K4G323222A Datasheet, PDF (14/49 Pages) Samsung semiconductor – 32Mbit SGRAM

K4G323222A
DEVICE OPERATIONS
CMOS SGRAM
BANK ACTIVATE
The bank activate command is used to select a random row in
an idle bank. By asserting low on RAS and CS with desired row
and bank addresses, a row access is initiated. The read or write
operation can occur after a time delay of tRCD(min) from the time
of bank activation. tRCD(min) is an internal timing parameter of
SGRAM, therefore it is dependent on operating clock frequency.
The minimum number of clock cycles required between bank
activate and read or write command should be calculated by
dividing tRCD(min) with cycle time of the clock and then rounding
off the result to the next higher integer. The SGRAM has two
internal banks on the same chip and shares part of the internal
circuitry to reduce chip area, therefore it restricts the activation
of both banks immediately. Also the noise generated during
sensing of each bank of SGRAM is high requiring some time for
power supplies to recover before the other bank can be sensed
reliably. tRRD(min) specifies the minimum time required between
activating different banks. The number of clock cycles required
between different bank activation must be calculated similar to
tRCD specification. The minimum time required for the bank to be
active to initiate sensing and restoring the complete row of
dynamic cells is determined by tRAS(min) specification before a
precharge command to that active bank can be asserted. The
maximum time any bank can be in the active state is determined
by tRAS(max). The number of cycles for both tRAS(min) and
tRAS(max) can be calculated similar to tRCD specification.
BURST READ
The burst read command is used to access burst of data on con-
secutive clock cycles from an active row in an active bank. The
burst read command is issued by asserting low on CS and CAS
with WE being high on the positive edge of the clock. The bank
must be active for at least tRCD(min) before the burst read com-
mand is issued. The first output appears CAS latency number of
clock cycles after the issue of burst read command. The burst
length, burst sequence and latency from the burst read com-
mand is determined by the mode register which is already pro-
grammed. The burst read can be initiated on any column
address of the active row. The address wraps around if the initial
address does not start from a boundary such that number of out-
puts from each I/O are equal to the burst length programmed in
the mode register. The output goes into high-impedance at the
end of the burst, unless a new burst read was initiated to keep
the data output gapless. The burst read can be terminated by
issuing another burst read or burst write in the same bank or the
other active bank or a precharge command to the same bank.
The burst stop command is valid only at full page burst length
where the output does not go into high impedance at the end of
burst and the burst is wrapped around..
BURST WRITE
The burst write command is similar to burst read command, and
is used to write data into the SGRAM on consecutive clock
cycles in adjacent addresses depending on burst length and
burst sequence. By asserting low on CS, CAS and WE with valid
column address, a write burst is initiated. The data inputs are
provided for the initial address in the same clock cycle as the
burst write command. The input buffer is deselected at the end
of the burst length, even though the internal writing may not
have been completed yet. The writing can not complete to burst
length. The burst write can be terminated by issuing a burst
read and DQM for blocking data inputs or burst write in the same
or the other active bank. The burst stop command is valid only at
full page burst length where the writing continues at the end of
burst and the burst is wrapped around. The write burst can also
be terminated by using DQM for blocking data and precharging
the bank "tRDL" after the last data input to be written into the
active row. See DQM OPERATION also.
DQM OPERATION
The DQM is used to mask input and output operations. It works
similar to OE during read operation and inhibits writing during
write operation. The read latency is two cycles from DQM and
zero cycle for write, which means DQM masking occurs two
cycles later in the read cycle and occurs in the same cycle dur-
ing write cycle. DQM operation is synchronous with the clock,
therefore the masking occurs for a complete cycle. The DQM
signal is important during burst interrupts of write with read or
precharge in the SGRAM. Due to asynchronous nature of the
internal write, the DQM operation is critical to avoid unwanted or
incomplete writes when the complete burst write is not required.
DQM is also used for device selection, byte selection and bus
control in a memory system. DQM0 controls DQ0 to DQ7,
DQM1 controls DQ8 to DQ15, DQM2 controls DQ16 to DQ23,
DQM3 controls DQ24 to DQ31. DQM masks the DQ′s by a byte
regardless that the corresponding DQ′s are in a state of Pixel
masking. Please refer to DQM timing diagram also.
PRECHARGE
The precharge operation is performed on an active bank by
asserting low on CS, RAS, WE and A8/AP with valid BA of the
bank to be precharged. The precharge command can be
asserted anytime after tRAS(min) is satisfied from the bank acti-
vate command in the desired bank. "tRP" is defined as the mini-
mum time required to precharge a bank. The minimum number
of clock cycles required to complete row precharge is calculated
by dividing "tRP" with clock cycle time and rounding up to the
next higher integer. Care should be taken to make sure that
burst write is completed or DQM is used to inhibit writing before
precharge command is asserted. The maximum time any bank
can be active is specified by tRAS(max). Therefore, each bank
has to be precharged within tRAS(max) from the bank activate
command. At the end of precharge, the bank enters the idle
state and is ready to be activated again.
Rev. 1.3 (Dec. 2000)
- 14