![]() |
Memory and I/O Interface
|
Need Cache ? 
Two types of clocked memories are
very common as either IP cores or stand alone chips. Both types read address
and control signals (RD, WR) at the beginning of a clock cycle, and output the
read out data before the beginning of next clock cycle. Some of stand alone memory
chips often make this reading out delay from the address input as a full clock
cycle (meaning the data is available on or after the next clock cycle.) for
faster clock frequency operation. Only the 'No latency' stand alone memories
or common IP memories read back the data in the same clock cycle as shown in
the figure. When seeing the write timing of these memories, most common IP memories
have the timing of either Type1 or Type2 figured. If your target memory is something
like Type1, then the CD24x0A DSP core can communicate with the memory in back-to-back.
No additional interfacing logics are necessary. However, if the target memory
timing is like Type2, where the writing data has to be available at the beginning
of the cycle, you need to add small cache logic (one or more words) to convert
the write timing from Type1 to Type2.
For the asynchronous I/O port, you can obtain necessary signals by latching the address and control signal at the beginning of the cycle before they are used. Control signals may need to be ANDed with inverse of Clock before using.
Return to Clarkspur HOME PAGE .