Digital voice is one of several traffic types supported by WAN switches. This chapter describes the technical concepts associated with digital voice traffic, including the following topics: 

• Analog-to-digital conversion
• Bandwidth optimization using ADPCM, CELP, and silence suppression
• Use and types of signaling including dialing formats
• Echo 

Voice Services in the WAN 

The telephone calls that you make from your home most likely use the public telephone network. You are provided an analog circuit between the local exchange and your house by the telephone company (also called telco, or PTT). When you dial a phone number the switch at the local exchange interprets the dialed digits and switches your call to the intended destination. 

The transmission facilities between switches carry digital voice signals multiplexed onto narrowband or wideband circuits. It is this part of the voice network that can be supported by a WAN. Most WAN voice applications are for private networks used for voice communications between sites of a private company. In this case, private branch exchanges (PBXs) replace the telco switches. 

The following graphic shows a phone call routed through a public or private network. 
 
The voice network could use either circuit or packet switching technologies. Packet switching networks include any of the following types: 

• Frame Relay
• ATM or other cell-based encapsulation
• IP
• X.25

The human voice is a low-frequency analog signal with a frequency range below 4000 cycles per second (Hz). Voice frequency (VF) circuits can carry a variety of payloads including data that uses voice-grade modems. 

The coder-decoder (CODEC) inside a channel bank or PBX converts the incoming analog signals to digital using the Pulse Code Modulation (PCM) standard. Eight-bit PCM samples digitally represent the analog signal for transport across the WAN. 

Multiplexing then combines multiple digital voice signals onto channels. Each channel is given a timeslot and the voice conversations all go out over a single T1, E1, or J1 line. The following graphic shows the function of  a digital PBX. 
 

The voice channel (called a DS0 in T1 systems) is the 64-kbps building block of time-division multiplexed (TDM) narrowband transmission facilities used to carry voice traffic. The first process in creating a digital voice channel from an analog voice signal is sampling. 

Sampling is the process of looking at a wave form at designated intervals and measuring the amplitude or height of the wave. The sampling rate determines the quality of the conversion. More frequent sampling results in more information stored about the signal and ultimately a better reconstruction of the signal at the receiving end. 

Basic communication theory requires that a minimum sampling rate of twice the frequency of the signal to be sampled will result in an accurate representation of the original signal. Because the human voice does not exceed 4000 Hz, a sampling rate of 8000 samples per second, or once every 125 microseconds, is used. 

The conversion process begins by analyzing each sample and converting it into an 8-bit word also called an octet. If there are 8 bits per sample and 8000 samples per second, then the product is 64,000 bits per second (64 kbps)—the digital voice channel. 

The following graphic shows a summary of the PCM process converting an analog signal to a 64-kbps digital signal. 
 

Companding is the part of the PCM process that determines the digital bits used to represent the voice signal. With companding, the analog signal values are logically rounded to discrete scale-step values on a nonlinear scale. The process is reversed at the receiving terminal using the same nonlinear scale. In other words, the amplitude measurement of the analog wave is converted into a digital code representing the nearest discrete value and then back again into an analog signal. 

Companding consists of a matched compressor at the sending end and an expander at the receiving end. The purpose is to apply greater emphasis to low-level speech components rather than the high-level components. The trade-off is lower signal-to-noise ratios for high-amplitude signals. The two companding laws are: 

• M-law (µ-law or mu-law) companding—The standard for North America and Japan, which uses a logarithmic scale.
• A-law—The standard for Europe, Asia, South America, Australasia, and elsewhere that uses a combination of linear at the lower levels and logarithmic at the higher input range.

An important point is that the two companding methods are incompatible. If a voice signal is coded into PCM samples using M-law, then it cannot be decoded using A-law. In fact, the resulting signal would be unintelligible. Many international voice systems must convert the digital signal from one companding method to the other. 

Multiplexing is the process of combining multiple voice channels onto one line. Each channel is given a timeslot, and the channel inputs all go out in two primary counts. 

• A single T1 frame uses 24 timeslots.
• A single E1 or J1 uses 30 timeslots for voice channel inputs, with two additional timeslots for framing and signaling. 

A detailed description of TDM is included in the "Narrowband Transmission" chapter of this guide. 

