AM3358 EDMA test failed

Lev Urusov

Other Parts Discussed in Thread: AM3358

I am attempting to set up a custom board based on AM3358 processor and K4B4G1646D SDRAM (DDR3-1866).

Parameters calculated using TI's spreadsheets are close to reference. But autoconfiguration gives me another values:

***************************************************************
   The Slave Ratio Search Program Values are...
***************************************************************
PARAMETER                       MAX | MIN | OPTIMUM | RANGE
***************************************************************
DATA_PHY_RD_DQS_SLAVE_RATIO    0x3ff | 0x0aa | 0x254 | 0x355
DATA_PHY_FIFO_WE_SLAVE_RATIO   0x3ff | 0x000 | 0x1ff | 0x3ff
DATA_PHY_WR_DQS_SLAVE_RATIO    0x078 | 0x000 | 0x03c | 0x078
DATA_PHY_WR_DATA_SLAVE_RATIO   0x3ff | 0x03c | 0x21d | 0x3c3
***************************************************************

And then EDMA test failed with results like this (lower 2 octets characteristically repeats higher one):

...

Failed@0x80000810   Expected=0x80808084    Read=0x80808080
CortxA8: GEL Output:
Failed@0x80000820   Expected=0x80808088    Read=0x80808080
CortxA8: GEL Output:
Failed@0x80000830   Expected=0x8080808C    Read=0x80808080
CortxA8: GEL Output:
Failed@0x80000840   Expected=0x80808090    Read=0x80808080
CortxA8: GEL Output:
...

However I can start Linux on this board with BeagleBoneBlack's memory parameters and it seems working.

What may be the reason?

over 10 years ago

0 Biser Gatchev-XID over 10 years ago

TI__Guru**** 393215 points

Hi Lev,

The software leveling seems to work fine on your board. Have you updated the configuration with the optimal values before running the test? I suppose you are following this wiki: http://processors.wiki.ti.com/index.php/Sitara_Linux_Training:_Tuning_the_DDR3_Timings_on_BeagleBoneBlack

0 Lev Urusov over 10 years ago in reply to Biser Gatchev-XID

Prodigy 140 points

Yes, I followed this article and updated GEL file with the optimal values before testing. By the way, when I edit DDR data via the Memory Browser, in some cells there is the same situation: editing in higher 16 bits duplicates in lower half, editing in lower half has no effect.

0 Biser Gatchev-XID over 10 years ago in reply to Lev Urusov

TI__Guru**** 393215 points

I have asked the DDR experts to help on this.

0 Lev Urusov over 10 years ago in reply to Biser Gatchev-XID

Prodigy 140 points

Need to add: U-Boot doesn't start with these "optimum" parameters, but when I restore original BBB values in ddr3_beagleblack_data structure (it contains RD_DQS, WR_DQS, PHY_FIFO_WE and PHY_WR_DATA), it works. I used CCS 5.3.0. My RatioSeed spreadsheet is in application.

0 PA Nilsson over 10 years ago in reply to Lev Urusov

Prodigy 200 points

I found the error in my config, my board does not use 24MHz input and the 'gel' files only supports 24MHz and does not terminate if any other value is used. It does detect the correct crystal frequency from the boot-sys pins, but I only has software to set it up for 24MHz.
e2e.ti.com/.../1579595

0 Lev Urusov over 10 years ago in reply to PA Nilsson

Prodigy 140 points

Thank you! It was the reason.

Processors

Processors forum

AM3358 EDMA test failed