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.

TMDSEVM572X: Design queries

Part Number: TMDSEVM572X
Other Parts Discussed in Thread: AM5728, SN75LVCP412A

Hi TI Support,

We have AM5728 evaluation kit (TMDSEVM572X) & further we intend to make our customized board based on EVM schematic. Pertaining to these we will appreciate following queries,

Application background

 We have the application of video analytics running continuously in AM5728 processor. We intend to MUX ethernet ports of AM5728 to receive image data from 2 X IP cameras, so both DSP & both ARM cores will be running at HIGH OPP. Apart from that there are some other peripherals in our application which would be running continuously besides mentioned above. Using power estimation spread sheet results in approx 14.5W processor consumption. I also gone through TI application note of thermal considerations.

Queries,

1. Is it practical to dissipate 14W with suitable heat sink & fan keeping junction temperature below 90'C? If yes, can TI provide some directions to achieve this target & possibly advice reference of processor heat sink manufacturer who could help us in this matter?

2. BTW, is it feasible in evaluation board to enable both DSP, both MPU, both IPU, IVA cores at HIGH OPP simultaneously? If feasible, it is advisable from TI to run evaluation board (with heat sink+fan) with portions enabled as mentioned above. Assume ambient to be 25'C.

3. For our custom design purpose, is it OK to use direct SATA port through 0.01uF capacitors without using MUX IC PI2DBS212 & SN75LVCP412A?

4. We don't require PCIe interface, can we remove differential clock IC CDCM9102RHBR & still keep both ethernet ports working on 1Gigabits/s speed from some internal clock of AM5728?

Thanks,

jagdish

 

  • Hi,

    3. The MUX IC can be removed if you don't need it. SN75LVCP412A is a SATA redriver, and it is entirely up to you to decide whether it's needed, depending on your use case.
    4. Ethernet ports are not clocked from the CDCM9102RHBR.

    I have asked the factory team to comment on the other questions. They will respond here.
  • Hi Biser,

    Thanks for your response.

    3. I think SN75LVCP412A is redriver & is used to compensate for losses in MUX switch. Reason for asking support was we are not using MUX & some old designs using other processors had direct connections through 0.01uF capacitor.

    Anyway, I would like to know personal view of TI in use of this IC without MUX. If there is recommendation we will use this.

    Kindly advice.

    Thanks,
    jagdish
  • Hi Jagdish, let's see if I can help you out! 

    1) I would like to review your power estimation spreadsheet. 14W is a very high estimate. Actual power is very likely below 7W. We don't have heatsink partners or reference partners. One company I am aware of is . It's good that you already reviewed the thermal considerations app note, the heatsink used in that analysis is CTS BDN10-3CB/A01

    2) See Table 5-3 for the maximum power-on hours by OPP. It is up to your requirements what target you want to meet. The boards in general are only designed for short-term evaluation at room temperature operation.

    3) Max trace length according to this doc is 3500 mil. If your sata device is too far from the chip it might be needed. 

    4) Should be OK. A 25 MHz osc is dedicated to the ethernet PHY already.

  • Hi Ahmed,

    1. Thanks for your response. Please find attached spreadsheet. The power consumption given in thermal application notes doesn't advice if both DSP cores, IPU cores are working or not. As I told before that our application will continuously scan image data from ethernet ports, run video analytics algorithms on data & upload data through other ethernet port. We anticipate both MPU, DSP cores, etc are required at full speed (OPP HIGH for all) & will be fully occupied (>90% loading) & hence pertaining loading has been assigned in power spread sheet. 

    Along with that peripherals SATA, USB, SPI, I2C & UARTs will be required to work continuously besides above mentioned cores, so I have just added what we require in our application & power consumption comes more than 14W. Please review the spread sheet & advice further if any optimizations/ assumptions could be planned.

    2. Actually, I saw POH table & we wish to keep temperature near 90'C (for +105'C part) with above mentioned peripherals & cores working simultaneously. Product range will be -40'C to +60'C & hence only margin of 30'C to keep 90'C junction temperature. So large heat sink may be needed here along with fan.

    3. OK, better we will keep redriver IC.

    4. OK.

    Thanks,

    jagdish

    AM572x_Power_Spreadsheet_v1p4p1.xls

  • Hi Ahmed,

    Can you please update for above mentioned email.

    Regards,

    jagdish

  • Hi Ahmed,

    Can you please respond to our queries. 

    Regards,

    jagdish

  • Hi TI support,

    Is there anybody who could respond our queries instead of Ahmad.

    Regards,

    jagdish

  • Hi,

    I've reviewed your spreadsheet settings. The I/O frequency of both GMAC is set to much higher frequency than the max freq supported by the device. When it's changed to the max freq 125 MHz supported by the device for 1Gbps, total power went down from 14547 mW to 10752 mW.
  • Hi Kshin,

    OK then let take the figure of 10.7W, so now can you please answers our below queries,

    1. Do you think this much power can be manageable practically, if yes some guidance in this respect would be appreciated like keeping 2 fans, one for processor  & other for exhaust, etc would be appreciated.

        We want to junction die temperature near 90'C at max ambient of +60'C to maintain required POH.

