Side_1_360

too slow to cope with the time sensitiveness of

real time voice transmissions. This forces VoIP

to incorporate other quicker solutions to solve

this problem. The implementations used are

either to use replay of the last packet, interpola-

tion or introducing redundant information in the

packet stream. Due to the redundant structure of

our language, these solutions could handle

packet losses of at least 5–7 % without major

reductions of the perceived quality.

Other ways of improving the speech is to use the

silent periods in a call, which could amount to

more than half of the time. The silent periods

include real silence, pauses made and breaths.

No packets are sent with information on

“silence”, this interruption of information will

be interpreted as total silence at the receiver.

To avoid this, the receiver side adds some

“nice sounding noise” to the output. Using

this approach the required bandwidth could

be reduced substantially.

VoIP Packet

The coded voice sample obtained by the voice-

receiving unit is encapsulated into the other nec-

essary protocols to be able to be sent on to the

network. Firstly, it is encapsulated by the RTP

(see section on VoIP Supporting Protocols), then

by the UDP and finally by the IP. The UDP

(User Datagram Protocol) is a transport protocol,

which coexists with TCP (Transmission Control

Protocol). For real time applications the UDP is

mostly used, while the TCP is more often used

for transport of text, sound and pictures in non

real time. A generic outlook of the encapsulated

packet is shown in Figure 4. A more detailed

format of a VoIP packet is shown in Figure 7.

As could be seen in the latter figure the 20-octet

voice sample becomes at least a 60-octet packet,

and many times more than that. (This size is fur-

ther enlarged before the information enters the

network, due to further encapsulation.) The

parameters in the headers of the encapsulated

VoIP packet format need some explanation. The

RTP header includes fields for which version of

RTP is used (V), if padding is applied (P), if an

extension exists (X), and the number of CSRS

identifiers that follow the fixed header (CC). The

RTP header also contains a sequence number

and time stamp. These two parameters assure

that the packets arrive in order and that informa-

tion is obtained on the actual round trip delay.

This is used to calculate the synchronization and

to minimize the effects of arrival delay varia-

tions (jitter). Finally, the SSRC and CSRC fields

are used to identify the different sources, in this

case audio sources, which are multiplexed

together to create this packet. The following

UDP header is self-explanatory.

The version of IP used is outlined in the first

header bits, followed by the IHL (IP Header

Length), and the type of service field. The latter

describes high or low precedence of delay,

throughput and reliability of this datagram. The

IP header also includes flags which indicate if

fragmentation is allowed and if so, how the pro-

cess should be handled. The fragment offset

field indicates where, in the reassembled mes-

sage, this specific fragment belongs. This is

done to be able to reassemble the entire packet

in a more efficient way, i.e. to be able to start

sorting the fragments before all of them have

arrived. The protocol field identifies the higher

layer protocol following the IP header, in our

case UDP (coded as 17 [3]).

Implementation

The implementation issue is quite complex and

several solutions exist to this problem. Basically,

there are four options to implement VoIP to an

existing network; IP PBX, converged appli-

ances, gateways and other solutions.

The IP PBXs are great for the design of the sys-

tem and have several features such as being able

to manage your phone from your PC, multiline

call control and automatic call distribution.

Using IP PBX also includes the possibility to

create a distributed system throughout an IP net-

work. This means that geographically distributed

and separated phones, with features such as

direct call, forwarding, conferencing, preset

numbers and voice mail, provide the appearance

of being connected directly to the local PBX.

Using converged appliances which join phone

and data networks provide the simplified man-

agement that fulfills the promise of VoIP. The

user gets voice PBX features with a full comple-

ment of data networking, messaging and Internet

functions.

The next option is to use a gateway. A VoIP-

gateway can be loosely defined as a mechanism

that takes circuit switched voice from a tradi-

tional PBX, converts it to IP and transfers it

across a LAN or WAN to another gateway

where it is reconstituted back into a format that

is understood by the receiving phone system.

Gateway functionality can be obtained through

stand-alone boxes, modules or chassis cards for

proprietary boxes, also expandable routers of

software and expansion cards for some servers.

It should however be pointed out that these are

voice packets running over IP. But the packets

are not running over the Internet, and none of the

features and capabilities gained by converging

voice and data networks are obtained. Finally,

Figure 4 The VoIP packet
structure, consisting of the IP,
UDP and RTP header in front
of the actual voice sample

IP-header UDP-header RTP-header Voice sample