GB2399982A - Adaptive frequency-hopping - Google Patents

Abstract

A method of selecting frequency channels for use by a frequency hopping communication device, from a set of channels, in the presence of interference from an interferer, comprises operating the communications device to select a first set of channels to collide with said interference; and then selecting a second set of channels for use by said communications device excluding said first set of channels. This facilitates rapid identification of the interferer.

Description

M&C Folio: GBP86968 Document: 9016g7 Communications Systems and Methods

This invention is generally concerned with apparatus, methods and processor control code for communications systems, particularly for adaptive frequency hopping techniques.

For convenience embodiments of the invention will be described with reference to Bluetooth (trademark) but the skilled person will understand that the applications of the invention are not limited to this and related standards but encompass other frequency hopping communication systems.

The Bluetooth group of standards is concerned with short range (up to around I Om) RF transmission as a replacement for cables and, for example, in personal area networks (PANs). The basic standard provides a frequency hopping spread spectrum (FHSS) link operating, for version 1.2 of the standard, at lOMbps. Bluetooth is primarily a European-initiated standard, and in the USA the IEEE 802.15 series of standards provides a convergent and a planned equivalent set of standards. Thus IEEE802.15.1 is almost identical to Bluetooth 1.1, 802.15.3 is a higher data rate variant, and 802.15.4 is a lower data rate variant (250kbps) intended for very low power control- type applications. Coexistence with IEEE802.11, discussed further below, is addressed by the 802.15.2 working group and thus IEEE802.15.2 relates to adaptive frequency hopping.

Figure la shows an example of a Bluetooth concept in which a computer 10, printer 12, and camera 14 are all in communication with one another by means of bidirectional Bluetooth radio links 16. Bluetooth can also be used for wireless connection to high speed voice/data access points.

The current version of Bluetooth in the public domain is version 1.1. This hops at a rate of 1600 hops per second (except during paging and enquiry when it operates at twice this speed) and thus a transmitter dwells at each carrier frequency or 625s (including guard and ramp timesthe actual dwell is approximately 4001ls). One packet is transmitted during each 625,us time slot.

Bluetooth has 79 lMhz wide channels in the 2.4Ghz ISM (Industrial Scientific and Medical) band and in Bluetooth version 1.1 the hopping scheme transmits equally across all these selecting a channel according to a predefined pseudo-random hop sequence. The hop sequence comprises 227 hops, sufficient for a whole day, without repetition, and is selected on the basis of a device address (BD_ADDR) of a master device in a Bluetooth link for network. The master's clock, to which all slave devices are synchronized, is then used to increment through this hopping sequence. The hopping sequences are devised such that all carriers are visited equally and consecutive hops cover a large frequency range so that, for example, a hop to an adjacent channel is less likely than a hop to a more distant part of the band. Figure lb illustrates in diagrammatic form a portion of the Bluetooth spectrum showing Bluetooth channels 20.

Bluetooth has to coexist with other technologies in the 2.4Ghz ISM band including baby monitors, microwave ovens, cordless telephones, and, especially IEEE 802.11/802.11b wireless local area networks (WLAN) systems, 802.1 lb often being known as "WiFi".

IEEE 802.11 and IEEE 802.11b both operate in the 2.4Ghz ISM band, the former using either FHSS or DSSS (Direct Sequence Spread Spectrum), the latter providing a higher data rate and using only DSSS. Adaptive frequency hopping (AFH) has been proposed both Bluetooth and IEEE 802.15, as a technique to reduce mutual interference between wide band systems, in particular 802.11 and Bluetooth. To all intent and purposes the evolving AFH standards of Bluetooth and IEEE 802.15 are the same and the techniques described herein are applicable to both these sets of standards.

In figure lb curve 22 diagrammatically illustrates an 802.1 lb signal. This overlaps the B1uetooth channels and has more power in the centre of its band than at the edges.

Interference between Bluetooth and 802.11 (b) is potentially a particular problem since it is envisaged that Bluetooth will be used, for example, to implement a wireless mouse and/or keyboard for a computer which may itself have a WiFi link to an access point.

Thus Bluetooth and WiFi systems are often likely to be co-located, potentially causing reduced throughput or both. One proposed solution to this problem is to time division multiplex Bluetooth and 802.11 transmissions, but this involves synchronising the two protocols. AFH instead works by adapting the hop set to avoid "bad" frequencies, that is those which are subject to damaging interference levels, and therefore does not rely upon accurate time synchronization. Furthermore by hopping around the interfering signals Bluetooth not only improves its own throughput, but also improves the throughput of neighbouring devices sharing the frequency band.

To preserve existing designs and silicon layouts for version 1.1 of Bluetooth, version 1.2, which includes AFH, should preferably not modify the existing (version 1.1) hopping kernel, thus implying that the adaptation for AFH should be performed using a I post-processing transform. Figure 2a shows a suitable AFH kernel 200 comprising a legacy hop selection kernel 202 followed an AFH post-processor 204 to perform frequency remapping, substituting "bad" frequencies that are output from kernel 202 or "good" ones, although preferably preserving the fundamental randomness of the original hopping sequence. In the kernel 200 of figure 2a hop selection kernel 202 operates on an input 204 comprising address or identity for the master device of the link or network to select a pseudo random sequence which is then stepped through in synchronization with a clock signal 206 of the master device. An input S 208 to post processing module 204 comprises the set of good frequencies, and output line 210 provides these in a pseudo random sequence.

Figure 2b shows the structure and function of AFH kernel 200 of figure 2a in more detail. Thus the remapping function is performed by block 204 using a mapping table 204a responsive to an input AFH channel map 208a and to the output of hop selection unit 202. Figure 2b incorporates functional aspects of the kernel illustrating that, for the Kth frequency fk the kernel determines 212 whether fk is in the set of used carriers and if not performs the remapping function 204 to output a signal 214 to use an alternative frequency fk 'prime instead of fk for the next channel slot, and otherwise outputting a signal 216 to use fk for the next slot.

Figure 3 shows an example of an AFH hop sequence remapping from an original sequence 300 to a reduced sequence 302. In this example channels "f" and "c" have been identified as bad and are therefore remapped, although the remapping is not fixed, in this example "f'' teeing variously remapped to "e", "g", and "b". A piconet is an adhoc Bluetooth network comprising one master and one or more slaves. The adaptive hopping piconet-wide, that is the entire piconet uses the same adaptive hopping sequence. The master device distributes updated channel information, that is any changes to the adaptive hop set such as the re- addition of"reformed" channels (that is channels that have ceased to be bad) and the rejection of ad channels. This information is communicated (reliably) by means of commands over the Link Manager Protocol (LMP) to all members of the piconet before the new set of frequencies is used (to prevent members getting lost). The changeover is referred to as the Hop Sequence Switch Instance (HSSI).

