Ramtron Article: 1 Meg FRAM memory chips production to begin.
Ramtron's FRAM revenues should begin to drastically improve as these larger sizes become available.
(FRAM is non-volatile memory that uses a miniscule amount of power and is ideal for wireless applications like cell phones, PDA's, and Smartcards. Very fast. Memory stays intact even with no power.)
Ticker: RMTR
Here's a new article about Ramtron and one of their partners, Fujitsu who have a licensing agreement with RMTR.
(I like the part about the Sony Playstation applications...
Move over Rambus. The next generation is upon us already.)
Link:
eet.com
Fujitsu sets production sked for
1-Mbit FRAM
By Anthony Cataldo
EE Times
(03/28/00, 3:42 p.m. EST)
TOKYO ? Fujitsu Ltd. is preparing to sample a 1-Mbit ferroelectric
RAM this August and to have the part in volume production this
December. The prospect of pushing FRAM beyond 256 kbits, along
with the promise of technical breakthroughs in voltage level and other
aspects of manufacturing, has revived confidence among the
technology's backers that FRAM will soon break out from its niche
status.
Fujitsu, which began volume production of FRAMs late last year, has
been working with Ramtron International Corp. (Colorado Springs,
Colo.) on a one-transistor, one-capacitor FRAM cell since 1996, and
the partners recently announced their first 1-Mbit prototype.
Ramtron is also working with Toshiba Corp. on a 1T/1C FRAM cell.
Toshiba has installed a 0.25-micron fab line for FRAM, but a
spokeswoman said the company is uncertain when mass production
will begin.
For its part, Fujitsu is making plans to migrate from a 0.5-micron
process technology to a 0.35-micron process. That will enable it to
produce FRAMs with densities of 2 to 4 Mbits, said a Fujitsu
spokesman.
The company is targeting IC cards used as bus passes and library
cards, applications where FRAMs are already being tapped due to the
technology's fast-write capability and high endurance. There's also
evidence that FRAM is being adopted by some high-volume consumer
applications, such as Sony Computer Entertainment's Playstation 2.
Masamichi Ogura, group president of the administrative and business
promotion section of Fujitsu's Electronic Devices Group, said that
Playstation is using Fujitsu's FRAM as part of the flash-memory card
slot, though the company declined to comment more specifically on
how it is being used.
Though FRAM proponents have long touted the technology's
non-volatility, high endurance and fast access speeds, manufacturing
problems have stalled its acceptance. The large cell size, additional
process steps and unstable nature of the high-k materials used for
FRAMs have kept yields low, forcing vendors to make low-density
devices to get a higher return on each wafer produced.
Another handicap is that FRAMs need a 5-V supply voltage, which
limits their penetration in mobile applications. Ramtron hopes to
correct this soon, and recently announced that all future products
based on its most advanced 0.5-micron ferroelectric process will be
3-V compatible.
However, to gain acceptance in high-volume portable devices such as
cellular phones, Ramtron and its partners will need to knock down the
voltage to at least 1.8 V, said Mike Alwais, director of marketing for
Ramtron's FRAM product line.
"[FRAM] inherently needs about a volt, but you have to achieve the
capability of the material in the end product," he said. "There's
manufacturing considerations in getting from what you get in the lab
to the end chip."
Having developed a reliable process for a 1T/1C FRAM cell, Fujitsu and
Ramtron hope to break the manufacturing jinx. FRAMs today use two
capacitors per cell to distinguish electrical levels, with one capacitor
acting as a fixed reference. With the 1T/1C cell, the FRAM uses one
reference for the entire chip. Digital data is then determined by the
differences in the electrical levels between the memory cell capacitor
and the reference electrical level, according to the companies.
Fujitsu, Ramtron and other companies like Toshiba have been trying to
perfect 1T/1C cells for years but making the process stable has
proved difficult. With the 2T/2C design, "each bit had its own
reference, and we had a good copy of the information next door so
we could make the comparison easily," said Alwais. "Now there's a
global reference and a need for a better process because there's half
as much signal. Each bit does not have its own reference; there's one
for the whole chip."
Yield hit
The new FRAMs still need additional deposition and etching steps in
order to form the high-dielectric PZT capacitors. Because of this,
wafer yields do take a hit, but not enough to hold back production.
"The technology has proven to be producible and reliable," Alwais said.
"The defect densities are a little worse than industry standard, but I
think you'll find them to be at production level."
Alwais said yields should improve as production levels rise. What's
more, FRAM makers will soon be able to piggyback on work being done
for 1-Gbit DRAMs, which many believe will need high-k dielectric
materials to form smaller capacitors.
"Our process has been one in which we've been using equipment
tailored to do other things. Now equipment vendors are coming on
board because they see the high-dielectric DRAM coming and can use
us as a proving ground," Alwais said.
Toshiba presented a research paper two years ago describing a
0.5-micron, 1-Mbit FRAM based on a 1T/1C cell, with a chain
architecture that promises an additional 60 percent reduction in chip
size.
Beyond bringing down voltage, the next step is to improve the FRAM's
endurance, or the number of read/write cycles that the chip can
perform over its life. "We'd like it to be unlimited," Alwais said. "Now
it's limited to 1012 but it turns out not to be enough." |
