Bearer Protocols in Detail

The bearer protocols are where the real work in voice gets done. The bearer channel carries the voice, sampled by microphones as digital data, compressed in some manner, and then placed into packets which need to be coordinated as they fly over the networks.

Voice, as you know, starts off as sound waves (Figure 1). These sound waves are picked up by the microphone in the handset, and are then converted into electrical signals, with the voltage of the signal varying with the pressure the sound waves apply to the microphone.

 

Figure 1: Typical Voice Recording Mechanisms

The signal (see Figure 2) is then sampled down into digital, using an analog-to-digital converter. Voice tends to have a frequency around 3000 Hz. Some sounds are higher—music especially needs the higher frequencies—but voice can be represented without significant distortion at the 3000Hz range. Digital sampling works by measuring the voltage of the signal at precise, instantaneous time intervals. Because sound waves are, well, wavy, as are the electrical signals produced by them, the digital sampling must occur at a high enough rate to capture the highest frequency of the voice. As you can see in the figure, the signal has a major oscillation, at what would roughly be said is the pitch of the voice. Finer variations, however, exist, as can be seen on closer inspection, and these variations make up the depth or richness of the voice. Voice for telephone communications is usually limited to 4000 Hz, which is high enough to capture the major pitch and enough of the texture to make the voice sound human, if a bit tinny. Capturing at even higher rates, as is done on compact discs and music recordings, provides an even stronger sense of the original voice.

 

Figure 2: Example Voice Signal, Zoomed in Three Times

Sampling audio so that frequencies up to 4000 Hz can be preserved requires sampling the signal at twice that speed, or 8000 times a second. This is according to the Nyquist Sampling Theorem. The intuition behind this is fairly obvious. Sampling at regular intervals is choosing which value at those given instants. The worst case for sampling would be ifone sampled a 4000 Hz, say, sine wave at 4000 times a second. That would guarantee to provide a flat sample, as the top pair of graphs in Figure 3 shows. This is a severe case of undersampling, leading to aliasing effects. On the other hand, a more likely signal, with a more likely sampling rate, is shown in the bottom pair of graphs in the same figure. Here, the overall form of the signal, including its fundamental frequency, is preserved, but most of the higher-frequency texture is lost. The sampled signal would have the right pitch, but would sound off.

 

Figure 3: Sampling and Aliasing

The other aspect to the digital sampling, besides the 8000 samples-per-second rate, is the amount of detail captured vertically, into the intensity. The question becomes how many bits of information should be used to represent the intensity of each sample. In the quantization process, the infinitely variable, continuous scale of intensities is reduced to a discrete, quantized scale of digital values. Up to a constant factor, corresponding to the maximum intensity that can be represented, the common value for quantization for voice is to 16 bits, for a number between −2¹⁵ = −32,768 to 2¹⁵ −1 = 32,767.

The overall result is a digital stream of 16-bit values, and the process is called pulse code modulation (PCM), a term originating in other methods of encoding audio that are no longer used.

SS7

Signaling System #7 (SS7) is the protocol that makes the public telephone networks operate, within themselves and across boundaries. Unlike Q.931, which is designed for simplicity, SS7 is a complete, Internet-like architecture and set of protocols, designed to allow call signaling and control to flow across a small, shared set of circuits dedicated for signaling, freeing up the rest of the circuits for real phone calls.

SS7 is an old protocol, from around 1980, and is, in fact, the seventh version of the protocol. The entire goal of the architecture was to free up lines for phone calls by removing the signaling from the bearer channel. This is the origin of the split signaling and bearer distinction. Before digital signaling, phone lines between networks were similar to phone lines into the home. One side would pick up the line, present a series of digits as tones, and then wait for the other side to route the call and present tones for success, or a busy network. The problem with this method of in-band signaling was that it required having the line held just for signaling, even for calls that could never go through. To free up the waste from the in-band signaling, the networks divided up the circuits into a large pool of voice-only bearer lines, and a smaller number of signaling-only lines. SS7 runs over the signaling lines.

It would be inappropriate here to go into any significant detail into SS7, as it is not seen as a part of voice mobility networks. However, it is useful to understand a bit of the architecture behind it.

SS7 is a packet-based network, structured rather like the Internet (or vice versa). The phone call first enters the network at the telephone exchange, starting at the Service Switching Point (SSP). This switching point takes the dialed digits and looks for where, in the network, the path to the other phone ought to be. It does this by sending requests, over the signaling network, to the Service Control Point (SCP). The SCP has the mapping of userunderstandable telephone numbers to addresses on the SS7 network, known aspoint codes. The SCP responds to the SSP with the path the call ought to take. At this point, the switch (SSP) seeks out the destination switch (SSP), and establishes the call. All the while, routers called Signal Transfer Points (STPs) connect physical links of the network and route the SS7 messages between SSPs and SCPs.

