GB2371438A - Telecommunications system - Google Patents

Abstract

A method of transmitting data packets in a packet-based communication system includes assigning respective communication links for data transfer between a master unit and a first slave unit, A, and a second slave unit, B. The link between the master unit and the second slave unit, B, is monitored so that during periods of inactivity, DTX, data packets can be transmitted between the master unit and the first slave unit, A, over the second communications link. Data packets are transmitted between the master unit and the first slave unit, A, on the second communications link, if a packet failure occurs on the first link.

Description

TELECOMMUNICATIONS SYSTEMS FIELD OF THE INVENTION The invention relates to telecommunications systems, and in particular to communication systems supporting synchronous or isochronous services like voice or video under error-prone conditions like in a radio environment.

BACKGROUND OF THE INVENTION In the last decades, progress in radio and VLSI technology has fostered widespread use of radio communications in consumer applications. Portable devices, such as mobile radios, can now be produced having acceptable cost, size and power consumption.

Although wireless technology is today focussed mainly on cellular communications where a user is connected to a fixed infrastructure via radio base stations and portable handsets, a new area of radio communications emerges providing short-range connectivity between nomadic devices like laptops, mobile phones, PDA and notebooks. Further advances in technology will provide very inexpensive radio equipment, which can be easily integrated into many devices. This will reduce the number of cables currently used. For instance, radio communication can eliminate or reduce the number of cables used to connect master devices with their respective peripherals. The aforementioned radio communications will require an unlicensed band with sufficient capacity to allow for high data rate transmissions. A suitable band is the ISM (Industrial, Scientific and Medical) band at 2.45 GHz, which is globally available.

The band provides 83.5 MHZ of radio spectrum.

By definition unlicensed bands allow all kinds of

radio systems to operate in the same medium. This gives rise to mutual interference. To reduce interference and allow a fair access to every user, signal spreading is usually applied. Spreading provides immunity to other systems and jammers sharing the band. In fact, the FCC in the United States currently requires radio equipment operating in the 2.45 GHz band to apply some form of spreading when the transmit power exceeds about OdBm. Spreading can either be at the symbol level by applying directsequence (DS) spread spectrum or at the channel level by applying frequency hopping (FH) spread spectrum.

The latter is attractive for the radio applications mentioned above since it more readily allows the use of cost-effective radios. A radio interface called Bluetooth was recently introduced to provide pervasive connectivity in particular between portable units like mobile phones, laptops, PDA, and other nomadic devices.

The Bluetooth system applies frequency hopping to enable the implementation of low-power, low-cost radios with a small footprint. The system supports both data and voice services. The latter is optimized by applying fast frequency hopping with a nominal rate of 800 hops/s through the entire 2.45 GHz ISM in combination with a robust voice coding. Devices based on the Bluetooth system concept can create so called piconets, which consist of a master device, and one or more slave devices connected via the FH piconet channel. The FH sequence used for the piconet channel is completely determined by the address or identity of the device acting as the master. The system clock of the master device determines the phase in the hopping sequence. In the Bluetooth system, each device has a free-running system clock. The slave devices add a time offset to their clocks such that they become

aligned with the clock of the master device. By using the master address to select the proper hopping sequence and by using the time offset to align to the master clock, the slave devices keep in hop synchrony to the master device; i. e. master and slave devices remain in contact by hopping synchronously to the same hop frequency or hop carrier. For more details, the reader is referred to US patent application IIFH piconets in an uncoordinated wireless multi-user system", by J. C. Haartsen, US 08/932,911 filed on Sept.

18,1997.

Although hopping in frequency, collisions between uncoordinated systems operating in the same band cannot be prevented. For asynchronous data services, retransmission schemes can be applied to retransmit failed data packets at different points in time and frequency. However, this may cause delay which depends on the number of retransmissions. For real-time services like voice, variable delay is unacceptable.

Therefore error correction protocols based on the retransmission of erroneous data cannot be applied.

Instead sufficient protection must be included on the voice connection to overcome interference. This can be accomplished by using robust voice coding schemes like Continuous Variable Slope Delta (CVSD) modulation.

