Vous êtes sur la page 1sur 13

High Capacity Hotspots Based on Bluetooth Technology

John Dunlop and Nathan Amanquah Mobile Communications Group University of Strathclyde Glasgow G1 1XW, Scotland Abstract Bluetooth is a short range wireless interface that offers data transmission rates of the order of 721 kb/s which is comparable with ADSL data rates, but which could be regarded as being too low for deployment as a limited range wireless access point. The capacity of a wireless access point based on Bluetooth technology can be increased significantly by co-locating a number of Bluetooth transducers in a hotspot scenario. This paper describes a technique which reduces mutual interference in co-located Bluetooth transducers by coordinating the hopping frequencies of the individual devices. Such a system can then provide attractive data transmission rates when deployed in wireless access point applications. 1 Introduction

The concept of service delivery over heterogeneous wireless networks is shown in Figure 1. In this case services may be provided via a 3G cellular link, via wireless hotspots, including infostation networks, or via a digital broadcast network (DAB and/or DVB). In fact services may be provided over a single network or over a combination of different (heterogeneous) networks including personal area networks (PAN), which is also indicated in this figure. To accommodate such service delivery options a concept known as the Personal Distributed Environment (PDE) has been introduced [1] in which the traditional user terminal is replaced by a distributed set of devices which may be local or remote to the user. Such a collection of devices will have a variety of radio interfaces available for connection to the core network including low power Bluetooth and its derivatives. Consequently there is growing interest in deploying short range wireless access points because they offer much higher throughput at reasonable cost than cellular counterparts. Wireless LAN access points based on 802.11 currently dominate this sector, but several scenarios can be envisaged where alternative access can be acquired via low range low power devices such as Bluetooth. In a Personal Distributed Environment, for example, a users devices are likely to be connected to the core network by means of several heterogeneous access technologies and access networks. This has the prospect of reducing service disruption and also facilitates cooperation between these networks for service delivery. Bluetooth is therefore a potential access point technology that could be combined with other technologies to provide a heterogeneous component for service delivery. However there are a number of problems associated with deployment of Bluetooth access points which must be addressed. The Bluetooth interface can support up to 7 clients per access point, which is limited compared to the number of potential clients that may be co-located, for example passengers in a railway carriage. Furthermore, an increasing number of concurrent users has a negative impact on the attainable throughput as the available capacity is shared between the number of connected clients. Capacity may be increased by co-locating access points, but as the Bluetooth radio interface is based on frequency hopping it is evident that interference will be present whenever two or more co-located Bluetooth devices choose the same hopping

-1-

frequency. This problem and a method of minimising its effect are the topics considered in this paper. 2 The Bluetooth Radio Interface

The basic unit of a Bluetooth system is a piconet, which consists of a master node and up to seven active slave nodes within a distance of approximately 10 metres. Figure 2 shows a scatternet consisting of two piconets, and an example of a slave device shared between two different piconets in the scatternet.. The node that initiates a connection, by sending either a PAGE or INQUIRY message, becomes the master of the piconet that is formed. Bluetooth devices operate in the ISM (Industrial, Scientific, and Medical) band at 2.4 GHz. Most of Europe and the USA have allocated the space between 2400 and 2483.5 MHz for this band and Bluetooth devices use 79 channels within this band, each occupying 1 MHz. Bluetooth uses a frequency hopping, time division duplexing (TDD) scheme for each channel, where the master device determines the frequency hopping scheme used and also sets the piconet clock. The frequency-hopping scheme is determined by a cyclic code of length 227 - 1, hopping at a nominal rate of 1600 hops per second. Transmissions are performed in 625 s slots, with a single packet being transmitted per slot; the frequency used for transmission does not change during the transmission of a packet. However, packets are allowed to be transmitted over multiple slots, allowing asymmetric connections. When a packet occupies multiple slots, the transmitting frequency stays the same, dropping the hopping rate below 1600 hops per second. The TDD mechanism is implemented by alternating the master and slave transmission slots, with the master transmitting in even-numbered slots, and slave(s) transmitting in odd numbered slots. In order to maintain this scheme, packets can last one, three, or five slots. A single Bluetooth unit may send/receive at a maximum data rate of 721 kb/s or a maximum of 3 dedicated voice channels of 64 kb/s each. Bluetooth has two link types, Synchronous Connection-Oriented (SCO) and Asynchronous Connection-Less (ACL) links. A SCO link is a symmetric, dedicated link between two devices. In effect, this is a circuit-switched connection, although the actual transmissions are based on a packet format. The ACL link is an asynchronous link that uses those slots in a piconet that are not dedicated to a SCO link, this is effectively a packet-switched connection. In ACL links, a slave is limited to transmitting to the master only in the slot directly after the slot where the master addressed this particular slave. The master controlled methodology, while it seems restrictive, can lead to an efficient use of the spectrum occupied by the piconet, since a single entity controls the usage of the channel and the probability of collisions between members of the piconet is zero. Also the dynamic nature of the master/slave architecture allows devices to establish themselves as masters in a piconet very easily. This dynamic behaviour provides devices that need to transmit information with the ability to establish a link and transmit its information when the application requires it. Broadcast messages to the whole piconet are possible using an ACL link. Another benefit of ACL links is that if the master has no information to send, and no polling is taking place, then the channel can be idle. In principle two SCO connections to an access point take up all available capacity. This indicates the need to deploy multiple access points in a single location if, for example, more than two audio sessions and other data sessions are to be supported. This paper considers the