    BTW, if device is not supporting the frequency of GMAC as you say don't you think there must be some limit checks in sheet to preven user entering unpractical data?

    Awaiting your response.

    Regards,

    jagdish

  • Hi Jagdish
    I see that Shin has replied to your earlier posts - sorry for the delays as Ahmad is out of office due to an unexpected issue.

    On your follow up question on thermal design - we will really not be able to provide additional guidance on how to design your thermal system
    as it really depends on your end system board design, enclosures , ambient, ability to use heat sinks/fans vs enclosures/fins or something.

    To help further you can look at the following simple thermal char we did on the EVM

    www.tij.co.jp/.../sprac53.pdf

    last page on this application note has a ~10 W use-case and you can see what the ambient to junction without heat sink, with heat sink (not the best heat sink), with heat sink and fan. Hopefully this gives you some additional guidance.

    On the feedback to add more limits to PET, that is a good suggestion, but we usually try to not do that as it is difficult to keep this maintained if the datasheet parameters changes. There is a lot of checks and balances that can be added in the tool, but that adds to complexity on the tool and its maintenance. We will keep your feedback in mind for future.

    Do also note that if your board / power/end application design is flexible, you could also consider using predominantly 1.8V IOs instead of 3.3V IO , you will see this save additional power for relevant interfaces ( I understand that for ethernet etc, that might not be a popular choice).

    Let us know if you have any follow up questions.

    Regards
    Mukul
  • Hi Mukul,

    => Thanks for your response, actually I gone through the thermal document before posting our queries to TI & hence I was expecting some direction to work on like only bigger heat sink or heat sink+fan, etc & not design details. So if you could suggest something that will be better.

    => Apart from this we have couple of queries like

    a. In thermal application document dual core dhrystone application (OPP_HIGH) is showing approx 8200W at 60'C ambient. Are both DSP cores running in this scenario or they are not included in this calculation? Can you please clarify.

    b. AM5728 has processing specs of 60fps. What is the OPP frequency at which this is specified like 1500MHz, 1000MHz, etc? Actually I want to see if we could reduce operating frequency keeping performance similar or less degraded & getting thermal dissipation lower. For ex. does AM5728 support 60 fps at 1000MHz?

    c.  How much fps AM5728 will be supporting assuming both ARM cores are 50% utilized & both DSP cores are 90% utilized. You can assume none of other application is running just pure frames processing. 

    I understand exact figures will be difficult but some directive values could be given & that will be helpful alot.

    Thanks,

    jagdish

  • Hi Mukul,

    Can you please respond to my trailing queries of 6-Dec.

    Regards,
    jagdish
  • Hi Jagdish
    Sorry for the delay in response

    On your queries
    >>=> Thanks for your response, actually I gone through the thermal document before posting our queries to TI & hence I was expecting some direction to work on like only bigger heat sink or heat sink+fan, etc & not design details. So if you could suggest something that will be better.

    Sorry I do not understand your request here. What would you like us to suggest? The thermal mitigation techniques for your application would require an understanding of your board size, layer count, processor orientation, whether or not the board is within an enclosure or not , what typical ambient temperature it gets exposed to, what is your space and price budget your application has, whether you can put a fan in your solution or just need to rely on the heat sink and chassis.
    So I will reiterate that this is likely something that you would need to decide based on various factors. Let me know if I am missing something here.
    If you are looking for another heat sink recommendation, on one of our upcoming AM572x EVMs we will have the following
    www.digikey.com/.../1284952
    Again, for us EVM is not an end product, it does not have an enclosure, and typically expected to operate in nominal temperature conditions, so the choice we make may not be directly applicable to your end product.


    For the additional questions
    a) For the dual Dhrystone test , the DSP are idle. No activity
    b) For this can you post another query as a separate post. I think this is best addressed by software team. I will like to reiterate that at 1500 MHz the POH profile is different , so please be aware of that for your end application. Details on POH in the datasheet.
    c) This will be , as you expected, hard to quantify. The simple definition of 100% ARM utilization is ARM doing Dhrystone , and 100% DSP utilization is DSP doing an IPC of 8. Even the most optimized video codecs will usually have any IPC of less than 4. So I think something like 70% ARM and 50% DSP utilization is fine from a power profile, but it is not a straightforward conversion to what it means in terms of frame per second.

    Hope this helps some.

    Let us know if you have additional questions.

    Regards
    Mukul
  • Hi Mukul,

    Thanks for your response. Yes I meant some directional guidelines (not in detail) for our application. OK, we will work it out for our application.

    b. OK we will post this query in software section.

    Regards,

    jagdish

  • Sounds good Jagdish.
    Here are some additional posts on thermal discussions you might find useful or may lead you to have more follow up questions

    e2e.ti.com/.../2385602
    e2e.ti.com/.../2029915

    When posting question b on a new post please do provide further clarity on camera resolutions, codecs etc.

    Let us know if you have additional questions. I plan to mark this as resolved, but feel free to post follow up questions as needed here or on a new thread.

    Regards
    Mukul
  • Hi Mukul,

    Thanks for useful suggestions. We will take care during posting b. Yes, will definitely ask you for any more queries.

    Regards,

    jagdish