<<

. 14
( 87 .)



>>

secondary problems (such as TCP instability and unfairness between nodes), reducing the
effectiveness of the proposed routing protocols.


2.3 WIRELESS PAN TECHNOLOGIES

Wireless Personal Area Networks (WPANs) are short to very short range (less than 10 me-
ters) wireless networks covering the immediate surroundings of individuals. WPAN tech-
nologies are not (and should not be) considered to be contenders of WLAN technologies,
but are destined to complement WLANs. The market segment of WPANs is different from
that of WLANs; not only is the required range shorter but the required service levels are
also different. A PAN is the next wireless networking paradigm in the ordered list of
WAN-MAN-LAN paradigms. To enable the embedding of WPAN technologies into gen-
eral, low-cost devices, theses technologies have to have small footprints, very low costs,
and relaxed power requirements. WPAN technology can be used, for example, to intercon-
nect portable computers/digital assistants and their peripherals, to connect sensors/actua-
tors, to connect devices worn by individuals establishing personal operating spaces
(POS), or to connect devices in cars without the need for cabling. Cost effectiveness is the
major keyword that one should associate with WPANs.


2.3.1 Short History
The term personal area network was forged and its standardization started by the estab-
lishment of an “Ad Hoc Group” within the IEEE Portable Applications Standards
Committee (PASC). In 1998, a Study Group inside the 802.11 Working Group was
formed to develop a project authorization request. In March 1999, the 802.15 Working
Group was established. Meanwhile, industrial interest groups were formed throughout
the world to address the same low-range, low-power, low-cost networking needs. The
HomeRF working group/consortium was formed in March 1998, focusing on the home
environment”a larger domain than personal area but smaller than local area, with needs
similar to PANs. The Bluetooth Special Interest Group (SIG) was formed in May 1998
with the goal of defining an industry standard to replace short-range data cables.
Bluetooth took the same route as the IEEE WPAN working group (strong overlap in in-
terested parties), overtaking the IEEE efforts, whereas HomeRF was getting more and
more away from WPAN.
The first publicly released version of the Bluetooth specification of the Bluetooth
SIG became available in the fourth quarter of 1999 but, due to disturbing imperfections,
a new version was released in February 2001. Meanwhile, the IEEE802.15 working
group had formed four Task Groups and a Study Group for different WPAN require-
ments. Task Group 1 (805.15.1) adopted the bottom layers of the Bluetooth specification
in June 2002, whereas Task Groups 2, 3, and 4 and the Study Group are concentrating
on coexistence with WLANs, and high-rate, low-rate, and alternative-high-rate versions
of the standard.
60 OFF-THE-SHELF ENABLERS OF AD HOC NETWORKS


2.3.2 Bluetooth Technological Overview
The Bluetooth SIG was formed in May 1998 by the so-called promoter companies, con-
sisting of Ericsson, IBM, Intel, Nokia, and Toshiba, and later on 3Com, Lucent, Mi-
crosoft, and Motorola. The SIG also contains associate members; participating entities
pay membership fees and, in turn, can vote or propose modifications for the specifications
to come. Adopter companies can join the SIG for free but can only access the oncoming
specifications if these have reached a given evolutional level.
The name Bluetooth supposedly comes from a Scandinavian history-enthusiast engi-
neer involved in the early stages of developing and researching this short-range technolo-
gy, and the name stuck; nobody being able to propose a better one. Bluetooth was the
nickname for Harold Blåtand”“Bluetooth,””King of Denmark (940“985 A.D.). Blue-
tooth conquered both Norway and Denmark, uniting the Danes and converting them to
Christianity. One of the major goals of the Bluetooth standard is to unite the “communica-
tion worlds” of devices, computers, and peripherals and to convert “the wired” into wire-
less; thus, the analogy.
The protocol stack of Bluetooth is depicted in Figure 2.3. Bluetooth is designed so that
a single chip can implement the bottom three layers with a serial (RS-232, USB, or simi-
lar) interface connecting the chip to the controller host through the so-called HCI (Host
Controller Interface).

2.3.2.1 The RF Layer. The physical or RF Layer (Radio Frequency) of Bluetooth is
built on a synchronous fast-frequency-hopping paradigm with a symbol rate of 1 Mbps
operating in the publicly available 2.4 GHz ISM band. In a normal operation mode, Blue-
tooth units will change the carrier frequency (hop) 1600 times a second over 79 different
carrier frequencies separated 1 MHz apart, starting with 2.402 GHz. (Since the 2.4 GHz
ISM band is not equally available in all countries, e.g., France and Spain, Bluetooth en-
ables the operation on a reduced band with only 23 different carrier frequencies.) The
modulation scheme employed is similar to that of GSM, that is, GFSK (Gaussian Fre-



Applications

Profiles

SDP RFCOMM Telephony

L2CAP
HCI Host
UART, USB
HCI Client
IEEE
Audio Link Manager 802.15.1
ACL
SCO
Baseband

RF


Figure 2.3. Simplified Bluetooth protocol stack.
61
2.3 WIRELESS PAN TECHNOLOGIES


quency Shit Keying). According to the transmitted power, Bluetooth devices can be classi-
fied into different power classes from 20 dBm to 0 dBm transmission power. Class-3 de-
vices are the most common, transmitting with 0 dBm, and not requiring external power
amplification or power control; thus, they can be integrated on a single chip.