-2-

transmissions in terms of packet length and thus the results presented are applicable to both SCO and ACL links. 3 The Coordinated Co-located Access Point Scheme

The co-located devices forming a Bluetooth access point are essentially the masters of the piconets which they form. The objective is to increase throughput in a hotspot whilst minimising interference between the co-located devices which form the access point. This may be achieved by synchronising the frequency hopping of the co-located devices which form the access point. This arrangement, is referred to as a Coordinated Co-located Access Point (CCAP) scheme. The basic principle is to establish a condition where no two devices use the same frequency at the same time, in a given location, and this can be achieved if the master nodes hop in tandem using the same hopping sequence, but separated in time by a constant offset, as shown in Figure 1. This is equivalent to applying a frequency offset between co-located devices which means that mutual interference can, in principle, be totally eliminated. An assessment of co-channel and adjacent channel interference in Bluetooth systems has been given by Sousissi [2], but this does not consider methods for interference reduction or elimination. Coexistence issues between Bluetooth and other ISM band devices has been considered by several authors [3,4,5], but this work concentrates on the potential for reducing interference within Bluetooth systems with co-located devices operating in realistic scenarios. 4 Coordination of Hop Frequency Selection

The frequency hopping sequence is unique for a Bluetooth piconet and is determined by two inputs to the Hop Frequency Selection Kernel (FHKernel) embedded in each device. These are the Bluetooth device address (BD_ADDR) of the master, and an input derived from the Bluetooth clock of the master which determines the position in the hopping sequence. The clock, which has a period of 312.5 s (0.5 slot duration), is applied to a 28 stage counter, which consequently has a cycle time of about 23.30 hrs. The input to the FHKernel is the 27 most significant bits or all 28 bits of the counter (depending on the hopping sequence or substate of the Bluetooth device), and least significant 28 bits of the 48 bit Bluetooth address. The requirement is to produce hopping sequences that will not mutually interfere, as illustrated in Figure 3. It would not be possible to generate the desired non interfering sets of hop sequences by supplying as input values to the FHKernel arbitrary sets of {address, clockcounter}. Two methods which have been considered to produce the appropriate hopping sequences for co-located devices. One is to use a single pair of values for {address, clockcounter}and then add multiples of a fixed frequency offset to each output. The other is to change the {address, clock-counter} input values by adding offsets to the clock counter value. 4.1 Use of a fixed offset In this case frequency offsets are added to the identical hop frequencies generated by {address, clock-counter}. It gives the distinct advantage that adjacent interference can be controlled, because the offsets can be added such that the frequencies chosen are fairly widely spaced. Another advantage is the ability to use an arbitrary Bluetooth device address as the Lead Master BD_ADDR (this is the BD_ADDR shared by all participants). The disadvantage of this method is that the resulting implementation will be incompatible with the large existing installed base of the Bluetooth wireless interface. In particular a new slave device

