DDR3 calibration tool write leveling issue

取消
显示结果 
显示  仅  | 搜索替代 
您的意思是: 

DDR3 calibration tool write leveling issue

2,095 次查看
leonardoprates
Contributor III

I have a custom board similar to SabreSD using latest Linux release (Yocto dizzy). I use a standard notebook memory module (204 pin SODIMM 2GB 1333MHz/1600MHz) connected to IMX6Q processor. I have tested with a memory module manufactured by SMART Modules (2GB 1333MHz) and everything works well. I have used the calibration and stress tool (1.0.3) and I got calibration data necessary to complete memory configuration with success.

I have tested another module, manufactured by Avant that uses NANYA memory chip and this module works too, but write leveling calibration procedure with this module finishes so fast and result for the four related registers are 0x001F001F.

MMDC_MPWLDECTRL0 ch0 after write level cal: 0x001F001F

MMDC_MPWLDECTRL1 ch0 after write level cal: 0x001F001F

MMDC_MPWLDECTRL0 ch1 after write level cal: 0x001F001F

MMDC_MPWLDECTRL1 ch1 after write level cal: 0x001F001F

I have tried some times and always the result is the same. For DQS calibration  the results are as expected. For SMART M, the write leveling gives more realist result. I have made some tests and write leveling values that I get for SMART M works better in Avant memory than 0x001F001F result from write leveling calibration process for this module.

Someone knows the possible reason for this default value return?

Thanks,

Leonardo

标签 (4)
标记 (5)
0 项奖励
回复
5 回复数

1,281 次查看
igorpadykov
NXP Employee
NXP Employee

Hi Leonardo

reason may be swapped  DDR bit connections, as described in

AN4467 i.MX 6 Series DDR Calibration sect.11 Write Leveling

i.MX 6 Series write leveling calibration senses the LSB of each DQx byte for the write leveling feedback.

That means, bits 0, 8, 16, and 24 (as well as 32, 40, 48, and 56, when 64-bits DDR3 or dual 32-bits

LPDDR2 is used) of the DQ bus are being used. This fact should be considered during board design.

Best regards

igor

-----------------------------------------------------------------------------------------------------------------------

Note: If this post answers your question, please click the Correct Answer button. Thank you!

-----------------------------------------------------------------------------------------------------------------------

0 项奖励
回复

1,281 次查看
leonardoprates
Contributor III

Hi Igor, the connection between my processor and SODIMM connector is not swapped. Maybe Avant DDR module can be some signal swap but if module does it they doesn't follow JEDEC, right?

I have a module from SMART M that write leveling calibration works well, than I believe that my connections are ok.

FOTO2.jpg

0 项奖励
回复

1,281 次查看
igorpadykov
NXP Employee
NXP Employee

Hi Leonardo

it does not relates to jedec and

from picture is is not possible to say that, since one

needs to look at schematic. More details one can find in

sect.3.5.1 Swapping data lines IMX6DQ6SDLHDG

~igor

0 项奖励
回复

1,281 次查看
leonardoprates
Contributor III

Unfortunately for memory module I haven't schematic, because it is a standard module sell to notebook market. The connection from processor to DDR connector I had care to maintain signals not swapped. Probably there's some signal swap in the AVANT PCB and in SMART M doesn't. Are there some way to estimate correct write leveling register value without use the calibration tool?

0 项奖励
回复

1,281 次查看
igorpadykov
NXP Employee
NXP Employee

If DDR tests run without errors one can leave default vlaues.

Also one can check AN4467 i.MX 6 Series DDR Calibration

it gives steps for write leveling calibration. Run code with jtag

and choose best option experimentally.

One can look at overview of DDR Stress Test Tool

https://community.freescale.com/docs/DOC-101708

~igor

0 项奖励
回复