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.

TDA4VL-Q1: EMMC HS400 failure w/ Longsys eMMC chip in high temperature environment

Part Number: TDA4VL-Q1

Tool/software:

Hi, TI Team:

      During the high-temperature operation test of our parking machine product, an error occurred when running the EMMC HS400.

1、Product configuration:SOC: TDA4VL21J5AALZRQ1  

                                     EMMC: Longsys   FEMDME008G-A8A93

2、Test conditions and problem reproduction methods:

      We placed the product in an environment temperature chamber set at 85 degrees, and powered it on for continuous operation for 30 to 60 minutes. At this point, the surface temperature of the chip reached around 98 degrees Celsius. Problems occurred under the eMMC communication HS400 rate. The probability of this issue recurring is 100%.

          When an abnormal situation occurs, the SOC log is as follows:

In addition,Under the same testing conditions, we used Samsung's eMMC part KLM8G1GEUF-B04Q and there were no issues.

3、The eMMC manufacturer Longsys used a logic analyzer to capture the CMD and Data0 when the problem occurred.

4、The test report for EMMC data reading signals is attached. It includes data from Longsys FEMDME008G-A8A93 and Samsung KLM8G1GEUF-B04Q.

The eMMC manufacture Longsys have two question need to be answered by the TI team:

Q1: Regarding the SOC read DS Delay Tuning, is it performed only during power-on startup or does it adjust in real time based on changes such as temperature and voltage?

Q2: Does the EMMC PHY of the SOC have the digital eye diagram function? During the tuning process, the offset of the Ds Delay is carried out step by step, resulting in a set of De Delay values that are available within the UI time frame. This forms a digital eye diagram.

Any ideas about how to solve this issue?

EMMC Signal test report 2025.6.11.xlsx

  • Hi zhu,

    Q1: Regarding the SOC read DS Delay Tuning, is it performed only during power-on startup or does it adjust in real time based on changes such as temperature and voltage?

    No, it does not adjust in real time based on temperature and voltage changes.

    Any ideas about how to solve this issue?

    Just a couple of questions to ask:

    • Was HS400 speed mode working at normal temperatures ?
    • Are you using Linux or RTOS?

    Suggestions:

    • Increasing the drive strength ideally improves the signal quality and could help. Can you try with that?
    Q2: Does the EMMC PHY of the SOC have the digital eye diagram function? During the tuning process, the offset of the Ds Delay is carried out step by step, resulting in a set of De Delay values that are available within the UI time frame. This forms a digital eye diagram.
    EMMC Signal test report 2025.6.11.xlsx

    I have requested the eMMC hardware expert to have a look into this . Thank you for your patience.

    Regards

    Gokul

  • Dear Gokul Praveen:

             Thanks you for your reply!

    Q1: Regarding the SOC read DS Delay Tuning, is it performed only during power-on startup or does it adjust in real time based on changes such as temperature and voltage?

    No, it does not adjust in real time based on temperature and voltage changes.

    If there is an abnormal deceleration to HS and then re-tuning is performed, will the DS Delay be updated?

    2、We found that when problems occurred, there were many IO error reports in the SOC log. What reasons can trigger IO errors?

    3、We found some information in the commands and responses when problems occurred captured by the logic analyzer.

    3、The eMMC manufacturer Longsys used a logic analyzer to capture the CMD and Data0 when the problem occurred.

    When problems occur, from the commands of the logic analyzer, we can find that after the emmc receives CMD46 and completes the data transmission, it receives CMD12 from the SOC and then CMD48. At this time, the response data of the EMMC is 00 40 09 00, indicating that the previous CMD12 is an illegal instruction.We captured the complete command interaction process before and after the problem occurred. As shown in the figure, after the problem occurred, the speed was reduced to HS, then re-tunned to HS200, and then switched to HS400. After running for a period of time, it would again slow down and re-tunning for the same reason, and this cycle would continue.


    Q1:What is the reason that the SOC sent the CMD12 instruction


    Q2:The CMD48 instruction 00 40 09 00 replied by emmc indicates that the previous CMD12 is an illegal instruction. Will this affect the SOC logic determination

    4、Regarding tunning, we can see from the command that tunning was performed when the device attempted to switch from HS to HS200. After success, it switched to HS400 for operation. Tunning was not carried out under HS400. Is there a need for Tunning under HS400? Because HS400 has more bilateral sampling than HS200 ;and uses DS as a clock reference when reading data.

    thanks

  • Hi Zhu,

    I will look into these questions and get back to you.

    Regards

    Gokul

  • Dear Xinfa.

    TI expert will check your latest post then feedback to you as mentioned by Gokul before.

    also, may I ask your help to help clarify more as below?

