Frequency division multiple access (FDMA) dedicated transmission system transmitter and receiver used in cush a transmission system

na Q>55aan o5pm nnezni ,ΐι’τιρ’ (FDMA) D’in’B nanyn

Frequency division multiple access (FDMA) dedicatedtransmission .system, transmitter and rece-iver usedin such a transmission system ALCATEL N.V- C. 100228 1

FREQUENCY DIVISION MULTIPLE ACCESS (FDMA) DEDICATEDTRANSMISSION SYSTEM, TRANSMITTER AND RECEIVER USED IN SUCH

A TRANSMISSION SYSTEM

The present invention relates to a frequencydivision multiple access (FDMA) dedicated transmissionsystem, a transmitter and a receiver as described in thepreambles of claim 1, claim 5 and claim 8 respectively.

Such a transmission system, transmitter andreceiver are already known in the art, e.g. from the opticaltransceiver arrangement described in the published EuropeanPatent Application EP0512642. Therein, the receiver includesan electrical splitting means to split the received signalover m communication lines, where m is the number offrequencies used for the data carriers. Each split signal isdemodulated by one of the demodulating means. The demodulatedsignals are then demultiplexed, if necessary, i.e. if theycontain signals in different time slots, by well known timedivision demultiplexing circuits. The known receiverarrangement thus comprises for each carrier frequency thecascade connection of demodulating means and time divisiondemultiplexing means. As a result, the complexity andcorrespondingly the cost of such a receiver grow linearlywith the number of carrier frequencies used.

An object of the present invention is to provide afrequency division multiple access (FDMA) dedicatedtransmission system of the above known type but wherein thedemodulation and demultiplexing can be performed by a lesscomplex and accordingly more cost-effective receiver.

According to the invention, this object is achievedby means of the transmission system described in claim 1, thetransmitter described in claim 5 and the receiver describedin claim 8.

In this way, by using a receiver of a discrete multitone type, the complexity of the receiver grows logarithmically with the number of carrier frequencies used. D. MESTDAGH - M. DE PRYCKER 10-5 2

If a large number of frequencies is used, the complexity ofthe receiver is reduced significantly when compared to thecomplexity of the known receiver.

Indeed, the complexity of the present receiver ismainly determined by the fast fourier transform processingunit thereof, and as is well known by a person skilled in theart, the complexity of such a unit grows logarithmically withthe number of samples applied to its input. For discretemultitone symbols carried by a set of carriers, this numberof samples applied to the input of the fast fourier transformprocessing unit is equal to the number of carriers in the setof carriers.

The discrete multitone symbols received by thereceiver in the present transmission system contain datatransmitted by the plurality of transmitters also includedtherein. Therefore, each transmitter is assigned apredetermined subset of carriers to modulate data on. Thesesubsets all together constitute a set of carriers withequidistant frequencies, which is a requirement of thediscrete multitone technique. Carriers used by differenttransmitters however, are generated by different carriergenerators. Their frequencies will not be perfectlyequidistant as a result of which the above subsets ofcarriers do not constitute a perfectly orthogonal set ofcarriers. A small frequency interference effect due to thisnon-orthogonality thus seems to be a penalty of the presenttransmission system. Nevertheless, this penalty is madenegligible provided that the frequencies of the clocks of thedistinct transmitters match within a specified range.

There are a variety of ways to implement thepresent transmitter and transmission system as can beconcluded from claim 2. A first implementation of a transmitter according to the present invention is a transmitter of the discrete multitone type, modulating data on a subset of carriers with D. MESTDAGH - M. DE PRYCKER 10-5 3 equidistant frequencies which forms part of the above set ofcarriers with equidistant frequencies. A more simple implementation of the presenttransmission system is based on a plurality of transmitters,modulating each one single carrier out of the above set ofcarriers with equidistant frequencies, with e.g. a QAMmodulation technique, as described in claim 2.

As will be described later, the modulationtechnique might even be different for each transmitter andfor each carrier.

An additional characteristic feature of the presenttransmission system is that it includes time divisionmultiplexing and demultiplexing means as described in claim3.

