Multicarrier and OFDM Systems

Introduction and Context

A frequency-selective wireless channel can attenuate different spectral components by substantially different amounts and can produce delayed replicas of a transmitted signal. When the symbol duration is short relative to the channel delay spread, these replicas cause intersymbol interference (ISI). Equalizing such a channel directly in the time domain may be computationally demanding, particularly at high data rates.

Multicarrier modulation addresses this problem by dividing a high-rate serial data stream into many lower-rate parallel streams. Each stream modulates a separate subcarrier. Because the symbol duration on an individual subcarrier is longer, the relative effect of delay spread is reduced. Orthogonal frequency-division multiplexing (OFDM) is a particularly efficient multicarrier technique in which adjacent subcarriers overlap spectrally while remaining mathematically orthogonal.

OFDM is central to wireless LANs, digital broadcasting, fourth- and fifth-generation cellular systems, broadband fixed access, and power-line communications. Its success results from a useful combination of spectral efficiency, simple frequency-domain equalization, compatibility with multiple-input multiple-output (MIMO) processing, and flexible resource allocation. Its principal disadvantages are sensitivity to frequency offset and phase noise, a high peak-to-average power ratio (PAPR), and the need for careful synchronization.

Learning Objectives

After completing this chapter, the reader should be able to:

1. Explain the motivation and operating principles of multicarrier modulation.
2. Derive the orthogonality condition for OFDM subcarriers.
3. Relate the continuous-time OFDM signal to the inverse discrete Fourier transform (IDFT) and discrete Fourier transform (DFT).
4. Explain the purpose and design of the cyclic prefix (CP).
5. Analyze OFDM performance in frequency-selective channels, including one-tap equalization.
6. Quantify PAPR and discuss practical reduction methods.
7. Evaluate synchronization, channel estimation, implementation, and system-design trade-offs.
8. Apply these concepts to a realistic broadband communication scenario.

1. Multicarrier Modulation

In a single-carrier system, a sequence of symbols is transmitted using one pulse-shaped carrier. If the channel impulse response has a duration comparable to the symbol interval, successive symbols overlap. A conventional equalizer must then estimate and compensate for the channel memory.

In a multicarrier system, the available bandwidth is divided into N closely spaced subchannels. The input symbols are grouped into blocks and mapped to complex values X₀, X₁, …, X_{N−1}, usually from a QAM constellation. Each value modulates one subcarrier. The aggregate signal is the sum of all subcarrier waveforms.

The lower symbol rate on each subcarrier increases the symbol interval. In addition, if the subcarrier spacing and symbol interval are chosen correctly, the subcarriers are orthogonal over one useful symbol interval. Orthogonality permits spectral overlap without mutual interference under ideal synchronization.

A useful comparison is given below.

2. Orthogonality of OFDM Subcarriers

Let the useful OFDM symbol duration be Tᵤ and let the subcarrier spacing be

The kth complex subcarrier is represented by

where j = √−1 and k is an integer subcarrier index. The inner product of subcarriers k and m is

For k = m, the integral equals Tᵤ. For k ≠ m, k − m is a nonzero integer and the integral is zero. Thus,

where the asterisk denotes complex conjugation. This is the mathematical basis of OFDM orthogonality. It assumes ideal timing, a constant frequency reference during the useful symbol, and a receiver integration interval aligned with the transmitter interval.

The frequency-domain spectra of individual rectangularly windowed subcarriers have sinc-shaped main lobes and sidelobes. At the centre frequency of one subcarrier, the sidelobes of all other subcarriers have zero crossings. Consequently, overlapping spectra do not produce interference at the ideal sampling points.

3. OFDM Signal Model and the DFT/IDFT

Let Xₖ be the complex data or pilot symbol assigned to subcarrier k. The complex baseband OFDM signal during one useful symbol is

Sampling at intervals Tₛ = Tᵤ/N gives tₙ = nTᵤ/N and, because Δf = 1/Tᵤ,

Depending on the DFT sign convention, this is the N-point IDFT, often implemented efficiently by an inverse fast Fourier transform (IFFT). A normalized version is

The corresponding receiver DFT is

The normalization convention does not change the underlying modulation, but it changes the numerical relationship between sample energy and subcarrier-symbol energy. A practical design must use one convention consistently in power, noise, and SNR calculations.

Not every DFT bin must carry data. Some bins may be reserved for pilots, guard bands, or a DC null. In a real passband transmitter, Hermitian symmetry can be imposed on the frequency-domain vector so that the time-domain samples are real-valued, as in some direct-current-biased optical or baseband systems.

Figure: Principal signal-processing operations in an OFDM link.

