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.

Heat problems with DM8168/DM8169

We're experiencing a runaway heat problem with the DM8169 and DM8168.  We're measuring at the heat sink and the temperature on some units will slowly creep up to about 75 C.  Then after some time the processor will lock up (we get a black video output or no video and the processor stops running) and the temperature will shoot up to over 100 C.  We only see this behavior on a few units, but it's a consistent problem for our customers.  On units that are working the typical temperature at the heat sink is about 54 C.

Is there a heat problem with the DM8168/9?  Is there a solution to this problem?

  • Hi Carl,

    The DM8168 EVM has 12V DC fan attached to the DM8168 processor for cooling:

    Regarding heat, this is what we have in datasheet:

    7 Device Operating Conditions

    A heat dissipation solution is required for proper device operation. Thermal performance of the overall system must be carefully considered to ensure conformance with the recommended operating conditions. Heat generated by this device must be removed with the help of heat sinks, heat spreaders, or airflow. SmartReflex can significantly lower the power consumption of this device and its use is required for proper device operation. A thermal model can be provided for thermal simulation to estimate the system thermal environment. Contact your local TI representative for availability.

    Also, we have two types of DM816x devices:

    default - operating junction temperature range Tj is 0 to 95 C

    extended temperature - operating junction temperature range Tj is -40 to 105 C

    See also the below e2e threads:

    http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/716/p/218165/768564.aspx#768564

    http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/717/t/128181.aspx

    Regards,
    Pavel

  • We are using the same fan that's on the DM8168 EVM.

    We are currently using the 0 to 95 C version of the processor.

    As I stated previously, we only have problems on a very few systems and on those systems we see a ramp up to 75 C at the heat sink where it stays for some time and then at the moment that the processor fails we see a sharp ramp up to over 100 C.  This behavior does not seem normal and I was hoping someone would know why this is happening.  There is no change to video nor is there a change in the software when this occurs.  Also, some systems seem to naturally run hotter (10 C) than others.

    Carl

  • I am not from HW team so cant provide useful input on debugging the issue but one thing that could cause such thermal run away on 816x is AVS. The SmartReflex sensor (AVS Controller) inside the chip will request the PMIC to adjust the voltage to the device so that the desired operating frequency can be maintained to compensate for variation across Process/Temperature/Voltage (PTV).816x is a SmartReflex Class 2 Device which means voltage adjustment will happen dynamically when the system is running.

    - if the smart relex sensor frequency reduces, then the operating voltage will be increased by AVS controller

    - if the frequency increases, then the operating voltage will be reduced by AVS controller

    When the temperature increases on die - the oscillator frequency will reduce because the transistors switching slows down with higher temperature along with increase in resistance for the network. So the operating voltage (by AVS Controller) will be increased with increase in temperature

    If the thermal design of the product isn't correctly evacuating the heat,  the increased operating voltage will result in die temperature to increase further causing a cascading effect resulting in thermal runaway and eventual crash.

    Simple way to check if issue you are seeing is due to AVS behavior is to disable AVS in kernel configuration and   supply fixed voltage. Note that disabling AVS violates device spec and this is only for debug. With this the device should not heat up but would crash at lower temperature since it would not receive sufficient voltage.

    Confirming issue is due to AVS doesn't actually help you resolve the issue though. You will anyway have to look at your product thermal design.

    DIfferent samples will have different SmartReflex target voltage depending on the process variation (Hot Vs Cold sample) . So they will reach steady temp at different values depending on how much voltage is required to cause the transistor to switch. Temperature variation you are seeing across samples is expected.

  • We are experiencing the same behaviour reported by Carl Blake. We have few tens of units across a few thousand manufactured that go in thermal runaway. Analysing those units we found that the core current consumption (cold start) is up to 60% more than average, moreover temperature increase from 30 degrees to 90 degrees cause the current draw to almost double.

    This means the more heat it ups, the more power consumption increases going into thermal runaway up to core lock up due to excessive temperature.

    Looks like a loosy silicon process control

    best regards

    Max

  • Max,

    Refer to the below links for more info regarding thermal management:

    processors.wiki.ti.com/.../DM816x_C6A816x_AM389x_Power_Estimation
    e2e.ti.com/.../160138
    e2e.ti.com/.../598875

    Regards,
    Pavel