• State TI Thinks Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 5 answers
  • Subscribers 62 subscribers
  • Views 299 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

Original question:

UCD90320U: Difference between UCD90320U and UCD90320

UCD90320U: UCD90320U

Mike Irvine
Prodigy 10 points
Part Number: UCD90320U
Other Parts Discussed in Thread: UCD90160, UCD90120, UCD90320

The UCD90320U has additional protection from Single Event Upsets compared to the UCD90320, the UCD90120 and the UCD90160, it has the ULA  molding and the SEU detection algorithm (8.4.20 Single Event Upset). My questions are:

  1. The ULA molding is not available on the UCD90120, or UCD90160 are there plans to add it.
  2. Was the ULA molding and SEU detection  added for a particular reason, were the other parts found to be susceptible to SEU events.
  3. For the non ULA molded, non SEU detecting devices, Did their internal fault management/watchdog timer reset the device in the event of a SEU event.
  4. Is there any data on creating upsets in any of the devices listed above based on LET or other data that can help with their evaluation? 

Regards,

-Mike

  • 0
    TI__Guru 60800 points

    Hi

    Please see my comments below.

    1. UCD90120/UCD90160 has much small size of SRAM and Flipflop therefore their SEU performance is much better. Therefore there is no plan to adopt ULA on those device.

    2. UCD90320 has large size of SRAM and flipflop but customer need its density for new reliable application. so we just adopt the ULA to give a extra protection. 

    3. No reset.

    4. we do have customer to perform a neutron beam test on the UCD90320U. The beam used in this case was "white" neutron spectrum (atmospheric like) produced from a spallation target downstream from a synchrotron produced pulsed proton beam with a max energy of 800MeV. The FIT data is about 10-15

    Hope this help

    Regards

    Yihe

  • 0 in reply to yihe
    Prodigy 10 points

    Thank you for your reply I would like to ask for a little bit of clarification as im unsure about some of the answers. 

    1. Are their plans to adopt the SEU Detection on the UCD90120/UCD90160? 
    2. What about the SEU detection,
      1. Was the U variant added because of TI or customer testing and the chip failing in a certain enviroment
    3. In this case what is the expected result of a SEU, Continued operation with errors in the configuration?
    4. Do you have testing information on a non U variant like the UCD90120 or the UCD90160
    5. Do you know if this was able to effect the stored configuration flash or just the run time SRAM memory.
  • 0 in reply to Mike Irvine
    TI__Guru 60800 points

    Hi

    Please see my answers below:

    1. No, since their FIT rate is relative low.

    2. No, it is just a customer ask for a more robust solution and adding ULA is the best for it.

    3. yes. continuous running

    4. No, their FIT rate is relative low due to small size of SRAM and flipflop

    5. SEU only impacts on the SRAM. Flash type is not changed.

    Regards

    Yihe

  • 0 in reply to yihe
    Prodigy 10 points

    Thank you for your answers Yihe they have been very helpful,

    Is there any way to get the neutron beam test information on the UCD90320U?

    Regards,

    -Mike

  • 0 in reply to Mike Irvine
    TI__Guru 60800 points

    Hi

    What provided in the original reply is what we know about it.

    Regards

    Yihe

  • 0 in reply to yihe
    Prodigy 10 points

    Ok understood. Hopefully one last question about the flash memory storage on the UCD90320U,

    Is the configuration ECC? And if so is there any correction or scrubbing at all of the non-volatile memory. (In the case where a correctable error is detected, will it correct it in the flash, or will the errors build up?)

  • 0 in reply to Mike Irvine
    TI__Guru 60800 points

    Hi,

    The flash has ECC.

    Comparison with SRAM, NVM is much less sensitive to the SEU. 

    The data of flash is copied to SRAM only after the device is power up. Moreover, if there is a bit flip in the flash, the self-checking will fail when device is power up again and device stops function due to a invalid configuration. 

    Regards

    Yihe

  • 0 in reply to yihe
    Prodigy 10 points

    Hey Yihe, 

    Unfortunately, our application doesn't allow us to reconfigure the device once it has been sent out into the world if the flash self-check fails. That is why we are concerned about both the configuration flash and the base program flash, either one failing would cause a total failure of the product.

    Regards,

    -Mike

  • 0 in reply to Mike Irvine
    TI__Guru 60800 points

    Hi

    As said, the flash is not as sensitive as SRAM besides the ECC. we haven't heard any report so far. If your system really need zero FIT rate, you may consider redundancy . 

    Regards

    Yihe