Various methods may be employed to identify good and bad channels for channel classification, including active, passive and other methods. One active method is to send data packets through a channel to see if they are received correctly; a passive method comprises, broadly speaking, tuning a receiver to a channel and listening to see if any interference is detected; alternatively external classification information signalled by a host may be employed. For example if the Bluetooth device is in a laptop also using 802.11, the laptop can tell the Bluetooth device what 802.11 frequencies are being used by the WEAN card. Channel measurements/classification may be made by any device, slaves as well as the master. Where slaves make a measurement the master may enable channel classification reporting from a slave, the slaves then periodically sending their measurement results back to the master. Whether the master or one or more slaves do the measuring, the master determines which channels are to be used.

The identification of interferers suffers from a number of difficulties and available techniques have drawbacks. An interferer may be identified by periodically performing an RSSI (Received Signal Strength Indication) measurement of each channel, but this disrupts the operation of the system and requires relatively expensive additional hardware. Another more straightforward approach is to observe which channels have high packet error or packet loss rates or high averages of these, which can optionally be enhanced using RSSI measurements taken at the time of packet reception. It will be understood since many packet errors are correctable a packet loss rate is lower than a packet error rate.

Packet-loss based observation techniques have two significant flaws. The more important of these is the length of time taken to detect an interferer when having to rely upon stumbling across the interferer at random. Furthermore during this discovery and evaluation phase there is a consequential loss in throughput resulting from the packet losses suffered. Alternative approaches are described in US 6,240,125, US 6,009, 332 and US application US2002 075,941, the latter involving cooperation between cohabiting systems to divide up the frequency band, which may suffer from regulatory problems. A paper "The Analysis of Coexistence Mechanism of Bluetooth" by B. Zhen, Y. Kim and K. Jang in Vehicular Technology Conference 2002, pages 419-423, vol l, VTC-9 May 2002, Birmingham AL, USA suggests using RSSI measurement techniques for improved Bluetooth AFH.

There therefore exists a need for improved methods of identifying interferers in the context of an adaptive frequency hopping communications systems, with the aim of then avoiding them. There further exists a need for more rapid identification of such interferers.

According to a first aspect of the present invention is therefore provided a method of selecting frequency channels for use by a frequency hopping communications device, from a set of channels, in the presence of interference from an interferer, the method comprising: operating said communications device to select a first set of channels to collide with said interference; and then selecting a second set of channels for use by said communications device excluding said first set of channels.

It is not essential to know the characteristics of interference to select the first set of channels to collide with this but if the main or most likely frequency bands of the interference are known this will facilitate such selection. Preferably the method further comprises a second stage during which the performance of the second set of channels is monitored and used to construct a third set of channels, based upon the second set of channels, and taking account of a prediction about the potential locations of other interfered channels made in response to the results of the monitoring. Such predicting is preferably based upon one or more known characteristics of the interference, and may comprise identifying candidate channels for removal from the second set of channels, either immediately or after further confirmation. In the latter case a candidate channel may be penalised responsive to the monitoring, and removed when a threshold penalty value is reached.

The selecting of the first set of channels may, likewise, be made by selecting an initial set of channels, monitoring these, and selecting the first set of channels responsive to the results of the monitoring. The monitoring may comprise RSSI-based or packet error or packet loss-based monitoring or some other form of quality monitoring. In a preferred embodiment the frequency hopping communications device comprises a Bluetooth version 1.2 or IEEE 802.15 compatible device, and in this case the method may be used to reduce the likelihood of interference from an IEEE 802.11 or TEEE 802.1 l b compatible device.

In another aspect the invention provides a method of selecting frequency channels for use by a frequency hopping communications device from a set of channels, in the presence of interference from one or more interferers each having a defined frequency band; the method comprising: selecting an initial predetermined subset of said channels, said predetermined subset including at least one channel in each said frequency band; monitoring a received signal quality measure for each of said predetermined channel to identify ones of said channels subject to interference; and selecting a second subset of frequency channels for use to avoid said channels identified as subject to interference.

The communications device may comprise a transmitter monitoring received signal quality, for example, by receiving reports from a receiver, for example via a back channel, or the communications device may comprise a transceiver and itself monitor received signal quality. As mentioned above any of a range of quality measures may be employed. Preferably, however, received signal quality is monitored over a period of time, that is for a plurality of received packets, to take account of interferers having "bursty" characteristics, as is the case for typical WEAN traffic. The initial predetermined subset of channels make comprise, for example, a set of channels programmed to collide with 802.11 or 802.1 lb frequencies, more particularly band centre frequencies.

In a preferred embodiment the initial subset of channels that is modified to, in effect, home in on the interference by monitoring which of the initial set of channels are subject to interference. Thus, one or more channels from the initial subset, which are determined to be 'good' may be removed to allow more time to be spent upon the suspect channels, thus generating data on these channels more rapidly. The initial subset of channels may further be modified to add one or more channels where interference is expected. For example, where a good channel is found at one end of one of a set of frequency bands known to be potentially occupied by interferers, one or more channels may be added at the other end of the frequency band to check for interference from an adjacent band. The modification of the initial subset of channels may continue until a pre-determined number of contiguous channels subject to inference is identified.

For example, this may comprise all the channels within a bandwidth within a portion of a known frequency band available for use by an interferer across which the majority of interference is expected.

Where the interferer has an expected duty cycle within a range, the monitoring may be continued for a period sufficient to detect the interference, that is for a period sufficient to detect at least one burst of interference. For example, in a packet data system, the monitoring may be continued to identify a plurality of erroneous or lost packets on a channel.

In a related aspect the invention provides a method of selecting frequency channels for use by a frequency hopping communications device, from a set of channels, the method comprising: selecting an initial predetermined subset of said channels; monitoring a received signal quality measure for each of said predetermined channels to identify ones of said channels subject to interference; modifying said initial subset of channels to add one or more channels within one or more predetermined frequency bands within which channels subject to interference are identified; monitoring a received signal quality measure for each of said modified subset of channels to identify ones of said channels subject to interference; and selecting a second subset of frequency channels for use to avoid said channels identified as subject to interference.

In this method any initial subset of channels may be selected, this then being modified to home in on channels on which interference is experienced, as outlined above.

Again, these methods may further comprise an additional phase in which the subset of frequency channels selected for use is modified in a response to a prediction of one or more suspect interfered channels, based upon results of single quality monitoring made on channels of the second subset.

