Improving MRAM Performance with Sparse Modulation and Hamming Error Correction.

Area of Science:

• Computer Engineering
• Materials Science
• Data Storage Technologies

Background:

• The Internet of Things (IoT) relies on smart sensors as edge processing units requiring continuously writable, low-power, fast-access memory.
• Magnetic Random-Access Memory (MRAM) offers advantages over DRAM/SDRAM but faces challenges like process variations and thermal fluctuations causing bit errors.
• These MRAM errors lead to imbalanced probabilities of 0 and 1 bits, impacting data integrity.

Purpose of the Study:

• To develop an effective error correction scheme for MRAM, addressing bit errors and probability imbalances.
• To enhance data reliability and decoding accuracy in MRAM for IoT edge computing applications.
• To propose a novel sparse coding method tailored to MRAM's inherent error characteristics.

Main Methods:

• A novel sparse coding scheme with a minimum Hamming distance of three was developed.
• Three check bits are appended to user data and processed via a generator matrix; codewords failing sparsity constraints are inverted.
• A dynamic threshold detection technique was introduced for real-time bit probability estimation during data transmission.

Main Results:

• The proposed sparse coding scheme significantly improves error resilience in MRAM.
• Dynamic threshold detection enhances decoding accuracy, especially at higher MRAM densities.
• Simulations demonstrated substantial gains in both error correction and data integrity.

Conclusions:

• The novel sparse coding and dynamic threshold detection effectively mitigate MRAM errors.
• These techniques are crucial for reliable MRAM deployment in demanding IoT edge applications.
• The methods show promise for future high-density, continuously writable memory systems.