As already mentioned above, the known receiver,used in frequency domain multiple access (FDMA) transmissionsystems, comprises, when combining frequency divisionmultiple access (FDMA) with time division multiple access(TDMA), for each carrier frequency a time divisiondemultiplexing means. Therefore in the alternative embodimentof claim 3, time division multiplexing and demultiplexing,allowing several transmitters to modulate data on carrierswith equal frequencies, if these data are transmitted indifferent time slots, are integrated in the transmissionsystem. Well known time division multiplexing means tuned byrespective control units are therefore included in thetransmitters, whilst a time division demultiplexing means,tuned also by a control unit, is included in the receiver. Atthe receivers side, for reasons already mentioned, thecomplexity of the time division demultiplexing means isrestricted in an analogous way as for the demodulating means. A further characteristic feature of the present transmission system is described in claim 4. This characteristic feature solves the problem of obtaining synchronization between the plurality of transmitters by D. MESTDAGH - M. DE PRYCKER 10-5 4 compensating propagation delay effects due to differentdistances from each individual transmitter to the receiver.

Indeed, by including delay units which delay thetransmission over a greater time period for transmitters atshort distances to the receiver and vice versa, the distancefrom each individual transmitter to the receiver is virtuallymade equal. A further characteristic feature of the presenttransmitter is described in claim 6 and has the advantage ofcompensating interblock interference due to impulse responseeffects of the transmission medium.

Indeed and as is described in the article 'AMulticarrier El-HDSL Transceiver System with CodedModulation', written by Peter S. Show, Naofal Al-Dhahir, JohnM. Cioffi and John A. C. Bingham and published in the issueNr. 3 May/June 1993 of the journal European Transactions onTelecommunications and Related Technologies (ETT), pages 257-266, the effective length of the channel impulse response ina discrete multitone transmission system may cause interblockinterference. To mitigate this interblock interference, thediscrete multitone receiver is equipped with a time domainequalizer, which is in fact a short adaptive digital filter,and a cyclic prefix extension is added to the modulatedsymbols in a discrete multitone transmitter.

Still another characteristic of the presenttransmitter is described in claim 7 and solves the problem ofsynchronization between the distinct transmitters without theuse of the plurality of delay units.

Indeed, when adding a number of cyclic prefix bitsto each block of data transmitted, the distances betweenrespective ones of the transmitters and the receiver areallowed to differ over a range smaller than (n.s)/(2.f)without synchronization problems, wherein n represents thenumber of cyclic prefix bits (in bits), s represents thepropagation speed of the transmission medium (in meter per D. MESTDAGH - M. DE PRYCKER 10-5 5 second) and f represents the number of bits transmitted eachsecond (in bits per second).

From the above formula it is seen that thesedistances may differ over a wider range when the number ofcyclic prefix bits, n, is enlarged. Enlarging the number ofcyclic prefix bits is thus an alternative for the abovedescribed use of delay units.

The above mentioned and other objects and featuresof the invention will become more apparent and the inventionitself will be best understood by referring to the followingdescription of an embodiment taken in conjunction with theaccompanying drawings wherein :

Fig 1 is a schematic representation of anembodiment of a transmission system TS according to thepresent invention;

Fig. 2 is a schematic representation of anembodiment of a transmitter T according to the presentinvention; and

Fig. 3 is a schematic representation of anembodiment of a receiver R according to the presentinvention.

Referring to Fig. 1, an embodiment of a frequencydivision multiple access (FDMA) dedicated transmission systemTS according to the present invention will be described.

This transmission system TS includes a plurality oftransmitters T1 ... Tn, a plurality of delay units DU1 . . .DUn, a multipoint to point transmission medium TM and areceiver DMT-R. Each transmitter, e.g. Ti of the plurality oftransmitters T1 ... Tn comprises a time division multiplexmeans TDM, a time division multiplex control unit TCU and amodulating part MOD, whilst the receiver DMT-R comprises ademodulating part DEM, a time division demultiplex means TDDand a time division demultiplex control unit RCU.

