392
`
`IEEE Transactions on Consumer Electronics, Vol. 41, No. 3, AUGUST 1995
`
`ORTHOGONAL FREQUENCY DIVISION MULTIPLEXING:
`A MULTI-CARRIER MODULATION SCHEME
`Yiyan Wu
`William Y. Zou
`Communications Research Centre
`Public Broadcasting Service
`Ottawa, Ontario, Canada
`Alexandria, VA, USA
`
`Abstract
`This paper presents a multi-carrier digital modulation
`technique - orthogonal frequency division multiplexing.
`A brief review of the technique and its recent develop-
`ment and implementation are provided. The advantages
`and disadvantages in comparison to other modulation
`techniques are also discussed.
`1. Introduction
`The orthogonal
`frequency division multiplexing
`(OFDM) is a form of multi-carrier modulation tech-
`nique that was first introduced more than three decades
`ago. Recent advances in digital signal processing (DSP)
`and very large scale integrated circuit (VLSI) technolo-
`gies have paved the way for the massive implementa-
`tion of OFDM techniques in the consumer electronics
`field. One recent successful implementation of OFDM is
`in digital audio broadcasting (DAB) [l], which was devel-
`oped in Europe for terrestrial and satellite broadcasting
`of multiple digital audio programs to mobile receivers.
`Another recent implementation is in asymmetric digital
`subscriber line (ADSL) technology that has been selected
`by ANSI for transmission of digitally compressed video
`signals over telephone lines [2].
`OFDM has been evaluated in Europe and elsewhere
`for digital television terrestrial broadcasting.
`In Eu-
`rope, a pan-European project Digital Video Broadcast-
`ing (DVB) was launched in 1993 with more than 130
`participants including broadcasters, manufacturers, car-
`riers and radio regulatory bodies. Its DVB-T project
`is expected to develop a terrestrial distribution system
`with data throughput up to 24 Mbps over an 8 MHz
`channel using OFDM. Under the umbrella of DVB-T,
`there are many projects under development. Most of
`them started before the formal launching of the DVB
`project. Among them are HD-DIVINE (DIgital VIdeo
`Narrowband Emission) developed by Nordic Countries,
`DIAMOND by Thomson-CFS/LER, STERNE (Systeme
`de TElevision en Radiodiffusion NumeriqE) by CCETT
`
`(a joint venture of France Telecom and TDF), SPEC-
`TRE (Special Purpose Extra Channels for Terrestrial
`Radiocommunication Enhancements) by NTL, DTVT
`(Hierarchical Digital TV Transmission) in Germany, and
`dTTb (Digital Television Terrestrial Broadcasting) by
`the Commission of the. European Communities (CEC)
`[3]. In Japan, OFDM is under investigation by many
`organizations for terrestrial broadcasting over a 6 MHz
`channel. In America, several broadcasting organizations
`of the United States, Canada and Brazil have also de-
`cided to investigate the COFDM for 6 MHz ATV terres-
`trial distribution.
`However, early development of OFDM can be traced
`back to North America in the late 50s [4]. A U S .
`patent was filed and issued in 1970 [5]. In the telecom-
`munications field,
`the terms of discrete multi-tone
`(DMT) modulation, orthogonal multi-carrier modula-
`tion, multi-channel modulation, and multi-carrier modu-
`lation (MCM) are widely used, and they are often inter-
`changeable with OFDM. COFDM means coded OFDM,
`where channel coding, or forward error correction (FEC),
`is used for error protection.
`This paper presents a brief review of the OFDM tech-
`nique. Ideas of using a guard interval to accommodate
`long delay ghosts and of creating spectrum notches to
`combat co-channel interferences, as well as of assigning
`different order of modulations or power levels to differ-
`ent sub-carriers for layered services are discussed. The
`effects of phase noise and the peak-to-average power ra-
`tio of the OFDM signal are also analysed.
`2. Basic Principles of OFDM
`In a conventional serial data transmission system, the
`information bearing symbols are transmitted sequen-
`tially, with the frequency spectrum of each symbol oc-
`cupying the entire available bandwidth. Figure l(a) is
`an unfiltered QAM signal spectrum. It is in the form of
`sin(x)/x, with zero crossing points at multiples of l/T,,
`where T, is the &AM symbol period.
