Sunday, December 16, 2012
Explaining Magnetoresistive Random Access Memory (MRAM)
What is MRAM?
MRAM is known as Magnetoresistive Random Access Memory. It is a different type of memory from the usual Dynamic Random Access Memory (DRAM) and Static Random Access Memory (SRAM). The very fundamental working principle of MRAM is to store the data in a pair of magnetic storage elements. One end is a permanent magnet while the other end can be varied to be parallel or anti-parallel to the permanent end and this will produce two types of configurations that can represent logical '0' and '1' for computational purposes.
What are the differences between DRAM, SRAM and MRAM?
Dynamic random access memory (DRAM), as the term "dynamic" suggests, needs to be refreshed at least 16 times per second, no matter the stored bit has changed or retains the same value. DRAM can only hold the data for at most 64 milliseconds before refreshing it, as defined by the universal standard. The main structure of DRAM is the capacitors which gives the RAM its dynamic properties. Capacitors can be charged and discharged quickly so the capacitor charge must be refreshed periodically before the current charge depletes which will produce memory errors. The structure of DRAM is very simple, one transistor and one capacitor are required for one bit, and this allows DRAM to be very high in density. The low cost per unit memory makes DRAM the ideal RAM for consumers, most computers and mobile phones are using DRAM as the main RAM.
Static random access memory (SRAM) can retain the data without refreshing it periodically. The typical SRAM is built using 6 transistors per bit. The basic reading operation of SRAM is to allow the memory cell to access two bit lines. The original stored logical value will cause the transistors to turn on and off and will cause the two bit lines to have a slight difference in voltage. The slight difference will be amplified by a sense amplifier and thus it can read the value stored in the memory cell. SRAM operates by the switching of transistors instead of the charging and discharging of capacitors which means SRAM is significantly faster than DRAM. But due to its low density, SRAM is only used in time-critical components like the L1, L2, L3 CPU caches.
Magnetoresistive random access memory (MRAM) is a type of RAM based on magnetic elements, instead of electric charges and voltages. It still uses electric current indirectly to read and write data. Current will pass through the cells and the resistance of the cells can be determined. A parallel configuration will produce a low resistance and a logical '1' while a anti-parallel configuration will produce a high resistance and a logical '0'. The conventional current-based MRAM is not power efficient as it still uses electric current and a newer type of MRAM, spin transfer torque MRAM (STT-MRAM) has been developed. The newer STT-MRAM is as fast as SRAM, as high dense as DRAM and it might replace SRAM and DRAM in all storage devices, memories and caches.
How does the newer spin transfer torque MRAM (STT-MRAM) work?
The STT-MRAM is almost similar with nuclear magnetic resonance, both are based on the magnetic property of subatomic particles, which will spin according to the magnetic field. The conventional current-based MRAM is using electric current to produce a magnetic field that will change the variable end in the pair of magnetic elements while the newer STT-MRAM uses a spin-polarized current to change the orientation of the variable end. The STT-MRAM consumes significantly less power than the conventional current-based MRAM and is more suitable for practical applications. A research team from University of California, Los Angeles has developed a new MRAM that uses voltage instead of current to access the data and it further reduces the power consumption of STT-MRAM.