4. Cyclic Prefix and Frequency-Selective Channels

Consider a discrete-time channel with impulse response h[ℓ], where ℓ = 0,…,L−1, and additive noise w[n]. The received signal is

If a guard interval of length Ncp is inserted before every OFDM symbol by copying its final Ncp samples, and if

then, after the receiver discards the CP, the channel convolution over the N useful samples is circular. The DFT diagonalizes circular convolution:

where Hₖ is the N-point DFT of the zero-padded channel impulse response, Yₖ is the received subcarrier value, and Wₖ is transformed noise. Each subcarrier can therefore be equalized independently.

For zero-forcing (ZF) equalization,

ZF removes the channel gain but can strongly amplify noise when |Hₖ| is small. A linear minimum mean-square error (MMSE) equalizer, under a commonly used complex Gaussian signal/noise model, is

and X̂ₖ = GₖYₖ. Here Hₖ* is the conjugate of Hₖ, Eₛ is the average symbol energy on the subcarrier, and N₀ is the corresponding noise spectral-density parameter under the adopted complex-baseband convention. The precise factor relating N₀ to two-sided real noise density depends on the convention used; consistency is essential.

The CP is not free: it consumes time and reduces useful throughput. If the useful symbol duration is Tᵤ and the CP duration is Tcp, the CP efficiency is

For example, Tᵤ = 3.2 μs and Tcp = 0.8 μs give ηCP = 0.8, or 80%, before accounting for pilots, coding, guard bands, and other overhead. A CP longer than necessary improves tolerance to delay spread but reduces efficiency and increases energy consumption.

5. Bit Loading, Power Allocation, and Spectral Efficiency

A broadband channel may provide substantially different SNRs on different subcarriers. OFDM permits adaptive modulation and coding: robust low-order modulation can be assigned to weak bins, while high-order QAM can be assigned to strong bins. Subcarriers affected by deep fades may be left unused.

For an ideal Gaussian-input parallel channel with subcarrier SNR γₖ, the theoretical capacity per OFDM symbol interval is

This is an upper bound, not the throughput of a practical coded-QAM system. It assumes ideal coding over sufficiently long blocks, accurate channel knowledge, and appropriate power allocation. Under a total power constraint, water-filling allocates more power to favorable subchannels and none to sufficiently poor subchannels. Practical systems use discrete modulation orders, coding-rate tables, pilot overhead, regulatory masks, and implementation margins instead of exact continuous water-filling.

The gross bit rate of an OFDM waveform can be approximated by

where Ndata is the number of data-bearing subcarriers, b is the number of coded bits per QAM symbol, and Rc is the code rate. This expression assumes the same modulation and code rate on all data bins; adaptive systems require a sum over subcarriers.

6. PAPR and Power-Amplifier Constraints

The OFDM time waveform is a sum of many independently modulated sinusoids. Their peaks can add constructively. The peak-to-average power ratio is

The definition applies to sampled samples; oversampling is normally required to estimate continuous-time peaks more accurately. A large PAPR forces the power amplifier to operate with substantial input back-off. Operating near saturation causes nonlinear distortion, spectral regrowth, and in-band error-vector magnitude degradation.

Common PAPR-reduction methods include:

• Clipping and filtering: simple but introduces in-band distortion and may cause out-of-band regrowth after filtering.
• Selected mapping: generates phase-rotated alternatives and transmits the lowest-PAPR candidate; side information or reliable blind detection is required.
• Partial transmit sequences: optimizes phase factors for subblocks, with increased computational complexity.
• Tone reservation: dedicates subcarriers to peak cancellation, reducing data efficiency.
• Tone injection: expands constellation points to reduce peaks, increasing power or decoding complexity.
• Coding: excludes symbol sequences with large peaks, at a rate and complexity cost.

PAPR reduction should be evaluated jointly with amplifier efficiency, error performance, spectral emissions, side-information reliability, and computational cost. A low measured sample PAPR is not automatically a system-level improvement.

7. Synchronization and Channel Estimation

OFDM is sensitive to synchronization errors because orthogonality depends on precise frequency and timing relationships.

7.1 Timing offset

A timing offset within the CP can often be tolerated without intercarrier interference (ICI), although it produces a subcarrier-dependent phase rotation. An offset outside the safe CP region causes ISI and ICI. Timing estimators commonly exploit a known preamble, repeated training sequences, or pilot correlations.

7.2 Carrier-frequency offset

Let ε denote the frequency offset normalized by the subcarrier spacing. A nonzero ε causes the received signal from one subcarrier to leak into others. This ICI becomes particularly severe when the offset is a significant fraction of Δf. Frequency offset may arise from oscillator mismatch, Doppler shift, or transmitter-receiver motion.

