O
`
`Terminals to digital and analog PBX
`interfaces. Packet networks supported are
`ATM, Frame Relay and the Internet.
`Let us take an example in which Voice
`over Packet software works with an exist-
`ing communication network, like a cellu-
`lar network, as shown in Fig. 2. The voice
`data in a digital cellular network is
`already compressed and packetized for
`transmission over the air by the cellular
`phone. Packet networks can then transmit
`the compressed cellular voice packet, sav-
`ing a tremendous amount of bandwidth.
`The IWF provides the transcoding func-
`tion required converting the cellular voice
`data to the format required by the Public
`Switched Telephone Network (PSTN).
`
`PBX
`
`VoP
`
`Voice Over
`Packet
`Software
`
`Telephone
`
`ATM
`Frame Relay
`Internet
`
`Fax
`
`Modem
`
`(voice over packet)
`Fig. 1
`
`Deepak Sharma
`
`When one cannot invent, one must at least improve.
`
`r g a n i z a t i o n s
`around the world
`want to reduce ris-
`ing communica-
`tions costs. The consolidation of sepa-
`rate voice and data networks offers an
`opportunity for significant savings.
`Accordingly, the challenge of integrat-
`ing voice and data networks is becom-
`ing a rising priority for many network
`managers. Organizations are pursuing
`solutions that utilize excess capacity on
`broadband networks for voice and data
`transmission, as well as utilize the
`Internet and company Intranets as alter-
`natives to costlier mediums.
`A Voice over Packet (VoP) applica-
`tion meets these chal-
`lenges by allowing both
`voice and signaling infor-
`mation to be transported
`over the packet network.
`This technology was con-
`ceived in the early 1960s.
`But its real potential only
`recently has been under-
`stood due to an enormous
`rise in Communication
`demands. This rise in
`demand saw the develop-
`ment of two fast technolo-
`gy standards in the 80s in
`the form of Frame Relays
`(FR) and Asynchronous
`Transfer Mode (ATM).
`Many corporations have
`long been using voice over
`Frame Relay to save
`money by utilizing excess
`Frame Relay capacity.
`However, the Internet Protocol’s rising
`dominance has shifted most attention
`from VoFR to VoIP.
`
`control and implement some sort of
`echo cancellation.
`Talker overlap (or the problem of
`one talker stepping on the other talker’s
`speech) becomes significant if the one-
`way delay becomes greater than 250
`msec. The end-to-end delay budget is
`therefore the major constraint and dri-
`ving requirement for reducing delay
`through a packet network.
`The following are possible delay
`sources in an end to end Voice over
`Packet call:
`1. Accumulation delay (sometimes
`called algorithmic delay): This delay is
`caused by the need to collect a frame of
`voice samples to be processed by the
`voice coders. It is related
`to the type of voice
`coders used and varies
`from a single sample
`time (.125 microseconds)
`to many milliseconds.
`Each algorithm requires
`that different amounts of
`speech be buffered prior
`to the compression. This
`delay adds to the overall
`end-to-end delay (see
`discussion below). A net-
`work with excessive end-
`to-end delay, often caus-
`es people to revert to a
`half-duplex conversation
`(“How are you today?
`Over…”) instead of the
`normal
`full-duplex
`phone call. (See Table 1)
`2. Processing delay:
`This delay is caused by
`the actual process of encoding and col-
`lecting the encoded samples into a
`packet for transmission over the packet
`network. The encoding delay is a func-
`tion of both the processor execution
`time and the type of algorithm used.
`Often, multiple voice coder frames will
`be collected in a single packet to reduce
`the packet network overhead. For
`example, three frames of G.729 code
`words, equaling 30 milliseconds of
`speech, may be collected and packed
`into a single packet.
`3. Network delay: This
`delay is caused by the physi-
`cal medium and protocols
`used to transmit the voice
`data, and by the buffers used
`to remove packet jitter on
`the receiver side. Network
`delay is a function of the
`capacity of the links in the
`network and the processing
`
`fortune cookie saying posted on blackboard at Univ. of Texas (‘92)
`
`What is VoP?
`“VoP” stands for Voice over Packet.
`Voice is carried as digital data, often
`compressed, along with non-voice data
`over a common packet-switched infra-
`structure. As shown in Fig 1, the legacy
`telephony terminals that are addressed
`range from standard two wire Plain Old
`Telephone Service (POTS) and Fax
`
`Characteristics of VoP
`Delay. Delay causes two problems:
`echo and talker overlap. Echo is caused
`by the signal reflections of the speak-
`er’s voice from the far end’s telephone
`equipment back into the speaker’s ear.