2.3.2.2 The Baseband Layer. The Baseband layer is in charge of controlling the RF
layer and providing the communications structure to the higher layers, thus taking on the
functions of the MAC sublayer of the OSI-7. The basic communication structure provided
by Bluetooth is a point-to-point link between two devices, each of them hopping along the
same pseudorandom sequence of frequency carriers. In order for the two nodes to agree
on the hopping sequence and on the control of the channel, one of the nodes will assume
the role of master while the other becomes a slave. (Nodes do not have to be different in
their capabilities, the master/slave roles are logical roles in the point-to-point communica-
tion link.) A point-to-multipoint Piconet can be established by a single master controlling
the channel for several slaves (the point-to-point communication structure in general is
also called a Piconet). A Piconet only has one master and can have several slaves hopping
along the pseudorandom sequence of the master of the Piconet, with a maximum channel
capacity of 1 Mbps shared by the members of the Piconet. As mentioned earlier, a func-
tioning Piconet makes 1600 hops in a second, thus having 1600 slots (each 625 s long)
in one second. In an odd-numbered slot, only the master of the Piconet is allowed access
(with a few exemptions); whereas, in an even-numbered slot, a slave that was polled in the
previous slot can gain access to the channel. To enable Bluetooth devices to tune to the
new frequency carrier and change their mode from reception to transmission, a 220 s
guard time is set aside at the end of transmission slots, thus reducing the goodput. Nodes
are also enabled to transmit during not only one but three and five slots, using different
packet types (with no hopping while in transmission) to increase efficiency by reducing
the “effective usage time“guard time” ratio. The effective data rate in a Piconet can be de-
termined according to the packet types and lies anywhere between 216 kbps and 780 kbps
per Piconet. Several Piconets can operate in the same space independently without caus-
ing a significant interference among each other, since all these Piconets will hop accord-
ing to different hopping sequences. The probability of interference between independent
Piconets grows by the number of Piconets covering the same area. It is also worth noting
that the 2.4 GHz band is also utilized by other (interfering) technologies such as
IEEE802.11b and microwave ovens.
There are two different types of classifications for the virtual links between nodes in a
Piconet: a link can be Synchronous Connection Oriented (SCO) or Asynchronous Con-
nectionless (ACL). If an SCO link is established between two nodes of a Piconet, then
slots are reserved at fixed intervals for the master and one of its slaves in the Piconet, en-
suring a deterministic assignment of slots to the traffic. SCO links provide a voice-type
quality of service provisioning, indeed designed for voice transmissions. ACL links, on
the other hand, are in sole control of the master polling the slaves in the order the master
desires. Slots assigned to SCO links have priority over ACL links as well as priority over
any other task a master may be performing (e.g., inquiring or paging).
As mentioned earlier, the basic communication structure of Bluetooth is a Piconet;
thus, Piconets need to be established over Bluetooth devices before they can exchange
data or communicate. The Piconet establishment process is a three-step process including
device discovery (or inquiry, in Bluetooth terms), device attachment (or paging, in Blue-
tooth terms), and Piconet parameter negotiations.
62 OFF-THE-SHELF ENABLERS OF AD HOC NETWORKS