-3-

would not know what the specific offset is and additional signalling would be required to pass the value to the client 4.2 Change in the Frequency Hop Kernel Input Values In this case, the inputs to the FHKernel are unrelated (but not arbitrary) sets of {address, clock-counter} values. A common {address, clock-counter} value is applied, with clock offsets added to the clock-counter value. Thus input sets such as {addr1, clock-counter + 1}, {addr1, clock-counter + 2}, {addr1, clock-counter + 3}etc., are applied to the different participating master nodes. The conceptual model for the modified FHKernel is given in Figure 4. This method preserves compatibility with existing nodes. When a client terminal (slave node) establishes a connection with a master, it will correctly evaluate the next hop frequency because the master passes to the slave the {addr1, clock-counter} value it has itself used, as dictated by the Bluetooth standard. However it should be noted that arbitrary {addr1, clock-counter} values cannot be used and that specific offsets for the clock counter values have to be determined beforehand. It should be further noted that only the nodes used as access points (or used in coordinated piconets) need to be modified in this way. All client nodes can maintain their existing Bluetooth implementation. The preferred method of determining hop frequencies is thus based on adding offsets to the clock counter value. By using a coordinated hop frequency selection scheme, guarantees can be provided that mutual interference will not occur, since no two frequencies will be the same at any point in time. A universally available BD_ADDR of 0x00000 was selected as the address to be used in such high capacity hotspots. Using this parameter, a set of clock offsets was determined, from computer based iterations, to ensure that no co-channel interference would be experienced between any two co-located piconets using identical packet lengths. 5 Evaluation of the Coordinated Co-located Access Point Scheme

A simulator was constructed to evaluate the CCAP concept in terms of the probability of occurrence of interference, rather than in terms of modelling the Bluetooth radio channel itself, the latter having been evaluated in other studies [6]. In the simulations conducted each piconet consisted of a one master and one slave. It should be noted that a single slave is sufficient in this case as all master frequency hops will be generated (for both master-to-slave and slave-to-master transmissions), thus covering all possible frequencies that can be used in any piconet, irrespective of the number of slave nodes. Results for the CCAP scheme are compared to both asynchronous and synchronised uncoordinated scenarios. 5.1 Scenarios Simulated The effectiveness of the CCAP has been evaluated by considering possible scenarios that could occur in a practical Bluetooth access point deployment, in particular scenarios which would involve the use of different packet lengths in the co-located piconets. The scenarios simulated in the evaluation were: All nodes use the same packet lengths All master nodes use one packet length and all slave nodes use a different packet length An increasing proportion of master-slave pairs (piconets) are introduced using a packet assignment different from the other pairs.

-4-

The simulations were carried out assuming steady CONNECTION state conditions. There are however transient conditions that arise during the CONNECTION state. Any deviation of a piconet from using the assigned packet type (unless all piconets switch to using the new packet type) will result in a measure of interference. Such a change from using an assigned packet type will nevertheless occur from time to time in a CCAP scheme, and indeed in any piconet arrangement, because of LMP signalling which defaults to using single slot packets. Also during a session, some interference may still occur from nodes performing the PAGE or INQUIRY procedures as these employ different frequency hopping sequences for the different states of the Bluetooth node. The effects of having different hopping sequences on the CCAP scheme have been evaluated. 6.0 Performance Results In this section comparisons are drawn between the simulation results obtained for the CCAP scheme and an un-coordinated synchronised scheme and an analytical evaluation, in order to determine the relative improvement attainable. In Bluetooth transmissions packets can last one, three, or five slots and the use of all three is considered. An asynchronous scenario has not been simulated but it is known to suffer more interference than the synchronous case [7,8]. 6.1 Synchronous Uncoordinated Piconets The synchronous scenario assumes all packet transmissions begin and end at time slot boundaries. An analytical evaluation of interference when 1-slot packets only are in use is presented in [2], where throughput is evaluated at 100% duty cycle. In the case where all master and all slave devices transmit using the same packet length, with packet types longer than single slot it is possible for packets to start from different slot boundaries and thus may overlap. In this case it can be shown that the probability of successfully transmitting a t-slot packet in the presence of synchronous t-slot packet interferers is Pt(s) where
1 t 1 1 1 Pt ( s ) = (1 ) + (1 ) 2 h t h t
n 1

(1)

and h is the number of frequencies used (h = 79 or 23, depending on the Bluetooth system). 6.2 Asynchronous Piconets In this case neither synchronisation nor co-ordination exits between nodes and an asynchronous scenario represents the realistic deployment of co-located access points. In addition account must be taken of the fact that packets do not completely occupy the nominal one, three, or five slots. The durations of 1, 3 and 5 slot packets are 366, 1622 and 2870s respectively. Thus it is necessary to consider packet duration as although timeslots may overlap there are instances where a packet in a particular timeslot will finish before the packet in an overlapping timeslot actually starts. This is shown in Figure 5. The packet error rate for an asynchronous system with only single slot packet transmissions is presented in [7,8,9]. It can be shown [8] that the probability of successfully transmitting a single slot packet in the presence of unsynchronised single slot packet interferers is -5-

