Raul-Adrian Cernea

A 34 MB/s MLC Write Throughput 16 Gb NAND With All Bit Line Architecture on 56 nm Technology

A 16 Gb 4-state MLC NAND flash memory augments the sustained program throughput to 34 MB/s by fully exercising all the available cells along a selected word line and by using additional performance enhancement modes. The same chip operating as an 8 Gb SLC device guarantees over 60 MB/s programming throughput. The newly introduced all bit line (ABL) architecture has multiple advantages when higher performance is targeted and it was made possible by adopting the ldquocurrent sensingrdquo (as opposed to the mainstream ldquovoltage sensingrdquo) technique. The general chip architecture is presented in contrast to a state of the art conventional circuit and a double size data buffer is found to be necessary for the maximum parallelism attained. Further conceptual changes designed to counterbalance the area increase are presented, hierarchical column architecture being of foremost importance. Optimization of other circuits, such as the charge pump, is another example. Fast data access rate is essential, and ways of boosting it are described, including a new redundancy scheme. ABL contribution to energy saving is also acknowledged.

A 5.6MB/s 64Gb 4b/Cell NAND Flash memory in 43nm CMOS

Today NAND Flash memory is used for data and code storage in digital cameras, USB devices, cell phones, camcorders, and solid-state disk drives. Figure 13.6.1 shows the memory-density trend since 2003. To satisfy the market demand for lower cost per bit and higher density nonvolatile memory, in addition to technology scaling, 2b/cell MLC technology was introduced. Recently, MLC NAND flash memories with more than 2b/cell [1,2] have been reported.

A 34MB/s-Program-Throughput 16Gb MLC NAND with All-Bitline Architecture in 56nm

In the diverse world of NAND flash applications, higher storage capacity is not the only imperative. Increasingly, performance is a differentiating factor and is also a way of creating new markets or expanding existing markets. While conventional memory uses, for actual operations, every other cell along a selected word line (WL) (Takeuchi, 2006), this design simultaneously exercises them all. A performance improvement of at least 100% is derived from this all-bitline (ABL) architecture relative to conventional chips. Additional techniques push performance to even higher levels.