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.

Looking for evaluation /development board with multiple outputs.

Other Parts Discussed in Thread: PMP9444, PMP7804, LM20333, LM21212-1, LMZ22008, LMZ22010, PMP

Hello, 

I'm looking for a some effective solution for powering our test cards to be tested under radiation. 
Each test card features 5 different FPGAs (Xilinx, Altera, Microsemi) and each FPGA counts 4 ... 6 different power rails (1.0 ... 3.3V, max 3A). So we need about 80 DC-DC regulators to power all this staff at distance of 50m.
Thanks in advance, 
Pavel 

  • Hello Pavel, Thank you for posting.
    TI's reference designs are located here:
    www.ti.com/.../index.html

    Most of these are not purchasable hardware, rather proven design examples that have been built and tested.
    Please look here for some ideas.

    Our IC's do have EValuation Modules, EVM's that may be purchased.
    Start at the Power Management homepage;
    www.ti.com/.../overview.html

    Use the Power Quick Search to see some solutions.
    In the product folder for each IC or Module will be a place to get the EVM, if there is one.

    Regarding radiation testing at 50 meter distance, that's something that TI's Space and HiRel group may be able to help with.
    I can loop them into this thread if needed.
  • Hello Ed,
    Thank you for feedback.
    Unfortunately with Quick Search I didn't find any product that fit our specs. Even simple request : input 5 ... 48V, 1st output: 1.0V, 3A, 2nd output: 1.8V, 3A gives 0 result.
    Concerning HiRel group, we don't plan using hardened devices for power solution.
    But anyway it will be nice to get some feedback from them also.
    Sincerely,
    Pavel
  • Hi Pavel, what are the exact specs you need here;

    Even simple request : input 5 ... 48V, 1st output: 1.0V, 3A, 2nd output: 1.8V, 3A gives 0 result.

    Vin = 5V or 48V?

    If that means 5 to 48V input, it is somewhat unrealistic.

    A design that has that wide input range is uncommon.

    I input 5Vin, 1st output: 1.0V, 3A, 2nd output: 1.8V, 3A

    10 results.

    If you need an input of 48V then you should convert 48V down to something lower first, like 5V.

    You can also input your specifications into Webench.
    Input requirements, then CREATE PROJECT.

    Here is a shareable design I just created. 5V to 1V 3A and 1.8V 3A
    https://webench.ti.com/wb5/OpenPublicSharedProject.jsp?id=BDD14ED63B06A40E&s=c 

    There are many ways to pull up and create designs on TI.COM.
    Will this help?

  • Dear Ed,

    Thank you for feedback.

    Finally I've found a product that matches perfectly our needs: Power Solution for Xilinx® Kintex® UltraScale™,

    www.ti.com/tools/PMP9444

    The only question: was this board tested over long cables (in our test case the loads are located at 50m from power supplies)

    Can we order several boards PMP9444 ?

    Our FPGA test set includes:

    • 3 Xilinx Kintex UltraScale
    • 3 Xilinx Zynq-7000
    • 3 Altera Cyclone-V SoCs
    • 3 Microsemi PolarFire
    • 3 Microsemi IGLOO2
    • 3 Microsemi ProASIC3

    Although power rails in these devices differ a little from Kintex Ultrascale, the rail voltage can be adjusted, isnt it ?

    Sincerely,

    Pavel

  • Hello Pavel, I am glad you found a design that may power your load.
    I can assure you that this was not tested with 50 meter cables, nor will it likely work with a 50m separation.
    There are many Low Voltage, High Current, rails that would suffer greatly from a long connection.

    Many of these rails do not have remote sense, so pushing 10A down the resistance of a cable will cause high losses and voltage drop at the load.
    The distance will also induce noise pickup on those rails.
    FPGA's need low noise, well regulated supplies to operate properly.

    Your end product must have the power supply on the same PCB or at least very close to the FPGA loads.
    Testing should be done with a setup closely resembling the actual configuration, otherwise results are invalid.

    As documented on the PMP9444 folder, Fully assembled board developed for testing and performance validation only, not available for sale.
    Everything is provided for customers to make their own, BoM, Schematic, Board files, however the actual hardware is not purchasable.

    Yes it is possible to modify the design for different voltages and currents.
    Customers are welcome to change the design as needed to meet their needs.