Thus, in a further related aspect, the invention provides a method of selecting frequency i channels for use by a frequency hopping communications device, from a set of channels, in the presence of interference from an interferer, the method comprising: monitoring a signal quality measure for received data on each channel of said set of channels to identify, of the channels on which data is received, one or more interfered channels subject to interference from said interferer; predicting one or more suspect interfered channels responsive to said one or more identified interfered channels and at least one characteristic of said interference; and selecting frequency channels for use by said frequency hopping communications device responsive to said prediction.

As before the method may be performed by a transmitter, with a back channel to monitor signal quality as reported, for example, by a remote receiver, or the method may be performed by a transceiver. The channels subject to interference may be those on which a packet has been erroneously received or lost or a count may be kept so that the channel is not considered subject to interference until some minimum number of packets has been erroneously received or lost on the channel.

By contrast with the previously described active interferer hunting methods, this method is generally slower to identify potentially interfered channels but has the advantage that embodiments of the method can run substantially continuously during operation of frequency-hopping communications, for example, without disturbing a Bluetooth piconet, finetuning the adapted hopset. Generally, however, the above-described methods help to achieve rapid identification of likely interfered channels, thus reducing the amount of data lost to bad transmissions (through unwittingly transmitting on a bad channel), thereby increasing overall performance, for example of a piconet, as well as being generally more friendly to co-existing systems. Improving co-existence of wireless communications within an operating band is also generally desirable from the point of view of a user, who may not appreciate the effects that mutual interference between competing systems may have on one another.

In the above-described prediction-based method, the selecting of frequency channels for use by the device may comprise starting with an initial set of channels and excluding those channels which are found to be interfered, and optionally also suspect interfered channels where, over time, confidence they are also interfered has increased to the point at which they can also be excluded. Thus, each channel they have associated penalty data, which is updateable to penalise a suspect channel such that, over time, it is more likely to be excluded from a selection (or less likely to be selected) or such that once a threshold penalty is reached the channel is no longer used.

Broadly speaking, the main sources of interference in a frequency band are identifiable and characterizable by one or more of a definable frequency band or bands, associated definable bandwidths and, in some cases, an expected duty cycle range where an interferer is of a bursty nature. Any one or more of these characteristics may be used to predict the suspect interfered channels. It will be appreciated that a complete characterization of an interferer is not necessary since some advantage can be gained by merely determining some range of frequencies over which interference is expected.

However, it is assumed that the bandwidth over which interference is occurring is generally broader than the frequency-hopping channel spacing so that once one interfered channel has been identified the likely location of another can be determined.

The predicting of suspect channels preferable comprises applying one or more prediction rules responsive to the frequency band(s), bandwidth(s) and/or duty cycle range(s) of the interference. Such rules may include an interfered channel location rule to determine when at least two interfered channels, and preferably a cluster of three or more channels, lie within a known interferer frequency band, the rule then predicting that other channels within said band are also likely to be interfered. An interpolation or gap-filling' rule may also be employed to determine when at least two, and preferably three or more, interfered channels lie within an interferer bandwidth, other channels between these channels then being suspect. However, where it is determined that at least one channel within this bandwidth is substantially free from interference, the rule is preferably not applied, since then the interference is unlikely to originate from the known interferer. A duty cycle rule may also be employed to determine when an interfered channel is subject to interference, having a duty cycle within an expected range. This can arise, for example, in the case of WEAN Traffic, which, in practice, is likely to have a duty cycle of between 30-40%, albeit not with any particular frequency, since traffic is generally user-dependent. When interference with a duty cycle in the expected range is found on a channel, this channel may, for example, be penalised more heavily so that it is more likely to be marked as bad or so that it will be marked as bad in a shorter time, although in other embodiments of the method it may be marked as bad immediately.

A signal level rule may be employed to suppress the operation of one or more of the above rules where, for example, it is known that an interferer of the known characteristics is unlikely or forbidden to have a signal strength greater than a maximum level. Such a rule can, for example, be used to suppress the application of the above- described prediction rules where, say, a microwave oven has a frequency which varies and slews around the band, hitting frequency-hopping channels which, if the rules were operating, could prevent the use of significant regions of available spectrum, particularly if no mechanism for reinstating channels was available.

The above-described prediction rules may be combined, for example with a logical AND or OR operation so that a channel is identified as suspect or bad in response to two or more rules. A fuzzy or weighted combination of rules may alternatively be applied based upon, for example, the certainty of identification of a bad channel, in embodiments giving more weight to rules predicting suspect channels which have already lost packets. Generally, a set of rules is hard-coded in a communications device, either in software or firmware, but rules could be selected based upon a training mode in which the rules are used for prediction but without acting upon their results, a comparison of predictions and actual performance then being used to evaluate a rule for selection.

The above-described prediction techniques may be used to restrict a set of channels, but channels may also be reinstated based upon a variety of techniques as described in the applicant's co-pending UK patent application number 0204093.9 filed on 215' February 2002, entitled Channel Management in Adaptive Hopping Systems. Thus, reinstatement may be based upon pilot or trial transmissions using either dummy or real packets, or upon passive listening during a lull in reception or upon a blind timeout procedure, optionally with a probationary period. For Bluetooth version 1.2 (and US AFH) equivalents regulatory authorities specify a minimum number of channels, currently 15 in the USA, and where necessary the restriction in the use of frequency- hopping channels may be limited, or channels may be reinstated accordingly.

In a first related aspect, the invention further provides a system for selecting frequency channels for use by a frequency hopping communications device, from a set of channels, in the presence of interference from an interferer, the system comprising: means for selecting a first set of channels to collide with said interference; and means for selecting a second set of channels excluding said first set of channels for use by said communications device.

In a second related aspect, the invention provides a system for selecting frequency channels for use by a frequency hopping communications device, from a set of channels, in the presence of interference from one or more interferers each having a defined frequency band, the system comprising: means for selecting an initial predetermined subset of said channels, said predetermined subset including at least one channel in each said frequency band; means for monitoring a received signal quality measure for each of said predetermined channel to identify ones of said channels subject to interference; and means for selecting a second subset of frequency channels for use to avoid said channels identified as subject to interference.

In a third related aspect, the invention provides a system for selecting frequency channels for use by a frequency hopping communications device, from a set of channels, the system comprising: means for selecting an initial predetermined subset of said channels; means for monitoring a received signal quality measure for each of said predetermined channels to identify ones of said channels subject to interference; means for modifying said initial subset of channels to add one or more channels within one or more predetermined frequency bands within which channels subject to interference are identified; means for monitoring a received signal quality measure for each of said modified subset of channels to identify ones of said channels subject to interference; and means for selecting a second subset of frequency channels for use to avoid said channels identified as subject to interference.