However, extra protection can be offered by transmitting additional information to correct for errors. In the provisional US application"Time and frequency diversity in FH/TDD systems"filed on May 17, 1999, a method was described to increase the robustness of a voice link by temporarily adding an additional link in a slotted FH system like Bluetooth to unconditionally retransmit the voice information on different times and different frequencies. Although an effective method, the drawback with this solution is

I

that the capacity of the system is reduced because the bandwidth per voice link doubles. As a result, the number of simultaneous voice links that can be supported decreases.

It is therefore the intention of this disclosure to describe a method and a means to provide robustness to voice links disturbed by interference without impact on the capacity of a multi-voice connection.

It is emphasised that the term"comprises"or "comprising"is used in this specification to specify the presence of stated features, integers, steps or components, but does not preclude the addition of one or more further features, integers, steps or components, or groups thereof.

SUMMARY OF THE INVENTION The method considered in this disclosure makes use of a slotted channel that consists of equal-sized time slots. Multiple voice links can be established by reservation of time slots spaced at a fixed interval.

The information to be transferred is divided into packets; each voice packet is sent in a time slot belonging to a dedicated voice link. Voice activity is applied to determine pauses and silence periods in the speed information. Discontinuous transmission (DTX) is applied to reduce the utilization of the voice channel.

The spaces left open by DTX on one voice channel can now be used by another voice channel to retransmit disturbed voice segments. If there is no silence period currently available, it may still be decided to steel a few packets from a first voice channel for packet retransmissions of a second voice channel, if the audio content of said second voice channel is more important than the audio content of said first voice channel. Voice activity detection is applied to

I

determine priority of current and retransmitted packets. Voice detection determines whether a disturbed packet of voice link A has to be retransmitted (i. e. that the packet with high probability contained important voice information), and whether it is preferred to retransmit this packet in a slot position reserved for voice link B. Due to silence periods in the voice channels and the resulting DTX, statistical multiplexing can be applied on the circuit-switched channels to provide more robustness to the voice streams. The method will improve as the number of voice channels over which statistical multiplexing can be applied increases.

According to one aspect of the present invention, there is provided a method of transmitting data packets in a packet-based communications system, the method comprising: assigning a first communications link for communication of data packets between a master unit and a first slave unit; assigning a second communications link for communication of data packets between the master unit and a second slave unit; monitoring the first communications link for periods of inactivity; and during periods of inactivity on the first communications link, temporarily assigning the first link for communication of data packets between the master unit and the second slave unit.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 illustrates a slot-based communication channel; Figure 2 illustrates a voice link established over a slot-based communication channel;

I

Figure 3 illustrates two voice links established over a slot-based channel ; Figure 4 illustrates two voice links with DTX established over a slot-based channel according to the current invention without interference; Figure 5 illustrates two voice links with DTX established over a slot-based channel according to the current invention in the presence of interference; Figure 6 illustrates a master station with multiple voice links; and Figure 7 illustrates a communications method for the system shown in Figure 6.

DETAILED DESCRIPTION OF INVENTION Many digital wired and wireless communication systems make use of a slot-based channel. The channel is divided into fixed-length time slots; in each time slot, a packet or burst is transmitted, see Figure 1.

The information to be transferred is first digitized and then separated into data packets, which are individually transmitted over the channel. In a packet-based system like Bluetooth, each packet includes the address or identity of the recipient.

This address is required to distinguish between the different recipients.

Synchronous information, for example voice information, is transmitted continuously. Usually the capacity of the channel is (much) larger than that required for the synchronous connection. Therefore, only a certain number of the slots have to be used; in the example of Figure 2, only 1 (A) out of 6 slots is used for a single synchronous link. The synchronous stream is compressed into packets which are sent at regular times with a fixed interval T. The reservation of the slots used to carry the synchronous information

I

can be accomplished in different ways by an agreement of all the units involved. The units that want to establish the synchronous link have to broadcast the reservation to all the participants on the channel. In such a case, each participant knows exactly which time slot is reserved for the synchronous link. From this

reservation, the units can then decide which time slots -. are left for other services like asynchronouslinks.

The packets sent on the synchronous links do not have to carry an address or identity of the recipient since the time slots are exclusively allocated to the recipient.

