• State TI Thinks Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 1 answer
  • Subscribers 31 subscribers
  • Views 564 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.

Original question:

TM4C1294KCPDT: Ethernet Stability Issue with TM4C1294KCPDT on custom board

TM4C1294KCPDT: Ethernet Stability Issue with TM4C1294KCPDT on custom board

Suchit Patel
Prodigy 95 points
Part Number: TM4C1294KCPDT
Other Parts Discussed in Thread: ENERGIA

Just to update. We could achieve very good stability with Tivaware firmware. The energia implementation seems to have some memory leak issue and hence is getting all these issues. We have successfully put the device on test run for a month without any issue, while the same could not be achieved with energia ethernet library.

  • 0
    Guru 117855 points
    Has your, 'month long test' included 'High Traffic' - which proves the 'usual suspect' - when stability becomes challenged?
    Indeed the 'definition of what constitutes such high traffic' comes into play - yet it seems (reasonable) to 'Test under DEMANDING Conditions."
  • 0
    TI__Guru**** 190496 points
    Hi Suchit,
    Glad your problem is solved. I guess you are reporting back your result in reference to your earlier thread e2e.ti.com/.../672044.
    cb1 has a good suggestion that you have "stress" test your network and conclude the TivaWare stability vs. Egergia implementation.
  • 0 in reply to Charles Tsai
    Guru 117855 points

    Hi Charles ... and thank you.    An "under-challenging" test (almost) qualifies as 'No Test' - and may prove even  WORSE!    

    HOW?    By generating 'false - likely "Premature Confidence" - causing a system to be "Released" - when in fact it is  NOT REALLY READY - for 'Prime-Time!')

    Any such 'test' - when  described ONLY by duration - is a high candidate for, 'Proving Suspect...'

  • 0 in reply to cb1_mobile
    TI__Guru**** 190496 points
    HI cb1,
    Thanks. I tend to think that under the same test, the Energia version is unstable at time (i.e. memory leaks) vs the TivaWare implementation. Hopefully the poster has done additional tests to conclude the TivaWare version is stable enough and ready for production.
  • 0 in reply to cb1_mobile
    Prodigy 95 points

    Test Setup Details:

    Protocol: Modbus TCP

    Clients: 8

    Scan Rate: 50ms

    Delay Between Polls: 20ms

    Timeout: 1000ms

    This is what we were targetting for our device to handle, and it worked without any issue for a month. I Hope it helps.

  • 0 in reply to Suchit Patel
    Guru 117855 points

    Thank you - much appreciated - and do understand - my goal is to,  'Highlight areas which amplify Test Success.'      My small firm plays a major Diagnostic role w/our clients - thus we often encounter Client, 'Test & Test Design Issues.'

    A common thought is that, 'Testing (alone) will aid quality.'    I don't believe that - and our findings reveal - testing often  to prove, 'Inadequate in that regard.'     Standard & recognized Quality Manuals note that 'Quality' stems primarily from very sound design - resulting from:  great planning - a  safe/conservative, methodical design objective & implementation - and the inclusion of (many) 'What If'  (i.e. Test Design scenarios.)

    In your specific case - how likely is it that 'each' of your 4 major test criteria:

    • Remain w/in your listed, boundary conditions?
    • Has included a 'guardband'?    (i.e. such that a '10% violation' - possibly for a prolonged period - is tolerated?)
    • As  ALL FOUR of your listed conditions near their respective limits - does your system (really) meet proper performance requirements - and remain stable?
    • As your boundary condition is neared - has your design provisioned for, 'Alert Signaling' - which 'commands' the, 'violating node' to 'better comply?'
    • Should the violating node  "Fail to comply" - is there some effective means to 'divorce it - from your network?'
    • And - how 'standard' are your listed test criteria?     Do they meet industry norms?     Might 'very helpful'  (other)  test measures - been bypassed - either through 'rush or oversight?'

    When such 'Test Programs' are incorporated from, 'DESIGN COMMENCEMENT' - and  Flow 'hand in hand' w/the (normal) Design Process - both the Design & Test tend toward optimization!     In stark contrast - too often - 'Test' is the  (very)  'Last Consideration' - is created under duress - and any (far earlier) 'feedback to the Design Process' - has 'Left the Building.'