`
`Manuscript received June 12, 1995
`
`0098 3063/95 $04.00 1995 IEEE
`
`Page 1 of 8
`
`

`
`Wu and Zou: Orthogonal Frequency Division Multiplexing: A Multi-Carrier Modulation Scheme
`
`393
`
`The concept of OFDM is to transmit the data in paral-
`lel QAM modulated sub-carriers using frequency division
`multiplexing. The carrier spacing is carefully selected
`so that each sub-carrier is located on all the other sub-
`carriers' spectra zero crossing points. Although there
`are spectral overlaps among sub-carriers, they do not
`interfere with each other, if they are sampled at the
`sub-carrier frequencies. In other words, they maintain
`spectral orthogonality. Figure l(b) is an OFDM signal
`spectrum, where the sub-carrier spacing is l/T,.
`Since an OFDM signal consists of many parallel QAM
`sub-carriers, the mathematical expression of the signal
`is,
`
`N - l
`
`(1)
`
`(2)
`
`S(t,) = c ( u , COS^',^, + b,sinw,t,)
`n=O
`where a, and b, are the in-phase and quadrature terms
`is the sub-carrier frequency.
`of the QAM signal, and w,
`For a large number of sub-carriers. direct generation
`and demodulation of the OFDM signal requires arrays
`of coherent sinusoidal generators that can become unrea-
`sonably complex and expensive. However, by observing
`Eq-1, one can notice that the OFDM signal is actually
`the real part of the Inverse Discrete Fourier Transform
`(IDFT) of the original data d, = a, + jb, [6], i.e.,
`(a, + jbn) . exp(-jw,t,))
`(a, + jb,)(cosw,t,
`
`N - l
`X(t,) = Re{
`n=O
`N- 1
`= Re{
`n=O
`N- 1
`= c ( a , COSW,~, + b, sinw,t,)
`
`- jsinwntm)}
`
`n=O
`where, w, = 2irn/(NAt), t, = m a t and At is the sym-
`It can be seen
`bol duration of the input serial data d,.
`that there are N sub-carriers each carrying the corre-
`sponding data a, + jb,, and the sub-carrier spacing is
`l/(At.iV). The inverse of the sub-carrier spacing, A t ' N ,
`is usually defined as the OFDM useful symbol duration,
`which is N times longer than that of the original data
`symbol duration At.
`Since IDFT is used in the OFDM modulator, the orig-
`inal data d, is defined in the frequency domain, while
`the OFDM signal X(t,)
`is defined in the time domain.
`The IDFT can be implemented via a computationally
`efficient fast Fourier transform (FFT) algorithm [7].
`3. Guard Interval and Its Implementation
`The orthogonality of sub-carriers in OFDM can be
`maintained and individual sub-carriers can be corn-
`pletely separated and demodulated by FFT at the re-
`ceiver when there is no inter-symbol interference (1%) in-
`troduced by transmission channel distortion. In practice,
`
`linear distortions such as multipath delay and micro-
`reflection cause IS1 between OFDM symbols, resulting in
`loss of orthogonality and an effect that is similar to co-
`channel interference [$I. However, when delay spread is
`small, i.e., within a few percentage points of the OFDM
`symbol length, the impact of IS1 is insignificant, al-
`though it depends on the order of modulation imple-
`mented by the sub-carriers.
`A simple solution to deal with multipath delay is to
`increase the OFDM symbol duration so that it is much
`larger than that of the delay spread. But this may be
`difficult to implement. When the delay spread is large,
`it requires a large number of sub-carriers and a large
`size FFT. Meanwhile the system might be sensitive to
`Doppler shift and carrier instability.
`Another way to manipulate multipath distortion is
`to create a cyclically extended guard interval, where
`each OFDM symbol is preceded by a periodic exten-
`sion of the signal itself [9]. The total symbol duration is
`Ttotal = Tg +T,, where Tg is the guard interval and T, is
`the useful symbol duration. When the guard interval is
`longer than the channel impulse response, or the multi-
`path delay, IS1 can be eliminated. However, the in-band
`fading will still exist [lo]. The ratio of the guard inter-
`val to useful symbol duration is application-dependent.