Outputs of respective ones of the transmitters T1 . . . Tn are coupled to access nodes of the multipoint to point transmission medium TM via respective ones of the delay units D. MESTDAGH - M. DE PRYCKER 10-5 DU1 ... DUn. Furthermore, an input IR of the receiver DMT-Ris connected to the output node of the transmission mediumTM.

In the following one transmitter, i.e. Ti, isdescribed in further details. However this description isequally well applicable to the other transmitters T1 .. Tn.

Between an input IT and an output OT of transmitterTi, the time division multiplex means TDM and the modulatingpart MOD are coupled in a series connection. Furthermore anoutput of the time division multiplex control unit TCU isconnected to a control input of the time division multiplexunit TDM. In a similar way, a series connection of thedemodulating part DEM and the time division demultiplex meansTDD is coupled between the input IR and an output OR of thereceiver DMT-R, whilst an output of the time divisiondemultiplex control unit RCU is connected to a control inputof the time division demultiplex means TDD.

The present transmission system TS is dedicated tofrequency division multiple access (FDMA), which implies thatthe distinct transmitters T1 ... Tn simultaneously transmitdata via the transmission medium TM to the receiver DMT-R.Therefore, each of them is assigned a subset of carrierfrequencies and correspondingly each of them has at itsdisposal a predetermined bandwidth capacity.

The receiver DMT-R according to the presentinvention is of a discrete multitone type and likewisedemodulates discrete multitone symbols D modulated on a setof carriers with equidistant frequencies. If necessary, i.e.if equal carrier frequencies are assigned to differenttransmitters during different time slots, the receiver DMT-Rsplits the data transmitted by these different transmitters.The discrete multitone demodulation at the receivers side isperformed by the demodulating part DEM, a particularembodiment of which will be described in further detail lateron, whilst the time division demultiplexing is executed bythe time division demultiplex means TDD. Thereto the time D. MESTDAGH - M. DE PRYCKER 10-5 division demultiplex means TDD is controlled by controlsignals applied to its control input by the time divisiondemultiplex control unit RCU. The technique of time divisiondemultiplexing is well known in the art, e.g. from thearticle "TPON - A Passive Optical Network for Telephony” fromJ.R. Stern et al from the proceedings of ECOC '88, pages 203-206. The time division demultiplex means TDD and its controlunit RCU will therefore not be described in more detail.

To enable the transmission medium TM to applydiscrete multitone symbols D to the receiver DMT-R,transmitter Ti is given access to the transmission medium TMand modulates data Di, applied to the input IT thereof, on asubset of carriers which forms part of the above set ofcarriers with equidistant frequencies. This modulation isperformed by the modulating part MOD included in transmitterTi.

Furthermore Ti is equipped with a time divisionmultiplex means TDM, controlled by a time division multiplexcontrol unit TCU, to assign time slots to data Di which aremodulated on a carrier whose frequency is also used byanother transmitter. Referring to the above cited article"TPON - A Passive Optical Network for Telephony”, the timedivision multiplex means TDM and time division multiplexcontrol unit TCU also will not be described in further detailhere.

In a first implementation (not shown) of thetransmitter Ti according to the present invention, the subsetof carriers modulated consists of one single carrier on whichall input data Di applied to this transmitter Ti aremodulated, e.g. by QAM modulation. In an alternativeembodiment of this transmitter Ti, the subset is constitutedby carriers with equidistant frequencies, in which case thetransmitter Ti, similarly to the receiver DMT-R, is of adiscrete multitone type. Such an embodiment will be describedlater on by means of Fig. 2. Still an alternative embodimentof the transmitter Ti (also not shown) even contains an D. MESTDAGH - M. DE PRYCKER 10-5 arbitrary collection of carriers out of the set of carrierswith equidistant frequencies. Each carrier of this collectionis modulated using different modulation methods and thuscarries a different number of databits. For example,transmitter T5 modulates carriers whose frequencies are equalto the second and fifth frequency out of the set ofequidistant frequencies respectively, whereby the carrierwhose frequency is equal to the second frequency is modulatedby a 4-QAM modulation method and the carrier whose frequencyis equal to the fifth frequency is modulated by a 16-QAMmodulation method. The second carrier of the above set ofcarriers thus carries 2 bits when the fifth carrier carries4 bits. Of course, the receiver DMT-R has to know whichcarrier is modulated by which transmitter and whichmodulation method is used thereto. Information of this kindis shared between the transmitters TI ... Tn and the receiverDMT-R when initializing the transmission system TS.

