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.

66AK2G12: DDR3 Clock drifting ?

Part Number: 66AK2G12

Hi,

My customer has a question about DDR3L differential clock. They said that it seemed like the clock drifting in Mhz order. I believe K2G does not have a feature of spreading spectrum on DDR3 clock, but could you please let me know if it is expected or not ? Please note DDR3L itself is working correctly. 

Best Regards,
NK

  • Hi Naoki,

    This is not normal or expected behavior. A drift in MHz order is not acceptable and will cause problems with your device operation.

    This could be caused by a number of things. First elaborate on how you measure this drift? Then start checking:
    1. Review your PCB
    2. Check if the crystal you use provides stable oscillations.

    I am also looping the design team to elaborate.

    Best Regards,
    Yordan
  • Hi Kawa-chan, I agree with Yordan that we need to find out EXACTLY how they're measuring this.  Are they looking at absolute max and min periods over some length of time?

    Since they are using the SYSOSC as a source, there's no external DDR reference clock to look at.  I would make sure that the PLL settings are OK.  It could be that something isn't configured right and they're losing lock.  This seems like the most likely thing.

    Have you/they ever seen anything like this on the K2G EVM?

  • Hey, Bobby-san

    Yes, I got the capture from the customer. What they are saying was actually eye pattern on DDR3 clock. To me, it looked not so worse, but the clock is spreading among +/- 1Mhz in ck2_drift.png. Can you see any problem or concerns on the attached ?

    Best Regards,
    NK 

    ck1_EyePattern.png:

    ck2_drift.png:

  • Hi, at a glance, this doesn't look bad; the eye is pretty well-defined.  How long of a time period were the data collected over?

    Per a colleague: The DDR3L spec by JEDEC does not measure clock jitter this way.  They specify the clock jitter using cycle-to-cycle jitter measurements.  Oscilloscopes with software to return these types of measurements should be readily available.  Can you have them provide measurements consistent with the JEDEC techniques?