DATA RECOVERY IN MULTIWAVEFORM TRANSCEIVER SYSTEMSThis invention relates to data recovery in multiwaveform transmission systems, particularly OFDM systems.
Wireless transmission systems based on orthogonal frequency division multiplexing (OFDM) modulation are attractive due to their high spectral efficiency and resistance to noise and multipath effects. OFDM is one kind of multiwaveform modulation technique. Though resistant to the data errors introduced by noise, multipath, and other effects, OFDM-based data transceiver systems still require extra signal processing steps compared to their spread spectrum counterparts to achieve comparable bit error rates (see references [1]- [6] at the end of the description).
It is desirable to reduce the bit error rate (BER) in transmission of data symbols. The incorporation of pilot subcarrier symbols in the transmitted OFDM symbol stream, either in addition to or as part of the OFDM symbols themselves, allows for the correction of such effects as transmitter/receiver carrier frequency and sampling frequency offsets, and fading due to multipath transmission. [3] [5] [6]. The pilot subcarrier symbols are known to the receiver ; hence the received symbols can be compared against the reference symbols and the result used to characterize the impulse response of the transmission channel. The impulse response can be used to provide channel equalization information at the receiver and sent back to the transmitter to pre-distort data prior to transmission. While known techniques do provide improvements in BER, this invention provides an enhancement to existing BER reduction techniques.
According to an aspect of the invention, information about the channel response of a channel over which multiwaveform multiplexed data is transmitted, as for example derived from pilot subcarrier symbols, can also be used to enhance the recovery of corrupted data due to errors in transmission. The general scheme proposed involves the derivation of information at the receiver about the transmission channel that is used to augment data recovery. This information can be utilized at the receiver and can be incorporated in messages sent back to the originating transmitter for use at that location.
Furthermore, the information derived can be used to augment data recovery in the current and later messages.
According to an aspect of the invention there is provided an OFDM system wherein information on the channel response, which may be derived from pilot subcarriers, is used to augment error correcting codes.
According to a further aspect of the invention, there is provided an OFDM transceiver with a Reed-Solomon forward error correction (RS FEC) employing erasures based on channel information.
According to a further aspect of the invention, there is provided an OFDM transceiver where RS FEC is used at the transmitter to encode the transmitted information, a packet containing pilot subcarriers is transmitted and is used at the receiver to estimate the channel frequency response, the subcarriers with amplitudes at the receiver below a certain level are determined and are inputted into a RS FEC decoder so that the particular subcarriers are not used in the decoder.
According to a further aspect of the invention, there is provided a wideband OFDM transceiver wherein the channel impulse response is not calculated explicitly.
According to a further aspect of the invention, there is provided OFDM with random subcarrier phase shifts chosen from a particular set of random phases. The set of random phases is chosen such that they whiten the signal (reduce the peak to average ratio). They are chosen with no particular relationship to the data but so as to whiten the largest possible number of data frames. These random phases should be changed arbitrarily from packet to packet.
According to a method of the invention, there is provided a method of recovering data from a multiwaveform data stream, such as an OFDM stream of data, transmitted over a channel, the method comprising the steps of receiving, at a receiver, data transmitted over a channel, finding an estimate of symbol distortion suffered by data symbols transmitted over the channel, determining from the estimate of the symbol distortion whether the data is considered to be corrupt, and correcting data, for example by erasing data, where the data is considered to be corrupt.
According to an aspect of the invention, there is therefore provided a multiwaveform signal processor, comprising a symbol distortion estimator connected to receive data symbols transmitted over a channel and having as output vectors whose values are determined at least in part by the channel impulse response of channels over which the data symbols have been transmitted, waveform correction location means connected to receive vectors from the channel estimator and being configured to output waveform correction parameters including locations corresponding to symbols for which the channel impulse response indicates the symbol is a suspect symbol, and data correction means connected to receive data symbols from the receiver section and to receive waveform correction parameters from the waveform correction location means for correcting data symbols that are identified as being suspect symbols.