During the inquiry process, the common objective of Bluetooth nodes is to discover
each other™s presence with some of the nodes listening or scanning the (reduced set) of
hopping frequencies while other nodes constantly transmit very short so-called ID pack-
ets. Since inquiry ID packets are extremely short and represent a unique bit pattern, the
number of hops can be increased to 3200 hops per second to reduce discovery times. If a
scanning node overhears an ID packet for the first time, it will refrain from replying im-
mediately but will wait a random (back-off) period of time to reduce the collision proba-
bility of scanning nodes replying to the same ID packet. When finished with the backlog-
ging, nodes return to the inquiry scan state, and, if they overhear another ID, packet they
will respond to the transmitter of that ID packet in exactly 625 s. The inquiring nodes
send two ID packets at two different frequencies and then listen to the corresponding re-
ply frequencies for the next 625 s if reply is received. The inquiring node will be aware
of the proximity and the identity of the scanning node.
The paging process can start if there are devices that are aware of the identities oth-
er devices in their proximity, most likely after a successful inquiry. Just like with the in-
quiry process, the frequency of the hopping is increased to 3200 and devices can be ei-
ther in a page scan or page mode. By definition, the node that initiates the paging (the
node in the page mode) will become the master of the Piconet, whereas the node that
was successfully paged will become the slave. The device in the paging mode will trans-
mit an ID packet with the address of the device it has discovered before. If the device
whose ID is transmitted is in the page scan mode and overhears the ID packet with its
own address, then it will respond to this “page” with the same ID packet. Note that the
paging node knows the identity of the paged device but not necessarily vice versa; thus,
the paging node that received a reply from the paged node will send an identification
packet with its own parameters to the paged node (the latter responding with another ID
packet). By the time this four-way handshake is executed, the slave (paged node) has
enough information to calculate the master node™s pseudorandom hopping sequence so
both the nodes can start using the hopping sequence of the master, establishing a con-
nection.
Reaching the connection state, the master will poll the slave to verify that the slave has
entered the Piconet. The third phase of the connection establishment is initiated by the
Link Manager layer to set up a control ACL link.
A Piconet can consist of a maximum of eight active nodes: a master and seven active
slaves. This is due to three-bit node addressing inside Piconets. Yet, a Piconet can con-
sist of much more devices in an inactive mode; indeed, the number of nonactive slave
devices in a Piconet is not constrained. Other than being actively participating in a
Piconet, slaves can go or be put into three different power saving modes: Sniff, Hold,
and Park. A slave in Sniff mode will not listen to the channel in every odd time slot, but
will negotiate a parameter with the master for periodic small time windows during
which it will wake up and check whether the master wants to transmit to it. The Sniff
mode can be used to reduce power consumption of rarely active nodes. In the Hold
mode (just like in the Sniff mode), a slave still does not give up its three-bit active-ad-
dress but will not be able to receive any ACL packets for a negotiated period of time.
The Hold mode may be used to perform inquiry and scanning operations while being
connected to a Piconet and to enable the participation of nodes in more than one
Piconet, as outlined later. Finally, slaves in the Park mode give up the three-bit active ad-
dress but will remain synchronized to the master by listening to the channel during so-
called Beacon intervals. If a master wants to wake up a parked slave, it will have to wait
63
2.3 WIRELESS PAN TECHNOLOGIES


for the negotiated Beacon window and address the slave to be awaked with the device
address or parked address. Parked slaves will also receive an opportunity during the
Beacon window to inform the master that they need to be woken up.
Although the main communication unit in Bluetooth is a point-to-multipoint Piconet,
the specification allows nodes to participate in more than one Piconet semisimultaneously
(note that a node can be a master in only one Piconet), switching between its roles of the
different Piconets acting as bridges between Piconets, likely using the Hold mode to
schedule between the several Piconets. Two or more overlapping Piconets interconnected
with bridges in such manner form a Scatternet. Although a Piconet™s topology is a star-
shaped point-to-multipoint structure with only a single link between a master and any of
its slaves (single-hop), a Scatternet can represent any type of the possible topologies and,
thus, can be used to establish a multihop or ad hoc network (a possible Scatternet is de-
picted in Figure 2.4). Other than describing the possibility of forming Scatternets, the
Bluetooth specification does not address how Scatternets or ad hoc networks should be
established; it solely provides the possibility to employ Bluetooth as the basis for ad hoc
networking.

2.3.2.3 Link Manager. The Link Manager (LM) layer of Bluetooth fulfils part of the
functionality of the Logical Link Control sublayer of the OSI-7 architecture. The main
functions of the LM are: Piconet management, link configuration, and providing security,
that is, authentication and encryption. Right after a slave has been put into a Connection
mode, an ACL link is established between master and slave to manage the Piconet. Man-
agement functions include the attachment and detachment of slaves, negotiating piconet
parameters, a possible change in the roles (when a slave becomes the new master of the
Piconet), the establishment of SCO or ACL links, and the handling of the low-power
modes. The management functions are based on a request“response communication
scheme between the master and the slave, whereby the master requests some parameter to
be changed and the slave either accepts it or challenges it.
The link configuration tasks consist of (i) quality of service negotiations, whereby the
maximum polling time is negotiated in a request“response manner and broadcast parame-
ters are set up; (ii) negotiation of power-control parameters; (iii) negotiation of accepted
packet types at both sides, with determination of whether multislot packets will be al-
lowed.


master/slave bridge
slave/slave bridge




master-slave relation:


Piconets

Figure 2.4. A Bluetooth Scatternet consisting of three Piconets.
64 OFF-THE-SHELF ENABLERS OF AD HOC NETWORKS


The security goals include (i) optional authentication to only allow devices that are
known or trusted to connect, and (ii) encryption to prevent eavesdropping by a third party
on the channel. Authentication is based on a common link key, whereby the verifier chal-
lenges the claimant to compute an answer that can only be computed by knowing the link
key. In order to distribute the link key, nodes go through a process called pairing. During
pairing, a link key is formed from a PIN code, a random number, and the claimants ad-
dress. For encryption of data, an encryption key length is negotiated between master and
slave and an encryption key is created using the same algorithm at both sides and the link
key.

2.3.2.4 Logical Link Control and Adaptation Protocol Layer. The Logical
Link Control and Adaptation Protocol Layer (L2CAP) is the other subprotocol of the Log-
ical Link Control sublayer of the OSI-7 protocol stack. The goals of L2CAP are to enable
several higher-layer protocols to transmit their protocol data units (PDU) over ACL links
(protocol multiplexing), segmentation and reassembly of higher layer PDUs into Base-
band packets, and quality of service negotiations for individual ACL links for the higher-

<<

. 14
( 87 .)



>>