Hardware Reference
In-Depth Information
the capacitor in the form of electric charges. The charges stored in the capacitor will leak away
over time, so periodic refresh operations are needed to maintain the contents in the DRAM. The
time interval over which each memory location of a DRAM chip must be refreshed at least once
in order to maintain its contents is called its
refresh period.
Refresh periods typically range from a
few milliseconds to over a hundred milliseconds for today's high-density DRAMs.
SRAMs
are designed to store binary information without needing periodic refreshes and
require the use of more complicated circuitry for each bit. Four to six transistors are needed to
store 1 bit of information. As long as power is stable, the information stored in the SRAM will
not be degraded.
MRAMs
were first developed by IBM. Several other companies were also involved in the
research, development, and marketing of this technology. MRAMs use a magnetic moment to
store data. A MRAM chip combines a magnetic device with standard silicon-based microelec-
tronics to achieve the combined attributes of nonvolatility, high-speed operation, and unlim-
ited read and write endurance. The first MRAM device from Freescale is the 4-Mbit MR2A16A.
This device is a parallel memory (8 or 16 bits can be accessed in one operation) and has a 35 ns
access time, reported in 2007.
FRAMs
use the property of ferroelectric crystal to store data bits. Much of the present
FRAM technology was developed by Ramtron International. The FRAM technology has already
achieved high maturity. Both the serial and parallel versions of FRAM chips are available. Ram-
tron even incorporates FRAM in some of its 8051 microcontroller products. The fastest access
time of FRAM from Ramtron is 55 ns, reported in 2007. However, the access time of FRAM
may improve in the future.
RAM is mainly used to store
dynamic
programs or data. A computer user often wants to
run different programs on the same computer, and these programs usually operate on different
sets of data. The programs and data must therefore be loaded into RAM from the hard disk or
other secondary storage, and for this reason they are called
dynamic
.
ROM is nonvolatile. When power is removed from ROM and then reapplied, the original
data will still be there. As its name implies, ROM data can only be read. If the processor attempts
to write data to a ROM location, ROM will not accept the data, and the data in the addressed
ROM memory location will not be changed. However, this statement is not completely true. For
some ROM technologies (EEPROM and flash memory), the user program can still write data into
the memory by following a special procedure prescribed by the manufacturer. However, it would
take a much longer time to write than to read from the flash memory.
Mask-programmed read-only memory
(MROM)
is a type of ROM that is programmed when
it is manufactured. The semiconductor manufacturer places binary data in the memory according
to the request of the customer. To be cost-effective, many thousands of MROM memory units,
each consisting of a copy of the same data (or program), must be sold. MROM is the major memory
technology used to hold microcontroller application programs and constant data. Most people
simply refer to MROM as ROM. The design of MROM prevents it from being written into.
Programmable read-only memory
(PROM) was invented in 1956 by Wen Tsing Chow. It
is a form of memory where the setting of each bit is locked by a fuse or antifuse. The memory
can be programmed just once after manufacturing by blowing the fuses (using a
PROM blower
),
which is an irreversible process. Blowing a fuse opens a connection whereas blowing an antifuse
closes a connection (hence the name). Programming is done by applying high-voltage pulses
that are not encountered during normal operation (typically 12 to 21 volts). Fused-based PROM
technology is no longer in use today.
Search WWH ::
Custom Search