`Echo becomes a significant problem
`when the round trip delay becomes
`greater than 50 milliseconds (msec).
`Echo is perceived as a significant quali-
`ty problem. Thus, Voice over Packet
`systems must address the need for echo
`
`Compression
`Scheme
`
`Compressed Rate Required CPU
`(Kbps)
`Resources
`
`Resultant
`Voice Quality
`
`G.711 PCM
`G.723 MP-MLQ
`G.726 ADPCM
`G.728 LD-CELP
`G.729 CS-ACELP
`
`64 (no compression)
`6.4/5.3
`40/32/24
`16
`8
`
`Not Required
`Moderate
`Low
`Very High
`High
`
`Excellent
`Good (6.4) Fair (5.3)
`Good (40) Fair (24)
`Good
`Good
`
`Added
`Delay
`
`N/A
`High
`Very Low
`Low
`Low
`
`14
`
`0278-6648/02/$17.00 © 2002 IEEE
`
`IEEE POTENTIALS
`
`Apple 1018
`U.S. Pat. 7,535,890
`
`

`
`circuit switched telephone network.
`However, it is acceptable because the
`round trip delays through the network
`are smaller than 50 msec. Also, the echo
`is masked by the normal side tone every
`telephone generates.
`Echo becomes a problem in Voice
`over Packet networks because the round
`trip delay through the network is almost
`always greater than 50 msec. Thus, echo
`cancellation techniques are always used.
`Echo is generated toward the packet
`network from the telephone network.
`The echo canceller compares the voice
`data received from the packet network
`with voice data being transmitted to the
`packet network. The echo from the tele-
`phone network hybrid is removed by a
`digital filter on the transmit path into the
`packet network.
`
`Technology developments
`& paradigms
`As mentioned earlier, the technology
`has its roots in the early 60s.
`Unfortunately, back then, the large cop-
`per analog and asynchronous infrastruc-
`ture caused huge amplitude and group-
`delay distortions and data conduction
`was done through “dumb” asynchro-
`nous terminals. But today we have
`smart and synchronous terminals as
`well as digital techniques that minimize
`the hassles of handling traffic on a node
`to node basis as well as increase the
`speed manifolds.
`Frame relay and asynchronous
`transfer mode. Two fast packet technol-
`ogy standards emerged in the late
`1980s: Frame Relay (FR) and
`Asynchronous Transfer Mode (ATM).
`
`Packet
`Network
`
`Branch I
`
`IWF
`Telephone
`
`Branch II
`
`IWF
`
`Telephone
`
`Home Office
`
`Mainframe
`
`Telephone
`
`IWF
`
`PBX
`
`that occurs as the packets transit the
`network. The jitter buffers add delay.
`They are used to remove the packet
`delay variation that each packet is sub-
`jected to as it transits the packet net-
`work. This delay can be a significant
`part of the overall delay since packet
`delay variations can be as high as 70-
`100 msec in some Frame Relay net-
`works and IP networks.
`Jitter. The delay problem is com-
`pounded by the need to remove jitter: a
`variable inter-packet timing caused by
`the network a packet traverses.
`Removing jitter requires collecting
`packets and holding them long enough
`to allow the slowest packets to arrive in
`time to be played in the correct
`sequence. This causes additional delay.
`The two conflicting goals of mini-
`mizing delay and removing jitter have
`engendered various schemes to adapt
`the jitter buffer size to match the time
`varying requirements of network jitter
`removal. This adaptation has the explic-
`it goal of minimizing the size and delay
`of the jitter buffer, while at the same
`time preventing buffer underflow
`caused by jitter.
`The approach selected will depend
`on the type of network the packets are
`traversing. Two approaches to adapting
`the jitter buffer size are:
`1. To measure the variation of the
`packet-level in the jitter buffer over a peri-
`od of time, and incrementally adapt the
`buffer size to match the calculated jitter.
`This approach works best with networks
`that provide a consistent jitter perfor-
`mance over time, such as ATM networks.
`2. To count the number of packets
`that arrive late and create a ratio of
`these packets to the number of packets
`that are successfully processed. This
`ratio is then used to adjust the jitter
`buffer to target a predetermined allow-
`able late packet ratio. This approach
`works best with the networks with high-
`ly variable packet inter-arrival intervals,
`such as IP networks.
`In addition to the techniques just
`described, the network must be config-
`ured and managed to provide minimal
`delay and jitter, enabling a consistent
`quality of service.