In a fourth related aspect, the invention provides a system for selecting frequency channels for use by a frequency hopping communications device, from a set of channels, in the presence of interference from an interferer, the system comprising: means for monitoring a signal quality measure for received data from each channel of said set of channels to identify, of those channels on which data is received, one or more interfered channels subject to interference from said interferer; means for predicting one or more suspect interfered channels responsive to said one or more identified interfered channels and at least one characteristic of said interference; and means for selecting frequency channels for use by said frequency hopping communications device responsive to said prediction.

Additional features corresponding to the features described above in detail with reference to the corresponding methods may be implemented in the above systems in preferred embodiments.

The above-described systems may be implemented in hardware, or in software, for example as code on a DSP (digital signal processor) or in a combination of both, such as a combination of dedicated hardware and micro-code for a processing engine or kernel.

Thus, the skilled person will understand that the above-described methods and systems may be embodied as processor control code, for example on a carrier medium such as a disk, CD- or DVD-ROM, programmed memory such as read-only memory (firmware) or on a data carrier such as an optical or electrical signal carrier. Embodiments of the invention may be implemented on a DSP, as mentioned, or on an ASIC (Application Specific Integrated Circuit) or using an FPGA (Field Programmable Gate Array). Thus, the processor control code may comprise conventional programme code or microcode or, for example, code for setting up or controlling an ASIC or FPGA. Similarly, the code may comprise code for a hardware description language such as Verilog (trademark) or VHDL (very high speed integrated circuit hardware description language). As the skilled person will appreciate, the code may be distributed between a plurality of coupled components in communication with one another.

The above-described features and aspects of the invention may advantageously be combined, as described further later, for improved performance.

These and other aspects of the invention will now be further described by way of example only, with reference to the accompanying figures in which; Figures la and lb show, respectively, an example of a Bluetooth piconet, and a diagrammatic illustration of a portion of a Bluetooth spectrum overlapped by an IEEE 802.11b signal; Figures 2a and 2b show an adaptive frequency-hopping kernel in outline and in detail, respectively; Figure 3 shows an example of hop-sequence remapping; Figure 4 shows an RF transceiver incorporating a channel assessment processor according to an embodiment of the present invention; Figure 5 shows a flow diagram of a bad channel identification process; Figure 6 shows a flow diagram of a first channel assessment process; Figure 7 shows a flow diagram of a passive heuristic-based channel assessment process; Figure 8 shows a flow diagram of a first phase of an active targeting interferer hunting procedure; Figure 9 shows a flow diagram showing details of the phase 1 active mode process of Figure 8; and Figure 10 shows a flow diagram of a transition from a first phase to a second phase of an active interferer hunting procedure.

Broadly speaking, we will describe methods of increasing the speed by which an interferer can be detected so that the interferer may then be avoided by an adaptive frequency-hopping procedure. We will describe a two-stage process in which, on system start-up, unless there are operational reasons why the system needs to begin passing traffic immediately, an active method is employed in order to quickly characterise the RF environment, this active method modifying the standard behaviour of a Bluetooth piconet in order to increase the likelihood of detecting interferers.

Following this, in the second stage of the process, a passive interferer detecting approach is applied using heuristic predictions based upon observations made during the usual operation of the piconet. This passive method preferably runs continuously in the background, fine tuning the hop-sets as the piconet operates. It will be appreciated that each of these two stages, active and passive, may be deployed independently of one another.

In order to either predict likely interferers or to explicitly identify them, it is helpful to take advantage of a priori knowledge, and for the purposes of the described examples we will assume that a Bluetooth frequency hopping system is used in conjunction with an IEEE 802.1 l b WiFi WLAN system, the latter being the predominant interferer. The skilled person will appreciate, however, that the principles described herein can readily be extended to other frequency-hopping systems co- existing with other interferers such as other WLAN systems, cordless telephones, microwave ovens and the like. In the initial, active stage, the hopset is modified to target or hunt for interferers and in the following passive stage statistics gathered from packet loss information and other sources are used in order to predict interfered channels related to those already identified.

In broad terms, the use of adaptive frequency-hopping carries with it the implication that interferers to be avoided are substantially frequencystatic in nature. Moreover, performance at adjacent frequencies tends to be relatively highly correlated due to correlation in frequency-selective multi-path and because, in the case of strong narrow band interference, adjacent channel suppression may be insufficient.

In the more specific instance of an IEEE 802.11 or IEEE 802.1 lb interferer, there are some further characteristics of the interference which it is helpful to consider. Firstly, in relation to channel occupancy, the specified channel wit is 22 megahertz but 99% of the emitted signal occupies the central 14 megahertz and 91% of the energy is contained within the central 10 megahertz. Secondly, in relation to duty cycle, WLAN systems are inherently bursty in nature. Although the duty cycle (and, by implication, frequency or bursts) is spread over a range related to usage of the WEAN network, it is a reasonable approximation to consider the duty cycle to be in the range 30% to 40%.

This implies that at a given sampling time there is a significant probability that interference will not be present, leading to a risk of misclassifying a channel as good.

The 802.11 series of specifications also predefines frequency bands for the system and a maximum transmit power (20 dBm) for WiFi.

Based upon the above observations, a broad conclusion is that if a channel is found to be consistently interfered then there is a high probability that adjacent channels will also be bad (assuming that the interferer is operating over a wider band than the 1 megahertz Bluetooth channel width). However, from the more specific observed properties described above, more detailed hypotheses can be constructed about the potential locations of 802.1 1 (b) interferers, which can then be confirmed or refuted as time goes on to slowly construct a model of interferers in the relevant band. This may be done by the application of rules to categorise the interfered channels that have been found and to draw inferences about the likelihood of other associated channels being subject to interference. Examples of such rules are given below; generally these will be hardcoded in a communications device.

1. If several almost-contiguous channels are interfered, it is likely that the intervening channels are also subject to interference. Here 'almost-contiguous' may be defined as channels within a predetermined bandwidth such as an expected bandwidth i or portion of bandwidth of the target interferer. For example, with IEEE 802.1 lb, if two interferers are found within l O megahertz of one another, all the channels between may be marked as suspect. Conversely, however, if clear channels have been reliably found close to a bad channel or, in the foregoing example, between the bad channels, this suggests that the interferer is narrow band in nature so that the rule should not be applied. Thus, a confirmed clear channel within the target bandwidth can be taken as an indication that there is no wide band interferer in that band. This may be referred to as a gap filling or interpolation rule.

2. If a channel appears to be experiencing interference with a duty cycle in a predetermined range, such as between 10% and 90%, or, more preferably for WiFi, 30% to 40%, this may be taken as an indication of a WLAN interferer. However, if a channel is experiencing extreme duty cycles of, for example, 1% or 100%, then that I indicates that the interferer is probably not WLAN traffic.