Finally, the transmission system TS of Fig. 1includes, between transmitter Ti and the corresponding accessnode of the transmission medium TM a delay unit DUi. In themultipoint to point transmission system TS, the effectivedistances between respective ones of the transmitters T1 ...Tn and the receiver DMT-R differ. Different distances howevercause synchronization problems. Indeed, data which aretransmitted simultaneously by different transmitters arriveat different instances at the receiver DMT-R. In the oppositedirection, signals which are transmitted by the receiver DMT-R to enable or disable the transmitters to transmit data, arereceived by these transmitters at different instances oftime. Therefore the delay units DUI . . . DUn, delaying thetransmission of data over a longer time period fortransmitters which are closer to the receiver DMT-R, areincluded in the transmission system TS. The delays introducedby these delay units DU1 . . . DUn are introduced to obtainvirtually equal distances from each transmitter T1 ... Tn tothe receiver DMT-R. The individual distances from these D. MESTDAGH - M. DE PRYCKER 10-5 transmitters T1 ... Τη to the receiver DMT—R therefore are measured at initialization of the transmission system TS.

An alternative technique to obtain synchronization between the transmitters T1 ... Tn, based on prefix extension, will be described later.

Referring to Fig. 2, a transmitter T according tothe present invention will be described and its working willbe explained. This transmitter T is a particular embodimentof the transmitters TI ... Tn in the transmission system TSof Fig. 1.

The transmitter T of Fig. 2 modulates databits DB,applied to an input IT' thereof, on a subset of carriers withequidistant frequencies, which forms part of a set ofcarriers with equidistant frequencies.

The carriers modulated by the transmitter Tconstitute necessarily a subset of carriers with equidistantfrequencies and therefore the transmitter T of Fig. 2 is onlya particular example of the transmitters T1 ... Tn shown inFig. 1.

The transmitter T of Fig. 2 comprises a timedivision multiplex means TDM' and a time division multiplexcontrol unit TCU', similar to those of the transmitter Ti inFig. 1, and a modulating part MOD' including a mapper M, aninverse fast fourier transform processing unit IFFTU, acyclic prefix adder CPA and a parallel to serial converterPSC.

The time division multiplex means TDM' and themodulating part MOD' are series connected between an inputIT' and an output OT' of the transmitter T. An output of thetime division multiplex control unit TCU' is connected to acontrol input of the time division multiplex means TDM' .Between an input and an output of the modulating part MOD',the mapper M, the inverse fast fourier transform processingunit IFFTU, the cyclic prefix adder CPA and the parallel toserial converter PSC are cascade connected. D. MESTDAGH - M. DE PRYCKER 10-5 10

The time division multiplex means TDM' and the timedivision multiplex control unit TCU' are known from theearlier mentioned article and thus will not be described indetails.

Databits DB applied to the modulating part MOD' viathe time division multiplex means TDM' are divided intosubgroups of databits by the mapper M. As already mentioned,different carriers may be modulated by different modulationmethods and thus may carry a different number of databits.The mapper M therefore generates subgroups of databits foreach carrier to be modulated and modulates the carriers withtheir respective subgroups.

The modulated carriers are additionally transformedinto a time domain sequence by the inverse fast fouriertransform processing unit IFFTU.