Coarse frequency synchronization is usually performed using a preamble, followed by fine tracking with pilots or decision-directed loops. The design must distinguish sampling-clock offset from carrier-frequency offset: the former causes a progressive timing drift across OFDM symbols, while the latter primarily causes phase rotation and ICI.

7.3 Channel estimation

A pilot-based estimate on a known subcarrier is obtained conceptually as

for a nonzero known pilot Xₖ. Estimates may be interpolated across frequency and time. Comb-type pilots occur in many OFDM symbols and are useful for time-varying channels; block-type pilots occupy many or all subcarriers in selected symbols and are useful when the channel varies slowly.

Interpolation must account for the channel’s coherence bandwidth and coherence time. If pilot spacing is too large relative to these quantities, the interpolated estimate is unreliable. MIMO systems additionally require orthogonal or otherwise separable pilot patterns for different transmit antennas.

8. OFDM Variants and Related Architectures

Cyclic-prefix OFDM is the conventional form in many broadband systems. Discrete multitone (DMT) is essentially OFDM used in wired channels, often with adaptive bit and power loading. Filter-bank multicarrier (FBMC) uses subcarrier filtering to reduce sidelobes and may avoid a conventional CP, but generally has more complex filtering and synchronization. Universal-filtered multicarrier (UFMC) filters groups of subcarriers. Generalized frequency-division multiplexing (GFDM) allows flexible block structures but may require interference management.

Single-carrier frequency-domain equalization (SC-FDE) uses an FFT-based receiver similar to OFDM while retaining a single-carrier-like transmit waveform with lower PAPR. This illustrates an important engineering principle: the FFT is a computational tool for equalization, whereas OFDM is a specific multicarrier waveform structure.

9. Worked Design Example

Consider an OFDM system with N = 1024 subcarriers, sampling rate Fs = 15.36 MHz, and a CP of 72 samples. Assume 600 active data subcarriers, 64-QAM, code rate Rc = 3/4, and one OFDM symbol per useful IFFT block.

The useful symbol duration is

The CP duration is

The total symbol duration is 71.354 μs. Since 64-QAM carries b = log₂64 = 6 coded bits per data subcarrier, the net coded payload per OFDM symbol is

Ignoring higher-layer overhead, the gross net bit rate is

The CP efficiency is

or 93.43%. The subcarrier spacing is

This design is suitable only if the effective channel delay spread is no greater than approximately 72 samples minus one sample, subject to implementation margins. If the channel requires a longer CP, the designer must increase the CP, reduce the useful symbol duration, or use a different waveform and equalization strategy.

10. Applied Case: OFDM in a High-Speed Indoor WLAN

Background

Consider an indoor wireless network serving laptops, industrial sensors, and video terminals. Reflections from walls, machinery, and people create a frequency-selective channel. The channel changes as users move, but not so rapidly that it becomes untrackable within a short frame.

Problem

A single-carrier design operating at the aggregate data rate requires a long adaptive equalizer. The equalizer must cope with multiple delayed paths and must update rapidly under mobility. Hardware complexity and power consumption become significant.

Analysis

The designer first measures the expected maximum excess delay and chooses a useful symbol duration sufficiently longer than that delay. A CP is selected to exceed the channel memory with a margin. The subcarrier spacing is then chosen as the reciprocal of the useful symbol duration. A larger spacing shortens the symbol and reduces sensitivity to phase noise, but provides less immunity to delay spread; a smaller spacing improves delay-spread tolerance but increases sensitivity to oscillator phase noise and Doppler over a symbol.

Pilot-bearing training fields estimate the channel on each spatial stream. After CP removal and FFT processing, the receiver performs one-tap equalization per subcarrier and feeds soft values to the channel decoder. Link adaptation selects QPSK, 16-QAM, 64-QAM, or higher-order modulation according to measured SNR and error statistics.

Approach and solution

The system uses convolutional or modern block coding, interleaving across frequency and time, pilot-assisted channel estimation, automatic gain control, coarse and fine synchronization, and a power amplifier with sufficient linearity. MIMO spatial multiplexing increases throughput when the channel provides multiple independent spatial paths. Subcarriers with deep fades are assigned lower-rate modulation or are excluded.

Outcome

Compared with a single-carrier equalizer at the same aggregate rate, the OFDM receiver achieves a simpler frequency-domain equalization structure and flexible adaptation. The price is CP overhead, high PAPR, and increased sensitivity to frequency offset. Performance therefore depends not only on the theoretical subcarrier SNR but also on synchronization quality, amplifier back-off, pilot density, and implementation loss.

