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.

Regarding CORE (AVS) POWER CONSUMPTION FOR 66AK2H14

Other Parts Discussed in Thread: 66AK2H14, UCD74120, UCD9244, UCD9222, UCD7242, LM10011, LM10010, TIDEP0026, TPS40428, TPS544C25

Hi,

     We are using 66AK2H14 in one of our Design. We have calculated  that the current consumption for both ARM & DSP( AVS )  core  supply  using  TI  66AK2H14_Power_Consumption_Model_r1p0p3 . The total current requirement  is 21.42 A . If we increase the SP & CC utilization to 100% for all DSP and ARM core PACS, the overall current requirement on CVDD crosses 25A.

     Is it practically possible to utilize 100% SP & CC of all cores. Since we are planning to use single UCD74120(25A) power driver instead of  two independent UCD74120 as existing in EVM.

   Please find attached our Power consumption chart. We need suggestions on using single UCD74120.

Regards,

Sakthi.A66AK2H14_Power_Consumption_Model_r1p0p3.xlsx 

  • Hi Sakthi,
    The UCD74120 will work as a solution for the CVDD supply for utilization that you have listed in your spreadsheet. As you stated, it is not possible to reach 100% utilization on all cores. A single UCD74120 should be enough. Note that the parallel UCD74120s implemented on the EVM were originally included to support separate power sources for the ARM core supply and the DSP core supply. Those supply rails were combined in the final version of the K2H but the EVM design could not be changed to support a single supply for that combined rail.
    Regards,
    Bill
  • Hi Bill,

    1. Please help me confirm one thing that if the design for DSP&ARM AVS power supply of the two combined UCD74120s in EVMK2H rev4.0 is enough for two 66AK2H14s or not.
    2. And if the 66AK2H14's ARM smartreflex is not work at this moment? So the EVMK2H connects the PWM singal controlled by the DSP smartreflex to the other UCD74120.
    3. And what do you mean about 'EVM design could not be changed to support a single supply for that combined rail' exactly?

    Regards,
    Feng
  • Hi Feng,

    1) If you look at the power supply pins for the K2H on page 23 of the EVM schematic, you will see that there are pins labeled CVDD and CVDDT which are both connected to power supply net CVDD. Initially, the K2H was designed to have separate AVS power supplies for the DSP core (CVDD) and the ARM core (CVDDT). We decided that this wasn't needed and that both CVDD and CVDDT could be supplied by the same power supply. The EVM was designed before this decision was made so it has two separate supplies using two separate UCD74120s. One of these would have been connected to the CVDD pins and one would have been connected to the CVDDT pins. A single UCD74120 could not supply sufficient current for both the CVDD and CVDDT pins so the outputs of the supplies were connected together.

    Note that the power supply design for the EVM was only designed to support a single K2H device. If you have two K2H devices, you will need two power supplies. Each of the power supplies should only be connected to one of the K2H components and each should be controlled by that device to assign the proper AVS voltage. An alternative power supply example can be found on the K2E EVM.

    2) AVS is required by the K2H products. The EVM was designed before the final AVS characterization was complete and does not have the AVS programming enabled but the EVM was design to operate in a lab environment for software development. A production system must use AVS.

    3) As I discussed above, the K2H was originally expected to have CVDD for the DSP and CVDDT for the ARM. These supplies were combined and are only referred to as CVDD in the K2H data manual.

    Regards,

    Bill

  • Hi Bill,

    Thanks for you replies with great details!! That helps me a lot!

    Regards,
    Feng
  • Hi Bill,

    One more question.

    In EVMK2H schematics, there are two VID control signals. I think that one is for DSP scaled core CVDD and the other is for ARM scaled core CVDDT.

    But in the lastest 66AK2H14's datasheet, the Pin VCNTL2T(AU37), VCNTL3T(AV36), VCNTL4(AU35) and VCNTL5T(AW36) are reserved.

    Does this mean that the 66AK2H14's ARM samrtreflex is not supported at this time?

    Regards,

    Feng

  • Hi Feng,

    That is correct. Since the CVDD and CVDDT voltages have been combined into a single CVDD rail, only one control interface was needed. The second interface is not supported. 