`Lost packet compensation. Lost
`packets can be an even more severe
`problem. It depends on the type of
`packet network that is being used.
`Because IP networks do not guarantee
`service, they will usually exhibit a much
`higher incidence of lost voice packets
`than ATM networks.
`
`In current IP networks, all voice
`frames are treated like data. Under peak
`loads and congestion, voice frames will
`be dropped equally with data frames.
`The data frames, however, are not time
`sensitive and dropped packets can be
`appropriately corrected through the
`process of retransmission. Lost voice
`packets, however, cannot be dealt with
`in this manner.
`Some schemes used by Voice over
`Packet software to address the problem
`of lost frames are:
`1. Interpolate for lost speech packets
`by replaying the last packet received
`during the interval when the lost packet
`was supposed to be played out. This
`scheme is a simple method that fills the
`time between non-contiguous speech
`frames. It works well when the inci-
`dence of lost frames is infrequent. It
`does not work very well if there are a
`number of lost packets in a row or a
`burst of lost packets.
`2. Send redundant information at the
`expense of bandwidth utilization. The
`basic approach replicates and sends the
`nth packet of voice information along
`with the (n+1)th packet. This method
`has the advantage of being able to exact-
`ly correct for the lost packet. However,
`this approach uses more bandwidth and
`also creates greater delay.
`3. A hybrid approach uses a much
`lower bandwidth voice coder to provide
`redundant information carried along in
`the (n+1)th packet. This reduces the
`problem of the extra bandwidth
`required, but still fails to solve the prob-
`lem of delay.
`Echo compensation. Echo in a tele-
`phone network is
`caused by signal
`reflections gen-
`erated by the
`hybrid circuit
`that converts
`between a 4-wire
`circuit (a sepa-
`rate transmit and
`receive pairs)
`and a 2-wire cir-
`cuit (a single
`transmit
`and
`receive pair).
`These reflections
`of the speaker’s
`voice are heard
`in the speaker’s
`ear.
`Echo is pre-
`sent even in a
`c o n v e n t i o n a l
`
`Telephone
`
`PSTN
`
`Fig. 2 Application 1: branch office
`
`OCTOBER/NOVEMBER 2002
`
`15
`
`

`
`Table 2
`
`Frame Relay
`
`Asynochronous
`Transfer
`Mode
`
`device. With this capability, the
`future fast packet-based telecommu-
`nications hierarchy has taken shape.
`In time, fast packet technology
`probably will replace the existing
`circuit-switched infrastructure.
`Furthermore, in the telecommunica-
`tions environment, FR and ATM
`exhibit attributes not seen in other
`network technologies: carriers
`around the world embrace FR and
`ATM. As a result, their respective
`growth rates are unprecedented. A brief
`comparison between these two tech-
`nologies is shown in Table 2.
`
`Packet Delay
`
`Packet Jitter
`
`Overhead
`
`Access Speed
`
`Variable
`
`Variable
`
`Low
`
`Deterministic
`
`Deterministic
`
`High
`
`Typical Nx64
`KBPS, T1/E1
`
`DXI, (> or =)
`45 MBPS
`
`While briefly there was confusion as to
`the relative positioning of these two
`standards, today each has a clearly
`defined position: FR has become the
`standard for low-speed access and feed-
`er networks; ATM has become the stan-
`dard for high-speed connection and
`backbone infrastructure.
`Interoperability between these two
`standards has also been established. An
`interface known as Frame User to
`Network Interface (FUNI) allows a sig-
`nal to enter the fast packet environment
`via a Frame Relay Access Device
`(FRAD) and be extracted by an ATM
`
`Office 1
`
`IWF
`
`Telephone
`
`PBX
`
`Telephone
`
`Packet
`Network
`
`Fig. 3 Application 2: interoffice trunking
`
`Base Station
`Subsystem
`
`Base Transceiver
`Station (BTS)
`
`Office 2
`
`IWF
`
`Telephone
`
`PBX
`
`Telephone
`
`Benefits of VoP networks
`Less bandwidth consumption:
`Packetized voice offers much higher band-
`width efficiency than circuit-switched
`voice because it does not take up any
`bandwidth in listening mode or during
`pauses in a conversation. Say we were to
`use the same 64 Kbps Pulse Code
`Modulation (PCM) digital-voice encoding
`method in both
`technologies. We
`would see that the
`bandwidth con-
`sumption of the
`packetized voice
`is only a fraction
`of the consump-
`tion of the circuit-
`switched voice.
