Control Engineering

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

What kinds of non-volatile RAM are there?
April 14, 2008

Off the top of my head, I can think of 3 non-volatile random access memory (NVRAM) technologies in current use. I’m here defining NVRAM as a bitwise readable, writeable, and erasable memory technology for typical general purpose computers. In NVRAM devices, each bit has its own memory address, and can be part of the computer’s addressable memory along with the volatile static and dynamic RAM (SRAM and DRAM) chips shipped with the computer.

Electrically erasable programmable read only memory (EEPROM) is not NVRAM because it can only be read by a computer. It takes a special device, called a programmer, to erase or write to EEPROMs.

While computers can write directly to hard magnetic discs (also called Winchester drives or hard drives) seemingly at will, they are actually sequential access memories like magnetic tapes, where large blocks of information are read out from the device and stored temporarily in the computer’s volatile RAM, where it becomes useful to programs. Compact discs (CDs) are out for a similar reason.

The non-volatile (NV) part of the NVRAM designation, of course, refers to the devices’ ability to retain information when electrical power is removed. SRAM is the earliest type of RAM, consisting of transistorized flip flops (more formally called bistable multivibrators) that latch into one state or the other so long as supply power is applied. When power is removed, however, the circuit de-energizes and all data is lost. In fact, the devices are typically biased such that they always “wake up” in the zero state.

DRAM cells hold information bits in the form of charges on capacitors. The charge constantly leaks away, however, and DRAMs must be periodically refreshed. There is a special circuit in the DRAM that constantly reads all the cells and rewrites the same data it found in each. This circuit refreshes all the cells fast enough to ensure that no data is lost. Of course, once power is taken from the chip, the data quickly evaporates.

Early NVRAM
The first NVRAM was magnetic-core memory. Tiny donut shaped beads of magnetic iron oxide (ferrite) were woven into a copper-wire mesh. The mesh was similar to window screen material and, in fact, was woven on similar looms. During weaving, however, the ferrite beads were added at each wire-crossing point. The beads stored information in the direction of their internal magnetization (clockwise for 0 and counterclockwise for 1, for example). Anisotropic ferrite was used because it is a “hard” magnetic material, whose magnetization cannot be changed except by forcing it with an external magnetic field strong enough to completely reverse its magnetization direction, in which case it immediately flips to the reverse direction and cannot be flipped back except by another strong enough field. 

Magnetic core memory was the earliest NVRAM.

To write a bit, the computer would send two current pulses down the two wires corresponding to the bit’s address. At the addressed bead, the magnetic fields surrounding the two wires would superpose to a field strong enough to flip the bead. Only the addressed bead would be threaded by both currents. Elsewhere, the magnetic field would not be strong enough to flip a bead.

To read out the bit, the computer would send programming pulses down the two address lines, in a sense to program a 1 into the bead. If the bead was already a 1, nothing would happen and the computer would know the bead already carried a 1. If, on the other hand, the bead carried a 0, energy would be pulled out of the current pulses to reprogram the bead to 1. This change in field induces a voltage spike in the sense/inhibit wire, which the computer senses. The computer would then have to go back and re-program the bead with a zero again, or the information would be destroyed!

Ferrite core memories stopped being used when semiconductor memories became available, because core memories were bulky, slow power hogs. Many computer users, however, continued to use the term “core memory” for decades after it became obsolete.

The first semiconductor NVRAM technology was battery-backed SRAM. It was created by the simple expedient of providing a rechargable battery to keep power applied to the SRAM when system power was removed. This is still in use and works well for limited time periods, but the batteries take up useful space, and eventually discharge. Computer users who store their computers without power for long periods of time find that the units lose their CMOS setup information because it is typically stored in battery-backed SRAM.

Flash to alphabet soup
Today, so-called flash memory takes the place of battery-backed SRAM in a number of applications. Most notably, flash has made possible compact “memory sticks,” which are just flash memory chips packaged along with a USB interface. When plugged into a USB port, they appear as a “removable drive.” They serve the same function as floppy disks did 10 years ago, allowing file sharing via “sneakernet.” Flash memory can be (and is) used for more reliable CMOS Setup storage and virtually any other NVRAM application. Flash technology was described in last week’s Ask Charlie. Flash’s main drawback is a limitation on the number of read/write cycles its cells can endure.

MRAM
A second type of NVRAM that is currently gaining popularity is magnetoresistive RAM (MRAM). MRAM’s greatest advantage over flash is a virtually unlimited number of read/write cycles. Frank Bartos covered advances in MRAM in his April 2008 column. 

FRAM
A third NVRAM technology currently in production is the ferroelectric RAM (FRAM or FeRAM). Like DRAM, FRAM stores information as voltage on a capacitor. Instead of using a linear dielectric, such as silicon dioxide (basically, glass), FRAM uses a non-linear ferroelectric dielectric, such as lead zirconate titanate (PZT). PZT is a crystalline material whose crystal unit cells have a permanent electric dipole moment. Applying an electric field (by putting a charging voltage across the capacitor) causes atomic rearrangements within the unit cells to align all the dipole moments with the impressed electric field. Removing the supply voltage leaves the dipoles still aligned, so the potential difference between the plates persists. 

PRAM, SONOS, RRAM, NRAM ... 
Other NVRAM technologies under development include phase-change RAM (PRAM), Silicon-Oxide-Nitride-Oxide-Silicon (SONOS), Resistive Random Access Memory (RRAM), Nano-RAM (NRAM), and perhaps others. All are under development as potential replacements for flash memory, but (to my knowledge) none are ready for commercialization.

Posted by Charlie Masi on April 14, 2008 | Comments (0)