routes around and between metropolitan areas were at their capacity, so the telephone companies set out to create capacity within the existing infrastructure but without having to install any more wiring between central offices. Analog carriers like “N” carrier helped, but they demanded high levels of routine maintenance.

Ultimately, T1 services were introduced that could serve 24 voice conversations on just four wires. The benefit to the telephone companies was enormous, because they were able to support digitally 24 voice trunks in place of every two analog circuits they replaced. PCM is more than an 8-bit voice-sampling technology; by itself it provided copper compression.

Changing Economics.  PCM created the market for digital PBXs, which in turn created the market for private T1s and E1s. Channel banks, T1/E1 multiplexers, and other such equipment then enabled data to be transported on those voice lines. Voice typically consumed 70% of the network’s capacity while paying for it all, hence the data could ride for free.

In today’s private networks, the trend has reversed itself and voice now rides for free, thanks to the growth of data applications. On a large wide area network (WAN), however, pure PCM will almost never be used.

Although it is a benchmark for voice quality, to assign 64K bps of capacity to one voice conversation wastes bandwidth, which is not acceptable on international WAN connections where costs are high and voice quality is usually sacrificed through the use of high-compression rates. In Mexico, an E1 from Mexico City to Monterey currently costs approximately $22,000 per month. This 450-mile circuit is equivalent to a T1 between Washington DC and Detroit, which typically costs less than $6,500 per month. In Brazil a 150-mile E1 from Sao Paulo to Rio de Janeiro currently costs $42,000 per month. A T1 from New York City to Baltimore costs less than $3,200 per month.





Copyright Manjor Inc.