2(t t ) 1 2t t ) 1 P ( s ) = s d (1 ) + d s (1 ) 2 t ts h h s

n 1

(2)

where ts is slot length, td is duration of packet and h is the number of frequencies. In general, for co-located systems employing a k-slot packet type, where k {1,3,5} , it is the case that with a second interfering piconet also employing k-slot packets then over an entire session either 1 or 2 k-slot packets will always overlap the desired packet. Thus the probability of a successful (interference free) transmission is

2(kt s t dk ) 1 2t kt s ) 1 (1 ) + dk (1 ) 2 P( s) = kt s h kt s h where tdk is the duration of a k-slot packet.

n 1

(3)

The curves in Figure 6 and Figure 7 show the relative performance of co-located systems. The curves for CCAP have been extrapolated. These show the trends for DH1, DH3 and DH5 packet types defined in the Bluetooth standard (connectionless data packets without FEC).
6.3 CCAP Connection State

For a CCAP scheme, it should be noted that when all master-slave pairs use the same combinations of packet types, namely when all master nodes use a packet type A and all slaves use a packet type B, where A, B {1,3,5} , co-channel interference is completely avoided. This represents a very significant result. All possible combinations of master and slave packet assignments are possible, such that no interference occurs. The throughput is the maximum attainable (neglecting out of band interference). A consideration of the maximum throughput per piconet indicates that the synchronous system marginally outperforms the asynchronous system for all packet types. However, the CCAP shows a much larger improvement, and the percentage improvement increases as the number of co-located piconets increases. This point is illustrated in Table 1 which lists the improvements obtained with 9, 20, 30 and 40 piconets deployed under these conditions, using DH1 and DH5 packets. In terms of packet loss (due to interference), this is essentially zero for CCAP but increases with the number of co-located piconets for the other deployments. The maximum aggregate throughput for a number of co-located piconets with CCAP compared to synchronous and asynchronous deployments is given in Figure 8. Of particular interest here is the indication that the impact of interference on the CCAP system with increasing number of co-located piconets is zero, whereas the impact of such an increase on the other cases causes a maximum throughput to be reached. This is an important indicator of the advantage of implementing a CCAP scheme and it should be noted that the highest improvement is obtained with the use of the 5 slot packets, as such packets are far more susceptible to interference when this does occur. As might be expected a significant improvement is obtained with CCAP for colocated systems in the CONNECTION state.

-6-

Co-located piconets Packet type Async throughput (kb/s) Sync throughput (kb/s) CCAP throughput (kb/s) % increase over Async % increase over Sync

9 DH1 96.6 98.3 108.8 12.7 10.7 DH5 599.8 602.1 732.2 20.6 20.1

20 DH1 82.0 85.4 108.8 32.7 27.4 DH5 463.7 467.9 732.2 56.0 54.6

30 DH1 70.6 75.2 108.8 54.1 44.7 DH5 367.0 372.1 723.2 97.0 94.4

40 DH1 60.8 66.2 108.8 78.9 64.3 DH5 290.5 295.9 723.2 149.0 114.4

Table 1 Improvement of CCAP with DH1 and DH5 packets


6.4 Synchronous System with LMP Signalling

Simulation of LMP signalling is represented by a continuous transmission of single slot packets by both the master and slave nodes involved in the LMP signalling. Figure 9 illustrates the packet loss figures for 9 co-located piconets. When there are no nodes involved in signalling, or when all nodes are involved in signalling, the percentage loss is zero. (not indicated on this graph). In Figure 9 curves labelled DHx-DHy represent results for the CCAP scheme when the master transmits x-slot packets and the slave transmits y-slot packets in the reverse direction. From Figure 9 it appears that, depending on packet type, the packet loss the CCAP case may be higher, comparable or lower than in the uncoordinated case. However scrutiny of the performance reveals that even when the packet loss is relatively high the majority of piconets have low or zero packet loss but a small number (e.g. 2) of the access points may exhibit interference from signalling packets in a large percentage transmissions. It has been determined that this behaviour is related to the clock offsets used in the CCAP, and to which master-slave pair is involved in signalling. It has been observed that when a wider separation between clock offsets is in place, a much lower packet loss is obtained, and the rate at which such mutual interference will occurs during signalling (if at all) is reduced to a random event. This is illustrated in Figure 10, where the simulation has been repeated for DH1 and DH3 combinations. The percentage of packet loss is now reduced to less than 2.5%. For the sake of comparison, the results for a CCAP with 6 co-located piconets are also presented. The scenario investigated represents the worst case when there is a continuous presence of at least one master-slave pair is engaged in LMP signalling. Typically in a particular session, LMP commands will constitute only a small percentage of the total number of transmitted packets.
6.5 Inquiry and Paging States