`Since the insertion of guard intervals will reduce data
`throughput, Tg is usually less than T,/4.
`the OFDM symbol duration T, is quite
`Since
`long, usually several hundred micro-seconds, and the
`VHF/UHF broadcast channel impulse responses are only
`in the order of 10-30 ps, inserting a guard interval of
`that range will not significantly reduce the data through-
`put. Meanwhile, as mentioned earlier, even if the im-
`pulse response is slightly longer than the guard interval,
`i.e., a few percentage points of T,, the impact on the sys-
`tem performance is limited. On the other hand, as the
`symbol duration for single carrier modulation system,
`such as QAM and VSB, is only about 0.1-0.2 ps, it is
`impossible to insert a guard interval comparable to the
`terrestrial broadcast channel response to eliminate ISI.
`Other techniques, such as adaptive equalization, must
`be used.
`An OFDM system diagram is shown in Figure 2. The
`incoming bit stream is first packed into z bits per sym-
`bol to form a complex number, where 2 determines the
`signal constellation, such as 32 QAM or 64 QAM. These
`&AM modulated data are, then, converted from serial
`to parallel with N complex numbers, or QAM symbols,
`per block. Each block is modulated by an IFFT pro-
`cess. The output of the IFFT forms an OFDM symbol,
`which is converted back to serial data for transmission.
`A guard interval is inserted between symbols to elimi-
`nate IS1 caused by multipath distortion. The discrete
`
`Page 2 of 8
`
`

`
`394
`
`IEEETransactions on Consumer Electronics, Vol. 41, No. 3, AUGUST 1995
`
`symbols are filtered and converted to analogue for RF
`up-conversion. The receiver performs the inverse pro-
`cess of the transmitter. An one-tap equalizer is usually
`used for each sub-carrier to correct channel distortion.
`The tap coefficients are calculated based on channel in-
`formation.
`From Figures l(b) and 2, it can easily be understood
`that the OFDM signal spectrum is close to rectangular.
`Actually, the spectrum is directly related to the original
`data. Since the OFDM modulator is an IFFT process, its
`physical meaning is to convert data from the frequency
`domain to the time domain and, then, transmit them in
`the time domain over the channel. If a spectrum anal-
`yser, which is a device that converts its input from the
`time domain to the frequency domain, is used to monitor
`an OFDM signal, what displayed is the original data.
`4. OFDM performance Expectation
`4.1 Multipath Distortion and Fading
`When there is multipath distortion, a conventional
`single carrier wide-band transmission system suffers fre-
`quency selective fading. A high speed adaptive equalizer
`has to be used to equalize the in-band fading. The num-
`ber of taps required for the equalizer is proportional to
`the symbol rate and the multipath delay. For VHF/UHF
`broadcasting channels, up to several hundred taps are re-
`quired.
`For an OFDM system, as mentioned in Section 2, if
`the guard interval is longer than the multipath delay, IS1
`can be eliminated and orthogonality can be maintained
`among sub-carriers. However, in-band fading will still
`exist [lo], therefore some of the sub-carriers are enhanced
`and others are attenuated. Since each OFDM sub-carrier
`occupies a very narrow spectrum, in the order of a few
`kHz, even under severe multipath distortions, they are
`only subject to flat fading. In other words, the OFDM
`can convert a wide-band frequency selective fading chan-
`nel into a series of narrow-band frequency non-selective
`fading sub-channels by using the parallel multi-carrier
`transmission scheme. There is no need to implement
`a sophisticated adaptive equalizer for each sub-carrier,
`when QPSK or MPSK modulation is used. When high
`prder &AM is implemented, channel estimation is re-
`uired. Each sub-carrier needs a well behaved one-tap
`qualizer. There is no requirement, however, to adapt to
`he instantaneous channel response (Section 4.3).