In many countries the cost of a transmission line, even one as low as 64 kbps, is very expensive. The goal is to maintain good voice quality while providing the bandwidth savings. To make better use of the bandwidth available for VF circuits, the following methods can be implemented: 

• Adaptive Differential Pulse Code Modulation (ADPCM)
• Code-Excited Linear Predictive (CELP) coding
• Silence suppression

ADPCM encodes only the difference between one encoded voice sample and the following sample, not the actual value of the sample amplitude. Because the range of values for the difference is much smaller than the actual value of the sample, it requires fewer bits for encoding this difference. In general, the lower the data rate, the lower the voice quality. In many countries ADPCM-compressed voice signals are not considered toll-quality services. 

ADPCM is defined in the ITU-T G.726 recommendation. The following compressed rates are available: 

• 40 kbps using 5-bit samples
• 32 kbps using 4-bit samples
• 24 kbps using 3-bit samples
• 16 kbps using 2-bit samples

 
With ADPCM, a formula is applied to the previous outgoing sample that predicts the range of values for the next sample. The transmitter encodes the difference in level between the previous sample and the present sample, on a scale set by the prediction, and transmits this information to the far-end receiver. Because the receiver has the same data to work with as the transmitter, it is able to reconstruct the PCM data input to the transmitter. 

ADPCM uses neither A-law nor M-law and can be used as a common coding method between unlike systems. 

High-speed modem and fax signals are distorted by ADPCM to the extent that operation above 4800 baud generally will not work reliably on ADPCM connections. A data-optimized 32-kbps ADPCM conversion exists to support modem and fax signals in some systems. 

At lower compression rates, ADPCM-compressed voice signals suffer from decreased quality. In an effort to increase bandwidth savings without decreasing voice quality, many voice systems have introduced CELP voice compression methods. In general, a CELP-compressed voice signal is much better than an ADPCM-compressed voice signal at the same or even higher data rate. 

CELP, unlike PCM and ADPCM, takes advantage of the known characteristics of human speech. A voice wave is not a random analog signal, and has characteristics that can be mathematically predicted. By modeling the speech wave based on known factors, the information that is required to recreate the signal is reduced. 

There are a wide range of CELP compression methods, some of which are standard (for example, ITU-T G.729 and G.729a CSACELP), and others that are proprietary. Regardless of the specifics, CELP methods are similar in how they encode and recreate voice signals. 

CELP compression techniques can reduce a 64-kbps voice signal down to 16 or even 8 kbps. 

Silence suppression, in conjunction with PCM, ADPCM, or CELP voice coding, can reduce the required bandwidth even more. Silence suppression is a method to reduce bandwidth by eliminating transmission during silent periods in the voice signal. The following graphic shows how silence suppression eliminates bandwidth. 
 
During a typical conversation only one person speaks at a time. During this time, the second person is silent. In the most extreme case, the PCM samples sent from the silent end would be all zeros. More likely, the signal would contain some low level of noise due to background sounds and transmission noise. Regardless, the information coming from the silent end is negligible and need not utilize network bandwidth. 

You can assume that in most cases any one person is speaking 50 percent of the time. In fact, the average utilization is less (as low as 38 percent), accommodating for gaps in speech. Of course, different languages will produce different results, but in all cases there are times when bandwidth is wasted. 

A silence suppression algorithm must examine the voice signal and determine if the signal represents speech or noise. Because human speech has defined characteristics such as power levels and frequencies, the signal is compared to these characteristics. If the signal is determined to be noise, then transmission is halted until speech is detected. 

Like the other compression methods discussed in this chapter, silence suppression provides a trade-off between bandwidth usage and voice quality. If silence suppression is too aggressive, the voice may sound choppy or the ends of words may even be cut off. You can expect in most systems up to 60 percent bandwidth savings by using silence suppression. 

Signaling is the process of transmitting status and control information between network elements. In a voice network, signaling is most often used to set up a call or a connection. Both analog and digital voice interfaces have defined signaling methods. Signaling occurs between either a subscriber and a switch or between switches. 

Signaling from the subscriber to the switch occurs whether the subscriber is calling from home or through the company's PBX. This signaling includes on-hook and off-hook states and dialing. In most cases, loop-start signaling is used on analog subscriber lines. An on-hook condition is indicated with an open loop (no current); an off-hook condition is indicated with a closed loop (current flowing). 