`For this reason,
`packetized voice
`has already been
`deployed in most
`Trans-Pacific and
`Trans-Atlantic
`Digital Circuit
`Multiplication
`E q u i p m e n t
`(DCME).
`I n d u s t r i a l -
`voice advan-
`tages: There is a
`number of voice
`products that are
`extremely price
`or cost sensitive.
`For example,
`800 numbers are
`paid for by one
`party, 100 call-
`ing is free to the
`public, and inex-
`pensive
`tele-
`phone directory
`service is key to
`retaining cus-
`tomer loyalty.
`
`Base Transceiver
`Station (BTS)
`
`Base Station
`Controller
`(BSC)
`
`ATM
`
`Base Transceiver
`Station (BTS)
`
`Base Station
`Controller
`(BSC)
`
`Base Transceiver
`Station (BTS)
`
`Transending
`Gateway
`
`PSTN
`
`Mobile
`Switching
`Center
`(MSC)
`
`Fig. 4 Application 3: cellular networking interworking
`
`Carriers could offer these types of voice
`products using their lower-cost fast
`packet infrastructure.
`Low-cost PSTN options: Remember
`the days of pulse-dial and touch-tone ser-
`vices? They were offered as alternatives to
`consumers at different pricing levels. The
`same strategy can be used to move voice
`service from circuit switched to packet
`switched. A consumer can, with the touch
`of a button at call initiation, decide
`whether to call on the existing PSTN or
`the new packetized PSTN.
`
`Concerns about VoP networks
`Some believe that fast packet tech-
`nology works well with data but not
`with voice. The common concerns sur-
`rounding voice over frame relay are
`voice quality, modem handling, network
`latency, congestion, switched connectiv-
`ity and the commercial implications of
`implementing fast packet systems.
`Voice quality: The earlier voice-
`compression systems did produce less-
`than-desirable
`voice
`quality.
`Consequently, use of these early sys-
`tems was restricted mainly to secured
`voice applications. This legacy took
`years to overcome. Today, however,
`advanced algorithms not only have
`reduced the complexity of the design
`but they also produce voice quality
`equal to the standard PCM and
`Adaptive PCM (ADPCM) so widely
`deployed in the digital network.
`Modem handling: The modem was
`invented to allow digital data to be trans-
`mitted across an analog network. It is
`therefore a
`transitional product.
`Ultimately, modems will be eliminated
`due to an ubiquitous digital infrastructure.
`Packetized voice systems handle
`modem signals by detecting signal pres-
`ence, demodulating the signal back to
`its original digital form, and then pass-
`ing it transparently across the network.
`At the remote end, this data is then re-
`modulated before presenting it to the
`receiving modem. A side benefit is
`bandwidth efficiency. Instead of a 64
`Kbps PCM channel to handle a modem
`signal carrying 9.6 Kbps data, only 9.6
`Kbps is needed.
`Network latency: The concern over
`network latency is the result of another
`legacy-related misconception. Fast
`packet systems offer faster transmission
`than do older packet systems.
`The old X.25 packet system uses
`node-to-node significance to transmit
`data: errors are checked every step of the
`way. The network, itself, can introduce a
`
`16
`
`IEEE POTENTIALS
`
`

`
`transmission delay of a few hundred mil-
`liseconds from source to destination.
`Fast packet systems, on the other
`hand, care only about where to route the
`traffic, not about the integrity of the
`transmission. They use end-to-end sig-
`nificance to transmit the data. The delay
`from source to destination is at least two
`orders of magnitude less than X.25.
`Congestion management: TDM/cir-
`cuit-switch network design is easy to
`understand because it is deterministic in
`nature. Once bandwidth is assigned to a
`given application, it is protected. Such
`simplicity, however, is also the cause of
`its inefficiency.
`Virtual circuit/packet switch network
`design, on the other hand, is statistical in
`nature. Bandwidth is not given until
`needed. This design technique is more
`difficult to understand and, therefore, has
`been less readily accepted. But, in reali-
`ty, the design is quite simple and logical.
`The proactive approach prevents con-
`gestion by understanding the traffic
`flow, properly sizing the network and
`constantly monitoring the bandwidth uti-
`lization. In addition, the virtual
`circuit/packet switch network removes
`the likelihood of a logjam by taking
`advantage of the congestion notification
`mechanism available in the fast packet
`infrastructure. Short-term bandwidth
`demand can be throttled back by a com-
`bination of reducing the voice encoding
`rate, slowing the data transfer speed and
`allowing certain packets to be discarded.