In the preferred embodiment, a centralized control is used; a single unit acts as a master of the channel, all other participants are slaves. The master unit controls the traffic over the channel. This master schedules the transmission on the synchronous links at fixed intervals. In this case, a packet address is required since the other recipients cannot deduce which time slot belongs to which slave. The advantage of a centralized control is that the master only has to agree with a single slave about the synchronous link.

No broadcasting is required. If the address in the packet does not match with the address of the recipient, this recipient is not interested whether the packet concerns a synchronous connection or an asynchronous connection. In addition, the synchronous link can at all times be interrupted by sending an asynchronous packet with the proper address in the time slot intended for the synchronous link. A method for providing synchronous and asynchronous services in on a slotted system has been described in a US patent application"Multi-media protocol for slot-based communication systems,"by J. C. Haartsen, filed July 7, 1999.

Over the slotted channel, several synchronous links can be established. Preferably, all links use the same packet interval T. They are staggered in order to avoid interaction between the channels. An example of two voice channels is shown in Figure 3. In this example, the information flow in only one direction is shown. In order to obtain a full duplex link, return channels (not shown) can either be established in a separate frequency band, as in a Frequency Division Duplex (FDD) system, or can be in the same band but on a different time, as in a Time Division Duplex (TDD) system. A TDD example is encountered in the Bluetooth system where transmit and receive actions take place on consecutive time slots.

Figure 3 also shows the digitized waveforms of the voice information carried over the voice channels. The voice information is transmitted in digital format.

The analog waveform is sampled and digitized.

Thereafter, a voice-coding scheme compresses the voice stream. The resulting binary stream is then packetized and sent in bursts over the channel. At the recipient the reverse processes take place.

In the channels illustrated in Figure 3, no Discontinuous Transmission (DTX) was applied. DTX is a common method in radio system to save power and reduce interference. When the amplitude of the speech is very small, close to the noise level, the speech is not transmitted. Instead, transmission is stopped for as long as the silence period lasts. At the recipient, comfort noise is inserted to give the user the experience that the link is still present. Voice activity detection is applied to determine whether voice is present or not, and whether DTX can be applied or not. In Figure 4, DTX is applied in the time periods with very little voice activity. It will be

I

seen from Figure 4 that the first sample of voice B and the last sample of voice A are deemed to have no activity so that data packets for those samples are not transmitted.

Interference in the radio channel may impact the voice data packets. As a result, voice information is disturbed or can be lost completely when the entire packet is discarded because of errors. When the radio operates in an unlicensed band like for example the ISM

band at 2. 45 MHZ, interference can be expected since

many""-----use of this band. Also micro 7 ate considerable amounts of ener ! e interference will have impact on t] channel. For data, the inter atted by applying a retro are the transmitter retransmits

a certain packet until its correct reception is confirmed by the recipient. Extra retransmissions will reduce the link throughput but will at least guarantee data integrity. Because of the real-time character of voice, voice segments that are lost cannot be retransmitted. On the other hand, voice does allow a certain amount of errors. In particular, robust waveform coding schemes like Continuous Variable Slope Delta (CVSD) modulation as applied in Bluetooth allows a considerable amount of random errors before it becomes annoying to the listener.

However, interference conditions can become so severe that the packet erasure rate on the voice channels exceeds the acceptable threshold and extra measures are required to maintain acceptable speech quality. One such measure has been described in a provisional US patent application :"Time and frequency diversity in FH/TDD systems"filed on May 17,1999.

This disclosure describes a method to increase the

I

robustness of a voice link by temporarily adding an additional link to unconditionally retransmit the voice information. Since the retransmission is unconditional, the delay is fixed and does not vary.

The recipient can take packets from either link and combine them to provide the best result.