For a better understanding of the invention, and to show how embodiments of the same may be carried into effect, reference will now be made, by way of example, to the accompanying diagrammatic drawings in which:Figure 1 shows an OFDM system model;Figure 2 shows a partial OFDM packet with a) an OFDM pilot symbol composed of only prescribed pilot subsymbols followed by an OFDM data symbol comprised of data subsymbols, and with b) two OFDM symbols composed of data subsymbols interleaved with prescribed pilot subsymbols;Figure 3 shows an exemplary receiver architecture.
Pilot subsymbols are extracted after OFDM demodulation (fast Fourier transform, FFT, operation) and used for channel response estimation. The estimated channel response is used to equalize demodulated OFDM data symbols and to provide information for subcarrier erasure location determination. The locations of the erasures are fed to the Reed Solomon decoder to augment error detection and correction; andFigure 4 is a graph showing bit error rate versus signal to noise ratio. The best performance is achieved by employing Reed Solomon coding with 20 erasures. The simulation parameters are N=256, FEC RS (255,223), the channel delay is 1000 nanoseconds with a reflection coefficient of 0.9, the carrier frequency is 2.410 GHz, the symbol duration is 5Q nanoseconds, the guard time is 1 microsecond (20 subsymbols), and the subsymbol modulation is 16 QAM.
Multiwaveform SystemsMultiwaveform systems are communication systems utilizing a form of multiplexing where data is spread over various waveforms. These waveforms can be effectively distinguihsed from each other. This distinguishing could be for example : i) mathematically through the use of some transform, ii) through the use of a filter bank, or iii) through spectral analysis. Examples of these waveforms could be wavelets, solitons, sinusoidal waves, wavefunctions and the basis functions of lapped orthogonal transforms. Sinusoidal waves are the basis of OFDM.
Orthogonal Frequency Division Multiplexing with PilotSubsymbolsAs the name suggests, the main principle in OFDM is the division of a channel of fixed bandwidth into N subchannels [1] Each subchannel is centred on a subcarrier frequency. Complex-valued data symbols, generated by QAM or PSK modulation, modulate the N subcarrier frequencies.
It is easily shown that this can be accomplished by applying the DFT operation to a block of N data symbols; the resulting block of N transformed data symbols is referred to as an OFDM symbol. The DFT and IDFT operations are normally performed using a fast Fourier transform algorithm. The spectral responses of the subchannels overlap. To maintain orthogonality of the subchannels and . eliminate intersymbol (OFDM symbol) interference (ISI), a cyclic prefix of L samples is prepended to the N-sampleOFDM symbol. L is chosen to provide a temporal extension of the symbol greater than the channel delay.
The input to the transmitter is a bit stream denoted by B (m) where m is the bit index. The bit stream is modulated, for example by QAM or PSK modulation, to produce symbols X (k). It is common to consider the data before OFDM modulation to be in the frequency domain; hence, following common notation for discrete-time systems, the input symbols are denoted by X (k), where k is the discrete frequency or subcarrier index. Each symbolX (k) is referred to as a subsymbol with respect to the N- point OFDM symbol. Every N complex-valued symbols, X (k), are converted from serial to parallel in a serial to parallel converter, inverse Fourier transformed in a transformer, cyclically extended by L samples (to eliminate ISI as described above), and converted back from parallel to serial, and converted from a digital to analog signal (if needed), for transmission. Finally, each sample is transmitted through the channel. At the receiver, the inverse operations are performed in the reverse order to yield the received bit-stream, J (m). This is shown inFigure 1; conventional devices may be used to achieve the various operations. Data processing operations described here, and throughout this patent document, may be carried out in a special purpose digital computer or application specific integrated circuit (ASIC) or may be carried out in software using a general purpose computerPilot subsymbolsIn the frequency domain, that is, prior to the IDFT operation, prescribed pilot subcarriers (subsymbols) can be inserted. The pilot subsymbols can be interleaved with data subsymbol. s at known subcarrier locations within each successive OFDM symbol. The pilot subsymbol locations can vary from symbol to symbol within the packet and between packets according to a prescribed scheme (e. g., consider the ETSI DVB-T Standard [3]). Alternatively, N pilot subsymbols can be selected to comprise a pilot OFDM symbol, or more simply, a pilot symbol. Pilot symbols can be sent prior to data symbols, or interspersed among the data symbols. Some possible OFDM packet formats employing pilot symbols and subsymbols are illustrated in Figure 2.