3. If bad channels are clustered within a known or standard frequency band, such as a WiFi channel band, and particularly if no significant interference is observed outside that band, this can be taken as an indication that the cluster of interfered channels probably arises from the interferer known to occupy the band, for example a WiFi interferer. This may be referred to as a 'known channel rule'.

4. If the interference power level on a channel is greater than a threshold level, 20 dBm for WiFi, then even assuming a fairly close range transmitter, for example 0.5 m, the interferer will appear to have a higher transmit power than permitted. This can be used as an indication that the interferer is not part of, for example, a WiFi carrier, I providing that the relevant transmit power regulations are being observed. This may be referred to as a 'power level rule'.

These examples of rules have been derived from characteristics of a WiFi interferer, but other similar rules may be derived for other interferers. These and any other rules derived from the characteristics of the interferer may be applied in various ways. Thus, the degree of suspicion with which the channels predicted as interfered by the rules are treated can be varied. Secondly, the outcomes of the rules may be combined in various ways. This is described further below.

Blue tooth adaptive frequency hopping may be implemented by means of a table of penalty counters, one for each channel. On reception of a bad packet over a channel, the penalty counter for the channel is decremented by a penalty value from an initial value of, say, 15 and when zero is reached the channel is marked as bad. In this context a bad packet is preferably a lost packet, that is a packet in which uncorrectable errors have been found, although it may alternatively be a packet in which correctable errors have been found. The penalty value may be varied between one and the initial value maximum, here 15, to provide a tradeoffbetween vast interference identification and false alarms. Thus, a standard penalty value of, say 3, may be adopted but if a channel is deemed to be suspect by one or more of the above-described soft prediction rules then a greater penalty is preferably imposed to accelerate identification of that channel as bad. Again, there is a trade- offbetween rapid identification and false alarms and, at one extreme, the penalty value may be set equal to the initial counter maximum (i.e. 15). In one embodiment with a penalty value of 15, the first packet loss marks the channel as suspect but a second packet loss is required to mark the channel as bad. Therefore, a further mode, which may be termed a 'draconian' mode, may be implemented in which as soon as a channel experiences a single packet loss the channel is immediately marked as bad.

Thus, referring back to the prediction rules, at one extreme the degree of suspicion with which a channel predicted as an interferer by a rule may be high by employing a draconian mode in which all channels identified as potential interferers by one or more rules are immediately marked as bad. Alternatively, a lesser degree of suspicion may be applied (under a so-called 'liberal' mode) in which channels identified as potential interferers are given an increased penalty, thereby causing them to be marked as bad more rapidly. The greater the penalty the greater the degree of suspicion and the more rapid the classification; conversely the less tolerant the system of transient interference.

A suitable penalty value can be determined by routine experiment but in tests a penalty value of 9, that is three times the standard penalty, was found to provide an acceptable detection time and false alarm rate.

There is also a range of possibilities for combining the results of interfered channel predictions from the prediction rules. Thus, any one of the rules may be employed on its own to predict interfered channels, for example if other rules are perceived to be unreliable. In another option a channel identified as suspicious by any one of the rules is treated as suspicious by the procedure as a whole (logical OR). Alternatively, for a channel to be treated as suspicious it must be identified as suspicious by all of the rules (logical AND) or by a majority of the rules (majority decision).

The rule combination process may be augmented with a degree of confidence being associated with each rule, for example by adopting a fuzzy logic rather than a Boolean process. In this way, the contribution from a rule that has made a tentative prediction of a potential interferer may be given less weight than a rule that has made a positive identification. In other variations different penalty values may be adopted depending upon the degree of confidence associated with the rules.

The operation of the active, interferer hunting approach will next be described. Broadly speaking use is made of knowledge of a frequency band or bands occupied by the interferer, the system adopting a hopset designed to collide with potential target interferers on start-up. Collisions may be detected by many methods, including the use of RSSI measurement circuitry and/or packet loss measurements. In the case of a WEAN, more particularly a WiFi interferer, because the interference is bursty, the collision channels should be monitored over a period of time in order to detect the expected 30% to 40% duty cycle. Thus, in this case, a single packet sent down each channel would not categorically confirm or deny the presence of a WiFi interferer on the channel and instead a minimum number of packets should be sent.

In the case of WiFi, the initial hopset may comprise the centre frequencies of standard WiFi channel positions or alternatively the 14 frequency raster positions, subject to the constraints on minimum hop size imposed by the regulatory authorities. After a number of hops, the potential interferers may be tentatively classified and the procedure may then end at this point, assuming (in the previously described draconian manner) that each tentative classification is correct. However, to reduce false alarms, a second phase is preferably employed in which having identified hop frequencies corresponding to WiFi centre frequencies which are subject to interference, hop frequencies in the regions of these WiFi carriers are now more closely examined. This may be done by adopting hopsets which restrict themselves to the regions in which the energy from those carriers is concentrated.

As time progresses the hopset may be periodically adjusted to fine tune the interferer identification by eliminating channels found to be "innocent" from the adapted hopset so that more time is spent on suspect channels, hence generating data on those channels more rapidly, and by expanding the hopset in the case of an earlier misdiagnosis, as described further below. Furthermore in order to improve the scalability of this approach with respect to the number of interferers it may be combined with the predictive techniques (that is, prediction rules) used in the above described passive approach to reduce the number of channels in the phase two active mode hopsets.

The active mode is preferably invoked on network (and in the specific example of Bluetooth, piconet) start-up but it may also be re-invoked if significant degradations are detected. If the active mode is to be reinvoked and it preferably starts on the basis of the currently adapted hopset plus any newly identified interferers rather than starting from the phase one hopset used initially.

In a combined approach, active mode on start-up followed by passive mode in use, if channels are reinstated when later found to be clear (for example as described in the applicants co-pending UK application number 0204093.9 filed on 215 February 2002) then the draconian penalty method may be adopted on system start-up with the more liberal passive mechanisms then running afterwards to fine tune the initial rough cut but very rapid estimate.

Table 1 shows the frequencies of the Bluetooth channels and their overlap with the fourteen WiFi channels, the WiFi channel numbers being shaded according to the WiFi signal energy.

