Today we have lots of options for storing information electronically like SRAM, DRAM, Flash and EEPROM. Random access memories (RAM) like SRAM and DRAM store information using capacitance which makes them very fast without a lot of extra complication. But their downside is that all of their information is lost when the memory isn’t powered. Nonvolatile memory like Flash or EEPROM keep their contents while powered off at the expense of speed and requiring more complex access methods like memory paging. Enter magnetic memory which combines the qualities of both kinds of memories.
One of the first incarnations was magnetic core memory that was used in the 50s and 60s because it was so much faster than other storage methods like Williams tubes which were based on CRTs. The name comes from the fact that the memory was constructed using magnetic toroids called cores with wires threaded through them for read and write operations. Each toroid would be set to a 1 or 0 by controlling its magnetic direction although some extra circuitry was required because every read operation would clear the core’s magnetization. This storage method proved to be very reliable and could even withstand an EMP pulse. But the downside of the toroids was that it was very difficult to manufacture and that made it relatively expensive so it was eventually superseded by SRAM which became available in the 60s. The Computer History Museum in Mountain View, CA, has a fine collection of magnetic-core memory devices on display, also check out their coasters with an MM image — buy one from the Gifts Shop (Ed.) A modern version of magnetic memory is magnetoresistive RAM (MRAM) which has been developed over the last 30 years. Early versions used ferromagnetic plates with an insulating layer between them.
One plate has a permanent magnetic field and the other plate’s magnetization is writeable to store a bit. The logic state of the cell is then determined by measuring its resistance which changes depending on the magnetic
orientation of the writeable plate to the permanent plate. Newer versions of MRAM use the spin transfer torque of electrons in their memory cells to lower their power consumption.
Another variation is ferroelectric RAM (FRAM) which has been developed in parallel with MRAM. Conventional DRAM memory uses one transistor and one capacitor per memory cell which can be made into a FRAM cell by adding a ferroelectric material in place of the regular dielectric. The ferroelectric material Comchanges the normally linear behavior of the cell to one that has magnetic hysteresis which gives FRAM its nonvolatile properties. Writing data is pretty straightforward but reading a FRAM cell requires that a transistor put the cell into a known state and then the cell is monitored to see if the ferroelectric material causes the current to flow. This will also clear the cell so it needs to be rewritten. Both MRAM and FRAM are being produced today by various companies. MRAM devices tend to focus on density and speed whereas FRAM concentrates on low power. Either way, they are very interesting parts!