• State Resolved
  • Locked Locked
  • Replies 11 replies
  • Answers 1 answer
  • Subscribers 97 subscribers
  • Views 1154 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.

C5502 DMA STOPS WORKING TILL NEXT HW RESET

Alexander51251
Prodigy 100 points

Hello, I use the DMA synchronized by McBSP0 receive event for an acquisition system. Suddenly, the flow of data stops and pointers are no longer changed and the only way to restore is the hardware reset. I've seen the Silicon Errata SPRZ020L Advisory DMA_1 and DMA_2 but despite having followed the mentioned workaround such problem still persists. My flow is:

1) DMA EN=1 in DMACCR (start DMA)

2) Enable McBSP0's frame rate generator

3)Remove reset to McBSP0 TX and RX sections

4) Enable IRQ from DMA channel 2

...data...data...data...

5) Put TX and RX sections of McBSP0 in reset

6) Disables DMA channel 2 interrupt

7) DMA EN=0 in DMACCR (stop DMA)

Am I doing something of wrong?

Are there additional circumstances causing such unexpected behaviour?

  • 0
    TI__Mastermind 23785 points

    Can you determine where it hangs? What condition is checked before step 5 is executed?

    Regards.

  • 0 in reply to Steve Tsang
    Prodigy 100 points

    Hello Steve,
    step 5 is performed upon an external asynchronous digital I/O status changing.
    It doesn't hang in the same point. As a true examples it hangs after 253, or 56, or 582, or 488, or ... 16bit elements received after filling the previous acquisition buffer which is of 1136 16bit elements.
    Further, I'm using both AUTOINIT and REPEAT features ON and handling both HALF and FULL frame IRQ for moving data from the acquiring buffer. I tried disabling REPEAT feature and handling DMA registers loading manually but without improvements.

    Thanks.

  • 0 in reply to Alexander51251
    Prodigy 100 points

    Hello, I don't know whether this could make sense or not.

    I have two different DSP markings for the boards in my hands:

    TMS 300/320VC5502PGF/CA-25A8NVW/G4  and TMS 300/320VC5502PGF/CA-77A356W/G4

    I can say that boards (verified on 6 samples) employing the first marking code (25A...) SHOW the problem

    while boards (verified on 5 samples) employing the second marking code (77A...) DON'T SHOW such DMA issue.

    Thanks and Regards

  • 0 in reply to Alexander51251
    TI__Mastermind 23785 points

    The second set of marking is for wafer lot but we cannot find 77A356W. Could you re-confirm?

    Do you have any input or IO pins unconnected/floating? The IO cell characteristic will be different between wafer lot materials.

    Regards.

  • 0 in reply to Steve Tsang
    Prodigy 100 points

    Absolutely confirmed : 77A356W.

    I have some unconnected pins in EMIF and McBSP2 (disabled) area. 

    Regards.

  • 0 in reply to Alexander51251
    TI__Mastermind 23785 points

    We cannot locate that lot code in our system. Can you contact your local FAE or sales for further assist?

    We recommend terminating all unused input or Hi-Z pins accordingly with pull-up or pull-down to guarantee reliable operation. A floating pin can float up or down to logic level hi or lo, depending on things like board/current supply design, process variation, temperature variations etc. and there could be several factors that can  make them see a different behavior from a lot to next lot.

    Having both AUTOINIT and REPEAT ON is the first step mentioned in Advisory. Can you try with REPEAT OFF?

    Regards.

  • 0 in reply to Steve Tsang
    Prodigy 100 points

    Does it mean that devices (both marking codes) may not be TI genuine? How can we verify genuineness ?

    We don't purchase ICs directly as it is the company who assembles our PCBs that does this for us, so can you arrange a mail to me explaining that such lot codes aren't present in your records?

    I tried with REPEAT OFF but unsuccessfully. Regarding unconnected pins I didn't pull them up or down where it is mentioned there's the internal pull up. Anyway will do it.

    Thanks.

  • 0 in reply to Alexander51251
    TI__Mastermind 23785 points

    We can locate one of the marking code, 25A8NVW. I will contact Quality Control Department for the proper steps.

    Regards.

  • 0 in reply to Steve Tsang
    TI__Mastermind 23785 points

    Sorry about the false alarm. Our Quality Control personnel found both lot in their system. He recommended sending three samples from the failing units for us to evaluate.

    Regards.

  • 0 in reply to Steve Tsang
    Prodigy 100 points

    Where should I send the three samples and how long I should expect to get the result of your analysis?

    Also, as I don't have ICs in their original packing is it acceptable for you to send samples unwelded from my boards ?

    Regards.

  • 0 in reply to Alexander51251
    TI__Mastermind 23785 points

    Please fill out attached form, mail units + hardcopy of form to Julie Zak (her address is in the attachment), and also e-mail electronic copy of form to Julie.

    Returned devices should be devices that have been observed to fail. Application note mentioned in attached document should be followed for removing devices from board to make sure devices aren’t damaged (http://www.ti.com/lit/an/slva439a/slva439a.pdf).6232.ASP_TI_customer_return_form_051612.doc

    Our target is to have final report completed in less than 25 days after receiving the units.

    Regards.