• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 99 subscribers
  • Views 889 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

High current draw for C6678 Evaluation board's UCD9222, UCD7242 power solution

Allan Law
Intellectual 340 points
Other Parts Discussed in Thread: UCD9222, UCD7242, UCD74111, UCD74106

Hi,

My current design makes use of two C6678 DSP with one UCD9222 and one UCD7242 supplying each DSP with its own CVDD voltage rail.

I did a power consumption estimate for the C6678 DSP and found out that the UCD9222 and UCD7242 10A per 1.0V rail power solution for the Evaluation board was at its limit for the default configuration of the "C6678 Power Consumption Summary Rev 2.0" document found in the TI website.

I used the "C6678 Power Consumption Summary Rev 2.0" to estimate the worst case power consumption and current draw for the C6678 DSP. However, the default configuration of the DSP in the excel sheet states that the current draw for CVDD is already 9455.26mA. However the UCD7242 can only supply 10A for each voltage rail. The Evaluation board uses only one rail to supply the 1.0V CVDD output. My board will also do the same with each DSP receiving a maximum of 10A. Hence, according to the excel sheet, the evaluation board is already running close to its limit for the given parameters.

I noticed that the Case Temperature make a huge difference to the Baseline current draw for the CVDD. If I were to run my DSP at the stated 85 degrees C, it will meant that the power solution of the Evaluation Board will be insufficient as it is drawing close to the limit of the UCD7242. If the core utilization is at 50%, the current draw will be over 10A. The only way that it it runs well under 10A is if the case temperature is lowered. This could be the case if the evaluation board is not meant to be run at higher temperatures.

Am I calculating the power consumption correctly?

If so, the evaluation board's power solution appears to be inadequate for intensive processing to be carried out on the board itself. May I find out TI's recommended power solution for intensive computation to be carried out on the C6678?
The hardware design guide gives a reference schematic for one UCD9222 and two UCD7242 with tied outputs. This solution will mean that I will have to use 4 large inductors and that will take up more of space in my design, in which I cannot afford.

The current UCD9222 and UCD2742 power solution recommendation appears to be only suitable lower case temperatures.

Regards,

Allan

  • 0
    TI__Mastermind 30825 points

    Allan,

    The numbers you provided seem a little high, but not out of the realm of possibility for high temps, high core usage.

    Please note that 50% SP Utilization on a core is very high processing, and even on high processing intensive applications 30% is usually about the max.  Note that 50% SP means that it's using all 8 Functional units on each core every single cycle for 50% of the time.  If it's actually using on average 5.8 functional units 50% of the time then this would need to be scalled back by 5.5/8*50%->35% SP.

    Also, as you mentioned the EVM is not intended for high temp environments.  It's intended for controlled lab environments with nominal room temp settings.  The temps at max processing will go up but they will not be in the 85C neighborhood given the thermal solution used.

    Depending on usage conditions (both environmental and device utilization), you are correct that the UCD7242 single rail 10A may not be sufficient for a board design.  You can gang two of the rails together and I believe we also have another higher current solution available.  I'll ping one of our board guys to follow up on this.

    Best Regards,

    Chad

  • 0 in reply to Chad Courtney
    Intellectual 340 points

    Hi Chad,

    Thanks for your reply.

    I will now have to reconsider the utilization percentage. The temperature factor does have a very great different on the current draw as well and I need to find out more about it.

    It will be nice if a small overall package and a high current output solution for the SmartReflex solution can be suggested.

    Regards,
    Allan

  • 0 in reply to Allan Law
    TI__Mastermind 46460 points

    Allan,

    The Known Issues document located on the C6678 EVM website hosted by Advantech discussed that issue.  It states:

    Note that the current CVDD design is limited to 10A maximum.  Operation of the C6678 at extended temperatures (up to 100C case) and at higher speeds (such as 1.25GHz) may require more than 10A.  Customer designs that require the C6678 to operate at these performance levels should implement a 15A using the UCD74110.

    The UCD74110 is the recommended solution for C6678 designs requiring more than 10A.  Note that the UCD74110 is limited to a VIN less than 7V.  You will need to use the UCD74111 to support a 12V input like is on the EVM.  The C6670 EVM schematic and layout design also done by Advantech contains the UCD74110.  It needs to be revised to contain the UCD74111 now that it is released for production.

    Pairing the UCD9222 with a UCD74111 for CVDD and a UCD74106 for CVDD1 as done on the C6670 EVM will provide sufficient current for a wide range of C6678 operating conditions and it should be the best fit for the space available.

    Tom