Comprehensive Overview of NAND Flash Technology, Manufacturing Processes, and Packaging
This article provides a detailed technical overview of NAND Flash technology, covering wafer fabrication, die and ink‑die testing, various packaging methods such as TSOP, BGA and LGA, stack‑die techniques, SLC/MLC/TLC characteristics, flash reliability, and the evolution of semiconductor process nodes.
Wafer, the base substrate for semiconductor chips, is a thin circular slice cut from a high‑purity silicon crystal ingot grown from quartz sand.
In wafer fabrication, photomasks create photoresist patterns, followed by precise etching and metal deposition to form individual dies; the wafer backside receives a gold layer for die‑attach.
Modern NAND Flash wafers are typically 8‑inch or 12‑inch, yielding only a few hundred chips per wafer.
Silicon is refined from quartz to polysilicon (>99.99% purity), melted, seeded, and pulled into single‑crystal ingots, which are then sliced, polished, and processed into wafers.
Each die (unpackaged chip) is laser‑cut from the wafer; good dies pass testing, while failing ones are marked as "ink die" on a mapping map.
Common NAND Flash package types include TSOP (Thin Small Outline Package), BGA (Ball Grid Array) and LGA (Land Grid Array), each offering different size, cost, and performance trade‑offs.
TSOP uses a simple lead frame and wire‑bonding, suitable for low cost but limited to two‑sided wire bonding, making stack‑die difficult.
BGA employs a substrate allowing four‑sided wire bonding, facilitating stacked dies at higher cost.
LGA provides an array of contacts on the package bottom, used for high‑speed logic LSIs despite higher manufacturing complexity.
Stack‑die technology stacks multiple dies within a single package, using wire bonding to interconnect them; current implementations can achieve up to 8 layers, enabling capacities of up to 32 GB per package.
NAND Flash cells are classified as SLC (Single Level Cell), MLC (Multi‑Level Cell), and TLC (Triple Level Cell), with SLC offering the highest endurance (~100 k program/erase cycles) and MLC/TLC providing higher density but lower endurance.
Flash memory wear is caused by high‑voltage tunnel injection and release during programming and erasing; bad blocks appear after the endurance limit is reached, and a small fraction of blocks are defective from manufacturing.
Semiconductor process nodes have progressed from 0.5 µm in the mid‑1990s to 34 nm today, with wafer sizes increasing from 5 inches to 12 inches, driving higher integration density and supporting Moore’s Law.
For further details, refer to the original article link.
Architects' Tech Alliance
Sharing project experiences, insights into cutting-edge architectures, focusing on cloud computing, microservices, big data, hyper-convergence, storage, data protection, artificial intelligence, industry practices and solutions.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.