Likely the confusion is coming from the fact that the Urukul methods advance the time cursor by the duration of the underlying SPI transfers.

@rjo Why did you do it this way instead of following the "zero delay" methods of the AD9914 driver?

    sb10q
    Thank you for your answer.
    What can I do or what code can I use to avoid this？
    I am using AD9910 now, and if I can avoid this situation by using AD9914？

      Put negative delays or reset the time cursor with now_mu/at_mu.
      Only NIST-internal hardware has AD9914 so far.

      sb10q the alternative leads to more confusion and frustration due to collisions. Iirc there was consensus. Maybe check the old discussions.