Modified Discrete Fourier Transform

Description

The Modified Discrete Fourier Transform (MDFT) is a specialized mathematical transform employed in the physical layer of 5G New Radio (NR). It is a variation of the standard Discrete Fourier Transform (DFT) that is optimized for generating sequences with specific properties crucial for modern wireless communication. Its primary application in 3GPP is for the generation of Demodulation Reference Signals (DM-RS) in the uplink, especially when using DFT-s-OFDM (Discrete Fourier Transform Spread Orthogonal Frequency Division Multiplexing) waveform, which is the 5G equivalent of the SC-FDMA used in LTE uplink. The MDFT is defined to create reference signal sequences that have low Peak-to-Average Power Ratio (PAPR) and good correlation properties.

Technically, the MDFT operates on a length-M input sequence (often a base sequence like a Zadoff-Chu sequence or a computer-generated sequence) and transforms it into a length-N output sequence, where M and N are specific integers defined by the standard, and typically M is less than or equal to N. The modification lies in the specific indexing and phase factors applied during the transform. The standard DFT formula is adjusted to ensure that the resulting frequency-domain symbols, when mapped to specific subcarriers and converted to the time domain via an Inverse DFT (IDFT) as part of the DFT-s-OFDM process, yield a time-domain signal with desirable properties. The key modification often involves a cyclic shift or a specific phase rotation in the transform definition to align the sequence optimally within the allocated resource blocks.

How it works in practice: For DM-RS generation, a base sequence is defined. The MDFT is applied to this base sequence to produce the reference signal sequence in the frequency domain. This sequence is then mapped to the designated DM-RS resource elements (subcarriers) within an OFDM symbol. The low-PAPR characteristic is critical for the uplink because it allows the User Equipment (UE) power amplifier to operate more efficiently, extending battery life and increasing coverage. The MDFT's design ensures that even for various sequence lengths and resource allocations, the generated DM-RS maintains consistent and optimal performance. Its role is embedded within the channel estimation process at the gNodeB, where the received DM-RS, transformed via the inverse process, is used to estimate the radio channel for accurate data demodulation.

Purpose & Motivation

The MDFT was introduced to address specific challenges in the 5G NR uplink waveform design. The primary purpose is to generate reference signal sequences (specifically for DM-RS) that possess a very low Peak-to-Average Power Ratio when used with the DFT-s-OFDM waveform. Low PAPR is a paramount requirement for uplink transmissions from mobile devices, as high PAPR forces the power amplifier to operate inefficiently in its non-linear region, leading to higher power consumption, reduced battery life, and unwanted out-of-band emissions. The standard DFT, when applied directly to some sequences, does not guarantee the lowest possible PAPR for all sequence lengths and resource allocations.

Historically, LTE used different mechanisms for uplink reference signals. With 5G's more flexible numerology (variable subcarrier spacing) and wider range of supported bandwidths, a more generalized and optimized method was needed. The MDFT provides this optimization. It solves the problem of efficiently generating a large set of reference signal sequences (to support many users and layers) that all maintain excellent PAPR characteristics. This was a key motivation for its creation in Release 18, as enhancements for uplink performance and efficiency were a focus area.

Furthermore, the MDFT enables better performance in high-mobility scenarios and for advanced multi-antenna techniques (MIMO). By providing reference signals with good auto-correlation and cross-correlation properties, channel estimation accuracy is improved, which is critical for achieving the high data rates and reliability targets of 5G-Advanced. Its specification in TS 26.253 (initially for other purposes) and adoption for physical layer signals exemplifies the cross-layer optimization in 3GPP, where mathematical tools are carefully selected to meet stringent system-level requirements for power efficiency, signal quality, and implementation flexibility.

Key Features

• Optimized variant of the DFT for generating low-PAPR sequences
• Primarily used for Demodulation Reference Signal (DM-RS) generation in 5G NR uplink
• Defined for specific input (M) and output (N) sequence lengths per 3GPP specification
• Enhances power amplifier efficiency in User Equipment by reducing signal peaks
• Supports the DFT-s-OFDM (SC-FDMA) waveform for uplink transmission
• Improves channel estimation accuracy through well-designed sequence properties

Evolution Across Releases

Defining Specifications