`f.2 Trellis Coded OFDM: COFDM
`Since OFDM can convert a wide-band frequency se-
`lective fading channel into a series of narrow-band fre-
`uency non-selective fading sub-channels, it is natural
`o implement trellis coded modulation (TCM) [ll], that
`s specially designed for frequency non-selective fading
`hannels, to sequentially code the OFDM sub-carriers.
`
`1
`f
`
`~
`
`At the receiving end, Viterbi decoding can be used to
`take full advantage of the soft-decision nature of the
`OFDM demodulator output. Substantial coding gain
`can be achieved for fading channels, such as Rayleigh
`and Ricean channels, especially for wide-band mobile
`communications. The trellis coded OFDM system is usu-
`ally called COFDM. In digital television implementation,
`which requires extremely low bit error rate (BER), an-
`other layer of channel coding, usually a Reed-Solomon
`code, is implemented to improve the BER performance
`and correct the burst errors created by the trellis de-
`coder. It should be noted that TCM coding increases
`only the constellation size and uses this additional redun-
`dancy to trellis code the signal. There is no bandwidth
`extension.
`4.3 Time Domain Short Term Distortion and In-
`t erference
`As indicated in Section 2, the OFDM symbol duration,
`which is equal to the inverse of the sub-carrier spacing, is
`quite long, usually in the order of 200- 1000 ps. There-
`fore, any short term distortion and interference caused
`by time domain impulse interference, amplitude clipping,
`short term fading and instantaneous change of channel
`response will be averaged out by the FFT process in the
`receiver over the entire OFDM symbol period. If dis-
`tortion and interference occurs during a small fraction
`of the OFDM symbol period, their impact is negligible,
`since all of the data sub-carriers would only be slightly
`affected, but still decodable. In a single carrier system,
`however, a few symbols could be destroyed completely
`causing a short burst of errors.
`4.3 Frequency Domain Interference
`OFDM systems are quite sensitive to the frequency do-
`main interference, such as tone interference and NTSC
`co-channel interference, where there are three carriers
`carrying luminance, chrominance and audio signals, re-
`spectively. For a single carrier modulation system, where
`one carrier occupies the entire channel, a tone interfer-
`ence will not cause eye-diagram closure and consequently
`no transmission error as long as its level is sufficiently
`lower than that of the carrier. For an OFDM system, the
`transmission power is divided among many sub-carriers.
`Therefore, even low level tone interference could destroy
`the corresponding sub-carriers and cause errors.
`One way to mitigate the tone and NTSC interference
`for an OFDM system is through spectrum shaping. From
`Eq.-2 and Figure 2, since the input data d, = a, + j b ,
`correspond to frequency domain sub-carriers, by sim-
`ply assigning some data zero value will create spectrum
`notches. When the notches are co-located with interfer-
`ing tones, there will be no impact on the OFDM signal.
`The major disadvantage of creating spectrum notches
`is the reduction of data throughput. Using trellis cod-
`
`Page 3 of 8
`
`

`
`Wu and Zou: Orthogonal Frequency Division Multiplexing: A Multi-Carrier Modulation Scheme
`
`395
`
`ing with soft-decision decoding might be a better ap-
`proach than spectrum shaping, since there is no loss of
`the throughput and the effect of tone interference on the
`OFDM signal is much like that of a fading channel case
`(Section 4.1 and 4.2) and can be solved in a similar way.
`4.4 Hierarchical Modulation to Provide Layered
`Services
`Similar to spectrum shaping, since the input data d, =
`a, + j b , correspond to frequency domain sub-carriers,
`the modulation type, constellation size and power level
`for each sub-carrier can be easily adjusted. Some sub-
`carriers can use low order &AM or MPSK modulation
`and/or higher power level to provide robust base layer
`service. Other sub-carriers can implement high order
`QAM constellations for other layers of services. As a
`result, a broadcast system with “graceful degradation”
`can be achieved. The idea behind graceful degradation
`is to arrange for the quality of the received signal to
`reduce gradually with decreasing signal strength, rather
`than failing abruptly as digital signals normally do.
