Control Engineering

Link This | Email this | Blog This | Comments (0)

What is an MRAM?
February 18, 2008

The acronym MRAM stands for magnetoresistive random access memory, or, alternatively, magnetic random access memory. MRAMs are a class of non-volatile random access memory (NVRAM) devices in which information bits are coded in the magnetic polarization direction of individual magnetic memory cells. Various schemes for using magnetoresistance to make NVRAMs have been proposed. Currently, however, only Freescale Semiconductor, manufactures MRAMs as commercial parts, so I’ll limit my description to that design.

As with ferrite core memories used in early computers, writing of information takes place via magnetic fields induced by pairs of wires laid orthogonal to each other and energized in such a way that only the memory cell at the intersection of the two selected wires undergoes a transition. Because the information storage mechanism relies on magnetization orientation in permanent magnet material, there is no need to keep the memory energized to retain information.

Similarities end there. Information is read out of an MRAM via the tunneling magnetoresistance (TMR) effect, rather than by magnetic induction as in the magnetic core memory. Because the read and write processes utilize different physical effects, reading has no effect on the stored data. Because MRAM memory cells can be fabricated using ordinary semiconductor fabrication process steps, the devices can be made very small and inexpensively.

TMR is a quantum mechanical effect observed in structures made of two ferromagnetic metal layers separated by an insulating layer, called a “magnetic tunnel junction” (MTJ). The ferromagnetic layers are anisotropic and may be aligned parallel or antiparallel. An electron moving through the MTJ aligns its spin with the magnetization in the first layer before tunneling across the ultra-thin insulating layer to the other ferromagnetic layer.

In practice, one of the ferromagnetic layers is pinned by an adjacent antiferromagnetic layer to ensure that its magnetic orientation stays fixed. The polarization of the other layer is free to move. When a bit is being written, two pulses separated in time by one half pulse width pass along the two crossed conductors (the “bit” line and “write” line). The first pulse rotates the magnetically free layer’s magnetization orientation by 90º, and the second pulse rotates it by an additional 90º to flip it completely around. Since the material is anisotropic, and can take on only parallel or antiparallel orientations, the magnetization either rotates all the way or not at all. This toggling system action prevents accidental writes to cells not at the crossing point.

MRAM memory cells consist of two components: a magnetic tunnel junction, and a field effect transistor.

To read a bit stored in the cell, the bit line is electrically grounded and a “read word line” is biased negatively. This bit line forms the gate electrode of a complementary-mode field effect transistor (FET), whose source is connected through the MTJ to the grounded bit line. In this configuration, electrons pass from the bit line through the MTJ to the FET’s source electrode. Negative charge on the read word line opens the FET’s channel to the drain electrode connected to the positive (Vdd) supply terminal.

In this mode, the FET acts as an electronic relay. When it’s “closed” (electrically conducting), a current passes from the bit line through the MTJ, through the FET and out through the drain terminal. Knowing which bit line and which read word line are energized identifies the memory cell being accessed. Measuring the current through the MTJ tells whether that bit is a “1” or “0.”

Having the MTJ’s ferromagnetic layers oriented parallel produces low resistance and high current, representing a “0.” Antiparallel orientation produces high resistance and low current, representing a “1.”

Since the only “moving parts” in the MTJ are intrinsic spins of the ferromagnetic atoms, there is no wearout mechanism. MRAMs can theoretically be cycled indefinitely. Read and write times are better than those of dynamic RAM (DRAM) devices and approach those of static RAM (SRAM). Since MRAM cells consist of only one transistor, while SRAM cells include several transistors, bit-packing density is much better for MRAM. In short, MRAM is a good choice for all RAM applications, but especially when non-volatility needs to be coupled with constant memory-content updating.

For more information about MRAM technology from Control Engineering, type “MRAM” into the search box on any page on the Website.

Posted by Charlie Masi on February 18, 2008 | Comments (0)