imx6q doesn't work in 1G, Could be the ripples of vddarm_cap?

[chuanyanyin]

Hi,

I have a custom board, the soc version is MCIMX6Q5EYM10AC, TO1.2, the software is jb4.3 GA(12/19/2013 released), I use dcdc converter (RT8070) to supply vddarm. The VDDARM_IN is 1.35V and the VDDARM_CAP is 1.225V. When I go to the android UI and click the settings,  the kernel crash ramdonly, It print oops errors. The kernel's log is different every time.I Search the  issue in the this community, patch the the some patches and got the the same result, so I change VDDARM_IN to 1.4V, chagne VDDARM_CAP to 1.25V,  see the code below:

.pll_rate = 996000000,

         .cpu_rate = 996000000,

         .cpu_podf = 0,

         .pu_voltage = 1250000,

         .soc_voltage = 1250000,

         .cpu_voltage = 1250000,},

I have used performance governor to disable dvfs already.

the system don't crash when I click Settings of Android, and it seems to be more stable. But when I run the display_tester_3.8.3.apk to  to check the lcd, the system crash again.   And this apk can run normally in my sabrelite board. When I add arm_freq=800 in the bootargs of u-boot, every thing is fine. Is it caused by the VDDARM_CAP's ripples? In my custom board, the trace length of  22uf bulk capacitor is more than 250 mils from the vddarm_cap pad, and the trace width is 5mil.

                       VOLTAGE    RIPPLES

VDDARM_IN     1.4V            <=80mv

VDDARM_CAP  1.25V          >=100mv

Original Attachment has been moved to: display_tester_3.8.3.apk.zip

[igorpadykov · Accepted solution]

Hi chuanyan,

yes, noise may be the cause,

IMX6DQ6SDLHDG IMX6DQ6SDLHDG, Hardware Development Guide

requires:

"requirement for ripple noise should be less than

5% Vp-p of supply voltage average value.

Related power rails affected: all VDD_xxx_IN and

VDD_xxx_CAP."

Actually big noise also may cause DDR errors.

Best regards

chip

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

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

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

[chuanyanyin]

hi chipexpert

I have used ddr-stress-test 1.0.3 to test the board with 1G and 528MHZ DDR for 4 hour without find any errors. The MCIMX6Q5EYM10AC is not very stable with worst vddarm_cap ripples than MCIMX6Q5EYM10AB?How about MCIMX6Q5EYM10AD?We are going to relayout the board next week,If the cause of unstable at 1

G could be sure,It would great help us.

[igorpadykov]

Hi chuanyan,

MCIMX6Q5EYM10AC,MCIMX6Q5EYM10AD,MCIMX6Q5EYM10AB

versions are identical regarding stability. I believe layout is the reason.

Best regards

chip

[chuanyanyin]

Hi ChipExpert:

    I have measure the ripples again, My DCDC work at 2MHZ,  I have comparee the sabrelites's ripples and my custom board's ripple,

Sabrelite's u-boot's VDDARM_CAP ripples

Sabrelite's Android  VDDARM_CAP ripples

My board's ripples is little higher than the sabrelite.

My board  u-boot's VDDARM_CAP ripples

My board  android's VDDARM_CAP ripples

With this, We can sure the layout cause the ripple issue?

[igorpadykov]

Hi chuanyan,

do you know arithmetic?

IMX6DQ6SDLHDG IMX6DQ6SDLHDG, Hardware Development Guide

requires:

"requirement for ripple noise should be less than

5% Vp-p of supply voltage average value.

Related power rails affected: all VDD_xxx_IN and

VDD_xxx_CAP."

Actually you can add some capacitors and check if this helps.

Best regards

chip