Post Go back to editing

AD9914 Output Distortion When Using 3 GHz Direct REFCLK (PLL Disabled)

Category: Hardware
Product Number: AD9914
Hello,
I am using an AD9914 DDS with a 3 GHz external reference clock applied directly to the REFCLK input. The internal PLL is disabled and all PLL-related pins are left unused.
I am observing an intermittent issue with the output signal:
The desired output is a stepped amplitude profile (looks correct when working properly).
However, sometimes the output appears distorted.
The distortion manifests as ripple and unstable amplitude levels within each step.
The frequency is correct, but the amplitude is not stable.
The noise floor looks normal.
The step shape is correct, but the flat tops are not clean.
Important observation:
If I disable and re-enable the 3 GHz reference clock, the output immediately becomes correct and stable again.
This strongly suggests that the issue is related to the REFCLK domain rather than the DAC output stage or reconstruction filter.
System details:
REFCLK frequency: 3 GHz
Internal PLL: Disabled
REFCLK applied directly to the device
Differential clock input
Proper termination implemented (100 Ω differential)
Power supplies appear stable (no obvious large ripple observed)
Questions:
Has anyone experienced unstable behavior when driving AD9914 directly with a 3 GHz reference clock (PLL disabled)?
Are there known sensitivity issues at the upper REFCLK frequency limit?
Could this be related to input amplitude window or common-mode biasing?
Is there a recommended reset or initialization sequence when operating at 3 GHz REFCLK?
Any insights or similar experiences would be greatly appreciated.
Thank you.
  • Hi  ,

    Thank you for posting on the Direct Digital Synthesis (DDS) community thread. I am now reviewing your post and will respond accordingly. 

    All the best,

    Jules

  • Hi  ,

    To answer the following questions: 

    1. Has anyone experienced unstable behavior when driving AD9914 directly with a 3 GHz reference clock (PLL disabled)? I have not encountered known issues with this specific unstable behavior. 
    2. Are there known sensitivity issues at the upper REFCLK frequency limit? I believe not. The limit of AD9914 is 3.5 GHz and have gone measurement characterization up to the maximum limit and have not experienced unstable conditions. 
    3. Could this be related to input amplitude window or common-mode biasing? Would you be able to provide me your setup? Is this with the evaluation board or you have used a custom one?
    4. Is there a recommended reset or initialization sequence when operating at 3 GHz REFCLK? Not really recommended but a good practice to have Power Supply -> Signal Generator -> Connect to USB Cable PC -> Configure Evaluation Software. 

    All the best,

    Jules

  • Thank you everyone for the suggestions.
    We eventually discovered that the issue was temperature related. When the device temperature exceeded approximately 59 °C, the output signal started to show the distortion seen in the screenshots. After adding a small heatsink to the AD9914, the device temperature stayed lower and the output became completely stable.
    So the problem appears to have been related to thermal conditions rather than the clock configuration. The system is now working correctly.
  • Hi  ,

    That's good to hear. Just let me know if there's anything else I can help you with. 

    All the best,

    Jules