`It should be pointed out that the implementation of
`such a hierarchical modulation system is subject to a
`CNR penalty. For example, if the aggregate data rate
`for a HDTV service is fixed, since part of the data are
`assigned to a low order of modulation for robustness, the
`rest of the data might be forced to use a higher order of
`modulation, i.e., higher than a stand-alone non-layered
`service system would use. Therefore the CNR require-
`ment for a hierarchical modulation system to provide
`HDTV service would be higher than that of a stand-
`alone HDTV system. Meanwhile, for a fixed data rate,
`present scalable video coding techniques would provide a
`picture quality slightly inferior to the quality of a stand
`alone HDTV system [12].
`4.5 OFDM Signal Peak-to-Average Power Ratio
`Since OFDM data sub-carriers are statistically inde-
`pendent and identically distributed, based on the cen-
`tral limit theorem, when the number of carriers, N , is
`large, the OFDM signal distribution tends to be Gaus-
`sian. Generally, when N > 20, which is the case for most
`of the OFDM systems, the distribution is very close to
`Gaussian; and it is, somehow, independent of the carrier
`constellation and the number of sub-carriers [lo]. On the
`other hand, the peak-to-average power ratio (PAR) for a
`single carrier modulation signal depends on its constel-
`lation and the pulse shaping filter roll-off factor a. For a
`Gaussian distribution, the PAR of 99.0,99.9 and 99.99%
`of time are 8.3, 10.3 and 11.8dB respectively.
`Since the OFDM signal has a high PAR, it could be
`clipped in the transmitter power amplifier, because of its
`limited dynamic range or non-linearity. Higher Output
`Back Off is required to prevent BER degradation and
`
`intermodulation products spilling into adjacent channels.
`However, clipping of an OFDM signal has similar ef-
`fect as impulse interference against which an OFDM sys-
`tem is inherently robust. Computer simulations show
`that, for a trellis coded OFDM system, clipping of 0.5%
`of the time results in a BER degradation of 0.2dB [lo].
`At 0.1 % of clipping, the degradation is less than 0.1 dB.
`The critical factor for the selection of the transmitter
`operating parameters may not be the BER performance.
`Instead, adjacent channel interference can be a decisive
`factor. This is especially important to protect analogue
`television service operating in a simulcasting environ-
`ment.
`4.6 Single Frequency Emission
`Since OFDM systems can be implemented with a long
`guard interval, they can accommodate high-level long-
`delay multipath. This might allow single frequency op-
`erations, such as single frequency networks (SFN), cov-
`erage extenders and gap fillers.
`A SFN is a cluster of transmitters (distributed topol-
`ogy) that are synchronized in both carrier frequency
`and signal transmission time. The guard interval of an
`OFDM system must be able to accommodate the “active
`ghosts” from all the nearby transmitters. The signal to
`be transmitted must be distributed to all of the trans-
`mitters within the SFN and delay adjustment (the rela-
`tive signal transmission time among transmitters) is re-
`quired. The spacing between transmitters, tower height,
`transmission power and frequency as well as the terrain
`environment are the key factors in the design of a SFN.
`A coverage extender is similar to SFN, but uses a cen-
`tralized topology. A main high power transmitter is used
`to cover a larger percentage of the service area. Several
`low power transmitters, or coverage extenders, that are
`frequency locked and delay adjusted to the main trans-
`mitter are implemented to extend or shape the service
`area to cover newly developed sub-urban communities
`and terrain-blocked areas. This approach is quite attrac-
`tive in the simulcasting environment, where the stand-
`alone high-power, high-tower transmitter approach could
`not be implemented because of possible interference into
`the existing NTSC service. It also has the potential for
`an UHF station to match the coverage of a VHF station
`without using extremely high power transmitter. The
`delay adjustment for a coverage extender depends on its
`tower height , transmission power, terrain environment
`and the distance to the main transmitter. These param-
`eters must be very carefully weighted and adjusted in
`the design and operation of a network.
`The concept of a gap filler, or on-channel repeater, is
`to use low power repeaters to provide service for terrain
`blocked areas. A gap filler is frequency locked to the
`
`Page 4 of 8
`
`

`
`396
`
`IEEE Transactions on Consumer Electronics, Vol. 41, No. 3, AUGUST 1995
`
`main transmitter. Delay adjustment might not be re-
`quired. Off-air signal pick up and retransmission might
`be possible, only if enough isolation can be achieved.