    You should also note that the data manual still includes the CVDD1 and CVDDT1 memory array voltage rails which are both 0.95V. On the EVM, these are driven by separate power supplies but they can be driven by the same fixed 0.95V power supply in customer designs. 

    Regards,

    Bill

  • Hi Bill,

    In EVMK2H's schematics, one UCD9244's channel 1A outputs one PWM signal that is Farm0_PWM-1A to controller two UCD74120s that are combined together to supply for CVDD. The signal Farm0_PWM-2A is not used really.

    And this UCD9244's channel 2A is reserved for ARM smartreflex. This UCD9244's channel 2A just monitors the voltage proportional to one UCD74120(part U44)'s output current.

    But as we discuss above, the ARM's smartreflex is not supported at this moment. So I want to modify channel 2A's signals to support another UCD chip.

    So my new question is,

    How can I use one UCD9244's channel signal CS1A to monitor the current summary of two UCD74120s(as showing in the EVMK2H's schematics, one is part U10 and the other is part U44)?

    As I did before when using UCD9222 to control two UCD7242s, the UCD7242's IMON pin outputs the current proportional to the actual load current and I just combined two UCD7242's IMON pins and connected them to one UCD9222's channel signal CS1A as the UD7242's datasheet shows.

    But the UCD74120's IMON outputs the voltage proportional to the load current. So I can not just combine these two IMON pin together.

    So do I need one voltage addr?? Could you give me advice to deal with the this situation using the simplest way?

    Or the IMON function is not imperative to generate the CVDD power supply and it is just a optional function to see how much the current load will actually be. 

    So can I ignore these two IMON pins?

    Regards,

    Feng

  • Hi Feng,

    You are attempting to copy the EVM design as a power supply solution for the K2H component. Due to the changes in the power strategy I've described in earlier posts, that power supply design may not be the best implementation for you. The first step for determining the best design is to populate the power consumption model with your expected needs. You may discover that your current needs for CVDD do not exceed the 25A that can be sourced by a single UCD74120. If you do need more than the 25A provided by that part I strongly urge you to consider a different solution. Attached is a presentation from the power supply group describing analog controller solutions for the K2H using the LM10011 for AVS control. After you have completed the power consumption model, you may find one of these solutions to be a better fit for your design.

    Regards,

    Bill

    TI Analog for Keystone 2 power.pdf 

  • Hi Bill,

    Your suggestion is perfect for me!!! Thanks!
    I see the EVMK2E uses the TPS+LM series to implement the CVDD AVS supply. Although the EVMK2H uses the UCD9244+UCD74120 to do so, I think it is deserve to try the TPS+LM series to implement the CVDD AVS supply of 66AK2H14 or TCI6638.
    And I will feedback the result of our own design board in this post. But it will be after 3 months or later because of the delay of schematics design, PCB fabrication and hardware debug.
    If possible, please show me more reference materials about the power supply and clock supply design of Keytone II SOC.

    By the way, this file attached above is only the power supply part of a fully file named 'TI Analog Solutions for Keystone II Family of SOCs'.

    Thanks again!

    Regards,
    Feng

  • Hi Feng,

    I didn't want to overload you with too much information so I trimmed the presentation to include just the power supply section. The full presentation is attached below. In addition you may want to read the TI Designs TIDEP0041 and TIDEP0026 which describe the power supply solution and the clock generation for the K2E EVM. The power and clocking solutions for K2E are very similar to the K2H and should provide some good examples. PMP7256 also gives an separate example of generating the core supply for the C6678 using the LM10010. The LM10011 should be used in place of the LM10010 in that design to meet all the specification for the KeyStone and KeyStone II components. 

    Regards,

    Bill

    TI Analog for Keystone 2.pdf

  • Hi Bill,

    Thanks for your so detailed replies! That really helps me a lot.

    Regards,
    Feng
  • Bill,

    Can the LM10011 be used with the TPS40428? Using that somewhere else and would like to use same supply for this.

    Thanks,

    Jay
  • Hi Jay,

    the TPS40428 does include the FB circuit similar to the TPS544C25 but I don't have any experience with that particular power supply controller. If you haven't already, I suggest you review the information in Generating AVS SmartReflex Core Voltage, PMBus for K2E Using TPS544C25 & LM10011 Reference Design and then post your question to the E2E forum supporting the TPS40428. 

    Regards,

    Bill