• State Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 1 answer
  • Subscribers 81 subscribers
  • Views 1462 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.

MSP430F6736: ACLK frequency drift

Samsai
Intellectual 825 points

Part Number: MSP430F6736

Hi Everyone,

We used MSP430F6736 in our project and use 32768HZ crystal for the RTC, we also select 32768HZ XT1CLK as the source of ACLK and then output the ACLK to the GPIO, when we use oscilloscope to test the ACLK and crystal frequency, we found that the crystal is 32767.452Hz and the ACLK is 32827.705HZ. We use ACLK to source the Timer A and generate an 125ms interrupt, we have a software counter in this 125ms interrupt, but we found the counter is faster than the RTC, that means the counter use Timer A is faster than the same clock based RTC. So does this make sense? If we select the ACLK to source from XT1CLK, will the ACLK be exactly the same frequency as XTICLK, or there may be some drift?

Thanks a lot.

BR,

Sam

 

  • 0
    Guru 26200 points
    Hello Samsai,

    <<< we found that the crystal is 32767.452Hz and the ACLK is 32827.705HZ.
    could your scope input capacitance affect crystal operations?

    <<< we also select 32768HZ XT1CLK as the source of ACLK
    <<< we found the counter is faster than the RTC
    <<< So does this make sense?
    <<< If we select the ACLK to source from XT1CLK, will the ACLK be exactly the same frequency as XTICLK, or there may be some drift?
    I do not know what you do.
    Double check how Timer_A is sourced.
  • 0 in reply to Tomasz Kocon
    TI__Mastermind 49692 points

    Hi Samsai,

    Perhaps you're using the wrong load capacitance for this crystal. Have you had a chance to read through the MSP430 32-kHz Crystal Oscillators app note yet? If not, please do. Figure 5 illustrates how the frequency changes with load capacitance.

    For measuring the XT1 frequency, we recommend outputting it to the ACLK pin (like you did). Measuring at this digital ACLK output does not influence the crystal oscillator in any way. ACLK still gives all necessary information to determine the stability and performance of the setup. A frequency counter with a resolution and accuracy of at least 0.1 ppm in the targeted frequency range should be used to measure the 32768-Hz clock signal.

    Regards,

    James

  • 0 in reply to James Evans
    TI__Mastermind 49692 points
    As Tom pointed out, it is strongly recommended not to measure the oscillator frequency directly at the crystal pins. The capacitance at the crystal pins is in the range of 10 pF, and the impedance on this signal line is several megaohms. A typical passive probe has a capacitance in the range of 10 pF and an input impedance of approximately 10 MΩ. Both values are in the range of the oscillator characteristics and heavily influence the behavior of the crystal oscillators.

    Regards,

    James
  • 0 in reply to James Evans
    TI__Intellectual 825 points
    Hi James,

    Thanks for your reply. We measured the ACLK output pin, some board is fine but some board had issue. To simulation the issue, we doing some measurement on the crystal pin and then remove the probe, then we will find that the ACLK frequency increased and will not restore unless we reset the device. This will make ACLK drift from RTC and accumulate over the time, we found the ACLK will count more 150 sec than RTC in 24 hours. One more thing is that the voltage on the crystal pin is 2.86V on these board, which is different from other board 3.0V.
    Do you have any advices?
    Thanks a lot.

    BR,
    Sam
  • 0 in reply to Samsai
    TI__Genius 14165 points

    Hey Sam,

    I think you see the REFO and not the XT1LFO clock when you measure the frequency on ACLK. The REFO is the fall back clock for ACLK if the OFIFG is set due to LFO fault flag.

    So please implement proper fault handling to ensure you always operate the ACLK from LFXT1 and not REFO.

  • 0 in reply to Dietmar Walther
    Guru 26200 points
    Dietmar,
    good point to start an investigation.
  • 0 in reply to Samsai
    TI__Mastermind 49692 points
    Hello Sam,

    Have you resolved your issue? Thanks.

    Regards,

    James
  • 0 in reply to James Evans
    TI__Intellectual 825 points
    James,

    Yes, we confirmed that the reason is XTICLK fault and ACLK switch to REFO internally.
    Thanks.
  • 0 in reply to Samsai
    TI__Mastermind 49692 points

    Nice work! Thanks for letting us know.

    Regards,

    James

**Attention** This is a public forum