SESSION SP24

SALON 8

NON-VOLATILE & SRAMS

Chair: T-S. Jung, Samsung, Seoul, Korea
Associate Chair: G. Atwood, Intel, Santa Clara, CA

24.1 - A 20MB/s Data Rate 2.5V Flash Memory with Current- Controlled Field Erasing for 1M-Cycle Endurance - 1:30 PM

M. Dallabora, C. Villa, F. Caser, S. Schippers, M. Sali, G. Ortolani, M. Defendi, M. Cane, L. Bettini, S. Bartoli, D. Cantarelli, R. Bez SGS-Thomson Microelectonics, Agrate Brianza, Italy

Flash memory endurance is improved to 1M cycles using circuit techniques that exploit the relationship between the source junction band-to-band current and erase gate current. The high initial erase field is eliminated. Erase time degradation is improved 2.5 times over conventional techniques.

24.2 - A 3.3V 16Mb Non-Volatile Virtual DRAM using NAND Flash Memory Technology - 2:00 PM

T-S. Jung, D-C. Choi, S-H. Cho, M-J. Kim, S-K. Lee, B-S. Choi, J-S. Yum, S-H. Kim, D-G. Lee, J-C. Son, M-S. Yong, H-K. Oh, S-B. Jun, W-M. Lee, S. Ali, E. Haq, K-D. Suh, H-K. Lim Samsung Electronics Co. Ltd., Kiheung, Korea

An 86.6mm2 16Mb non-volatile memory is DRAM-compatible except for an extended RAS precharge time after write. The device can be used as an execute-in-place memory while sharing the same data bus with DRAMs. 63ns tRAC, 21ns tAA, and 11ns tCAC are achieved using 0.5um CMOS technology with a 2um2 cell.

24.3 - A 300MHz Dual-Port Graphics RAM using Port-Swap Architecture - 2:30 PM

Y. Nakase, K. Mashiko, T. Tokuda Mitsubishi Electric Corporation, Itami, Japan

A 3.3V color palette SRAM uses 0.5um CMOS. The port-swap architecture realizes asynchronous single bitline read/write reducing memory cell size by >24%.

BREAK 3:00 PM

24.4 - A 2ns Access, 285MHz, 2-Port Cache Macro using Double Global Bitline Pairs - 3:15 PM

K. Osada, H. Higuchi, K. Ishibashi, N Hashimoto1, K. Shiozawa1 Hitachi, Ltd., Kokubunji/1Kodaira, Tokyo, Japan

A two-port 16kB cache macro uses 0.35um four-level-metal CMOS. The two-port cache utilizes a single-port, 6-device 21.67um2 SRAM memory cell with a hierarchical bit-line architecture to achieve 285MHz operation and a 2ns access time.

24.5 - A 350MHz 3.3V 4Mb SRAM in a 0.3um CMOS Process - 3:45 PM

D. Evans, G. Braceras, J. Sousa, J. Connor IBM Microelectronics Division, Essex Junction, VT

A 145mm2 4Mb SRAM using an 18.77um2 cell and organized as 128kx36 or 256kx18, achieves 4.7ns flow-through access using self-resetting, low-voltage-swing circuits. A voltage regulator supplies internal 2.5V with 1.4mV/deg temperature coefficient to compensate for performance degradation.

24.6 - A 500MHz 4Mb CMOS Pipeline-Burst Cache SRAM with Point-to-Point Noise Reduction Coding I/O - 4:15 PM

K. Nakamura, K. Takeda, H. Toyoshima, K. Noda, H. Ohkubo, T. Uchida, T. Shimizu, T. Itani, K. Tokashiki, K. Kishimoto NEC, Kanagawa, Japan

A 132mm2, 4Mb pipelined cache SRAM with 12.77um2 cell achieves 2GB/s bidirectional I/O. Controlled impedance push-pull output drivers and minimum-transition encoding reduce bus noise and power. Source resetting and bitline self equalization allow 3.5ns core access time.

CONCLUSION 4:45 PM


Go back to Table of Contents