Nonvolatile RAM

Electronic musicians depend on random access memory (RAM) to hold samples, synth-patch edits, sequences, and other types of data that must be instantaneously
Publish date:
Updated on

Electronic musicians depend on random access memory (RAM) to hold samples, synth-patch edits, sequences, and other types of data that must be instantaneously accessible and changeable. But for long-term storage, another form of memory must be used, because anything in RAM immediately disappears when the power is turned off. Read-only memory (ROM) retains data without power, but that data can't be changed. (The data in so-called flash ROM can be changed, but this type of memory works too slowly to be an effective replacement for RAM.)

The concept of nonvolatile RAM that requires no power to hold its data seems like the Holy Grail of computer technology, and a number of companies are on the verge of finding it. Conventional RAM chips use tiny capacitors, which depend on a constant supply of power to retain the charges that represent 0s and 1s, but many new designs use magnetic techniques to store data.

There are several approaches being taken in the effort to create powerless, nonvolatile RAM. Among the magnetic possibilities is a phenomenon discovered only ten years ago called giant magnetoresistance (GMR), in which a magnetic field changes the electrical resistance of a thin conductive film. Honeywell has developed experimental chips based on this concept with capacities of more than 1 megabit. However, GMR devices consume a lot of current, which tends to burn them out when the memory elements are reduced to submicron size. Motorola is working on a way around this problem by doubling the strength of the GMR effect (which reduces the power requirements) in a device called a pseudospin valve.

IBM and Motorola are working on another technique, which depends on the quantum-mechanical ability of electrons to tunnel through a thin insulator sandwiched between two magnets. The current through the insulator depends on the orientation of the two magnetic fields. A team of IBM engineers has demonstrated such tunnel junctions in which individual bits are as small as 200 nanometers (nm) wide and can be switched in 5 nanoseconds (ns) or less. However, this effect is very sensitive to the depth of its thinnest layer, which is typically a film of aluminum just 0.7 nm (approximately four atoms) thick, making large-scale manufacturing difficult.

Yet another approach exploits a phenomenon discovered by Edwin Hall 120 years ago, in which a current moving in a thin film can be deflected to one side by a magnetic field. The Hall Effect is used in MAGRAM (magnetic RAM) devices being developed at Honeywell and the University of Utah, which have built 1-micron devices that can write bits in just 8 ns.

A different, nonmagnetic technique utilizes ferroelectric devices, in which the position of an atom within a crystal lattice represents the value of a bit. A company called Ramtron uses ferroelectric crystals in conjunction with conventional CMOS semiconductor technology to build what they call FRAM chips. Each crystal consists of only a few atoms; when an electric field is applied, the central titanium or zirconium atom moves in the direction of the field (see Fig. 1). To change the state of the crystal, you simply reverse the electric field. If the electric field is removed, the atom stays in place, preserving the data.

Clearly, nonvolatile RAM would be a great boon to electronic musicians of all persuasions. The ability to retain data in RAM without power would provide enhanced security against power failures and propel portable music devices light-years ahead of their current limited applications. It will be very exciting to see this technology emerge from the lab and onto the commercial market for all to use and appreciate.