In a particular, data transmission system implementation, it may be advantageous to employ one scheme or the other, or a combination of both.
Enhanced Data Recovery using Pilot SubsymbolsIt is well known that pilot subsymbols and symbols can be used to aid in receiver and transmitter synchronization and to reduce the effects of carrier frequency offset and sample clock frequency and phase mismatch (see [6]).
Additionally, pilot subsymbols and symbols can be used to recover information about the received signal distortions due to the transmission channel including the effects of noise and multipath transmission. This is true not only ofOFDM, but of other modulation techniques as well (see [7] [10]). This patent document describes a novel use of pilot subcarriers.
As described above, in an OFDM system the pilot subsymbols are inserted into the packet symbol stream in the frequency domain prior to OFDM modulation. That is, each pilot subsymbol is modulated onto a unique (within the OFDM symbol) frequency subcarrier. Let the kth pilot subsymbol be represented by the complex-valued subcarrierP (k). The entire OFDM packet requires r seconds to be transmitted. The pilot subsymbol is transmitted after OFDM modulation during the temporal interval 7. Assuming the channel is wide-sense stationary over the interval, this means that at the receiver and following OFDM demodulation, each pilot subsymbol will reflect the effect of the transmission channel at the frequency of its subcarrier during the interval r. If the channel response during is denoted Hr (k), the received pilot subsymbol is given by Pt (k) = Ht (k) P (k).
As the pilot subsymbol is known at the receiver, a single complex division is required to compute the estimated channel response at the kth subcarrier frequency; that is, H (k) P (k)P (k) If P (k) is inserted S times in the OFDM packet, the ensemble of pilot subsymbols can be used to give an improved estimate of the channel response at the kth subcarrier frequency. Simple averaging of the S pilot subsymbols is one approach. If there is more than one pilot symbol in a packet, then averaging can be accomplished in the time or frequency domains by averaging the pilot symbols. More sophisticated channel response estimation schemes are possible including adaptive algorithms (e. g., least mean square or recursive least squares estimation) employing uniform or non-uniform sampling (depending on the distribution of pilot subsymbols) both within an OFDM packet and across severalOFDM packets.
Whether the pilot subsymbols are contiguous (as inFigure 2 (a)), interleaved in a fixed pattern within OFDM symbols (as in Figure 2 (b)), interleaved in a varying pattern from OFDM symbol to symbol, or some combination of all three possibilities, computing the channel estimates for each value of k yields an estimate of the entire channel response during the temporal interval r of theOFDM transmission. Clearly, at each discrete frequency k, the channel response gives an indication of the distortion suffered by the corresponding data subcarriers (i. e., from symbol to symbol) at that frequency throughout the OFDM packet.
Information from the estimated channel frequency response derived at the individual discrete subcarrier frequencies, and from the estimate as a whole, provides a priori knowledge of possible data subsymbol errors. For example, the magnitude response IHr (k) l may be highly attenuated at the kth subcarrier, indicating that data subsymbols modulated on that subcarrier will be highly distorted at the receiver. If forward error correction (FEC) coding is employed at the transmitter, the relative position and distortion of the distorted subsymbols can be exploited at the receiver during decoding to correct the data, as for example by erasing data, as described in more detail below. Assuming that the channel estimate is valid for a TE seconds, all or part of the channel estimate can be returned to the transmitter and used to augment FEC coding of further messages to the receiver.
If the magnitude of all pilot subsymbols, P (k), is set to unity, the estimates, Pr (k), can be used directly for the purpose of channel equalization. It is not necessary to compute the channel response Hr (k) ; received data subsymbols need only be divided by the corresponding Pr (k).