Due to the effective length of the impulse responseof the transmission medium TM interblock interference canoccur. As will be described later, interblock interference iscompensated by an adaptive digital filter included in thereceiver. In practical applications however, such a digitalfilter technique is usually combined with cyclic prefixextension to obtain complete compensation of interblockinterference. In the present transmitter T, the cyclic prefixextension is performed by the cyclic prefix adder CPA. Eachdatablock to be transmitted is supplemented with n prefixbits whose values are equal to the values of n bits at theend of this datablock. In the extended datablock generated bythe cyclic prefix adder CPA the first n bits thus are equalto the final n bits. In this way, fast fourier transformationexecuted on a block of bits, whose length is equal to thelength of the non extended datablock and which is arbitrarilytaken out of the extended datablock will give similar resultsbecause the fast fourier transform is implicit periodically.The cyclic prefix extension performed by the cyclic prefixadder CPA is kncwn from point to point discrete multitoneapplications but involves, when used in a multipoint to point D. MESTDAGH - M. DE PRYCKER 10-5 environment, an additional advantage: synchronization betweentransmitters can be guaranteed if the condition that theirindividual distances to the receiver do not vary over a rangegreater than (n.s)/(2.f) is fulfilled. In this formula nrepresents the number of cyclic prefix bits (in bits) addedto each datablock, s represents the propagation speed overthe transmission medium (in meter per second) and frepresents the number of bits transmitted per time unit (inbits per second) from the distinct transmitters to thereceiver. In the following, this will be proved.

Consider for example in a transmission system TS asis drawn in Fig. 1, a first transmitter T1 and a secondtransmitter T2 at distances d1 and d2 from the receiver DMT-Rrespectively, and suppose both transmitters, Tl and T2 to beof the type of the transmitter T shown in Fig. 2. A datablockA transmitted by transmitter T1 and a datablock B transmittedby transmitter T2 need to arrive synchronously at thereceiver DMT-R, to form part of the same discrete multitonesymbol D. The receiver DMT-R previously transmits a controlsignal to both the first and second transmitter, Tl and T2,to make them transmit their datablocks, A and B respectively.After receipt of this control signal both transmitters, Tland T2, start transmitting their respective datablocks A andB. Suppose d2 is a greater distance than d1 . Consequently thetransmitting of datablock B will be delayed with respect tothe transmitting of datablock A because the control signalhas to travel a longer distance before it arrives at the

I second transmitter T2. At the receiver DMT-R, the delay ofdatablock B with respect to datablock A will be enlargedbecause datablock B also has to cover the longer distance d2.If s represents the propagation speed of the transmissionmedium TM (in meter per second), the arrival at the receiverDMT-R of datablock B will be delayed over a time period (d2-d1).2/s with respect to the arrival of datablock A. If thisdelay is smaller than the time needed to transmit the cyclicprefix bits added by the cyclic prefix adder CPA of D. MESTDAGH - M. DE PRYCKER 10-5 12 transmitter T1 to datablock A, full synchronization is stillmaintained. If f represents the number of bits transmittedper time unit (in bits per second) and n represents thenumber of prefix bits added to datablock A (in bits), thenthe time needed to transmit the cyclic prefix bits ofdatablock A is given by n/f. Thus, the condition to obtainsynchronization is given by the inequality: (d2-d1).2/s <= n/f which can be rewritten as: d2-d1 <= (n.s)/(2.f)

Summarizing, if n cyclic prefix bits are added to eachdatablock, no synchronization precautions should be foreseenas long as the distances from the transmitters T1 ... Tn tothe receiver DMT-R do not vary over a range greater than(n.s)/(2.f). Taken into account this condition, enlarging thecyclic prefix extension can be an alternative technique toavoid the use of the above mentioned delay units DU1 ... DUn.

Finally, referring again to Fig. 2, the extendeddatablocks are applied to the parallel to serial converterPSC which generates a serial data stream that can be appliedto the transmission medium TM via the transmitter output OT'.

Fig. 3 shows an embodiment of a receiver R of adiscrete multitone type, according to the present invention.This receiver R is a particular embodiment of the receiverDMT-R, included in the transmission system TS of Fig. 1.

The receiver R in Fig. 3 includes a demodulatingpart DEM', a time division demultiplex means TDD' and a timedivision demultiplex control unit RCU'. The demodulating partDEM' consists of a time domain equalizer TEQ, a serial toparallel converter SPC, a fast fourier transform processingunit FFTU, a frequency domain equalizer FEQ and a demapperDM. D. MESTDAGH - M. DE PRYCKER 10-5 13

The demodulating part DEM' and the time division demultiplex means TDD' are serial connected between an input IR' and an output OR' of the receiver R. An output of the time division demultiplex control unit RCU' is connected to a control input of the time division demultiplex unit TDD'.

