Has FRAM's time come?

Has FRAM's time come?
As the alert reader will remember, I was reared in Dayton, Ohio, home of the Wright brothers. When faced with a technology that is struggling to achieve liftoff, I often recall the story of an editor who received word of the Wrights' successful flight of a motor-powered aircraft near Kitty Hawk, N.C., but shrugged it off, commenting that blimps had stayed aloft far longer.

It would be equally easy to give scant attention to several papers about ferroelectric memories, scheduled for the International Electron Devices Meeting Dec. 8-11 in San Francisco. According to the IEDM abstracts, three presentations mention 4-Mbit embedded-FRAM densities, a shrug-provoking number when you consider that the same 2002 IEDM will feature a discrete 2-Gbit flash device from Samsung Electronics.

A joint paper from Texas Instruments, Agilent Technologies and Ramtron International describes a 1.5-volt embedded-FRAM technology that requires only two additional mask layers. TI has stuck by embedded SRAM thus far, rejecting embedded DRAM and flash because they require an extra six to eight mask layers.

TI's IEDM abstract describes a 130-nanometer process that co-integrates eFRAM, with a cell size of 0.58 micron2, with SRAM that requires 1.95 micron2 per cell.

Fujitsu researchers also will go to IEDM to present work on a 4-Mbit embedded FRAM aimed at system-on-chip designs. An Oki Electric-Sony paper will describe another 4-Mbit embedded-FRAM technology that can be combined with logic. The Oki-Sony group developed a pedestal-type capacitor that brings particular advantages to merging logic with embedded FRAM.

Samsung researchers are scheduled to present their thoughts on what it will take to move FRAM technology beyond low bit densities. And another Samsung paper will discuss integrating magnetoresistive RAM with logic.

Are companies as varied as Agilent, Fujitsu, Oki, Samsung, Sony and TI ready to take discrete and embedded FRAM to the market in a big-time way? Having seen the hype over embedded DRAM run up against hard manufacturing realities, it surely is too early to declare embedded FRAM as a sure thing. At this point, let's just say that creating a nonvolatile embedded memory with only two additional mask layers is worth noting.

In a few years, when powerful computing and communications reach the mobile-systems market, the industry will see whether embedded FRAM really flies.

Please send feedback to dlammers@cmp.com.