The interesting part of this is that the SCP has this mapping of phone numbers to real, physical addresses. This means that phone numbers are abstract entities, like email addresses or domain names, and not like IP addresses or other numbers that are pinned down to some location. Of course, we already know the benefit of this, as anyone who has ever changed cellular carriers and kept their phone number has used this ability for that mapping to be changed. The mapping can also be regional, as toll-free 800 numbers take advantage of that mapping as well.

ISDN and Q.931

The ISDN protocol is where telephone calls to the outside world get started. ISDN is the digital telephone line standard, and is what the phone company provides to organizations that ask for digital lines. By itself, ISDN is not exactly a voice mobility protocol, but because a great number of voice calls from voice mobility devices must go over the public telephone network at some point, ISDN is important to understand.

With ISDN, however, we leave the world of packet-based voice, and look at tightly timed serial lines, divided into digital circuits. These circuits extend from the local public exchange—where analog phone lines sprout from before they run to the houses—over the same types of copper wires as for analog phones. The typical ISDN line that an enterprise uses starts from the designation T1, referring to a digital line with 24 voice circuits multiplexed onto it, for 1536kbps. The concept of the T1 (also known, somewhat more correctly, as a DS1, with each of the 24 digital circuits known as DS0s) is rather simple. The T1 line acts as a constant source or sink for these 1536kbps, divided up into the 24 channels of 64kbps each. With a few extra bits for overhead, to make sure both sides agree on which channel is which, the T1 simply goes in round-robin order, dedicating an eight-bit chunk (the actual byte) for the first circuit (channel), then the second, and so on. The vast majority of traffic is bearer traffic, encoded as standard 64kbps audio. The 23 channels dedicated for bearer traffic are called B channels.

As for signaling, an ISDN line that is running a signaling protocol uses the 24th line, called the D channel. This runs as a 64kbps network link, and standards define how this continuous serial line is broken up into messages. The signaling that goes over this channel usually falls into the ITU Q.931 protocol.

Q.931's job is to coordinate the setting up and tearing down of the independent bearer channels. To do this, Q.931 uses a particular structure for their messages. Because Q.931can run over any number of different protocols besides ISDN, with H.323 being the other major one, the descriptions provided here will steer clear of describing how the Q.931 messages are packaged.

Table 1 shows the basic format of the Q.931 message. The protocol discriminator is always the number 8. The call reference refers to the call that is being referred to, and is determined by the endpoints. The information elements contain the message body, stored in an extensible yet compact format.

Table 1: Q.931 Basic Format 

Protocol Discriminator	Length of Call Reference	Call Reference	Message Type	Information Elements
1 byte	1 byte	1-15 bytes	1 byte	variable

The message type is encompasses the activities of the protocol itself. To get a better sense for Q.931, the message types and meanings are:

• SETUP: this message starts the call. Included in the setup message is the dialed number, the number of the caller, and the type of bearer to use.
• CALL PROCEEDING: this message is returned by the other side, to inform the caller that the call is underway, and specifies which specific bearer channel can be used.
• ALERTING: informs the caller that the other party is ringing.
• CONNECT: the call has been answered, and the bearer channel is in use.
• DISCONNECT: the phone call is hanging up.
• RELEASE: releases the phone call and frees up the bearer.
• RELEASE COMPLETE: acknowledges the release.

There are a few more messages, but it is pretty clear to see that Q.931 might be the simplest protocol we have seen yet! There is a good reason for this: the public telephone system is remarkably uniform and homogenous. There is no reason for there to be flexible or complicated protocols, when the only action underway is to inform one side or the other of a call coming in, or choosing which companion bearer lines need to be used. Because Q.931 is designed from the point of view of the subscriber, network management issues do not need to be addressed by the protocol. In any event, a T1 line is limited to only 64kbps for the entire call signaling protocol, and that needs to be shared across the other 23 lines.

Digital PBXs use IDSN lines with Q.931 to communicate with each other and with the public telephone networks. IP PBXs, with IP links, will use one of the packet-based signaling protocols mentioned earlier.

