Post Go back to editing

Occasional unlock of ADF4106 in an FSK modulation circuit

Thread Summary

The user is experiencing PLL lock loss in ADF4106-based FSK modulation circuits, particularly with PN 15 and PN 23 modulation data. Increasing the Anti Backlash Pulse Width (ABPW) to 6.0ns prevents the issue, but the user seeks to understand the underlying cause and potential side effects. The final answer requests an update on the problem's status and offers to provide further assistance if needed.
AI Generated Content
Category: Hardware
Product Number: ADF4106

Hi, 

I have a few questions about PLL ADF4106 used in an FSK modulation circuit.

Fig.1 Block Diagram of FSK modulation circuit

- Used in our FSK transmitter products manufactured for over 10 years.
- A few defective products were found of which transmission modulation spectrum was abnormal.
- Solved by replacing the PLL ICs.
- Spectrum abnormality does not occur when the modulation data is PN 9, but occurs when the data is PN 15 or PN 23, which has a series of 0s and 1s.
- In the most recent lot, approximately 40% of the products were defective with the same symptoms.

Please see Fig.2, 2A, 2B, and 2C.

Fig. 2 VTune & LD under the modulation signal of 0.5Hz square wave

Fig. 2A   Magnified part A in Fig. 2

Fig. 2B  Magnified part B in Fig. 2

Fig. 2C  Magnified part C in Fig. 2


- Occasionally the lock comes loose. Loss of lock occurs after a series of "H" or "L" in the modulation signal.
- Under the modulation signal of a 0.5Hz square wave, VTUNE may drop to 0V 3.5ms after the falling edge of the modulation signal and become unlocked. (Fig.2B)
- However, the drop doesn't happen every time after the falling edge. (Fig.2A)
- VTUNE changes immediately after the rise and fall of the modulation signal, and the pull-in ends in 1ms to become a steady state.
- If Anti Backlash Pulse Width (ABPW) of 2.9ns or 6.0ns is set for ADF4106, the unlock phenomenon do not occur.
- When CW is generated, unlock does not occur even if ABPW is 1.3ns.

My Surmise
- When the PLL locks and continues to stay in the PFD dead zone, a reset will occur.
- The reset takes only once after entering the dead zone from outside the dead zone.

Questions
- Is the above reasoning correct?
- What is the purpose of resetting (or kickoff)? Unlike PD, PFD does not cause false lock in which the phase differs by pi from the correct phase (Fig.3, and Fig.4). So a reset doesn't seem necessary.

Fig. 3  Phase Detector (PD) chararacteristics

Fig. 4  Phase Frequency Detector (PFD) chararacteristics


- What are the conditions for applying a reset?
- How does the PLL distinguish between false lock and correct lock?
- Why doesn't the reset (kickoff) work in CW?
- What are the side effects of increasing ABPW? What kind of performance will be degraded?
- Are there other solutions besides increase of ABPW?
- Some devices occur a reset and some do not. Why does it vary?

Edit Notes

The figure has been enlarged for clarity.
[編集者:Rokko、編集時刻: 21 Feb 2025 日 2:26 AM 時 (GMT -5)]
  • Hi Rokko,

    Apologies for the delayed reply. We have missed this. Is this problem ongoing? Can you update the status about how can ı help you?

    Regards,
    Burhan

  • Hi Burhan,

    Thank you for your reply. The situation has not changed.

    As mentioned in the previous post, setting the Anti Backlash Pulse Width (ABPW) to 6.0ns for ADF4106 turned out to prevent the PLL from losing lock. Therefore, the ABPW, which was originally 1.3ns, changed to 6.0ns during manufacturing.


    We have received inquiries from our quality control department regarding whether this change of ABPW is an effective measure and whether it has any side effects. However, as stated in the post, we do not know the true cause of this issue, so as the designer, I am unable to explain the effectiveness of the measure or any side effects.
    If you are able to answer the questions stated in the post, we would appreciate it.

    Ryoji