`Terrain environment, retransmission tower height and
`power level are the main design factors.
`It should be pointed out that the frequency gen-lock of
`all the transmitters is crucial to the single frequency op-
`eration, since the OFDM system is sensitive to frequency
`offset. The motivation of single frequency operation is to
`preserve valuable and crowded spectrum resources and
`to take advantage of topography. Further investigation
`is required, since many factors have to be takcn into
`consideration in the design and operation of a single fre-
`quency emission system.
`4.7 Effects of Phase Noise in OFDM
`The impact of local-oscillator (LO) phase noise on the
`performance of an OFDM system has attracted a lot
`of attention and has raised some doubts on the pos-
`sible use of long OFDM symbol durations (in the or-
`der of 1, OOOps), which correspond to close sub-carrier
`spacings. A long symbol duration is required for im-
`plementing a long guard interval that can accommodate
`long delay multipath in single frequency operation with-
`out excessive reduction of data throughput. A recent
`study has shown that phase noise in OFDM can result
`in two effects: a common sub-carrier phase rotation on
`all the sub-carriers and a thermal-noise-like sub-carrier
`de-orthogonality [13].
`The common phase error, i.e., constellation rotation,
`on all the demodulated sub-carriers, is caused by the
`phase noise spectrum from DC up to the frequency of
`sub-carrier spacing. This low-pass effect is due to the
`long integration time of the OFDM symbol duration.
`This phase error can in principle be corrected by using
`pilots within the same symbol (in-band pilots). The SNR
`caused by the common phase error can be quantified as:
`S
`1
`-
`- -
`I(cr)nl f s
`N p h a s e - r o t a t i o n
`where f s is the sub-carrier spacing, n1 is the phase noise
`spectral mask upper limit, cr = F / fs, F is the equivalent
`spectrum mask noise bandwidth, and
`
`(3)
`
`( 4 )
`with I(0.5) = 0.774, 1(1) = 0.903, and I(c0) = 0.774.
`It can be seen that, when cr > 1, or F > fs, the com-
`mon phase error decreases as the sub-carrier spacing de-
`creases.
`The thermal-noise-like phase error causes sub-carrier
`constellation blurring rather than rotation.
`It results
`from the phase noise spectrum contained within the sys-
`tem bandwidth. This part of the phase noise is more
`
`(6)
`
`(5)
`
`crucial, since it cannot be corrected. For sub-carriers
`in the middle of the channel, the SNR caused by the
`thermal-noise-like phase error can be quantified as:
`-
`S
`1
`- -
`n1 f S [ 2 0 - I(cx)]
`N t h e r m a l - n o i s e - l i t e
`For sub-carriers at the band-edge, there is an improve-
`ment of about 3dB. For (Y > 1, the thermal-noise-like
`phase noise remains more or less constant, when fs de-
`creases.
`Combining Eq-3 and Eq-5, the total SNR caused by
`phase errors is:
`S
`1
`1
`-
`-- -
`- -
`2721 F
`N total-phase-error
`2an1 fs
`It is independent of sub-carrier spacing, but directly re-
`lated to the LO phase noise performance n l and F .
`4.8 Time and Frequency Domain Dualities
`There are interesting frequency/time domain duali-
`ties between OFDM, which is a multi-carrier modula-
`tion (MCM) scheme, and a conventional single carrier
`modulation (SCM) scheme. MCM can be thought as
`a frequency domain technique, since all the signal pro-
`cessing is implemented in the frequency domain (in the
`transmitter, it is before the IFFT process, such as chan-
`nel coding, spectrum shaping, sub-carrier power level
`and constellation adjustment; in the receiver, it is af-
`ter the FFT process, such as channel decoding, channel
`equalization and phase noise tracking), while the signal
`is transmitted in the time domain. MCM is robust to
`time domain impulse interference (Section 4 . 2 ) , but it
`is vulnerable to frequency domain impulse interference
`(Section 4 . 3 ) . In contrast, SCM is a time domain tech-
`nique. It is more robust to frequency domain impulse
`interference, and by duality it is vulnerable to time do-
`main impulse interference.