Polycom SpectraLink Voice Priority (SVP)

Early in the days of voice over Wi-Fi, a company called SpectraLink—now owned by Polycom—created a Wi-Fi handset, gateway, and a protocol between them to allow the phones to have good voice quality, when Wi-Fi itself did not yet have Wi-Fi Multimedia (WMM) quality of service. SVP runs as a self-contained protocol, for both signaling and bearer traffic, over IP, using a proprietary IP type (neither UDP nor TCP) for all of the traffic.

SVP is not intended to be an end-to-end signaling protocol. Rather, like Cisco's SCCP, it is intended to bridge between a network server that speaks the real telephone protocol and the proprietary telephone. Therefore, SCCP and SVP have a roughly similar architecture. The major difference is that SVP was designed with wireless in mind to tackle the early quality-of-service issues over Wi-Fi, whereas SCCP was designed mostly as a way of simplifying the operation of phone terminals over wireline IP networks.

Figure 1 shows the SVP architecture. The SVP system integrates into a standard IP PBX deployment. The SVP gateway acts as the location for the extensions, as far as the PBX is concerned. The gateway also acts as the coordinator for all of the wireless phones. SVP phones connect with the gateway, where they are provisioned. The job of the SVP gateway is to perform all of the wireless voice resource management of the network. The SVP performs the admission control for the phones, being configured with the maximum number of phones per access point and denying phones the ability to connect to it through access points that are oversubscribed. The SVP server also engages in performing timeslice coordination for each phone on a given access point.

 

Figure 1: SVP Architecture

This timeslicing function makes sense in the context of how SVP phones operate. SVP phones have proprietary Wi-Fi radios, and the protocol between the SVP gateway and the phone knows about Wi-Fi. Every phone reports back what access point it is associated to. When the phone is placed into a call, the SVP gateway and the phone connect their bearer channels. The timing of the packets sent by the phone is such that it is directly related to the timing of the phone sent by the gateway. Both the phone and the gateway have specific requirements on how the packets end up over the air. This, then, requires that the access points also be modified to be compatible with SVP. The role of the access point is to dutifully follow a few rules which are a part of the SVP protocol, to ensure that the packets access the air at high priority and are not reordered. There are additional requirements for how the access point must behave when a voice packet is lost and must be retransmitted by the access point. By following the rules, the access point allows the client to predict how traffic will perform, and thus ensures the quality of the voice.

SVP is a unique protocol and system, in that it is designed specifically for Wi-Fi, and in such a way that it tries to drive the quality of service of the entire SVP system on that network through intelligence placed in a separate, nonwireless gateway. SVP, and Polycom SpectraLink phones, are Wi-Fi-only devices that are common in hospitals and manufacturing, where there is a heavy mobile call load inside the building but essentially no roaming required to outside.

Skype | Signaling Protocols in Detail

Skype is mentioned here because it is such an intriguing application. Famous for its resiliency when running over the Internet, or any other non-quality-of-service network, as well as for its chat feature and low-cost calls, questions will always come up about Skype. Undoubtedly, Skype has helped many organizations reduce long distance or international phone bills, and many business travelers have favored it when on the road and in a hotel, to avoid room and cell charges for telephone use.

Skype is a completely proprietary peer-to-peer protocol, encrypted hop-by-hop to prevent unauthorized snooping. There are plenty of resources available on how to use Skype, so it will be appropriate for us to stick with just the basics on how it applies for voice mobility.

The most important issue with Skype is that it is not manageable in an enterprise sense. Not only is it a service hosted outside the using enterprise, but the technology itself is encrypted to prevent even basic understanding or diagnosis. Furthermore, it cannot be run independent of Internet connectivity, and it is designed to find ways around firewalls. As a primarily consumer-oriented technology, Skype does not yet have the features necessary for enterprise deployments, and thus is severely limited in a sense useful for large-scale voice mobility.

Another main issue with Skype is that it does not take advantage of quality-of-service protocols to provide reliable or predictable, or even prioritized, voice quality. Traffic engineering with Skype is incredibly difficult, especially if one tries to predict how Skype will consume resources if large portions of the networked population choose to use it, inside or outside the office.

On the other hand, Skype comes with better, high-bitrate codecs that make voice sound much less tinny than the typical low-bitrate codecs used by telephones that may have to access the public switched telephone network (PSTN). Skype's ability to free itself from PSTN integration as the standard case (Skype's landline telephone services can be thought of more as special cases) has allowed it to be optimized for better voice quality in a lossy environment.

Skype is unlikely to be useful in current voice mobility deployments, so it will not be mentioned much further in this book. However, Skype will always be found performing somewhere within the enterprise, and so its usage should be understood. As time progresses, it may be possible that people will have worked out a more full understanding of how to deploy Skype in the enterprise.

