• State TI Thinks Resolved
  • Locked Locked
  • Replies 10 replies
  • Answers 1 answer
  • Subscribers 80 subscribers
  • Views 2394 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.

MSP430F437: Slow LFXT1 startup at low temperature

David M. Alter
Guru 61235 points
Part Number: MSP430F437
Other Parts Discussed in Thread: MSP430F5529, MSP430FR5994,

A customer is using F437, LFXT1 in LF mode with 32kHz crystal.  They are seeing long oscillator startup times at -40C, anywhere from 15 seconds to 1 minute.  They have looked at appnote SLAA322, but it doesn't seem to help.  In fact, the technique of toggling the XTS_FLL bit seemed to make things worse (longer startup times).  The voltage supply is 3.3V.  The problem doesn't occur at room temperature.

Are these startup times to be expected at -40 C?

They have tried several different crystals with little difference.  Anything else they should look at?

Regards,

David

  • 0
    TI__Mastermind 49692 points

    Hello David,

    I suspect that the crystal is not rated for this temperature or they are not using the correct crystal load capacitance.

    The default value of XCAPxPF is 0, providing a crystal load capacitance of ~1 pF. Reliable crystal operation may not be achieved unless the crystal is
    provided with the proper load capacitance, either by selection of XCAPxPF values or by external capacitors. Keep in mind that low-frequency crystals often require hundreds of milliseconds to start up, depending on the crystal.

    Regards,

    James

    MSP Customer Applications

  • 0
    Guru 88490 points
    Yes, crystals start up slower at low temperature. At -14C (kitchen freezer temperature) I was seeing close to a minute. Lowering the temperature after the crystals were jiggling did not cause them to stop.

    These were 32kHz crystals, so the temperature effect was coupled with the naturally slow startup.

    I never found a remedy, I just accommodated it in my code.

    I'm not aware of low-temperature crystals (i.e. "works well" vs "works eventually"), so I defer to James Evans on this.
  • 0 in reply to Bruce McKenney47378
    TI__Guru 61235 points
    James and Bruce, thank you for your responses.

    Customer responded that "The crystals we use are rated for -40C. We also tried crystals from a different manufacturer and with lower ESR rating. The issue seems to follow the oscillator. We tried all the available values for the internal cap setting and shutting off the internal cap and using an external one."

    Bruce states that he saw 1 minute startup times too at low temperature (-14C for him). James, do we know what the expected startup time is at low temperatures? I want to make sure we're not trying to fix something that isn't broken!

    Regards,
    David
  • 0 in reply to David M. Alter
    TI__Mastermind 49692 points

    Hello David,

    First, do their crystals meet our recommendations? As mentioned in Section 2 in SLAA322D, the most important parameters when choosing a crystal are as follows (please provide their values):

    • Crystal's required effective load capacitance (for 32-kHz crystals, typically 6 pF to 15 pF)
    • Crystal's ESR (for 32-kHz crystals, typically 30 kΩ to 100 kΩ)
    • Tolerance (typically 5 ppm to 30 ppm)

    Notice how the frequency deviation varies across temperature in Figure 2, but this will depend on the crystal selected. I would recommend they contact the manufacturer to provide the real curve for the crystal. Section 4.1 explains a test for oscillator frequency and a method to adjust the oscillator frequency

    Regarding start-up times, according to Section 2.4,  start-up times between several hundred milliseconds and a few seconds are normal values for low frequency tuning-fork crystals, like 32768-Hz crystals. I'm assuming this is at room temperature. At -40C, a startup time of one minute doesn't sound unreasonable.

    Ultimately, I would recommend they carefully read through the MSP430 32-kHz Crystal Oscillators (SLAA322D) app note again. If they compare their design to what's recommended, they'll find the answer more quickly than me posting each recommendation manually.

    Regards,

    James

    MSP Customer Applications

  • 0 in reply to David M. Alter
    TI__Mastermind 49692 points
    Also, the XOSC5 errata could be contributing to the extended startup time, since the oscillator fault error detection may not work reliably.

    Regards,

    James

    MSP Customer Applications
  • 0 in reply to David M. Alter
    Guru 88490 points
    I dug up my notes. I measured 15sec at -14C. I projected ~60sec at -40C.

    The crystal was an ABS06, rated to -40C.

    The MCU was from [Other Guys]. Their data sheet helpfully(!) documented this phenomenon in a Table. Since I wasn't paying attention, I [ahem] "independently rediscovered" it when I took my devices outdoors in the winter. My subsequent observations roughly matched their table (but I couldn't go below -14C).
  • 0 in reply to Bruce McKenney47378
    TI__Mastermind 49692 points

    I found that our newer devices (e.g. MSP430F5529, MSP430FR5994) include a typical LFXT crystal startup time in their datasheets. For the MSP430FR5994, the times at 25C are between 800ms and 1000ms (shown below). Again, the crystal manufacturer will know more about the behavior at various temperatures than we will.

    Regards,

    James

    MSP Customer Applications

  • 0 in reply to James Evans
    Guru 88490 points
    This was a few years ago, but I recall being surprised at the time that -- aside from that one Table -- no one seemed to acknowledge this symptom. As mentioned, I wasn't able to cause the oscillator to stop by dropping the temperature (self heating?), so maybe it wasn't a Thing -- unless you're the one standing in the snow wondering why your RTC isn't starting (:-)).
  • 0 in reply to Bruce McKenney47378
    TI__Mastermind 49692 points

    Hello Bruce,

    That's an interesting comment. It seems like the crystal is like a vehicle's internal combustion engine. At 25C, the engine cranks over easily and runs fine even if the temp drops to -40C. If that same engine sits all night and then tries to crank over at -40C, the starter and battery may not be able to do it or may take many cranks to fire up the engine. In this case, it's the oil that's thick at low temps instead of a slow-moving tuning fork.

    I did some more digging and found an interesting app note for the XOSC8 errata (called XOSC8 Additional Information) which the MSP430F437 does not have. In Section 2.2, it mentions that a higher load capacitance decreases the crystal failure rates at -40C. In Section 2.1.1, it mentions that a crystal with a lower ESR has a shorter startup time and better reliability.

    Regards,

    James

    MSP Customer Applications

  • 0 in reply to Bruce McKenney47378
    TI__Mastermind 49692 points

    Hello David,

    Does the customer have the ability to use external high precision, high tolerance load capacitors? Are their traces for the crystal very long? Have they compared their layout with what we recommend? Also, keep in mind that crystal manufactures typically provide a maximum ESR value in their datasheets whereas the typical ESR value is around 15kOhms less (according to Section 2.1.2 in the XOSC8 Additional Information app note).

    Regards,

    James

    MSP Customer Applications

**Attention** This is a public forum