Adding an extra link means a doubling in the required bandwidth for each voice channel. Although meant to be used temporarily when severe interference occurs (for example from a microwave oven), the capacity of the system is temporarily halved. In the current disclosure, a method is described which, instead of establishing a second separate link for the retransmissions, another existing voice link on which DTX is applied is used for the retransmissions. That is, two voice links use each others silence periods to retransmit failed data. Failed packets can be detected in several ways: 1) additional parity bits in the packet that can be used for error detection (can be in packet payload, packet header, or both); 2) analyzing signal content of waveform. For the first item, the Bluetooth system, for example, provides packet detection by the access code which must match the expected access code preceding each packet; there are FEC and CRC bits in the packet header that can detect failures in the header; FEC and CRC bits may be added to the payload to enable detection of errors. For the second item, voice waveform has predictable characteristics. Large deviations can be detected by statistical estimation methods thus marking a packet as failed (=highly unlikely payload content). The method is illustrated in Figure 5. When the packet failure on one voice link coincides with a silence period on the other voice link, the failed packet on the first voice link is repeated in the second voice link. This is

I

happening in the first interference event X on link A in Figure 5. Each voice packet has an address or identity which identifies the voice channel (or a slave as is applied in Bluetooth). In that way, the recipients can distinguish between normal and retransmitted information. If the failed packet does not coincide with a silence period, the system may still choose to steal a slot from the second voice channel if it is expected that the impact of the lost packet on the first voice link will be more noticeable than a lost packet on the second voice link. This is happening in the second interference event Y on link A in Figure 5. Although DTX is not applied on link B, the audio content on link B is much less than on link A. Therefore, a failed packet on link A is repeated on link B. Link B can carry out a lost frame filling or placement well-known to those skilled in the art. The voice activity processor has to consider the two voice streams simultaneously to decide whether time slots can be stolen or not. A voice waveform has predictable characteristics. Voice is semi stationary over about 10-20ms. The packets in Bluetooth only cover 1.25ms.

Therefore, there is a lot of correlation between consecutive packets. An activity detector can detect and predict fairly accurately how long the waveform levels will be low (below the audible threshold) so that the voice interval can be marked as silence (DTX).

Voice Activity Detection (VAD) is closely related to the speech coding algorithms and is a well-known technique in wireless communications. The recipients have to monitor two voice links to pick a packet from either of them, or have to insert a filler or replacement if no packet is delivered on any of the two links.

The proposed method is best applied if the voice I

streams originate from the same source. This is typically the case for a cellular or a cordless application with a base station and a plurality of terminals, an illustration of which is shown in Figure 6. Usually, the base station is more sensitive for interference for several reasons. Its location is usually on elevated positions, which reduces the propagation loss between the jammer and the base station receiver. In addition, the base station usually has a better antenna with more antenna gain which also reduces the propagation loss. Therefore, errors caused by external interferers most probably occur in the uplink direction from terminal to base station. If the system uses a piconet structure with the base station acting as a master which centrally controls the traffic for uplink and downlink, the method as described in the current disclosure can be optimally implemented. In such a method, the master schedules the traffic on the channel by sending packets to different slaves on different slots. A slave address in the packet header reveals to the slaves for which slave the packet is intended. To avoid collisions in the slots, a polling method is applied which implies that only the slave that is addressed in one time slot is allowed to return a packet in the next time slot. The polling scheme uses a central controller (like the base station in a star network) that controls the slot allocation. The retransmission required in the uplink can only be induced by the downlink (since the uplink responds to the downlink).

For optimal voice performance, these polling rules are adapted for the voice channels. The adapted polling scheme works as follows. The base station is

indicated as master M, the terminals as slaves Sl, S2, etc. Master M sends a voice packet to slave SI. It I

does this on downlink time slot TS1, reserved for transmission to slave Si. The packet contains a slave address Sl (although strictly not necessary due to the reserved slot TS1, but the address is exploited later on). All slaves receive the packet, but only slave Sl recognizes its slave address and processes the packet.

Only slave Sl is allowed to return a packet in the uplink time slot TS1'. Now the master can send a packet (incl. slave address S2) to slave S2 in TS2, etc. However, it may also send a packet to slave Sl (with slave address Sl) in time slot TS2. Although TS2 was reserved for slave S2, the master steals this time slot to allocate to slave Sl. Slave S2 will not respond in TS2'since its slave address did not match, but slave Sl will. When slave S2 cannot read the packet address (Sl or whatever) due to errors, it will assume it was S2 and will return a packet in slot TS2'.

This is because packet erasures in the downlink (from master to slave) would always result in packet erasures in the uplink. Instead, a voice slave is allowed to return a packet in the reserved uplink slot, unless it received a packet addressed to another slave in the preceding time slot. However, if slave Si did get the packet and the address was Sl, both slaves will return a packet and a collision will occur. This should not happen too often.