Cisco SCCP: "Skinny" | Signaling Protocols in Detail

Cisco has a proprietary Skinny Client Control Protocol (SCCP), which is used by the Cisco Unified Communications Manager and Cisco phones as their own signaling protocol. SCCP requires the Cisco Unified Communications Manager or open-source PBXs to operate. Given the downside of proprietary protocols, the main reason for discussing SCCP within the context of voice mobility is only that Cisco's Wi-Fi-only handsets support SCCP, and so SCCP may be seen in some voice mobility networks. Unfortunately, SCCP internal documentation is not widely available or as well understood as an open protocol is, and so enterprise-grade implementations tend to lock the user into one vendor only.

SCCP runs on TCP, using port 2000. The design goal of SCCP was to keep it "skinny," to allow the phone to have as little intelligence as needed. In this sense, the Cisco Unified Communications Manager (or older Cisco Call Manager) is designed to interface with other telephone technologies as a proxy, leaving the phone to deal with supporting the one proprietary protocol.

SCCP has a markedly different architecture from what we have seen already. SCCP is still an IP-based protocol, and there is the one point of contact that the phone uses for all of its signaling. However, the signaling design of SCCP has the remarkable property, unlike with SIP or H.323, that the phone is not self-contained as an extension. Rather, SCCP is entirely user event—based. The phone's job is to report back to the call manager, in real time, whenever a button is pressed. The call manager then pushes down to the phone any change in state that should accompany the button press. In this way, the entire logic as to what buttons mean is contained in the call manager, which locally runs the various telephone endpoint logic. In this way, SCCP has more in common with Remote Desktop than it has with telephone signaling protocols: the phone's logic really runs in some centralized terminal server, which is called the call manager. To emphasize this point, Table 1 lists a typical sequence of events between a phone and a call manager, from when the phone is taken off the hook.

Table 1: Example SCCP Call Setup Event Flow 

#	Direction	Event Name	State	Meaning
1	Phone → Call Manager	Offhook	Dialing	User has taken the phone off the hook.
2	Call Manager → Phone	StationOutputDisplayText		Displays a prompt that the phone is off hook and waiting for digits.
3	Call Manager → Phone	SetRinger		Turns off the ringer.
4	Call Manager → Phone	SetLamp		Turns on the light for the line that is being used.
5	Call Manager → Phone	CallState		Sets the phone up so that the user can hear audio and press buttons.
6	Call Manager → Phone	DisplayPromptStatus		The phone is not connected to any other extension yet.
7	Call Manager → Phone	SelectSoftKeys
8	Call Manager → Phone	ActivateCallPlane
9	Call Manager → Phone	StartTone		Starts a dial tone.
10	Phone → Call Manager	KeypadButton (dialed 7)		The user dialed the number 7.
11	Call Manager → Phone	StopTone		Stops the dial tone, acknowledging that a digit has been dialed.
12	Call Manager → Phone	SelectSoftKeys		Changes the keys of interest to just the number pad (no redial buttons, etc.).
13	Phone → Call Manager	KeypadButton (dialed 0)		The user dialed the number 0.
14	Phone → Call Manager	KeypadButton (dialed 2)		The user dialed the number 2.
15	Phone → Call Manager	KeypadButton (dialed 0)		The user dialed the number 0.
16	Call Manager → Phone	SelectSoftKeys	Ringing	Changes the keys of interest.
17	Call Manager → Phone	CallState		Changes the state of the phone.
18	Call Manager → Phone	Callinfo
19	Call Manager → Phone	DialedNumber		Reports that 7020 has been dialed.
20	Call Manager → Phone	StartTone		Starts playing a ringback tone.
21	Call Manager → Phone	DisplayPromptStatus		Changes the prompt to show that the other side of the phone is ringing.
22	Call Manager → Phone	Callinfo		The call is still ringing.
23	Call Manager → Phone	StopTone	Connected	Stops playing the ringback tone.
24	Call Manager → Phone	DisplayPromptStatus		Displays that the phone call was answered.
25	Call Manager → Phone	OpenReceiveChannel		Prepares for the downward leg of the call.
26	Phone → Call Manager	OpenReceiveChannelAck		Acknowledges the downward leg.
27	Call Manager → Phone	StartMediaTransmission		The call's bearer channel starts flowing.
28	Phone → Call Manager	OnHook	Hanging Up	The caller hung up.
29	Call Manager → Phone	CloseReceiveChannel		Tears down the receive leg.
30	Call Manager → Phone	StopMediaTransmission		Stops the bearer channel entirely.
31	Call Manager → Phone	SetSpeakerMode		Restores the phone to the original state.
32	Call Manager → Phone	ClearPromptStatus
33	Call Manager → Phone	CallState
34	Call Manager → Phone	DisplayPromptStatus
35	Call Manager → Phone	ActivateCallPlane
36	Call Manager → Phone	SetLamp		Turns off the light for the line that was in use.