    It is possible that, with proper consideration and verification, your FPGA loads may work with the power source 50 meters away.
    However that will be an exercise in itself to get the system to work properly before you hit it with radiation.
  • Hello Ed,

    I can assure you that this was not tested with 50 meter cables, nor will it likely work with a 50m separation.
    There are many Low Voltage, High Current, rails that would suffer greatly from a long connection.
    Please, see in attachment the piblication on a test case that is similar to ours (altough much simpler concerning testin staff).
    In this test case Kintex FPGA was powered over 40m cable (please see slide 7).
    Well, they used  "generic" 4-channel power supply GW-Instek GPD4303S.
    So, why what it was possible for a "generic" supply isn't for dedicated supply which is PMP9444 board ?
    Moreover,  PMP9444 assures right power sequence, that is probably not the case for GW-Instek GPD4303S.
    Many of these rails do not have remote sense, so pushing 10A down the resistance of a cable will cause high losses and voltage drop at the load.
    I've taken a look at tidr850 document (schematics) and, for example,  for core VCCINT - the most important rail, current sensing is present.
    Your end product must have the power supply on the same PCB or at least very close to the FPGA loads.
    Testing should be done with a setup closely resembling the actual configuration, otherwise results are invalid.
    Unfortunately it's impossible due to radiation environment ... and space-graded regulators are very expensive and well beyond our budjet.
    As documented on the PMP9444 folder, Fully assembled board developed for testing and performance validation only, not available for sale.
    Everything is provided for customers to make their own, BoM, Schematic, Board files, however the actual hardware is not purchasable.
    Yes, i'm avare of this but probably we can find some solution ?
    Overall test staff must be prepared for October, now we are concentrated on test procedure and for power supply cards we have no resources.
    Thank you for comprehension.
    Sincerely,
    Pavel.
  • Hello Pavel,
    The opportunity for noise induced operational errors is high due to the long cables.
    If your goal is to have it operating, running code, you may see errors.
    If your goal is to simply have the system powered while you dose the radiation then you have more options.

    Like that generic supply. (the attachment didn't make it)

    You could make your supplies with modules.
    The PMP7804 reference design provides all the power supply rails necessary to power the Xilinx ® Kintex ® 7 series family of FPGAs. This design utilizes Simple Switcher power modules along with the LM2121x low voltage synchronous buck regulators for "ease of use" and shorter design cycles. This design is optimized for a 12V input.

    It uses:
    LM20333
    LM21212-1
    LMZ22008
    LMZ22010

    http://www.ti.com/tool/PMP7804


    You can buy evaluation boards for the modules used here.
    Modules are much easier to use.
    I think this approach would work.
  • Hello Ed,

    Thanks for feedback.

    The opportunity for noise induced operational errors is high due to the long cables. 
    If your goal is to have it operating, running code, you may see errors.
    If your goal is to simply have the system powered while you dose the radiation then you have more options.

    Our goal is, for sure, have our design operating.

    But the test staff will likely provide shielded cables, so impact of noise should be considerabely reduced.

    Like that generic supply. (the attachment didn't make it)

    You mean test case presented in attachment slides ?

    Yes, in the 1st version they just readback FPGA configuration, but in 2nd version they plan to test transceivers, so some kind of design shuld bu run.

    The PMP7804 reference design provides all the power supply rails necessary to power the Xilinx ® Kintex ® 7 series family of FPGAs.

    Frankly speaking, I didn't understand why PMP7804 is better for our case ?

    We need current monitoring ... and I didn't see this option on PMP7804, whereas PMP9444 provide PMBus interface.

    Moreover buy multiple evaluation boards with one regulation per board will result in more than 100 eval. boards ... dificult to manage.

    Sincerely,

    Pavel

  • Dear Ed,

    We've checked BOM list: some components are obsoletes.

    Could we find an arrangement on acquiring PMP9444 boards from TI.

    Thanks in adavnce.

    Pavel.

  • Hello Pavel,
    The PMP designs are not available for sale.
    They are reference designs intended to show how power can be designed.
    Since some of the devices in that one PMP may be obsolete, you would need to find suitable replacements.

    I proposed several PMP's that show the design approach.
    The easiest path is to find a reference design that has gerbers and design files, and make it yourself.

    As shown on the web, the PMP's will meet specifications when built according to design files.
    However the users take full responsibility for the design when using it.
    Customers are welcome to use these to build their own.

    Evaluation boards may be purchased for some devices.
    Power IC's have EVM's to demonstrate their function.
    You may find some purchasable EVM's that can fill in the major power requirements.