This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
Dear all,
we have designed a custom board based on BeagleBone Black with AM3358 and DDR3L ( MT41K256M16HA-125).
The board is, at this point, a copy of BBB when it comes to MCU + SDRAM connection, so the schematics is
accurate and verified. Surely, PCB design is slightly modified. I mean lines DATA [0...15], ADDR [0...15], BA [0..2] are
traced and matched accurately (I hope...) but it is not a full copy of BBB.
However during board bring up we stuck into a weird issue with RAM. We have started U-boot 1st and 2nd stages
successfully, DRAM is found correctly but doing memtest on a large memory addresses range reboots or
hangs up the board, we can though write and read occasional memory addresses (one by one).
Of course, checked address space does not overlap with U-boot, so we are accurate here.
But when we do not stop U-boot , the board hangs up on "Linux kernel uncompressing...".
So my question is, can it be a AM3358 DDR registers problem since we see the RAM during boot or it is clearly a hardware?
Generally, how do you solve this kind of problem?
Thank you,
Iurii
Hello Biser,
thank you for your reply. Indeed it helps a lot with DDR configuration.
We changed SDRAM drive strength and resistor termination value by setting SDRAM_CONFIG.reg_sdram_drive = 1 (RZQ/7), SDRAM_CONFIG.reg_ddr_term = 5 (RQZ/8) and we managed to run the board's u-boot with no hangs up. It works just as it should.
However when we do DRAM memtest throughout the whole memory range some errors occur. Typically, 1-2 errors per iteration. We didn't match the length of DQS / DQSN lines with respect to DATA, ADDR, BANK lines. Is it a serious error in the design?
Regards,
Iurii
Hi Iurii,
DQS/DQSn need to be matched to each other (100Ohm differential traces) and to the corresponding DATA byte (see AM335X Datasheet Rev G. Table 7-67 for details). The same is valid for CLK/CLKn (100Ohm differential traces) and Address/Control signals (Table 7-66).
Hi Biser,
thanks a lot for explanation.
Seems like we need a second revision of our board with all this input taken into account.
Best regards,
Iurii
Please make sure that you follow the guidelines in the AM335X Datasheet Rev. G, section 7.7.2.3.