Real Fast Fourier Transform

Description

The Real Fast Fourier Transform (RFFT) is a specialized implementation of the Fast Fourier Transform algorithm optimized for processing real-valued input data sequences. In the context of 3GPP 5G New Radio (NR), it is a fundamental signal processing block used extensively within the physical layer, particularly for Orthogonal Frequency Division Multiplexing (OFDM) waveforms. The OFDM system relies on transforming time-domain signals into the frequency domain for subcarrier modulation and demodulation. Since many baseband signals, such as in-phase (I) and quadrature (Q) components from an analog-to-digital converter, are real-valued, using a standard complex FFT would be computationally wasteful as it would process redundant imaginary parts set to zero.

The RFFT algorithm exploits the Hermitian symmetry property inherent in the Fourier transform of real-valued data. This property states that for a real input sequence of length N, the resulting complex frequency-domain output exhibits symmetry: the value at frequency bin k is the complex conjugate of the value at bin N-k. Consequently, an RFFT algorithm only needs to compute and store approximately half of the output points (N/2 + 1 for even N), significantly reducing memory requirements and computational operations. The typical implementation involves pre-processing the real input, performing a complex FFT of half the length, and then post-processing the results to reconstruct the full set of unique frequency bins.

Within the 5G NR architecture, the RFFT is a core component of the channel estimation and equalization process in the receiver chain. After the cyclic prefix is removed from an OFDM symbol, the time-domain samples are fed into an RFFT block to convert them into subcarrier values in the frequency domain. These values are then used for tasks like demodulating Quadrature Amplitude Modulation (QAM) symbols, estimating the channel frequency response via reference signals (e.g., DM-RS, CSI-RS), and performing interference cancellation. Its efficiency directly impacts the power consumption and processing latency of User Equipment (UE) and gNodeB base stations, making it vital for meeting 5G's high-throughput and low-latency requirements.

The specification of RFFT characteristics, including supported transform sizes and numerical precision, is detailed in 3GPP TS 26.118, which focuses on performance requirements for the UE's radio transmission and reception. The transform sizes are aligned with the supported OFDM numerology (subcarrier spacing) and bandwidth configurations in NR. By standardizing these processing requirements, 3GPP ensures interoperability and consistent performance benchmarks across different chipset vendors and network equipment manufacturers, enabling a robust and efficient global 5G ecosystem.

Purpose & Motivation

The RFFT was introduced to address the specific computational demands of 5G New Radio's physical layer, which employs wide bandwidths and high-order OFDM numerologies. Processing these wideband signals with standard complex FFTs would impose an unnecessary computational burden on devices, leading to higher power consumption, increased silicon area, and potential thermal issues, especially in battery-powered User Equipment. The primary motivation was to define a standardized, efficient signal processing method that leverages the inherent properties of real-valued baseband signals to reduce complexity without sacrificing performance.

Historically, as wireless standards evolved from 3G to 4G LTE and then to 5G, the channel bandwidths and data rates increased exponentially. LTE primarily used a 20 MHz maximum bandwidth, while 5G NR supports bandwidths up to 400 MHz in FR2 (mmWave) and 100 MHz in FR1 (sub-6 GHz). This massive increase in sampling rates and FFT sizes made computational efficiency a critical design constraint. The RFFT provides a mathematically optimal solution, cutting the required number of multiplications and additions nearly in half compared to a full complex FFT of the same length.

By specifying the RFFT in Release 15, 3GPP provided a clear performance target for baseband processor designers. It solved the problem of defining how a UE should efficiently perform the essential Fourier transform operation, ensuring that all compliant devices achieve a baseline level of efficiency. This standardization prevents fragmentation where different vendors might implement proprietary optimizations, potentially leading to interoperability issues or inconsistent performance metrics in network testing and certification.