Learning points

1. OFDM is valuable because it converts a difficult broadband equalization problem into many simpler narrowband problems.
2. The CP must be designed from measured or specified channel delay spread, not selected arbitrarily.
3. Adaptation and coding are as important as the waveform itself.
4. A waveform comparison must include RF efficiency and synchronization, not only spectral efficiency.

11. Design Trade-offs, Limitations, and Best Practices

The central design trade-off is between subcarrier spacing, delay-spread tolerance, Doppler tolerance, phase-noise sensitivity, and implementation complexity. Increasing N while holding sampling bandwidth constant reduces subcarrier spacing and increases useful symbol duration, but may make the system more sensitive to time variation and phase noise. Reducing N has the opposite effects.

Best practices include:

• Select CP length from a statistically justified channel-delay model, including deployment margin.
• Reserve adequate guard bands and comply with spectral-emission regulations.
• Use windowing or filtering when sidelobe leakage is important, while checking its effect on orthogonality.
• Estimate PAPR using oversampled waveforms and evaluate actual power-amplifier error-vector magnitude.
• Design synchronization and channel estimation jointly with the numerology.
• Use pilot patterns matched to coherence bandwidth, coherence time, and MIMO structure.
• Report throughput after coding, pilots, CP, control information, retransmissions, and implementation losses.
• Validate simulations with fixed-point, RF impairment, and hardware-in-the-loop testing.

12. Contemporary Developments and Research Directions

Modern systems use flexible OFDM numerologies, scalable subcarrier spacing, beamforming, massive MIMO, channel coding based on LDPC or polar codes, and sophisticated link adaptation. Research continues on reducing PAPR, improving operation under severe phase noise, designing waveforms for high mobility, and integrating sensing with communications.

Machine learning can assist in channel estimation, nonlinear amplifier compensation, PAPR reduction, and resource allocation. However, learned methods require representative training data, robustness testing, explainable failure analysis, and protection against distribution shifts. A neural receiver that performs well in simulation may fail when oscillator phase noise, clipping, hardware imbalance, or an unseen channel profile is introduced. AI should therefore complement, rather than replace, physically grounded synchronization, estimation, and verification.

Regulatory and professional responsibilities include compliance with spectrum masks, control of interference to adjacent systems, electromagnetic-exposure requirements where applicable, secure firmware and update mechanisms, and honest reporting of measured rather than ideal performance.

Chapter Summary

Multicarrier modulation divides a high-rate stream among parallel lower-rate subchannels. OFDM selects subcarrier spacing as the reciprocal of the useful symbol duration, making the subcarriers orthogonal over the FFT interval. The IDFT generates the time-domain waveform, while the receiver removes the cyclic prefix and applies the DFT.

When the CP is at least as long as the channel memory, linear convolution becomes circular over the useful interval. The FFT then converts the frequency-selective channel into approximately independent subchannels, enabling simple one-tap ZF or MMSE equalization. OFDM supports adaptive modulation, coding, pilot-assisted estimation, and MIMO, but it incurs CP overhead and high PAPR and is sensitive to timing error, carrier-frequency offset, phase noise, and Doppler.

Successful OFDM design is therefore a system-level exercise involving waveform parameters, channel statistics, synchronization, channel estimation, coding, RF linearity, regulatory constraints, and implementation validation.

Key Terms

Orthogonality: A zero inner product between distinct subcarrier waveforms over the useful symbol interval.

Multicarrier modulation: Transmission using multiple parallel subchannels or subcarriers.

OFDM: Orthogonal frequency-division multiplexing with spectrally overlapping but orthogonal subcarriers.

Useful symbol duration: The FFT interval excluding the cyclic prefix.

Subcarrier spacing: Frequency separation between adjacent subcarriers, normally 1/Tᵤ.

IFFT: Efficient implementation of the inverse DFT used to synthesize an OFDM symbol.

FFT: Efficient implementation of the DFT used to recover subcarrier values.

Cyclic prefix: A copied portion of the OFDM symbol inserted as a guard interval.

Channel impulse response: Time-domain representation of multipath channel behavior.

Circular convolution: Convolution operation diagonalized by the DFT.

Intercarrier interference: Leakage between subcarriers caused by loss of orthogonality.

Inter-symbol interference: Overlap between successive transmitted symbols.

Pilot: A known symbol used for synchronization or channel estimation.

PAPR: Ratio of maximum instantaneous power to average signal power.

ZF equalizer: Equalizer that divides by the estimated channel response.

MMSE equalizer: Equalizer balancing channel inversion and noise enhancement.

Bit loading: Assignment of modulation orders to individual subcarriers.