`Spectrum shaping and channel coding can be used to
`improve the performance of a MCM system against fre-
`quency domain impulse interference. Adaptive equaliza-
`tion and channel coding can be used for a SCM system
`to combat time domain impulse interference.
`Another frequency-time duality is that, to prevent
`inter-symbol interference, SCM needs to reserve part
`of the spectrum for pulse shaping (frequency domain),
`while MCM needs to insert guard intervals (time do-
`main).
`For channels with multipath distortion, a SCM system
`may insert a training sequence (time domain) to assist
`adaptive equalizer convergence and system synchroniza-
`tion. A MCM system usually sends out reference sym-
`bols/carriers (frequency domain) to obtain channel state
`information for frequency domain equalization, phase
`noise tracking and channel decoding.
`
`Page 5 of 8
`
`

`
`Wu and Zou: Orthogonal Frequency Division Multiplexing: A Multi-Carrier Modulation Scheme
`
`391
`
`Both SCM and MCM have comparable BER perfor-
`mance over a Gaussian noise channel. However, since a
`MCM signal has a Gaussian distribution, while a SCM
`signal usually uses an equal probability signal constella-
`tion, there might be a few tenths of dB C/N advantage
`for a MCM system. This is because that, theoretically,
`to attain the channel capacity bound requires that the
`signal points have a Gaussian distribution [14].
`5. Conclusions
`An OFDM system has the following advantages:
`
`It is very flexible in meeting various design require-
`ments, such as complexity, bandwidth efficiency,
`spectrum shaping, performance and sensitivity to
`various impairments.
`
`It requires no adaptation to instantaneous channel
`responses. It is robust to impulse interference and
`channel variations. Channel estimation is, however,
`required for high-order QAM modulations.
`
`As a parallel transmission technique, OFDM is less
`sensitive to shift in sampling time in comparison to
`serial transmission techniques.
`
`The bandwidth efficiency of an OFDM system ap-
`proaches the Nyquist rate as the FFT size increases
`(this however increases system complexity and is
`subject to an upper-limit due to channel variations
`and carrier frequency offset).
`
`For a fixed bandwidth efficiency, the complexity
`(multiplications per symbol) of a FFT-based OFDM
`system grows logarithmically with the increase of
`the channel multipath spread, in comparison with
`a linear increase of complexity for equalizers for a
`single carrier system.
`
`A properly coded and interleaved OFDM system
`can exceed the BER performance of many other
`practical systems, especially for wide-band mo-
`bile reception where there are strong and dynamic
`ghosts.
`
`The disadvantages of an OFDM system:
`
`Higher transmitter output back off is required in
`comparison to a single carrier system, because of
`the high peak-to-average power ratio of an OFDM
`system.
`
`As a parallel transmission technique, OFDM is more
`sensitive to carrier frequency offset and tone inter-
`ference than that of a single carrier system.
`
`This paper has discussed the OFDM technique and its
`recent development and implementation. Its advantages
`and disadvantages in comparison to the conventional sin-
`gle carrier modulation techniques have also been pre-
`sented.
`References
`[l] European Telecommunication Standard, “Radio
`broadcast systems: digital audio broadcasting (DAB) to
`mobile, portable and fixed receivers,” ETSI final draft
`pr ETS 300 401, Nov. 1994.
`[2] W. Y. Chen and D. L. Waring, “Applicability of
`ADSL to support video dial tone in the copper loop,”
`IEEE Comm. Mag., vol. 32, pp. 102-109, May 1994.
`[3] J. G. N. Henderson, et. al., “Trip report and recom-
`mendation regarding COFDM,” by the Task Force on
`COFDM of the ACASTS Transmission Expert Group,
`Jan. 1994.
`[4] R. R. Mosier and R. G. Clabaugh, “Kineplex, a
`bandwidth-efficient binary transmission system,” AIEE
`Trans., vol. 76, pp. 723-728, Jan. 1958.
`[5] “Orthogonal frequency division multiplexing,” U.S.
`Patent 3,488,445, filed Nov. 14, 1966, issued Jan. 6,
`1970.