Frequenry/MHz CbaneI D Cbanne O _ _ _ _ _ _ _ _ _ _ _ _ _ 240 = _ _ _ _ _ _ _ _ _ _ _ _ _ _ 2Z44o _ __- _t_ E 2241 1, , -. , , = = = = = = = = 241 1.1. _ _ _ _ _ _ _ _ _ _ _ 241, .. -1 _ -' -4 _ _ _ _ _ _ _ _ _ _ 241E =-1 _ _ -4 _ _ _ _ _ _ _ _ _ _ 242. = -1 _ -. -4 _ _ _ _ _ _ _ _ _ 242! ! _. .-4. _ _ _ _ _ _ _ _ _ 242( = _ -s -4 _ f = = = _ _ = = = 224421 = . . 4. E = = = _ = = = = 243( _ _.-4. f _ _ _ _ _ _ _ _ 22443d ic = _ -. -4 _ -f _, = = _ = = = _ 243' s = _ 4 E _ = = _ = = = = 22443f: i = _ = - 44. E 7 _ = _ = = = _ 243 3- _ _ _-4. f _, f _ _ _ _ _ _ 2433f. = -= 4 E _ E = _ = = = = 244( ., _ _ _ _ f f _ _ _ _ _ _ 244. I. = _ = = E-/ f_ = = = = 24 4 1 _ _ _ _. e _, fS _ _ _ _ _ 2444- = _ = = = E _ E S 1 = = = = 2441 _ _ _ _ _ E _, E c 1 _ _ _ _ 245( = _ = = = =-/ E 1 = = _ = 245. ( = _ = = = = _, E f 1 11 = = = 245 _ _ _ _ _ _-7 c c 1C 1 _ _ _ 224 55. = _ = = = = = E E 1 C 1 = = _ 245 ry = _ = = = = = E c 1C 1 12 = _ 245f ( _ _ _ _ _ _ _ E C 1 C 11 12 _ _ 224465f ' = = = = = = = = _t -'c -. -.i = _ 246 i( _ _ _ _ _ _ _ _ 1C 1l 12 1. _ 246 f i( 1 = = = = = _ = = 1 C 11 12 1: = 22446, f.i = = = = = = = = = -1C -1l -12 1: _ 2464 _ _ _ _ _ _ _ _ _ 1C 11 12 1: _ 246f f) 1. = = = = = _ = = = 1C 1 12 1s _ 246f [i _ _ _ _ _ _ _ _ _ _ 11 12 1. _ 2247{ If)l = = = = = = = = = = -1l -12 -1< _ 247: .f 1. _ _ _ _ _ _ _ _ _ _ 14 12 1: _ 247 = = = = = _ = = = = -1 -12 1: -14 247 I. _ _ _ _ _ _ _ _ _ _ _ -12 1 -14 247' .t _ _ _ _ _ _ _ _ _ _ _ 12 1: 14 224E7,, = = = = = _ = = = = F = . -.4 2244.3 _ = = = = _ = = = = = = = _ 24E3| Jal m c _ _ _ _ _ _ _ _ _ _ _ _

Table 1

In one embodiment of the active mode procedure a group of packets is sent down the centre frequencies of the fourteen WiFi channels raster frequencies using the WiFi hunting hopset { 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 78}, although channel 78 may be omitted outside Japan since the fourteen WiFi channel, centred on 2484Mhz, is only used in Japan. This constitutes phase one of the active mode.

Having identified Bluetooth channel corresponding to WiFi centre frequencies which are subject to inference in phase one, the Bluetooth in the regions of those WiFi carriers are now more closely examined by adopting hopset which restrict themselves to the regions in which the energy from those carriers is concentrated. This is shown in table 2 below.

1 __ 1 1 1 _ i l 1 __1 1 1 1 1 l I I __' I I I I _ I i! I 1 1 __ 1 1 _ _ 1 1 _ 1 1 __ _ 1 1 __ 1 1 _ 1 1 _- _ 1 1 __ 1 1 _ 1 C 1 T 81 lot 11 11 1 41 151

_

1 101 11 1 131 1' 151 16 171 181 191 20 220 1 15t 1t 1' 18| 1' 201 21 22T 231 241 25 1 201 21 2 231 24 251 26 271 28| 291 30 1 251 26 27 281 2'. 301 31 32T 331 341 351 1 301 31 3z 331 34 351 36 371 381 391 40 T 351 36 37 381 3' 401 41 42T 431 441 451 1 401 41 42 431 44 451 46 471 481 491 50 r45T46 47 484 501 55315415 1 501 51 52 531 54 551 56 571 581 591 60 t55T56 57 5815 60161 663164165 1 601 61 62 631 64 651 66 671 68| 691 701 r651 66 67 681 69 70t 71 721 73T 741 751 78 1 781 79 = I l I I

Table 2

Table 2 shows a set of active phase two hopsets to include for each centre frequencies the Bluetooth channel identified as suspect in phase one. Thus the phase two hopset comprises the union of the sets shown in table 2 according to which channels were found to be bad in phase one. As time progresses in phase two channels later found to be good can be removed from the adaptive hopset, and the hopset can also be expanded in the case of an earlier misdiagnosis. For example if the phase two hopset comprised channels 15 to 25 but channels 20 to 25 appeared to be good then channels 20 to 25 could be eliminated from the hopset, the hopset then expanded to include channels 10 to 15. This recovers from a situation where the procedure initially identifies WiFi channel 3 as the potential interferer when in fact channel 2 was occupied. Such a misdiagnosis is possible as only 91 percent, not 100 percent of the WiFi energy is concentrated in the middle 10Mhz of the WiFi channel and thus the outlying Bluetooth channels on either side ofthe 10Mhz band are also subject to interference (see table 1).

After a predefined period, or once all of the Bluetooth channels have been unambiguously classified, the active mode ends and the piconet then operates using an adapted hopset which avoids all of the channels identified as bad. If, during normal operation, the piconet observes high packet losses active mode may be re-invoked with a phase 1 active hopset based upon those observations.

Table 3 shows a comparison of mean and standard deviation values (in seconds) of interferer detection times for various combinations of rules, against a benchmark with no interferer prediction but merely a standard penalty value of three, or one, two and three WiFi interferers (the disjunctive combination being a logical OR combination). It is preferable that a procedure with a low mean detection time and a low standard deviation is employed as simply aiming for an improved average detection time could result in unacceptably long times being taken under certain conditions. It can be seen that the draconian method is very quick but has no tolerance to transient or non-WiFi interferers. The targeted known channel rule seems the most successful and the three combination methods provide increasing degrees of certainty, traded off against increased detection time. With the active interferer hunting approach, an increase in the number of interferers results in an increase in the number of phase 2 hopset channels and thus an increase in the time taken to confirm all the interferers. Selection of a rule or of a combination of rules, and of active/passive mode or both may be made according to the intended application and RF environment.