Between an input and an output of the demodulatingpart DEM', a cascade connection of the time domain equalizerTEQ, the serial to parallel converter SPC, the fast fouriertransform processing unit FFTU and the demapper DM iscoupled.

Discrete multitone symbols, modulated on a set ofcarriers with equidistant frequencies and applied to thereceiver R via the input IR', are applied to the time domainequalizer TEQ which is in fact an adaptive digital filterdedicated to compensate impulse response effects of thetransmission medium. This digital filter is shaped aftermeasurement of the impulse response characteristic of thetransmission medium, which is represented by TM in Fig. 1 .This measurement is performed at initialisation of thetransmission system TS and is executed for the worst case,i.e. the most remote transmitter seen from the receiver DMT-R. Afterwards, the shape of the digital filter is designed asto reduce the length of this worst case impulse responsecharacteristic to an acceptable number of samples.

The in time domain processed discrete multitonesymbols additionally are supplied to the serial to parallelconverter SPC which generates a parallel bit stream andapplies this stream to the fast fourier transform processingunit FFTU to be transformed from time domain into frequencydomain.

Once the frequency domain representation isobtained, a correction of amplitude and phase for eachfrequency component or carrier is executed by the frequencydomain equalizer FEQ. Indeed, frequency dependent amplitudeand phase distortion caused by the transmission mediumaffects the amplitude and phase of each carrier. If the D. MESTDAGH - M. DE PRYCKER 10-5 14 frequency dependent amplitude and phase distortion of thetransmission medium are known, the original amplitude andphase of each carrier can be regained. The above citeddistortion characteristics of the transmission medium aremeasured at initialization of the transmission system. Thefrequency equalizer FEQ, which is in fact also an adaptivedigital filter, then can be adapted to compensate thefrequency dependent distortion of the transmission medium.

Finally, the demapper DM, knowing the modulationmethods used to modulate each carrier, recovers thetransmitted databits by demodulating the set of carriers.

It is to be noted that carriers of the abovementioned set of carriers with equidistant frequencies, whichare modulated by different transmitters, are generated bydifferent carrier generators (not shown), as a result ofwhich a perfect equidistant set of carrier frequencies cannotbe obtained. Due to this limitation, small frequencyinterference effects will appear in the received discretemultitone symbols. Consequently, compared with the knownpoint to point discrete multitone applications, the signal tonoise ratio of the present transmission system TS would belower. To remedy this disadvantage, the clocks are designedin such a way that the frequencies of the distincttransmitters T1 ... Tn match within a specified range.

Referring to Fig. 1 it is further to be noted thatthe time division multiplex means TDM and the modulating partMOD in the transmitters T1 ... Tn may be interchanged.Correspondingly, in the receiver DMT-R, the demodulating partDEM and the time division demultiplex means TDD may beinterchanged. If the transmitters T1 ... Tn are not allowedto use carriers with equal frequencies, the time divisionmultiplex and demultiplex means, TDM and TDD, and theirrespective control units, TCU and RCU, can be omitted.

It should be mentioned also that a central controller might be included in the transmission system TS of

Fig. 1 to coordinate the upstream transmission in the D. MESTDAGH - M. DE PRYCKER 10-5 15 transmission medium TM, i.e. the transmission from thetransmitters T1 ... Tn to the receiver DMT-R. Such a centralcontroller therefore transmits polling signals to thedistinct transmitters T1 ... Tn, thereby enabling these 5 transmitters Tl . . . Tn to supply their modulated carriers viatheir time division multiplex means TDM or respective delayunits DU1 ... DUn to the transmission medium TM.

It is finally also noted here that the mentionedmultipoint to point transmission medium TM should not 10 necessarily be a star-shaped transmission line as is shown inFig. 1 . It might for example be based on a bus configuration.Furthermore, it can be constituted by electrical transmissionlines as well as by optical waveguides.

While the principles of the invention have been 15 described above in connection with specific apparatus, it isto be clearly understood that this description is made onlyby way of example and not as a limitation on the scope of theinvention. D. MESTDAGH - M. DE PRYCKER 10-5