As you can see, the phone's entire personality—the meaning of the buttons, what the display states, which lights are lit, the tones generated—are entirely controlled by the call manager.

Overall, this is a marked difference from true telephone signaling protocols. In this sense, then, one can consider SCCP to be mostly a remote control protocol for phones, and the call manager is thus left with the burden of implementing the true telephone protocol. Unfortunately, however, when SCCP is used with a packet-based voice mobility network, the protocol going over the wireless or edge network is going to be SCCP, and not whatever protocol the call manager is enabled with.

Bearer traffic, on the other hand, still uses RTP, as do the other protocols we have looked at so far. Therefore, most of the discussion on bearer traffic, and on voice traffic in general, holds for SCCP networks.

H.323 | Signaling Protocols in Detail

H.323 is an International Telecommunication Union (ITU) specification that defines how to establish both the signaling and the bearer channels. The goal was to use the call signaling from ISDN, a call control protocol for renegotiating calls as they are ongoing, and a registration protocol, to provide for an all-encompassing solution. (Compare how SIP defines only registration, call signaling, and merely basic call control.) However, the process was different, and the H.323 technology suffers from the typical emphasis on layering and precise botanical definitions of technologies that haunts the world of telecommunications.

H.323's major advantage, compared to SIP, is that it contains the complete protocol definition for the application, covering features such as media reservation and conference negotiation that SIP leaves alone. H.323 is also able to pull together a number of other ITU definitions and technologies, into one larger umbrella. Because of the ITU protocols' amount of definition for media applications, H.323 is still the signaling protocol of choice for many videoconferencing applications.

That being said, H.323's relevance for voice is waning. For that reason, we will stick to H.323 at a higher level than we did with SIP.

H.323 Architecture

H.323 has a somewhat similar architecture to SIP. Figure q shows this architecture.

Figure 1: H.323 Architecture

The endpoints are known as terminals. The terminals must register with the registrar, which is now in a function known as the gatekeeper. The gatekeeper is the PBX, and has complete responsibility for administration, user definitions, registration, and routing. Gateways are now special devices that are specifically called out for bridging signaling and media between two different networks.

H.323 uses a protocol known as H.225.0 for call setup signaling. H.225.0 itself is a package that refers to Q.931 for call signaling definitions. (This sort of nesting is typical for telecommunications definitions). The good news is that it stops at Q.931, and we can identify it. ITU Q.931 is the call signaling protocol used in ISDN lines. H.225.0 also includes the Registration, Admission, and Status (RAS) protocol, used by the client to register with the network.

The registration function is, therefore, defined by RAS. When a phone comes online, its first task is to use RAS to find a gatekeeper (from a known or discovered list) that is willing to let it register. Once it does that, it then requests to register. After the registration is complete, the phone is ready to send or receive calls.

To place a call, the phone sends an admission request to the gatekeeper. The gatekeeper's job is to find out where the other endpoint is, by looking up in its extension or routing tables. The result will be an acceptance or rejection. If the result is an acceptance, the gatekeeper will also respond with the contact information for the other endpoint. Notice that the model here is based on admission control. The gatekeeper is allowed to monitor voice resources, and reject calls purely on the basis of there not being enough resources. In any event, the caller now has the contact information of the called party, so the caller contacts the called party directly to attempt to establish the call. This direct contact is done using Q.931 signaling over IP. If the called party is willing to accept the call, the called party must contact its local gatekeeper with an admission request. If that is granted, the call is ready to be finalized.

H.245 plays the role of establishing what the bearer channel will hold. H.245 was designed to provide the information necessary to set up the bearer channels over RTP, and so takes the place of SIP's SDP. H.245 exchanges the codec and bearer capabilities of each endpoint, and is used to negotiate what bearer technology to use. This can be done in a manner that works for multiple-party calls, and in this way is useful with teleconferencing.

It is still possible to find softphones and open source technology that supports H.323, especially because of the videoconferencing aspect. However, voice mobility networks are unlikely to see much of H.323.