Cypress CY7C1166V18 Manual De Usuario

Document Number: 001-06620  Rev. *D

Page 8 of 27

The CY7C1166V18, CY7C1177V18, CY7C1168V18, and
CY7C1170V18 are synchronous pipelined Burst SRAMs
equipped with a DDR interface. 

Accesses are initiated on the rising edge of the positive input
clock (K). All synchronous input and output timing are referenced
to the rising edge of the Input clocks (K/K).

All synchronous data inputs (D

[x:0]

) pass through input registers

controlled by the rising edge of the input clocks (K and K). All
synchronous data outputs (Q

[x:0]

) pass through output registers

controlled by the rising edge of the input clocks (K and K) also. 

All synchronous control (R/W, LD, BWS

[0:X]

) inputs pass through

input registers controlled by the rising edge of the input clock
(K/K). 

CY7C1168V18 is described in the following sections. The same
basic descriptions apply to CY7C1166V18, CY7C1177V18, and
CY7C1170V18.

The CY7C1168V18 is organized internally as a single array of
1M x 18. Accesses are completed in a burst of two sequential
18-bit data words. Read operations are initiated by asserting
R/W

 

HIGH and LD LOW at the rising edge of the positive input

clock (K). The address presented to address inputs is stored in
the read address register. Following the next two K clock rise, the
corresponding 18-bit word of data from this address location is
driven onto the Q

[17:0]

 using K as the output timing reference. On

the subsequent rising edge of K the next 18-bit data word from
the address location generated by the burst counter is driven
onto the Q

[17:0]

. The requested data is valid 0.45 ns from the

rising edge of the input clock (K/K). In order to maintain the
internal logic, each read access must be allowed to complete.
Read accesses can be initiated on every rising edge of the
positive input clock (K).

When read access is deselected, the CY7C1168V18 first
completes the pending read transactions. Synchronous internal
circuitry automatically tri-states the outputs following the next
rising edge of the negative Input clock (K). This enables for a
seamless transition between devices without the insertion of wait
states in a depth expanded memory. 

Write operations are initiated by asserting R/W

 

LOW and LD

LOW at the rising edge of the positive input clock (K). The
address presented to address inputs is stored in the write
address register. On the following K clock rise the data presented
to D

[17:0]

 is latched and stored into the 18-bit Write Data register

provided BWS

[1:0]

 are both asserted active. On the subsequent

rising edge of the Negative Input Clock (K) the information
presented to D

[17:0]

 is also stored into the Write Data register

provided BWS

[1:0]

 are both asserted active. The 36 bits of data

is then written into the memory array at the specified location.
Write accesses can be initiated on every rising edge of the
positive input clock (K). This pipelines the data flow such that 18
bits of data can be transferred into the device on every rising
edge of the input clocks (K and K). 

When write access is deselected, the device ignores all inputs
after the pending write operations are completed. 

Byte Write operations are supported by the CY7C1168V18. A
Write operation is initiated as described in the 

section. The bytes that are written are determined by BWS

0

 and

BWS

1

 which are sampled with each set of 18-bit data word.

Asserting the appropriate Byte Write Select input during the data
portion of a write enables the data being presented to be latched
and written into the device. Deasserting the Byte Write Select
input during the data portion of a write enables the data stored in
the device for that byte to remain unaltered. This feature can be
used to simplify read/modify/write operations to a Byte Write
operation.

The CY7C1168V18 enables high-performance operation
through high clock frequencies (achieved through pipelining) and
double data rate mode of operation. The CY7C1168V18 requires
two No Operation (NOP) cycle when transitioning from a read to
a write cycle. At higher frequencies, some applications may
require a third NOP cycle to avoid contention.

If a read occurs after a write cycle, then the address and data for
the write are stored in registers. The write information must be
stored because the SRAM cannot perform the last word write to
the array without conflicting with the read. The data stays in this
register until the next write cycle occurs. On the first write cycle
after the read(s), the stored data from the earlier write is written
into the SRAM array. This is called a Posted Write.

If a read is performed on the same address on which a write is
performed in the previous cycle, the SRAM reads out the most
current data. The SRAM does this by bypassing the memory
array and reading the data from the registers.

Depth expansion requires replicating the LD control signal for
each bank. All other control signals can be common between
banks as appropriate.

An external resistor, RQ, must be connected between the ZQ pin
on the SRAM and V

SS 

to enable the SRAM to adjust its output

driver impedance. The value of RQ must be 5x the value of the
intended line impedance driven by the SRAM. The allowable
range of RQ to guarantee impedance matching with a tolerance
of ±15% is between 175

Ω and 350Ω

, 

with V

DDQ

= 1.5V.  The

output impedance is adjusted every 1024 cycles upon power up
to account for drifts in supply voltage and temperature.

Echo clocks are provided on the DDR-II+ to simplify data capture
on high-speed systems. Two echo clocks are generated by the
DDR-II+. CQ is referenced with respect to K and CQ is refer-
enced with respect to K. These are free-running clocks and are
synchronized to the input clock of the DDR-II+. The timings for
the echo clocks are shown in the 

QVLD is provided on the DDR-II+ to simplify data capture on high
speed systems. The QVLD is generated by the DDR-II+ device
along with data output. This signal is also edge-aligned with the