• State TI Thinks Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 1 answer
  • Subscribers 98 subscribers
  • Views 35 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.

AM4378: Crystal stops working - any insight ?

Geoffrey Ficara
Expert 3385 points
Part Number: AM4378

Tool/software:

Hi, 

Customer is facing an issue. On the field he notices that the crystal suddenly stops working.

- Turning off and on the system doesn't systematically solve the issue. 

- Changing the xtal to new ones systematically solves the issue. 

- Appliying heat and cold they observe that the voltage of the 24 MHz signal goes up and down and in some case stops working. 

- Customer has the following xtal circuit. He uses the ECS-F-240-18-28AX - datasheet here  3007.csm_4ax-14522.pdf

Customer measures a Cshunt of 3 pF and Xtal resistance is also within the recommended operations. 

Customer is of course checking with the crystal supplier but have you seen such issues before and is there something to modify on the capacitors used around the crystal ? Or any intern cap calibration to modify ? 

Thanks

Geoffrey 

  • 0
    TI__Guru 63125 points

    It is not unusual to find a crystal circuit that never begins to oscillate when power is first applied, but it very unusual to find a crystal circuit that stops oscillating after it has started.  This occurs because it takes more gain in the oscillator and crystal circuit to start oscillation than maintain oscillation.

    The values of C151 and C152 are not correct for the crystal being used.  The crystal Y1 and capacitors C151 and C152 make up a resonate tank circuit, where C151 and C152 are connected in series and the two series capacitors are connected in parallel to the crystal. Therefore, the capacitive load applied to the crystal by capacitors C151 and C152 crystal is [(C151×C152)/(C151+C152)] = 9pf rather than the crystal requirement of 18pf.  However, it is not possible to increase the values of C151 and C152 to 36pf because the maximum capacitance allowed by the AM437x device is 24pF. This means you need to select a different crystal that has a lower capacitive load requirement.

    I'm not sure how you are measuring the amplitude of the crystal circuit, but this can be tricky. You need to measure the amplitude with a non-intrusive scope probe that has very low capacitance and very high DC impedance. Most scope probes that have very low input capacitance will be an active probe that has 25k to 50k of DC input impedance. A 10x passive probe will have very high DC input impedance but is likely to have 8 - 10pF of input capacitance. Both of these options can be intrusive to the measurement. Using the lowest input capacitance 10x probe you can find will be the best option for measuring the amplitude of oscillation.

    The oscillator already has an internal feedback resistor, so there is a good chance the 1M resistor R39 is not required to start oscillation.

    Have you checked the crystal power dissipation to see if the oscillator is over-driving the crystal circuit and causing it to fail after being exposed to an over-drive condition?  If so, you may need to include the optional resistor RD shown in the "OSC0 Crystal Circuit Schematic" found in the datasheet. See the Pxtal parameter in the "OSC0 Crystal Circuit Characteristics" table in the datasheet. 

    Regards,
    Paul

  • 0 in reply to peaves
    TI__Expert 3385 points

    Hi Paul, 

    1. Shouldn't you take into consideration the shunt capacitance of the crystal, the capacitance between the crystal pads, etc... to calculate the values of C151 and C152?  

    My customer has estimated a Cshunt capacitance of approximately 8.5pF, such that C1 and C2 should be around 19pF. Is this correct?

     

           2. To answer your question - I submitted a post : https://e2e.ti.com/support/processors-group/processors---internal/f/processors---internal-forum/1493194/am4378-formula-power-dissipation which is related to how the DL is calculated 

    Regards

    Geoffrey 

  • 0 in reply to Geoffrey Ficara
    TI__Guru 63125 points

    The formula for calculating crystal load capacitance provided in the AM437x datasheet doesn't properly represent all of the contributions to crystal load. It also should not include the shunt capacitance of the crystal since this capacitance has already been accounted for by the crystal manufacture. I suggest using the formula provided in the AM64x datasheet. Please refer to the "Load Capacitance" section of the AM64x datasheet.

