Orthogonal Frequency Division Multiplexing (OFDM) is a multicarrier modulation scheme that splits a high-rate data stream across many narrowband, mutually orthogonal subcarriers transmitted simultaneously. The orthogonality condition allows subcarrier spectra to overlap without inter-carrier interference, making efficient use of available bandwidth.
In practice
OFDM is the physical-layer foundation of Wi-Fi (802.11a/g/n/ac/ax/be), LTE, 5G NR, DVB-T/T2, DAB, and many other standards, though some of these use OFDM-based variants or apply OFDM to only part of the system. Embedded developers working on these protocols typically interact with OFDM through a modem IC or a baseband SoC (such as Qualcomm, MediaTek, or Broadcom Wi-Fi chipsets) rather than implementing the full modulation chain themselves. However, understanding OFDM internals is necessary when tuning link-layer parameters, diagnosing RF issues, or developing SDR-based systems.
The core DSP operation in OFDM is the Inverse FFT (IFFT) at the transmitter and the FFT at the receiver. The IFFT maps frequency-domain symbols (one per subcarrier) to a time-domain waveform for transmission; the receiver reverses this with an FFT after synchronization. A cyclic prefix (CP) -- a copy of the tail of each OFDM symbol prepended to its beginning -- is inserted to absorb multipath delay spread and convert linear convolution with the channel into circular convolution, which the FFT can handle cleanly. CP length must exceed the channel's maximum excess delay; choosing it too short causes inter-symbol interference (ISI), while choosing it too long wastes spectral efficiency.
Multipath resilience is OFDM's primary advantage in wireless channels. Because each subcarrier's symbol period is long relative to the delay spread, individual subcarriers experience approximately flat fading even when the wideband channel is frequency-selective. The "A brief look at multipath radio channels" post on EmbeddedRelated gives a useful grounding in why this matters. The tradeoff is sensitivity to frequency offset and phase noise: any carrier frequency error or oscillator noise destroys orthogonality, causing inter-carrier interference (ICI) across all subcarriers simultaneously.
A significant practical challenge for OFDM transmitters is high Peak-to-Average Power Ratio (PAPR). Because many subcarriers can add coherently, instantaneous power can peak far above the average, requiring power amplifiers with large linear headroom or driving them into nonlinear compression -- a problem explored in "Radio Frequency Distortion Part II: A power spectrum model." Techniques such as clipping, tone reservation, and digital predistortion are used to manage PAPR in hardware-constrained designs.
Discussed on DSPRelated
Frequently asked
Why is the cyclic prefix necessary, and how is its length chosen?
The cyclic prefix (CP) makes the linear
convolution of the OFDM symbol with the multipath channel appear circular to the receiver's
FFT, which preserves subcarrier orthogonality and prevents inter-symbol interference (ISI) from preceding symbols -- provided the CP is longer than the channel's maximum excess delay and receiver timing is correct. In practice, standards define fixed CP lengths -- for example, 802.11a/g uses a CP of 16 samples (0.8 us at 20 MHz sampling) for a 64-point FFT -- balancing ISI protection against the overhead of transmitting non-data samples.
What does 'orthogonal' mean in OFDM, and why does it matter?
Two sinusoids are orthogonal over an interval T if their inner product (integral of their product) is zero. In OFDM, subcarriers are spaced at integer multiples of 1/T, which satisfies this condition exactly. Orthogonality means subcarrier spectra can overlap in frequency without one subcarrier corrupting another at the receiver's
FFT output, enabling much denser packing than conventional FDM with guard bands between carriers.
What is PAPR and why is it a problem for embedded hardware?
Peak-to-Average Power Ratio (PAPR) measures how much the instantaneous signal power can exceed the average. With many subcarriers adding in
phase, OFDM signals can have PAPR values of 10 dB or more. Power amplifiers must either be backed off to keep these peaks in the linear region -- reducing efficiency -- or they clip the peaks, introducing spectral regrowth and EVM degradation. This is a key constraint in battery-powered Wi-Fi or cellular devices where PA efficiency directly affects battery life.
Can an embedded MCU or DSP implement OFDM from scratch?
A full real-time OFDM PHY is computationally demanding: a 64-subcarrier 802.11a symbol requires a 64-point complex
FFT every 4 us, plus synchronization, channel estimation, and equalization. High-end DSPs (e.g., TI C66x, C7x) or FPGAs can handle this; general-purpose Cortex-M MCUs typically cannot sustain the required throughput for standard-compliant Wi-Fi or LTE rates.
SDR experimentation at reduced bandwidths (e.g., a few hundred kHz) is feasible on faster MCUs or Cortex-A class processors. The EmbeddedRelated post 'A Beginner's Guide to OFDM' covers the algorithmic building blocks at an accessible level.
How does OFDM differ from OFDMA, and where does OFDMA appear in embedded systems?
OFDM assigns all subcarriers to a single user per symbol period. OFDMA (Orthogonal Frequency Division Multiple Access) partitions the subcarriers among multiple users simultaneously, allowing the medium to be shared in the
frequency domain. OFDMA is used in LTE and 5G NR downlink/uplink scheduling and in Wi-Fi 6/6E/7 (802.11ax/be) to serve multiple stations concurrently -- a major efficiency gain for dense deployments. From a PHY implementation standpoint, the
FFT/
IFFT core is the same; the difference is in resource allocation and MAC-layer scheduling.
Differentiators vs similar concepts
OFDM is sometimes confused with OFDMA and with classic FDM. Classic FDM separates carriers with guard bands to prevent overlap, wasting spectrum; OFDM uses mathematical orthogonality to allow overlapping subcarrier spectra without interference. OFDMA extends OFDM by dynamically assigning subsets of subcarriers to different users within the same symbol period, enabling multiple-access; it is a multiple-access scheme built on top of OFDM rather than a separate modulation family. OFDM is also distinct from spread-spectrum schemes (DSSS, FHSS): at the modulation level, OFDM transmits independent symbols on each narrowband subcarrier simultaneously rather than spreading a signal across the band, though practical OFDM systems typically layer coding and interleaving on top of this structure.