Interswitch signaling occurs between the local exchange carrier (LEC) and interexchange carrier (IXC) in a public network to signal each other to complete a call. Interswitch signaling is also used between PBXs attached to a private or public network. Interswitch traffic is often digital, and therefore the signaling is also digital. Digital signaling is either sent with the voice traffic or on a separate, dedicated signaling channel. If the interface between two switches is analog, E&M signaling is the most common method used. 

It is the interswitch signaling that must be interpreted or passed transparently through the WAN. The WAN is rarely involved with subsciber-to-switch signaling. 

Interswitch signaling uses one of the following two methods: 

• Channel Associated Signaling (CAS)—CAS uses a well-known location for the signaling bits for each voice channel, which is sent on the same path with the voice traffic. CAS allows for interpretation of signaling bits by all network devices responsible for switching the call. The number of signaling states supported by CAS is limited to the number of bits used.
• Common Channel Signaling (CCS)—CCS uses an independent signaling channel that carries signaling information for the voice channels as needed. CCS allows for a wide range of signaling states. CCS is often more complex than CAS, and often network devices are unable to interpret the signaling states of the voice channels. 

The following table lists the key distinctions between CAS and CCS. 

Refer to the "Narrowband Transmission" chapter of this guide for more information on T1, E1, and J1 signaling. 

Regardless of the type of circuit line (T1, E1, or J1), most voice switches recognize the following two types of dialing formats: 

• Dual-tone multifrequency (DTMF)—Where dialed digits are sent inband to the switch from the telephone. Each number on a telephone key pad produces two tones that in combination uniquely identify the digit. A voice switch, after receiving an off-hook condition, listens for DTMF tones and either interprets them or passes them through the network.
• Pulse—Signaling bits are used to convey dialed digits. Usually, a transition from off-hook to on-hook and back again represents a single pulse. Each dialed digit created a series of pulses equal in number to the digit. For example, a dialed 4 would create four pulses. The signaling bits will change every 40 to 60 milliseconds (ms) within a dialed digit, with pauses between digits of greater than 100 ms. A voice switch must differentiate between a pulse-dialed digit and a series of independent signaling transitions based on the time between the pulses.

In a voice network, echo is a common problem that occurs because of impedance mismatches between 2-wire and 4-wire systems. A telephone is attached to a PBX or switch using a 2 wires but the wiring inside the PBX or switch is often 4 wires. The impedance mismatch causes a portion of the original signal to be reflected to the source as shown in the following graphic. 
 
For echo to be heard, the following elements must be present:

• A signal source
• A reflection point
• A delay between the original signal and the reflected signal
• An audible reflected signal

Echo is always present in a voice network, although it is not always a problem. Depending on the reflection delay and the strength of the reflection, the echo may be acceptable. The following graphic shows the dividing line between acceptable and unacceptable echo. 
 
The primary cause of unacceptable echo in a voice network is delay. If two PBXs were installed side by side with a cable between them, the delay would be so short that the echo would go unnoticed. But, if you were to move one of the PBXs to a remote site with a private or public network between them, the delay may increase enough to make the echo noticeable. A packet-switched network adds more delay than a circuit-switched network.

The following can cause delay in a voice network:

• Transmission—Propagation of signals across long distances
• Packetization—The time required to receive enough voice samples to create a packet
• Compression—Processing delay associated with compression algorithms
• Congestion—Queue delay in a packet-switched network
• Silence suppression—Silence is often determined over a period of time, which requires that voice samples are held for that length of time

To remove echo, echo cancellers should be installed into the system. An echo canceller is a digital signal processor that continuously monitors the digitized speech going coming from the network (receive direction) and compares it to the signal in the reverse direction. By comparing the two signals, the echo canceller can model the echo by determining the loss and the delay of the reflected signal. Based on these measurements, the echo canceller calculates the expected echo power level and subtracts it from the received signal.

Copyright Statement

I take no responsibility of the material accessible threw this site.None of the book is hosted on this site . I am only providing link to the books hosted on other site which i found by doing extensive search on google.I take no responsibility for the legality of these books.You can check the websites from where the link has takes for the term and conditions for downloading these books.If any one is suffering financial loss due to this then drop a mail to me I will remove the link of that book.