Coherence bandwidth: Frequency range over which a channel response is strongly correlated.

Coherence time: Time interval over which a channel remains significantly correlated.

Spectral regrowth: Out-of-band emissions caused by nonlinear amplification.

Review Questions

Short Answer Questions

1. Why does multicarrier modulation reduce the equalization difficulty of a frequency-selective channel?
2. State the relationship between useful OFDM symbol duration and subcarrier spacing.
3. What condition should the cyclic-prefix length satisfy relative to channel impulse-response length?
4. Why can adjacent OFDM subcarrier spectra overlap without causing ideal mutual interference?
5. What is the principal cause of high PAPR in OFDM?

Long Answer Questions

1. Derive the orthogonality condition for two OFDM subcarriers and explain its practical assumptions.
2. Explain how the cyclic prefix converts linear convolution into circular convolution and enables one-tap equalization.
3. Compare ZF and MMSE frequency-domain equalizers, including their behavior in deep fades.
4. Discuss the principal sources of OFDM synchronization error and methods for estimating and correcting them.
5. Evaluate PAPR-reduction methods with respect to distortion, complexity, side information, and power-amplifier efficiency.

Analytical / Application Questions

1. An OFDM system has 2048 samples per useful symbol, a sampling rate of 30.72 MHz, and a CP of 144 samples. Calculate the useful symbol duration, CP duration, subcarrier spacing, and CP efficiency.
2. A channel has four significant taps whose earliest and latest taps are separated by 80 samples. Explain the minimum CP requirement and the consequences of using a 40-sample CP.
3. Design a qualitative link-adaptation strategy for an OFDM system in which some subcarriers experience deep fades, the channel varies moderately with time, and the power amplifier has limited linearity.

Multiple Choice Questions with Answers

Question 1

For ideal OFDM subcarriers, what condition guarantees orthogonality over the useful symbol interval Tᵤ?

A. The subcarrier spacing is Tᵤ

B. The subcarrier spacing is 1/Tᵤ

C. The subcarrier spacing is N/Tᵤ²

D. All subcarriers have different amplitudes

Correct Answer: B. The subcarrier spacing is 1/Tᵤ

Question 2

What is the primary purpose of the cyclic prefix in CP-OFDM?

A. To increase constellation size

B. To reduce the number of subcarriers

C. To make channel convolution circular over the FFT interval

D. To eliminate thermal noise

Correct Answer: C. To make channel convolution circular over the FFT interval

Question 3

An OFDM system has N = 1024 and sampling rate Fs = 15.36 MHz. What is the useful symbol duration?

A. 6.667 μs

B. 33.333 μs

C. 66.667 μs

D. 1024 ms

Correct Answer: C. 66.667 μs

Question 4

If the channel impulse response has length L samples, which CP condition avoids ISI under the ideal quasi-static model?

A. Ncp = 0

B. Ncp ≥ L − 1

C. Ncp ≤ L − 1

D. Ncp = N + L

Correct Answer: B. Ncp ≥ L − 1

Question 5

Which equalizer is most likely to amplify noise on a subcarrier experiencing a deep channel fade?

A. Zero-forcing equalizer

B. Matched filter with no channel estimate

C. A source encoder

D. A cyclic-prefix inserter

Correct Answer: A. Zero-forcing equalizer

Question 6

What is the main RF consequence of high OFDM PAPR?

A. The amplifier can always operate at saturation without distortion

B. The amplifier requires back-off to limit nonlinear distortion

C. The channel impulse response becomes shorter

D. Pilot symbols become unnecessary

Correct Answer: B. The amplifier requires back-off to limit nonlinear distortion

Question 7

A system uses 600 data subcarriers, 64-QAM, and code rate 3/4. How many net coded bits are conveyed per OFDM symbol under the stated assumptions?

A. 1800

B. 2400

C. 2700

D. 3600

Correct Answer: C. 2700

Question 8

Which impairment directly causes leakage of energy between OFDM subcarriers by destroying frequency orthogonality?

A. Carrier-frequency offset

B. Correct CP insertion

C. Ideal channel estimation

D. Gray labeling of QAM symbols

Correct Answer: A. Carrier-frequency offset

Question 9

Which pilot arrangement is generally suitable when the channel changes slowly and a detailed channel estimate is needed at selected instants?

A. Block-type pilots

B. No pilots under all conditions

C. Random CP deletion

D. Only a DC subcarrier

Correct Answer: A. Block-type pilots

Question 10

For an OFDM symbol with useful duration 3.2 μs and CP duration 0.8 μs, what is the CP efficiency?

A. 20%

B. 25%

C. 80%

D. 400%

Correct Answer: C. 80%