OFDM Symbol Timing Recovery

The fundamental goal is to place the FFT window entirely within the ISI-free region of the cyclic prefix. If the window starts too early, it captures a portion of the CP that is contaminated by the previous symbol's multipath tail. If it starts too late, it truncates the current symbol and includes samples from the next symbol. The ISI-free region is defined as the CP duration minus the maximum channel delay spread. Correct alignment preserves subcarrier orthogonality and prevents both inter-symbol interference and inter-carrier interference.

A foundational blind estimation technique exploits the redundancy introduced by the cyclic prefix. The CP is a copy of the last Ng samples of the OFDM symbol, prepended to the beginning. By computing the autocorrelation of the received signal with a lag equal to the useful symbol length (N), a correlation peak emerges at the start of each symbol. The maximum likelihood (ML) estimator derived by van de Beek uses this principle, correlating over a window of CP length and normalizing by signal energy to produce a robust timing metric.

A widely implemented data-aided algorithm that uses a dedicated training symbol with two identical halves in the time domain. The receiver computes an autocorrelation between the first and second halves of the received preamble. The timing metric reaches a plateau rather than a sharp peak, providing coarse timing. A second training symbol with a known differential sequence between subcarriers refines the estimate. This method is robust to frequency offset because the phase difference between the two halves is proportional to the carrier frequency offset, enabling joint estimation.

In cellular standards like LTE and 5G NR, timing recovery relies on cross-correlation with known synchronization sequences. The Primary Synchronization Signal (PSS) is a Zadoff-Chu sequence mapped to the central 62 subcarriers. The UE performs a sliding cross-correlation in the time domain against three candidate PSS sequences. The correlation peak identifies the 5 ms half-frame boundary and the sector identity (N_ID2). This method provides high processing gain and works reliably at low SNR, but requires prior knowledge of the transmitted sequence.

A timing offset within the ISI-free region causes a linear phase rotation across subcarriers proportional to the subcarrier index. This rotation is corrected by the channel equalizer. However, an offset outside this region introduces ISI and ICI, which manifest as a noise-like cloud around constellation points. The error vector magnitude (EVM) degrades sharply. For a timing error of δ samples, the phase rotation on subcarrier k is 2πkδ/N. This predictable rotation is exploited by pilot-based fine timing tracking loops that measure the phase slope across pilot subcarriers.

The channel delay spread directly reduces the available timing margin. In a multipath environment, the effective start of the received symbol is smeared over the delay spread duration. The optimal FFT window placement is a trade-off: starting earlier captures more of the early-arriving paths but risks ISI from the previous symbol; starting later avoids ISI but loses energy from the earliest paths. Maximum ratio combining of paths requires the window to capture the entire channel impulse response. Typical LTE CP lengths of 4.7 µs (normal) and 16.7 µs (extended) define the maximum tolerable delay spread for a given deployment.

A blind parameter estimation technique that identifies the number of subcarriers in an unknown OFDM signal before demodulation can occur.

• Analyzes the cyclostationary signature generated by the cyclic prefix at the symbol rate
• The spectral correlation function exhibits peaks at cyclic frequencies equal to multiples of the symbol rate
• Autocorrelation lag profiles reveal the useful symbol duration (Tu), from which FFT size is derived given the sampling rate
• Critical for spectrum monitoring and signal intelligence (SIGINT) applications where waveform parameters are unknown

A technique that discriminates between normal and extended cyclic prefix modes in unknown OFDM transmissions by analyzing the correlation lag structure.

• Computes the autocorrelation function over a range of candidate CP lengths
• The true CP length produces a distinct plateau in the correlation magnitude profile
• Extended CP mode is used in LTE MBMS and specific 5G NR numerologies for large cell deployments
• Enables automatic waveform classification between CP-OFDM variants without protocol-level decoding