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.

DS90UB949A-Q1EVM: DS90UB949A-Q1EVM :I2c read/write failure

Part Number: DS90UB949A-Q1EVM
Other Parts Discussed in Thread: DS90UB949A-Q1, DS90UB949-Q1

Dear Ti :

           I encountered some problems when using the ds90ub949a-q1 EVM. I hope I can get help from ti.

           Background: I have a project that uses ds90ub949-q1 as the serializer on the soc side, in conjunction with the screen's 948 deserializer, to light up the lvds screen. 

           Problem: I have a problem when I use i2c and ti949-q1 on the soc side to communicate: when I first supply power to the ti949 adapter board and then to the soc development board, I cannot read the ti949 registers. But when I cut off the power to the ti949 and then re powered it, the registers that I used the i2c tool to operate it can be read again.

           However, in my project, the i2c communication between soc and ti949 is to be placed in the driver code, but I cannot guarantee the loading order of ti949 and driver. This also leads to the fact that the 949 cannot interact with the i2c code in the driver to truly light up the screen.

Best Regards,

Xinyu

  • Dear TI:

    Add a new one: To exclude the impact of writing other registers, I only read the 0x00 register. Read/write failure is displayed after about 10 times. At this time, reading any register will display failure.

  • Hello Xinyu,

    I am not sure I understood your problem! Are you using 949 EVM or it is your own design?

    What is the adapter board you are talking about and how it is connected to the 949 and SoC?

    Does your SoC supports multi-Master arbitration? 

    Is your SoC and 949 using the same I2C bus?

    Make sure the I2C bias voltage and pull-up resistors are datasheet compliant.

    Also, please make sure your 949 power-up sequence is datasheet compliant.

  • Hello Jaradat,

    Thank you for your reply,

    I am not sure I understood your problem! Are you using 949 EVM or it is your own design?

    >>>>  I use the EVM of 949, not a product designed by ourselves

    What is the adapter board you are talking about and how it is connected to the 949 and SoC?

    >>>> There is no adapter board. The above "adapter board" refers to 949 evm

    Does your SoC supports multi-Master arbitration? 

    >>>> Yes

    Is your SoC and 949 using the same I2C bus?

    >>>>  Use the same I2c

    Make sure the I2C bias voltage and pull-up resistors are datasheet compliant.

    >>>>  I don't know what part of the data table is in this part? Can you point it out?

    Also, please make sure your 949 power-up sequence is datasheet complian

    >>>>  What is the power on sequence of this part?

    Best Regards,

    Xinyu

  • I believe your issue is the interaction between the I2C Master (Microcontroller) on the 949 EVM and your SoC. Once you power up the 949 EVM first, this microcontroller become the Master on the bust, but once you turn ON your SoC first, or power cycle the 949 EVM, then your SoC becomes the Master on the bust.

    Please try to diactivate the I2C Master on the 949 EVM and test again.

  • Hello Jaradat:

    Allow me to elaborate on my question:
    As shown in the figure below

    i2C Master attached to Serializer
    In this configuration, DS90UB949 evm acts as a slave on the local I 2 C bus and DS90UB948 will act as a master proxy on the remote i2c bus
    I need to communicate with the 948 serializer on the screen through the 949evm serializer. Realize functions such as lighting the screen and controlling touch
    The screen is now lit, but there is a problem with the touch function. My touch screen can be understood as the remote slave device in the figure, and I can already communicate with him. However, since my touch is implemented in the cyclic mode (and only in the cyclic mode), that is, I will continuously read the data of the touch device on the screen through the i2c link of 949+948. It was normal at first, but i2c read failed after a period of time.
    At this time, both the 949 EVM reading and the remote slave reading failed.
    I can only read the 949 evm register again by re powering down and re powering on the 949 evm

    This is all my current registers:

    Fullscreen
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    0x00 -> 0x18
    0x01 -> 0x00
    0x02 -> 0x00
    0x03 -> 0xda
    0x04 -> 0xa0
    0x05 -> 0x00
    0x06 -> 0x58
    0x07 -> 0x60
    0x08 -> 0x62
    0x09 -> 0x00
    0x0a -> 0x00
    0x0b -> 0x00
    0x0c -> 0x05
    0x0d -> 0x20
    0x0e -> 0x00
    0x0f -> 0x00
    0x10 -> 0x00
    0x11 -> 0x00
    0x12 -> 0x00
    0x13 -> 0x88
    0x14 -> 0x00
    XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

    Best Regards,

    Xinyu

  • Hi,

    Let me look into this and get back to you!

  • Hello Xinyu,

    Do you mean that after powering your system up you can communicate locally between the I2C Master and the UB949, and through that to the UB948 into the Touch screen, but after some time you can't communicate anymore, either locally or remotely?! Is my understanding correct?

    If the above is correct, how long is the waiting time from working to non-working?

    Are you doing any changes on the system during this time before it fails? Or you just keep cyclic reading? How often you do the reading?

    Once there is no more reading is possible, do you still see LOCK on the UB948?

    How do you recover this issue? You said resting the UB949 will allow local communication again, but what about UB948 and Touch screen?