Interference is also possible within a CCAP system when nodes are in an inquiry or paging. state. The worst case scenario has been considered where a node continually transmits the sequence of frequencies associated with an inquiry or paging message. The impact of this interference on the CCAP scheme is illustrated in Figure 11 and it is evident that transmissions of longer packet sizes are affected most. As packet lengths increase, this increases the probability that an inquiry packet will be encountered using the same transmission frequency as frequencies used in the inquiry state are not under the control of the CCAP system. However, the impact of this interference on nodes in a co-located AP can be reduced if the least loaded node of the AP, in the CCAP scheme, is required to enter the inquiry or page scan state frequently, in order to respond to potential clients in the shortest possible time. This will have a net effect of shortening the duration over which any potential interference might occur. -7-

For comparison the effect of inquiry packets on a deployment of uncoordinated synchronous piconets is shown in Figure 12. It should be noted that there is a much higher incidence of interference compared to the CCAP scheme. And that is because there is interference caused by the nodes of the AP themselves, in addition to the interference produced by nodes in the inquiry state.
6.6 Platform Specific Benefits of the CCAP Scheme

The previous results presented show the advantage of the CCAP scheme, with regard to interference reduction, under various conditions. In practice there would be an additional advantage in terms of connection time and session set-up time because the known 0x0000 address can be paged directly, instead of a lengthy inquiry followed by the page procedure. Therefore the CCAP concept may be useful in the support of hand-off in the situation where the location of the nodes of an AP are actually distributed. In such circumstances the use of the CCAP scheme can decrease handover times considerably.
7 Conclusions

Interference reduction may be regarded as a significant component in a QoS management scheme. This paper has presented a technique for minimising co-channel interference in co-located Bluetooth piconets. The technique operates by coordinating hop frequency selection between co-located nodes which form the AP itself. Under these conditions, in the CONNECTION state, no co-channel interference occurs when all master nodes use packets of the same length, and all slave nodes also use packets of the same length. The packet length used by the slave nodes does not have to be the same as that used by the master nodes. Results have shown that some interference can occur when LMP signalling takes place. However it has been demonstrated that, in such circumstances, by using clock offsets that are not closely spaced (e.g. offsets of 4, 6 or greater) less than in an un-coordinated synchronous arrangement. Even when this small amount of interference is present, the CCAP scheme outperforms a conventional Bluetooth deployment. The overall conclusion is that there can be a significant gain in capacity and throughput when Bluetooth networks are operated as a CCAP, rather than being operated independently. Further, a high capacity hotspot may be deployed using Bluetooth technology by co-locating multiple access point nodes. An improvement of between 20% and 97% can be obtained when a number of nodes between 9 to 30 nodes employing DH5 packets are co-located using the CCAP scheme.
Acnowledgements

The work reported in this paper formed part of the Core 2 Research Programme of the UK Virtual Centre of Excellence in Mobile & Personal Communications, Mobile VCE, www.mobilevce.com whose funding support is gratefully acknowledged. More detailed technical reports on this research are available to Industrial Members of Mobile VCE.

-8-

References

1)

2)

3)

4)

5)

6)

7) 8)

9)