Preferred EmbodimentWireless OFDM systems on their own do not yield extremely low bit error rates (BERs); consequently, some form of forward error correction (FEC) must be used for obtaining the extremely low bit error rates. The recentIEEE 802.11a draft standard [5] recommends the use of convolutional coding and Viterbi decoding. In general, this FEC scheme has been shown to yield very low overallBERs in wireless data transmission applications. However, when information about the transmission channel is available, as described above, convolutional coding may not offer the best solution.
Instead, block coding may be more appropriate. ReedSolomon coding is known to deliver good results for systems in which errors are generated in bursts [11].
Consider OFDM transmission in a multipath channel environment. A deep null in the frequency response results in bit errors in one or more consecutive data subsymbols.
The errors are generated in the same subsymbols in one or more OFDM symbols for the duration of the null. The positions of the subcarriers (subsymbols) affected by the null are estimated from the estimated channel response. A simple threshold is applied in a subcarrier erasure location mean$ to determine the subcarrier frequency indices in which data is suspected to be corrupt. These indices are matched to the corresponding locations in theReed Solomon codeword to indicate words that are to be erased. By erasing data known (or suspect) to be in error, the error correcting power of the Reed Solomon code is doubled. Although a simple threshold works, more complicated ways of determining corrupt data may be employed by the subcarrier erasure location means. For example, the data may be fitted to a curve to assess whether the low amplitude is due to fading or to noise. If the low amplitude is due to noise, a decision may be made that subsequent data subsymbols are not corrupt and may be kept.
An exemplary receiver architecture employing this approach is illustrated in Figure 3. The receiver includes a conventional amplifier section 10, a conventional I andQ demodulator 12 connected to receive signals from the amplifier section 10, a conventional guard interval remover 14 connected to receive demodulated signals from the demodulator 12 and remove guard intervals, a conventional Fast Fourier transformer 16 connected to receive signals from the guard interval remover 12, a channel estimator 18 connected to receive output symbols from the Fast Fourier transformer 14 and having as output an estimate of the channel response of channels over which the OFDM receiver receives signals, an equalizer 20 following the Fast Fourier transformer, and a decoder section 24, which includes a deinterleaver (if required), a subsymbol demodulator (if required), and a decoder, for example a Reed Solomon decoder. These components are substantially known in the art, but require modification as described here to carry out the invention. A waveform or subcarrier correction (erasure) location means 22 is connected to receive an estimate of the channel response from the channel estimator 18 and provide a signal either to the equalizer 20 or to a decoder 24, or both, to instruct the equalizer 20 or decoder 24 (or both) to discard symbols for which the channel response indicates the symbol may be a suspect symbol. The signal from the channel estimator 18 may also in conventional fashion output an estimate of the channel directly to the equalizer 20.
The decoder 24, which follows the equalizer 20, is largely a conventional decoder (for example a Reed Solomon decoder) for decoding equalized symbols output from the equalizer 20, and is modified according to the invention described here to receive a list of locations where data is to be discarded and to discard data at those locations.
By carrying out data discarding at the equalizer 20, some unnecessary computations may be avoided. Whether a symbol is suspect may be determined in the means 22 by comparing the amplitude of the signal with a threshold. For example, all symbols whose amplitude is more than, say, 10 dB below the running mean of the magnitude frequency response of the channel, may be considered to be within a null and labelled for erasure. The threshold may be set so that suspect data is considered corrupt. Particularly where there is redundancy in the transmitted data, as for example occurs in a Reed Solomon decoder, erasure of data may not affect the data transmission significantly.
As used herein, the term data correction means covers both an equalizer and a decoder, and any other device that may be used to correct, modify or erase data that is determined to be suspect. For general multiwaveform systems, the subcarrier equivalents may be wavelets or solitons for example. Further, the location of the data correction as determined from the impulse response may be temporal or spatial or temporal spatial. A receiver need not incorporate the amplifier or demodulator, and may in its most general sense mean only a wire along which signals may be transmitted from a source. The waveform location correction means may also supply to the data correction means more than the location of the corrupt symbol, as for example an actual value of the corrupt symbol. The channel estimator is an exemplary symbol distortion estimator. The symbol distortion estimator may also detect distortion due to other factors such as the transmitter and receiver. In addition, the symbol distortion estimator may produce a vector whose values are determined at least in part by the channel impulse response of the channel over which the symbols are transmitted. The vector may, in its purest form, be the channel impulse response, but its values may also be determined by other operations such as phase whitening, phase correction, modulation and forward error correction.