  • Hello Jaradat:

    1.Do you mean that after powering your system up you can communicate locally between the I2C Master and the UB949, and through that to the UB948 into the Touch screen, but after some time you can't communicate anymore, either locally or remotely?! Is my understanding correct?

    If the above is correct, how long is the waiting time from working to non-working?

    >>>> Your understanding is correct. It took about 1 minute from working to not working

    2.Are you doing any changes on the system during this time before it fails? Or you just keep cyclic reading? How often you do the reading?

    >>>> During this period, I will not make changes to the system, but read the data returned from the touch screen circularly.
    The reading time of each cycle is 20ms.

    3.Once there is no more reading is possible, do you still see LOCK on the UB948?

    >>>> If I want to control the UB949 again, I can only power off the UB949 and then power it on again

    4.How do you recover this issue?

    >>>> Because the above UB949 suddenly loses control, in order to control the UB948 and touch device again, when the UB949 is powered off, the entire system needs to be powered on again.
    This is because when the EVM is powered off, it means that the video signal and the entire control link of I2c are interrupted

    5.You said resting the UB949 will allow local communication again, but what about UB948 and Touch screen?

    For your convenience, the following figure shows my overall display architecture

    >>>> As shown in the figure above

    UB948 is designed in the screen and controlled by its own MCU architecture. All communication with UB948 and Touch devices should be through UB949. After the UB949 is disconnected, it can no longer communicate with the UB948 and Touch devices

    Best Regards,

    Xinyu

  • Hello Xinyu,

    Thanks for the provided details. That makes things easier to understand!

    I have some questions to you:

    - Can you list all the registers from 949 and 948, that you are changing/configuring during your initialization?

    - Please check if the 949 is the one holding the bus or if something else on the SoC side, or the SoC itself is holding the bus on the 949 side when this happens?

    - When the issue happens, can you just PDB the 949 while keeping all the rest of the power? Does that restore the I2C communication to the 949?

    - If you can manage access to the 948 board, can you probe the I2C bus of the 949 and the I2C bus of the 948 with a logic analyzer, along with the LOCK pin of the 948, all on the same window so we can see what exact sequence of events led to the lockup?

  • Hello Jaradat:
    Please forgive me for my late reply
    - Can you list all the registers from 949 and 948, that you are changing/configuring during your initialization?
    Fullscreen
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    boardI2c.write( 0x04,0xa0)
    boardI2c.write(0x5B,0x20)
    boardI2c.write(0x16,0x02)
    boardI2c.write(0x07,0x60)
    boardI2c.write(0x08,0x62)
    boardI2c.write(0x1e,0x01)
    boardI2c.write(0x03,0xDA)
    boardI2c.write(0x5c,0x42)
    XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
    -Please check if the 949 is the one holding the bus or if something else on the SoC side, or the SoC itself is holding the bus on the 949 side when this happens?
    >>>>On the Soc side, I2c used by 949 will connect multiple slave devices
    - When the issue happens, can you just PDB the 949 while keeping all the rest of the power? Does that restore the I2C communication to the 949?
    >>>>Only PDB 949 can resume communication
    - If you can manage access to the 948 board, can you probe the I2C bus of the 949 and the I2C bus of the 948 with a logic analyzer, along with the LOCK pin of the 948, all on the same window so we can see what exact sequence of events led to the lockup?
    >>>> Please post this after I grab it

    Best Regards,

    Xinyu

  • Hello Xinyu,

    I guess these registers are configured on the 949, correct?

    What do you re-configure the BCC Watchdog timer to only 2ms? Can you please use the default value, i.e., 0x7F and see if there is any improvement?!

    >>>>On the Soc side, I2c used by 949 will connect multiple slave devices

    Can you monitor this I2C bus using a scope or Logic analyzer and monitor all the I2C logs to see when is the bus getting stuck, after what transaction!?

  • Hello Jaradat,

    I'm sorry to reply so late. Due to the overall design problem, 948 is designed on the screen, and I can't measure it. Because the I2c used by 949 EVM on the soc side is connected with multiple devices, the logic analyzer is not able to measure accurately.
    But I found a way to improve. After the J9 jumper of EVM was pulled out, the above problems were improved a lot


    This is the default setting. What is the reason?

    Best Regards,

    Xinyu

  • Hello Xinyu,

    thanks for your feedback! That was actually one of the first questions I have asked you, if all I2C pull-ups and voltages are compliant!

    It looks like your SoC or other I2C device is using different I2C voltage! You need to make sure all devices using the same I2C bus are using same I2C voltage

  • Hello Jaradat,

    My SoC side I2c bus connects multiple devices.Each device on the I2c bus on the Soc side is connected with a 2.7k resistor and pulled to 3.3v. Does this affect EVM?

    Unplug the jumper of J9, will it have a bad impact on the subsequent I2c communication?

    Best Regards,

    Xinyu

  • Hello Xinyu,

    that Jumper gives you the option to select either 1.8V or 3.3V for the EVM I2C. So, in this case, you shall choose the 3.3V settings.