Dunlop J, Atkinson R, Irvine J and Pearce D A Personal Distributed Environment for Future Mobile Systems, Proceedings of the IST Mobile Summit, Aveiro, Portugal, June 2003. Souissi S, and Meihofer E F Performance Evaluation of a Bluetooth Network in the Presence of Adjacent and Co-Channel Interference, IEEE Emerging Technologies Symposium: Broadband, Wireless Internet Access, pp. 6-11, April 2000. N. Golmie N, Dyck R E V, and Soltanian A, Interference of Bluetooth and IEEE 802.11 Simulation Modelling and Performance Evaluation, Proceedings of the ACM International Workshop on Modelling, Analysis, and Simulation of Wireless and Mobile Systems, Rome, Italy, July 2001. Batra A, Ho J M and Anim-Appiah K, Proposal for Intelligent Bluetooth Frequency Hopping for Enhanced Coexistence, IEEE P802.15 Working Group for Wireless Personal Area Networks Coexistence Task Group Technical. Report, IEEE 802.15-01/082, January 2001. Punnoose R I, Tseng R S and Stancil D D, Experimental results for interference Between Bluetooth and IEEE 802.11b DSSS systems, Proc. of 54th IEEE Vehicular Technology Conference Fall 2001 , vol. 1, pp. 67-71, 2001. Cordeiro C, Sadok D and Agrawal D Piconet Interference Modelling and Performance Evaluation of the Bluetooth MAC Protocol Proc. of IEEE Global Telecommunications Conference GLOBECOM2001, San Antonio, USA, 2001. El-Hoiydi A Interference Between Bluetooth Networks - Upper Bound on the Packet Error Rate, IEEE Communications Letters, vol 5, no 6, pp245-247, June 2001 Decotignie D and El-Hoiydi A Soft Deadline Bounds for Two-way transactions in Bluetooth Piconets Under Co-channel Interference Proceedings of Emerging Technologies and Factory automation Conference, France, 2001. Sevillano J, Daz F, Jimnez G, Cascado D and Civit_Balcells A An Analytical Model of Inter-channel Interference in Bluetooth-Based Systems, Proceedings of the 4th IEEE International Workshop on Mobile and Wireless Communication Networks, pp 384-388, Stockholm, Sweden, 2002.

-9-

Figures
FN = Forwarding Node

Internet

3G Cell BS DxB Coverage Area

WLAN AP PAN 3G Cell FN DxB Transmitter

WLAN AP FN

low tier sensors

Figure 1 Service Delivery Over Heterogeneous Wireless Networks

Scatternet
Slave
Piconet A

Slave
Piconet B

Master Slave Slave

Master

Figure 2 Example of a Bluetooth Scatternet

Co-located Bluetooth device

fa

fb

fc

fd

fe

ff

fg

fh

fi

fj

fk

time offset n+1 fx fy fz fa fb fc fd fe ff fg fh

time offset n+2 fu fv fw fx fy fz fa fb fc fd fe

hopping sequence

Time

Figure 3: Fixed hopping sequence with time delay between devices

- 10 -

Clock counter offset

Shared address

FHKernel Block
Clock count (native) Clock count (corrected)

Frequency

Figure 4 Conceptual Model for the modified FHKernel

td Desired packet 2-packet overlap 1-packet overlap ts ts

Figure 5 Packet overlap in asynchronous piconets


CCAP DH1 CCAP DH3

Sync DH1 Sync DH3 800

Sync DH5 Async DH1

Async DH3 Async DH5

CCAP DH5

Throughput (kb/s)

600

400

200

0 0

10

20

30

40

Number of Piconets

Figure 6 Comparison of throughput for CCAP, synchronous and asynchronous scenarios

- 11 -

Sync DH1 Sync DH3


0.6

Sync DH5 Async DH1

Async DH3 Async DH5

CCAP

Probability of Loss

0.4

0.2

10

20

30

40

Number of Piconets

Figure 7 Comparison of packet loss for CCAP, synchronous and asynchronous scenarios

Sync DH1 Async DH5 30

Sync DH3 CCAP DH1

Sync DH5 CCAP DH3

Async DH1 CCAP DH5

Async DH3

Throughput (Mb/s)

20

10

0 10 20 30 40

Number of Piconets

Figure 8 Comparison of aggregate throughput (connection state)

DH1-DH3 30

DH3-DH3

DH5-DH5

DH1-DH3c

DH3-DH3c

DH5-DH5c

Percentage of Packets lost

20

10

0 2 4 6 8

Number of piconets engaged in signalling

Figure 9 Comparison of packet loss in CCAP and uncoordinated piconets

- 12 -

Widely spaced 30

Closely spaced

Probability of packet loss

20

10

0 2 4 6

Number of nodes signalling

Figure 10 Comparison of the effect of offset value on packet loss (DH3 packets)

DH1-DH3 DH1-DH5
45

DH3-DH1 DH3-DH3

DH3-DH5 DH5-DH1

DH5-DH3 DH5-DH5

Percentage of packet loss

30

15

0 2 4 6 8

Number of Inquiry signallers

Figure 11 Impact of inquiry procedure on CCAP

DH1-DH3 DH1-DH5 80

DH3-DH1 DH3-DH3

DH3-DH5 DH5-DH1

DH5-DH3 DH5-DH5

60

Probability of Loss

40

20

0 2 4 6 8

Number engaged in Inquiry

Figure 12 Impact of inquiry procedure

- 13 -

Vous aimerez peut-être aussi