Referring to Figure 4, an antenna 400 serves as both a receive antenna for a receiver 402 and as a transmit antenna for a transmitter 404. An adaptive frequency-hopping kernel 406, similar to those previously described, performs hop selection and channel remapping functions. Kernel 406 provides hopping channel data to transmitter 404 to control the channel tuned to for the next transmission and to receiver 402 to control the channel tuned to for the next incoming packet. Data packets from receiver 402 are provided to an error detector/corrector module 408, which provides an output 410 comprising successfully received packets, i.e. packets with no errors or with correctable errors. Error detector/corrector 408 has a second output 412, carrying data comprising a notification of channels on which packets were received which could not be recovered, that is "bad" channels with uncorrectable errors. A channel assessment module 414 receives this data from error detector/corrector 408 and provides an output 416 to the AFH kernel 406, output 416 carrying data comprising a definition of an adapted hopping set, that is data defining channels to use as a subset of the available channels in the adaptive frequency-hopping process. A second output 418 from channel assessment module 414 provides adapted hopping set information for publication to the slaves within the piconet via Link Manager Protocol (LMP) messages. Dashed line 420 indicates an interface between the front end processing system of the transceiver of Figure 4 and higher layers of the Bluetooth protocol stack implemented, for example, by a microcontroller (not shown in Figure 4). These higher layers provide interfaces with equipment within which the Bluetooth transceiver is installed, such as a laptop computer, mobile phone or the like. Thus, these higher layers of the protocol stack receive successfully received packets from output 410 and also provide an output 422 to transmitter 404 comprising data packets to be transmitted.

The channel assessment module 414 and/or AFH kernel 406 may be implemented in hardware, or in software, or in a combination of the two. For example, a software implementation may comprise a processor such as an ARM (trademark) processor from ARM Limited of Cambridge, UK, coupled to working memory (RAM) and stored programme memory (firmware ROM). A Bluetooth transceiver typically incorporates around 200k bytes of ROM and 50k bytes of RAM, the ROM storing compiled C-code.

In embodiments, channel assessment module 414 may be implemented with less than approximately I k bytes additional ROM and less than approximately I k bytes additional RAM where channel remapping and penalty-counter tables are already available. In embodiments a bad channel list is implemented with one entry per channel, in the range zero to two, where zero is "innocent", one is "suspect" and two is "bad", and a table in ROM is used to map between Bluetooth hopping frequencies and WiFi channel numbers. To implement an embodiment of the above-described passive mode method intermediate lists of bad channels from each of the rules employed may be held on the processor stack, and time-based bad channel information used to identify patterns for the duty cycle rule may be implemented using an historical data sliding window. Thus a sliding window of, for example, ten entries per channel may be employed to provide history data for each channel or, in a simpler implementation, a counter may be employed for each channel which accumulates when successive bad receptions occur, to trigger marking of the channel at a predetermined value, the counter zeroing out when a good reception occurs. In an embodiment of an active mode implementation, the active hopset may be stored in RAM and a "pseudo-correlator" that looks for contiguous bands of interference, if implemented, may be accommodated within the processor stack since it needs no static data (alternatively, a simpler, time-out based approach may be employed).

In a simple embodiment of the channel assessment module 414, 79 counters are employed, one for each Bluetooth channel, each counter recording the number of bad packets received on its respective channel. When the number of bad packets on a channel reaches a predetermined threshold, that channel is marked as bad and from then on avoided.

Referring now to Figure S. this shows a flow diagram of a bad channel identification procedure, as illustrated a relatively simple embodiment which only relies upon packet error detection. Thus, the procedure of Figure 5 for interference detection may be implemented as part of the error detector/corrector module 408 of Figure 4.

Referring to Figure 5, the procedure begins at Step 500 with a receiver weight state until a packet is received 502 on a selected channel. The procedure then checks (Step 504) whether the packet has been received correctly (optionally counting correctable errors as correct packet reception). If, at Step 504, the packet was received correctly the channel status value "good" for the channel is passed to a channel assessment algorhythm (Step 506), otherwise the channel status "bad" is passed to the channel assessment algorhythm (Step 508); the procedure then ends (Step 510).

Figure 6 shows a flow diagram of a basic channel assessment procedure which, in a simple embodiment, merely counts the number of "CHANNEL_STATUS (BAD)" messages the procedure receives for each channel and, when a threshold is reached, bars that channel from use. Thus, in more detail, the procedure begins with an idle state at Step 600 waiting for reception of channel status information (Step 602), and then checking whether the status of the channel is "good" (Step 604). If the status is "good" a counter for the channel is incremented 606; if bad the counter is decremented 608 and a further check is made on whether the counter is less than a threshold value (Step 610).

If the counter was incremented or if the counter, following decrement, is not less than a threshold value, the procedure ends (Step 612). If the counter is less than the threshold value, the procedure checks, at Step 614, whether the adapted hopset is already at a minimum size and, if so, again the procedure ends. Thus, for this simple procedure, if the hopset becomes too small an interferer on a channel may not be avoided. For example, the present minimum number of channels is 15 in the USA (regulated by the FCC).

However, if the hopset is not at its minimum size, then the interfered channel is removed from the adapted hopset (6 l 6) and data defining the adapted hopset is passed to the link manager protocol (Step 618) and to the hopping kernel (Step 620) in preparation for the HSSI (Hop Sequence Switch Instant). The procedure then ends. It will be appreciated that with the procedure of Figure 6, no provision is made for reinstating a channel and reception of good packets on a channel merely delays the marking of that channel as bad.

lt will further be understood that the minimum hopset size may, if desired, be set at a greater minimum value than that specified by the relevant regulatory authority. Figure 7 shows a procedure which extends the basic scheme described with

reference to Figure 6 to consider a pattern of interference, that is information from adjacent and other channels related to the interfered channels, to provide a mechanism for passively analysing information generated by the interference detection process in order to more quickly identify suspected interferers. Thus Figure 7 shows a flow diagram of a passive, heuristic-based channel assessment procedure which may be employed in place of the basic channel assessment procedure of Figure 6.

The procedure begins in an idle state 700 waiting for channel status information 702, which is checked 704, if the status is good incrementing a channel counter 706 and, if bad, decrementing this counter 708. Following a counter decrement, a check is again made as to whether the counter is less than a threshold value 710 and, if it is less than a threshold the channel is marked as bad 712. Following increment or decrement of the channel counter, whether or not the counter is less than the threshold, the procedure continues to Step 714 at which the heuristic prediction rules are applied to the current bad channel list. A determination at Step 716 of the mode of application of the rules is then made, either a draconian mode in which all channels identified as potential interferers by the applied rules are immediately marked as bad, or a liberal mode in which channels identified as potential interferers are merely penalised, thereby causing them to be marked as bad more rapidly. The mode determination may be either preselected, for example built in to the fimnware, or maybe made in response to a software switch or other configuration data, or in response to some aspect of the determined RF environment. Thus with the draconian mode at Step 718 all the suspect channels are marked as bad, and with the liberal mode, at Step 720, the penalty values for the suspect channels are increased (that is, in this example, the counter is further decremented, as indicated dashed feedback path 722 to decrement counter Step 708 so that a check can again be made as to whether the decrement of counter is less than a threshold.