    1. Which side of signal that you captured by logical analyzer for CLK and Data0? close to eMMC? or close to SOC?

    2. is it enough to capture only Data0 to analyze the data communication between SOC and eMMC?

    3. CMD12 is to do STOP transition, and CMD48 is to discard a specific task or entire queue as JESD84-B51A required.

    we may know more about why HOST(SOC) send CMD12 to Device(eMMC) assuming we can parse the command/response before that.

    So would you please check if you can share the complete log captured by logical analyzer, not a piece of that?  

      

    4. Again would you please share how to interpret the argument and CRC for each command and response that captured by logical analyzer?

    let me take CMD44 as example.

    I think I can parse the command CMD44.

    CMD44: 40050018
    DD: 1
    task id: 5
    number of blocks: 0x18

    R1: 00000B00

    How to parse the response?

    here is the definition in JESD84-B51A.

    thanks a lot!

    yong

  • Dear yong:

    1. Which side of signal that you captured by logical analyzer for CLK and Data0? close to eMMC? or close to SOC?

    Close emmc

  • Dear yong:

    CMD12 is to do STOP transition, and CMD48 is to discard a specific task or entire queue as JESD84-B51A required.

    we may know more about why HOST(SOC) send CMD12 to Device(eMMC) assuming we can parse the command/response before that.

    So would you please check if you can share the complete log captured by logical analyzer, not a piece of that?  

    We  shere the complete logic analyzer log to you.  

    You can use the Acute TravelLogic ver 1.6.12 ; you can download by www.acute.com.tw

    In addition ,We will add more logs to identify the cause of the problem and reproduce the following two log entries that have issues.

    The first CQE recovery and CMD12  -110

    The second CMD 35 error

    the CMD35 Normally log is below

  • log capture by logical analyzer, sent from TTE.

    error_log_run_at_hot_temp_85-1.zip

    Dear Gokul.

    have asked customer xinfa to upload the logical analyzer tool, it can parse above log. Please help check the log.

    BTW, custeomer Xinfa that the log captured on the side that close to eMMC. please check if they need capture the log on the other side that close to TDA4VE SOC.

    thanks a lot!

    yong

  • HI Zhu,

    Can you reply to these questions as well:

    • Was HS400 speed mode working at normal temperatures ?
    • Are you using Linux or RTOS?
    • Which it the eMMC IP you are using?

    Also can you try the suggestion mentioned above:

    1. Which side of signal that you captured by logical analyzer for CLK and Data0? close to eMMC? or close to SOC?

    Close emmc

    Since, it is signals coming from the eMMC which is failing,it is better you discuss this with this eMMC IP vendor as well regarding the same

    Regards

    Gokul

  • Dear Gokul:

    Was HS400 speed mode working at normal temperatures ?

    Yes,HS400 speed mode working at normal temperatures  is OK;

    Are you using Linux or RTOS?

    we ues Linux 

    Which it the eMMC IP you are using?

     EMMC: Longsys   FEMDME008G-A8A93

    Also can you try the suggestion mentioned above:

    1. Which side of signal that you captured by logical analyzer for CLK and Data0? close to eMMC? or close to SOC?

    Close emmc

    Since, it is signals coming from the eMMC which is failing,it is better you discuss this with this eMMC IP vendor as well regarding the same

    The EMMC manufacturers are currently conducting synchronous analysis on this issue.

  • Hi Zhu,

    Suggestions:

    • Increasing the drive strength in Linux ideally improves the signal quality and could help. Can you try with that?

    Regards

    Gokul

  • Hi,Gokul:

    we try change drive strength in dts;

    we Configuring DR_TY at 44 ohms or 30 ohms  does improve things, but it doesn't solve the problem.

    The other values have not improved.

  • Hi zhu,

    You have changed this property in dts to 33 ohms right .

    • Also have you got any feedback from the eMMC vendors on this?
    • Can you also provide the SDK version you are using so that we can look at the otap delay parameter as well in the device tree.

    Regards

    Gokul

  • Hi,Gokul:

    You have changed this property in dts to 33 ohms right .

        Yes,we have changed.

    Can you also provide the SDK version you are using so that we can look at the otap delay parameter as well in the device tree.

    we use SDK Version is 8.5

  • correct we use SDK is 8.05

  • Hi Zhu,

    I will look into this and get back to you.

    Regards

    Gokul

  • Hi Zhu,

    We cross verified the otap delay values as well.They are the same.

    I am highly suspecting the eMMC IP to be causing issues as the response is not coming from the eMMC but the command is going from the SOC.

    Regards

    Gokul

  • Dear Xinfa.

    would you please help provide the status on your side?

    thanks a lot!

    yong

  • Dear Xinfa,

    please let us know if need support from TI on this. We will try to close this ticket this week.

    thanks a lot!

    yong