For packet allocation with this polling scheme, the master voice control unit has to monitor both the uplink and the downlink. For retransmissions on the downlink, it may receive a retransmit request on uplink from the slave. For retransmissions on the uplink, it can assess the quality of the uplink packet.

With the master unit acting as a central controller, the currently disclosed method of voice packet retransmission can be applied. Preferably, the

I

master unit (base station in the considered application) takes care of the voice activity detection. When an uplink failure has been detected, the master can decide on which voice channel it is desired to receive a retransmission. This decision is influenced by the voice activity on the channels in both uplink and downlink. On the most suitable channel (i. e. the channel with the least activity in both uplink and downlink), a retransmission is carried out in the downlink, which automatically will result in a retransmission in the uplink since the appropriate slave is polled.

The master may not have knowledge about failures in the downlink (at least not on a per packet basis).

However, since certain jammers are time varying, like the microwave oven whose duty cycle corresponds to the mains cycle, uplink and downlink interference may be correlated. In such a case, the extra downlink retransmission applied to trigger the uplink transmission helps in improving the overall performance. In addition, since it was explained that uplink transmission is more disturbed than downlink transmission, the missing knowledge about downlink transmission failures is of less importance. An example of the complete method for three voice channels between one base station and three terminals is shown in Figure 7. Both uplink and downlink transmissions are shown. In that case, a TDD scheme is used. Since the retransmitted packets may be transmitted on different frequencies, additional diversity is obtained. In Figure 7, the base station transmits on the even-numbered slots whereas the slaves transmit on the odd-numbered slots. The address in the downlink packet indicates for which slave the packet is intended; the address in the uplink packet indicates

I

from which slave the packet originates. Depending on the voice activity, a link is overruled for retransmissions. In Figure 7, the link to slave A is hit twice. The first time, a retransmission takes place on the B voice because, the B link was in DTX mode (i. e. silent, no packet transmitted); the second time the voice content (not shown) on the B link is more important than on the C voice link and so the C link is used for retransmission of the lost voice packet from the A link.

The described method is similar to statistical multiplexing as is applied for data traffic on packetswitched channels. The pauses in the bursty data transmission of one link are exploited by another link to transmit its data. This improves efficiency. In this case, variable delay occurs since one link may have to wait until another link has finished its burst transmission. In the current disclosure, the statistical multiplexing concept is applied in a circuit-switched network where voice links use reserved channels. Capacity on one channel is stolen to be used by another voice link for retransmissions of failed information. Since variable delay cannot be tolerated, a voice activity detection must decide which part can be stolen, or that the retransmission (s) may be cancelled.

Claims (6)

CLAIMS :

1. A method of transmitting data packets in a packet-based communications system, the method comprising: assigning a first communications link for communication of data packets between a master unit and a first slave unit; assigning a second communications link for communication of data packets between the master unit and a second slave unit; monitoring the second communications link for periods of inactivity; and during periods of inactivity on the second communications link, temporarily assigning the second link for communication of data packets between the master unit and the first slave unit.

2. A method as claimed in claim 1, wherein the first and second links are time-division multiplexed.

3. A method as claimed in claim 1 or 2, wherein data packets are transmitted between the master unit and second slave unit on the first link, if a packet failure occurs on the second link.

4. A method as claimed in claim 1,2 or 3, wherein data packets are transmitted between the master unit and the second slave unit in place of data packets for transmission to the first slave unit.

5. A method as claimed in any one of the preceding claims, comprising: sending a first data packet from the master unit to the first slave unit; the first data packet being sent on a time slot reserved for communication with the first slave unit, and including a slave address corresponding to the first slave unit; receiving, at the master unit, a return data

packet from the first slave unit ; during a period of inactivity on the second communications link, sending a second data packet from the master unit to the first slave unit, the second data packet being sent in a time slot reserved for communication with the second slave unit, and including a slave address corresponding to the first slave unit.

6. A method as claimed in any one of the preceding claims, wherein the master unit initiates retransmission of a data packet on the second communications link in response to an error in reception of a data packet by the master unit on the first communications link.