The distortion due to the transmission channel is corrected for each received OFDM symbol in the equalizer.
Distortion removal may be carried out in conventional fashion, and the equalizer may be a conventional device to this extent. However, subsymbols too severely distorted to be equalized are erased and the erasure location is passed to the Reed Solomon decoder.
The bit error rate curve from a simulation of the transmission system that shows the improvement in error performance with various Reed Solomon erasure schemes is shown in Figure 4.
Clearly, without FEC, the bit error performance of the system is the worst. With Reed Solomon coding, the performance is improved, but the addition of coding with erasures yields the best BERs of all.
The transmitted OFDM signal may be modulated with random phases chosen from a particular set of random phases. This may be accomplished in the QAM or PSK modulator, or by a separate modulator provided for that purpose. The set of random phases is chosen such that they whiten the signal (reduce the peak to average ratio).
They are chosen with no particular relationship to the data but so as to whiten the largest possible number of data frames. These random phases should be changed arbitrarily from packet to packet. When the transmittedOFDM signal is provided with random phases, identification of nulls and corrupt data may be carried out more effectively.
Immaterial modifications may be made to the invention described here without departing from the scope of the invention.
The reader's attention is directed to all papers and documents which are filed concurrently with or previous to this specification in connection with this application and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference.
All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and/or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and/or steps are mutually exclusive.
Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features.
The invention is not restricted to the details of the foregoing embodiment (s). The invention extend to any novel one, or any novel combination, of the features disclosed in this specification (including any accompanying claims, abstract and drawings), or to any novel one, or any novel combination, of the steps of any method or process so disclosed.
References [1] L. J. Cimini, Jr.,"Analysis and simulation of adigital mobile channel using orthogonal frequencydivision multiplexing,"IEEE Trans. Commun., vol.
COM-33, pp. 665-675, July 1985.
[2] M. Fattouche and H. Zaghloul,"Method andapparatus for multiple access between transceiversin wireless communications using OFDM spreadspectrum,"U. S. Patent 5,282,222, January 25,1994.
[3] European Telecommunications Standards Institute(ETSI) ETS 300 744: Digital Video Broadcasting;Framing structure, channel coding and modulationfor digital Terrestrial television (DVB-T). SophiaAntipolis, France, 1997.
[4] P. H. Moose,"A technique for orthogonal frequencydivision multiplexing frequency offsetcorrection,"IEEE Trans. Comm., vol. COM-42, pp.
2908-2914, October 1994.
[S] Hitoshi Takanashi, Masahiro Morikura and RichardVan Nee,"OFDM Physical Layer Specification for the 5 GHz Band"Doc. IEEE P802.11-98/72-r5,Submission to the IEEE802.11 committee.
[6] S. Kapoor, D. J. Marchok and Y. Huang,"Pilotassisted synchronization for wireless OFDM systemsover fast time varying fading channels,"Proc.
VTC'98, pp. 2077-2080, Ottawa, May 1998.
[7] J. K. Cavers,"An analysis of pilot symbol assistedmodulation for Rayleigh fading channels,"IEEE Trans. on Veh. Tech., vol. 40, No. 4, pp. 686-693,November 1991.
[8] J. K. Cavers,"Cochannel interference and pilotassisted modulation,"IEEE Trans. On Veh. Tech.,vol. 42, No. 4, pp. 407-413, November 1991.
[9] J. K. Cavers,"Pilot assisted modulation anddifferential detection in fading and delayspread,"IEEE Trans. Commun., vol. COM-43, No. 7,pp. 2206-2212, July 1995.
[103 J. K. Cavers,"Pilot signals improve theperformance of a Reed-Solomon errors and erasuresdecoder, in Rayleigh fading channels,"submitted toIEEE Trans. Commun., October 1997..
[11] J. D. Gibson, ed., The Mobile CommunicationsHandbook. Boca Raton, Florida: CRC Press, 1996.