The main attraction

The main attractionWill this become the ideal memory? Fast to read and write, dense, non-volatile, low ower and infinite erase/write cycles… magnetic RAM seems to have it all. David Manners reports
Government laboratories in Europe, America and Japan are co-ordinating moves by commercial companies and universities to develop a new kind of memory technology called Magnetic RAM (MRAM) which could solve potential bottlenecks in PCs threatened by the slow access times of hard disk drives.
Boot-up time in PCs could be eliminated, and data download times from the Internet could be drastically reduced if MRAM was used in PCs instead of hard disk.
As well as hard disk replacement, MRAM would be almost the ideal semiconductor memory – fast to read, fast to write, dense, non- volatile, low power and with infinite erase/write cycles.
It would allow the manufacturers of every kind of portable equipment to design in much more memory using much less power than the flash memory now used in such equipment. What’s more, MRAM lasts forever.
MRAM technology is based on the same ‘spin electronics’ – the manipulation of an electron’s spin – which is used for giant magneto-resistive (GMR) heads in disk drives.
GMR technology is well-understood by most leading electronics companies but a new approach called TMR (Tunnelling Magnetic Resistance) is the technique being used by most research-ers now looking to develop an MRAM.
It has the potential to be made in densities above 1Gbit. A 1Gbit MRAM would have power consumption of under 10mW, access speeds of 40/50ns and have an unlimited number of erases.
In Japan, Toshiba says it has been pursuing the technology for two years and has developed some advantages on the materials side. However it is looking for a government lead in co-ordinating a national project.
In Germany, Siemens, Bosch and government laboratories are developing MRAM technology. “We are working on MRAMs in the Central Research Department at Siemens and at the BMBF – the government’s research department,” said Hans Pieter Bette, executive vice- president of the memory division at Infineon Technologies, Siemens’ semiconductor arm. “It’s a most interesting technology which has the potential to deliver much higher integration than ferro.
“We have several years of experience in the area which started with our non-volatile memory activity then, in 1998, a project got started where we worked on magnetic effects,” added Bette. “It all comes from a good understanding of magnetic materials – TMR is the future.”
Bette would not be drawn as to when Infineon could deliver a commercial MRAM product based on TMR. All he would say is that he expects that the research projects “will run for a few years.”
Naturally, people are comparing MRAMs with FRAMs – ferroelectric RAM which has been pursued for many years without becoming more than a tiny niche of the memory business. “The difference between ferro and MRAM is that ferro does not have such a high integration capability as MRAM,” said Bette.
Whereas researchers currently believe that the density potential of ferro is up to 10Mbit, they believe that MRAMs will be able to be fabricated in 1Gbit+ densities.
Philips is also involved in pursuing TMR technology but not yet, it says, for MRAMs. “Philips Research is involved in a European project researching GMR and TMR technology and materials,” said a company spokesperson, “this technology can be used for MRAMs, but also for other applications such as magnetic heads for HDDs. The latter is the reason for the Philips’ interest, not the MRAMs as such.”
In the US, at the University of Utah, researchers are building prototype 8kbit MRAMs under contract for a subsidiary of Avanticorp of Toronto. Other US universities pursuing MRAMs are Carnegie-Mellon University (Pittsburgh) and the University of Arizona.
The leader of MRAM research in the US is Defense Advanced Research Projects Agency which began, in 1994, a $50m MRAM-development project involving several US chip companies.
DARPA has produced a 16-kbit device with sub-100ns access and intends to have a 1Mbit device this year.
DARPA has given contracts to various US semiconductor companies of which Honeywell is said to be the most advanced.
Although GMR is now being replaced by TMR as the main focus for researchers, Honeywell is considering setting up for a production line for MRAMs based on GMR technology.
IBM and Motorola are also pursuing TMR research, with IBM leveraging its Josephson Junction experience for the activity, while H-P is said to be anticipating products in two years and sees MRAM as a disk drive replacement technology.
The beauty of replacing a hard disk drive with MRAM in a PC is that it will significantly reduce boot up time and will allow for much quicker downloading of material over the Internet which will become increasingly required as bandwidth rises.
With consumers likely, over the next few years, to get wideband connections capable of delivering data rates of a megabit per second or more, the frustration of waiting for hours while data squeezes though the limited data interface of today’s hard disks would become intense.
So the most exciting potential use for MRAM technology would be as a hard disk replacement device. Military users are said to have had their operations adversely affected by boot up delays, which explains DARPA’s initial funding, but the other big application will be in portable equipment where MRAM’s qualities will make it a welcome replacement technology for flash memory which needs on-board charge pumps to generate the 15V+ programming voltages needed to erase and write data.
For instance high density MRAMs in cell phones would allow you to record conversations and store them on the hard drive of a computer.
In Japan, Toshiba, Fujitsu, Hitachi, NEC, Matsushita and the Electrotechnical Laboratory at MITI (the Ministry of International Trade and Industry) are all said to be drawing up plans to pursue MRAM research.
With R&D budgets under pressure in Japan, following large losses by the semiconductor companies, it is believed that the Japanese companies won’t go ahead without a co-operative effort led and – in part at least – financed by government.
While Japanese companies have been reluctant to risk the wrath of US trade disputes by pursuing collaborations, in the case of MRAM they will be able to point to the US DARPA initiative to justify government intervention to set up a MITI-financed consortium.
At this stage, say researchers, no one knows whether MRAMs will become commercially feasible, but there is no doubt their potential is causing intense excitement within the companies which are pursuing the research.
After years of having to combine the various disadvantages and inconveniences of traditional memories: the volatility and refresh requirement of DRAM; the high programming voltage and block erase of flash; the low density of SRAM, and the inflexibility of EPROM – the world could have a mainstream semiconductor memory which does the job without any of memory’s traditional downsides.

Leave a Reply

Your email address will not be published. Required fields are marked *