    I also suggest confirming the combined shunt capacitance of the crystal circuit including PCB and package contributions is less than the 5pF maximum defined in the AM437x datasheet. The "Shunt Capacitance" section of the AM64x datasheet describes how to determine the shunt capacitance applied to the oscillator. I'm concerned with your comment above that says the customer estimated a shunt capacitance of about 8.5pF. If so, they are violating the 5pF maximum. Excess shunt capacitance effectively decreases the oscillator gain, which can cause oscillator startup issues. Note: The max shunt capacitance value found in most crystal datasheets represent the worst-case shunt capacitance of the entire family of crystals represented by the datasheet. Your customer may need to request a part number specific crystal datasheet to determine the max shunt capacitance of their specific crystal.

    Based on my experience, they are going to need to select a lower load capacitance crystal to be compliant to the maximum C1 and C2 capacitance parameter value of 24pF. Note: The C1 capacitance parameter limits should be applied to the combination of capacitor C1 and the XTALIN signal capacitance relative to ground, and the C2 capacitance parameter limits should be applied to the combination of capacitor C2 and the XTALOUT signal capacitance relative to ground.

    Regards,
    Paul

  • 0 in reply to peaves
    TI__Expert 3385 points

    Hi Paul, 

    Thank you for the feedback.

    Do you recommend the crystal used on the EVMs (SCM18B-24.000MHz TR) ? Its load capacitance starts from 7 pF 

    Regards

    Geoffrey 

  • 0 in reply to Geoffrey Ficara
    TI__Guru 63125 points

    No. The "18" in this crystal part number indicates this crystal was cut to be resonate with a 18pF load which creates the same problem as the ECS crystal. You need to select a crystal that is cut to be resonate with an 8pF - 10pF load.

    Regards,
    Paul

  • 0 in reply to peaves
    TI__Expert 3385 points

    Hi Paul, 

    Which crystal can I recommend to my customer? 

    Regards

    Geoffrey 

  • 0 in reply to Geoffrey Ficara
    TI__Guru 63125 points

    TI should never make component recommendations for a customer's system. The customer's system may have requirements beyond those defined for the AM2437x device that need to be considered, and TI may not be aware of them. For example, their system may require a reference clock that is more accurate than the AM437x limits of +/-50PPM. The customer's system designer should be selecting components that meet all of their system requirements.

    I have no problem answering questions about AM2437x requirements, but do not feel comfortable making firm component recommendations. However, we could potentially provide a suggestion to be considered.

    The 24 MHz crystal they select should have a ESR less than 48 ohms.

    The max shunt capacitance allowed by the oscillator is 5.0 pF, so the combined PCB mutual capacitance between the XTALIN and XTALOUT signals and the crystal shunt capacitance needs to be less than 5.0 pF.

    The crystal selected should be cut for resonance when the load is in the range of 8 pF to 10 pF, such that the combined capacitance relative to ground of each XTALIN and XTALOUT signal and the respective C1 and C2 capacitance remains within the range of 12.0 pF to 24.0 pF while the series combined capacitance from each creates the appropriate load for the crystal. For example: Let's assume they select a crystal that requires 8pF of loading. Each of the XTALIN and XTALOUT signals would need to have 16 pF of capacitance relative to ground. The 16 pF of capacitance required on each signal would be well within the range of 12 pF to 24 pF limit defined in the datasheet. The C1 and C2 capacitor values will need to be 13 pF if their PCB introduces 3 pF on each of the XTALIN and XTALOUT signals.

    They also need to make sure the crystal they select has a worst-case frequency error of less than +/-50 PPM when combing the frequency error introduced by the initial frequency accuracy of the crystal, the frequency variation due to temperature variations, and frequency variation due to crystal aging.

    We recently selected ABM11W-25.0000MHZ-8-D1X-T3 for one of our EVMs. They may want to consider a 24MHz version with similar characteristics to this crystal.

    Regards,
    Paul