Memory Interfaces
4-21
Memory Interface Traffic Controller
4.3.2.7
Burst Read Operation
The synchronous read mode is selected for each device by setting the
RDMODE configuration bit field to 100.
In this mode of operation, FLASH.CLK is driven on the OMAP5910 device pin.
Both AMD burst flash and Inter burst flash have three modes of operation:
-
Asynchronous single read mode (device startup mode)
-
Synchronous single read or burst read mode
-
Asynchronous write
Asynchronous single read mode and asynchronous write modes are compat-
ible with operation described in Section 4.3.2.5, Asynchronous Read
Operation, and Section 4.3.2.8, Asynchronous Write With WE Operation.
Figure 4–6 shows the timing view of synchronous burst read mode operation.
On the AMD device, LBA is directly connected to the FLASH.ADV OMAP5910
pin.
The address is latched on the rising edge of FLASH.ADV with a specified hold
time of 3 ns. This is easily met by maintaining the address during two cycles.
Data output of the device is stable on the rising edge of FLASH.CLK (specified
with a setup and hold time referenced to this edge).
Two configuration registers are used in this operating mode:
-
FCLKDIV. Specifies the frequency ratio between the TC clock and
FLASH.CLK (see Table 4–13, EMIF Slow Chip-Select Configuration
Registers).
-
RDWST. Specifies the number of FLASH.CLK cycles between the falling
edge of FLASH.ADV and the edge at which first data is valid (see
Table 4–13, EMIF Slow Chip-Select Configuration Registers).
The FLASH.RDY signal is not used in this mode: however, it is used during
flash program and erase operations.
Note:
Intel Burst Flash Operation
For proper operation in applications that combine OMAP5910 with Intel burst
flash (examples include Intel 28FxxxK3, 28FxxxK18, and 28FxxxW18), the
flash WAIT signal must not be connected to the OMAP5910 FLASH.RDY in-
put. Instead, the FLASH.RDY input pin in the OMAP5910 must be tied active
high through a pullup resistor. The OMAP5910 traffic controller properly han-
dles all accesses across Intel burst flash boundaries (where WAIT could be
asserted) without any input from WAIT, and without performance penalty.