Following Steps 716 and 718/720, all the channels determined to be bad are removed 722 from the adapted hopset and adapted hopset data is then passed to the LMP 724 and to the hopping kernel 726 ready for the HSSI; the procedure then ends at Step 728.

Flow diagrams for the above-described active mode of operation will now be described which, as previously mentioned, is in preferred embodiments a two phase process. The active mode may be used on system start up for a set period of time to quickly characterize the system/AS environment before changing to the passive, heuristic-based mode of Figure 7 for longterm maintenance of the AFH configuration.

Figure 8 shows a flow diagram for the first phase of the active mode in which, in this example, the centre frequencies of the internationally recognised standard 802.11 carrier positions are targeted. Thus, the procedure begins in an idle state 800 and, at Step 802, sets the adapted hopset to the targeted centre frequencies, if hunting for 802.11, channels 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70 and, optionally, 78. The adapted hopset data is then passed to the link manager protocol 804 and to the hopping kernel 806 in preparation for the HSSI. The phase 1 procedure then waits 808 and whilst in this WAIT state, errors detected during operation are recorded using the procedure illustrated by the flow diagram of Figure 9, which therefore begins with WAIT state 808.

Thus, referring to Figure 9, which shows the operation of Phase 1, when channel status information is received 902 a check is made 904 as to whether the channel status is "good", if so the relevant channel counter being incremented 906, and if not decremented 908. Following a decrement, the procedure determines 910 whether the counter is less than a threshold value and, if it is, the channel is marked as suspect 912.

Following increment or decrement of the channel and any consequential actions the WAIT state 808 is then re-entered.

In one embodiment Phase 1 of this active mode procedure ends after a predetermined number of hops, for example determined by routine experiment. In one embodiment 500 hops was determined to be a suitable trade-off between reliability and speed of suspect channel identification. Following the Phase 1 procedure, the hopset is modified to focus on the 802.11 wideband carriers identif ed as potentially present, as shown in the flow diagram of Figure 10, and which has previously been described with reference

to Table 2.

Figure 10 shows a flow diagram of the transition from Phase 1 to Phase 2 of the active mode procedure. Thus, the procedure begins with the Phase 1 WAIT state 808 and is triggered 1000 by the reception of a "Phase I period elapsed" message. Then, for each channel identified as suspect in Phase 1, the appropriate surrounding channels are added to the hopset as previously described with reference to Table 2. Thus, for example, if channel 15 was the only channel marked as suspect, the hopset becomes { 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 203. Again, this adapted hopset data is then passed to the LMP 1002 and to the hopping kernel 1004 ready for HSSI, and the procedure then enters a Phase 2 WAIT state 1006.

Whilst in Phase 2, packet reception is handled in a corresponding manner to that described with reference to Figure 9 above for Phase 1, although preferably in Phase 2 the procedure also substantially continuously checks to see whether there are ten contiguous bad channels (assuming a characterization of, in this example, a WiFi channel as having a ten megahertz wide central band). (In other embodiments different numbers of contiguous bad channels may be employed.) In embodiments, as soon as a plurality of contiguous bad channels has been identified, that is a set of bad channels with no intervening good channels, the procedure leaves the active mode and begins adaptive hopping on the other channels, preferably all of the other channels. Thus, for example, if channels 10-20 are bad, once active mode is left the hopset is changed to { I - 9, 21-79}. If desired, the active mode may be applied periodically to re- evaluate the bad, and by implication good channels.

It can happen that ten contiguous bad channels are not found, which can imply, for example, that either the interferer is not of the type expected, for example not 802.1 1, or that the Phase 2 hopset is "incorrect", for example because a transient error has lead to a misidentification in Phase I. To address this, action may be taken after a predetermined number of hops, such as 1000 hops, without a termination. In such a case, the Phase 2 hopsets may be widened, for example by five channels in either direction. For example, if {10-20} doesn't find ten contiguous channels by 1000 hops, the procedure may then widen the hopset out to {5-253, and so forth.

The skilled person will understand that the piconet does not remain in the active mode, and it is merely a way of identifying interferers more quickly than would otherwise be the case. Generally, very large packet losses will be suffered during Phase 2 because the worst channels have deliberately been selected for the channel hopping. Thus, in embodiments after a predetermined time period, for example, 3000 hops, the active mode procedure ends and the passive mode procedure begins, to further fine-tune the hopset and to react to future changes in the RF environment.

-I l l l I I l l I I l l r I l _- . _ _ T. . =m _ _ _ _ _ 3 _ _ _ - _ N '. . . . '. _ _ _ _ _ _. ,. ., ., _ _ _. _ N. . . . . _ 'I '' ---': - N ' '' -' -------"' ' ' 2 ----- N, , ,, ,, , ,. _, , _ _ _,, , , , _ _ _,, _ N _ "' .' ,, _ _ _. . _ N, ,, ,,, _ _ _ _ _ _ _ _ s _,- _ _ _,, _ N 9. . . \ _ _ _. ., _ _ _.. , ., ., _ _ _ t.,,. _ N. .' . <. . . l '. ' ,.' _ _ N.. '' . .'. _ _ _. _ _ _.,. _ '. . _ _, .

_ _ _ _ _ _ C _ _ _ _

a, t. Y., -, _ _ _ t51 _ _, _ _ _ c _ CO V' . ,\ _ _ _ _ _ _ _,. _ _ C _ t. . _ _ _.' _ _ _.' _ _ O _ =I-D.' -.,, .\ _ _ _ _ _ _ _ Àc I -O _ _ _ C _ _ i _ _ _ _.,., c _ =m _, _ _ _ _ -2 _ _ _ -.', _ _ _ _ C = -O.! _ _ _ _ _ _ _, -O _ _ _ C _ = (, , W.,' _ _ _ ''';'' _ _, ''' .N _ _ _,,, _ .,.' ', . _ _ _ 2.t _. \. ,,,.. _ _ _ 'N' _ (D. '., .'' _ _ _. \,\ 1';' W. 5, _ = .,.' ,,w, -.,, _ _ _.',.2 _ _ -., ,.,' 2., _ _ _ 'n2 _ 't,, .N.'. _. _ _ x, ,.,. .,. _ _ _,. _ a N _ _ _ _ _ _ _ " __ _ _ :, r r'i: ii