`[6] S. B. Weinstein and P. M. Ebert, “Data transmis-
`sion by frequency division multiplexing using the discrete
`Fourier transform,” IEEE Trans. Comm. Technology,
`vol. COM-19, No. 15, Oct. 1971.
`[7] E. Bidet, et al., “A fast 8K FFT VLSI chip for large
`OFDM single frequency networks,” Proceedings of the
`Intl. conf. on HDTV 94, Turin, Italy, Oct. 1994.
`[8] D. Castelain, “Analysis of interfering effects in a sin-
`gle frequency network,” CCETT report, Sept 8, 1989.
`[9] A. Ruiz, et al., “Discrete multiple tone modulation
`with coset coding for the spectrally shaped channel,”
`IEEE Trans. Comm., vol. COM-40, No. 6, June 1992.
`[lo] Y. Wu, et al., “OFDM for digital television ter-
`restrial distribution over channels with multipath and
`non-linear distortions,” Proceedings of the Intl. conf. on
`HDTV 94, Turin, Italy, Oct. 1994.
`[ll] G. Ungerboeck, “Trellis-Coded Modulation with Re-
`dundant Signal Sets Part I: Introduction,” IEEE Comm.
`Magazine, vol. 25, pp. 5-11, Feb. 1987.
`[la] DVB Project Office, “Television for the third mil-
`lennium,” DVB Project Brochure, pp. 25, Aug. 1994.
`[13] J . H. Scott, “The effects of phase noise in COFDM,”
`BBC R&D Department Technical Note No. R&D
`0127(94), Nov. 1994.
`[14] G. D. Forney, Jr., et al., “Efficient modulation for
`band-limited channels,” IEEE Journal on Selected Areas
`in Comm., vol. SAC-2, No. 5, pp. 632-647, Sept. 1984.
`
`Page 6 of 8
`
`

`
`398
`
`IEEE Transactions on Consumer Electronics, Vol. 41, No. 3, AUGUST 1995
`
`(a>
`Figure 1: (a) An unfiltered QAM signal spectrum.
`
`(b)
`(b) OFDM signal spectrum.
`
`serial
`
`packing&
`channel coding
`
`persymbol
`
`conversion
`
`-
`
`'
`
`IFFT
`(OFDM
`modulator)
`
`-
`
`DIA&
`filtering
`
`parallelherial
`conversion
`&guard
`----j interval
`insertion
`
`-
`
`R F U P -
`conversion
`
`-
`-
`c- - -
`
`channel
`: equalizer :
`
`channel
`decoding &
`symbol-@bit
`unpacking
`
`parallellserial
`conversion
`
`guard -
`interval -
`
`-
`
`noise &
`
`channel
`
`RF down
`conversion
`
`- F F T
`removal&
`(OFDM
`* : seriallparallel
`demodulator) + conversion
`
`Figure 2: OFDM system diagram.
`
`Page 7 of 8
`
`

`
`Wu and Zou: Orthogonal Frequency Division Multiplexing: A Multi-Camer Modulation Scheme
`
`399
`
`Dr.
`Yiyan Wu received
`the B. Eng.
`degree from
`the Beijing University of Posts
`and Telecommunications, Bei-
`jing, China in 1982, and the
`M. Eng. and Ph.D. Degrees in
`electrical engineering from Car-
`leton University, Ottawa, On-
`tario, Canada, in 1986 and 1990
`respectively.
`He was a member of technical staff in the Research Insti-
`tute of Telecommunications, Beijing, China, from 1982
`to 1984, and worked on video transmission, digital mi-
`crowave radio and video quality assessment. In 1990,
`he joined Telesat Canada and worked on digital TV and
`digital audio/video data compression and transmission.
`He is now a research scientist with the Communica-
`tions Research Centre, Ottawa, Canada. His research
`interests include digital video compression and trans-
`mission, high definition television (HDTV) , signal and
`image processing, satellite and mobile communications.
`He is couurently involved in the North America HDTV
`standard development and ITU-R digital television ter-
`restrial broadcasting (dTTb) study. He is a member of
`the IEEE Broadcast Techn