`Switched connectivity: Packet systems
`offer a far more flexible way of forming
`switched circuits across a network than
`circuit systems. Each packet has a header
`containing the originating and destination
`addresses. This allows connections to be
`set up on a packet-by-packet basis, hence,
`the term virtual circuit.
`Two types of virtual circuits are
`available: Permanent Virtual Circuit
`(PVC) and Switched Virtual Circuit
`(SVC). PVC is equivalent to today’s
`leased line, whereas SVC is equivalent
`to today’s switched line. SVC is now a
`standard; equipment vendors for switch-
`es and access devices have begun offer-
`ing such a facility in their products.
`
`Software architecture
`Two major types of information must
`be handled in order to interface telepho-
`ny equipment to a packet network: voice
`and signaling information (Fig. 5).
`Voice over Packet software interfaces to
`both streams of information from the
`telephony network and converts them to
`
`a single stream of packets transmitted to
`the packet network.
`The software functions are divided
`into four general areas:
`Voice packet software module: This
`software, typically runs on a digital sig-
`nal processor (DSP), prepares voice
`samples for transmission over the packet
`network. Its components perform echo
`cancellation, voice compression, voice
`activity detection, jitter removal, clock
`synchronization and voice packetization.
`Telephony signaling gateway software
`module: This software interacts with the
`telephony equipment translating signaling
`into state changes used by the Packet
`Protocol Module to set up connections.
`These state changes are on-hook, off-
`hook, trunk seizure and so forth.
`Packet protocol module: This module
`processes signaling information and con-
`verts it from the telephony signaling proto-
`cols to the specific packet signaling proto-
`col used to set up connections over the
`packet network (e.g. Q.933 and Voice
`over FR signaling). It also adds protocol
`headers to both voice and signaling pack-
`ets
`before
`they are trans-
`mitted
`into
`the packet
`network.
`N e t w o r k
`management
`module: This
`module pro-
`vides
`the
`Voice man-
`a g e m e n t
`interface to
`configure and
`maintain the
`other modules of the voice over packet
`system.
`The software is partitioned to pro-
`vide a well-defined interface to the DSP
`software usable for multiple voice pack-
`et protocols and applications. The DSP
`processes voice data and passes voice
`packets to the microprocessor with
`generic voice headers.
`The microprocessor is responsible
`for moving voice packets and adapting
`the generic voice headers to the specific
`Voice Packet Protocol called for by the
`application, such as Real-Time
`Protocol (RTP), Voice over Frame
`Relay (VoFR), and Voice Telephony
`over ATM (VTOA). The microproces-
`sor also processes signaling informa-
`tion and converts it from supported
`telephony signaling protocols to the
`packet network signaling protocol (e.g.
`
`H.323 (IP), Frame Relay, or ATM sig-
`naling). This partitioning provides a
`clean interface between the generic
`voice processing functions, such as
`compression, echo cancellation, and
`voice activity detection.
`
`Conclusion
`All the major communication com-
`panies are pouring money into the
`research of carrying voice in a digital
`compressed form along with the non-
`voice data over a common packet
`switched network. Much work still needs
`to be done. Better technology is required
`for, among other things, minimizing
`delay and reducing transmission errors.
`Voice quality alone cannot guarantee
`the success or failure of voice over
`packet networks. All the same, speech
`quality remains one of the prime hur-
`dles affecting such networks.
`
`Acknowledgements
`Thanks to my teachers who helped
`me prepare this paper, especially my
`project guide, Prof. S.L. Haridas and
`
`Network
`Management
`Module
`
`SNMP
`
`Telephony
`Signaling
`Module
`
`Packet
`Protocol
`Module
`
`Microprocessor
`
`Voice &
`Signaling
`Packets
`
`Prof. M.A.Gaikwad. Thanks to the col-
`lege authority for allowing me to access
`the college Internet for my research.
`Finally, thanks to the librarian who
`arranged the encyclopedias on IEEE
`Spectrum magazine for me .
`
`Read more about it
`• IEEE Spectrum, February 2000
`• IEEE Spectrum, March 1998.
`• IEC ONLINE EDUCATION
`(www.iec.org/online/tutorials).
`• www.protocols.com/papers/voip 2.htm
`• www.telogy.com/our_products/
`golden_gateway/VOPwhite.html
`
`About the author
`Deepak Sharma received his
`Bachelors in Electronics Engineering
`from BD College of Engineering,
`Nagpur University in India.
`
`Signaling
`
`Voice
`
`Voice
`Packet
`Module
`
`DSP
`
`Fig. 5 Voice over Packet software architecture
`
`OCTOBER/NOVEMBER 2002
`
`17