Hi,
Because of the obsolescence issues I replaced the current 2x32 MByte devices with ISSI IS42S16320F -6BLI.
However only 32 MB address space in each device is accessed using CS2.
Different prototype boards based on Renesas RZ/A1H were produced.
Their SDRAM configuration settings are identical as given below
BSC.CS2BCR = 0x00004e00;
BSC.CS2WCR = 0x00000480;
BSC.RTCOR = 0xa55a0080;
BSC.RTCSR = 0xa55a0008;
BSC.SDCR = 0x00110811;
#define CS2_SDRAM_MODE_32BIT_CAS2_BR_BW (*(volatile uint16_t*)0x3FFFD080) // burst read / burst write
CS2_SDRAM_MODE_32BIT_CAS2_BR_BW = 0x0000;
Running memory retention tests for the entire address space i.e., 0x0800.0000 to 0x0bff.ffff,
fail on two boards but have no issues on the third board.
The test writes a pattern (e.g., 0xa5a5a5a5 or 0x5a5a5a5a5a) on 16MByte of memory at a time and verifies at the end of writing.
When the test fails, the read data seem to be corrupted (i.e., not same as the write data) in two 4-bit nibbles (bits[4:7] and bits[12:15]).
Rest of the data were correct.
Incidentally, another board that fails show corruption in bits[20:23] and bits[28:31].
Raising the temperature of the SDRAM devices by exposing to slightly warm air makes the board to fail more quickly.
CKIO was measured to be 66.67 MHz on the scope.
I was wondering if the timings are at the border line. If so, tweaking the configuration settings might resolve the issue.
Adjusting the RTCOR and RTCSR values but had no luck. Maybe i haven't got the sweetspot yet!
Or it could be some dodgy chips procurred during these hard times.
Any suggestion and advice based on privous experience with these devices are highly appreciated.
I am reluctant to replace these chips with another sets until I am convinced its a SDRAM configuration issue.
Thanks in advance
Best Regards
Suresh