• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 4 answers
  • Subscribers 20 subscribers
  • Views 464 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

LMK00308: Output skew between LVDS and LVPECL outputs

Torstein Sunde
Prodigy 40 points
Part Number: LMK00308
Other Parts Discussed in Thread: LMK00304, , LMK00304EVM

The output skew specification t_SK(O) for the LMK00308 and LMK00304 is only valid for the same buffer type, e.g. LVDS to LVDS. Do you have any skew data for LVDS out to LVPECL out on the same device, that's better than one may calculate from the worst-case difference between t_PD_LVDS and t_PD_LVPECL, i.e.  [-340..420] ps, versus as little as 50 ps for t_SK(O)? As long as the temperature is the same (same die) and the load is similar, one might expect somewhat better than this.

Regards,

Torstein Sunde

  • 0
    TI__Expert 8585 points

    Hello Torstein,

    Unfortunately we do not have any skew data for different output formats, but it is safe to assume that the t_SK(O) between an LVDS and LVPECL output will be:

    min = 20 ps

    typ = 40 ps

    max = 60 ps

    These values are found by finding the difference between the LVDS propagation values and LVPECL propagation values (as you stated). We can make this assumption that the output skew will be a difference of these values because the propagation delays for both output formats were found under the same test conditions.

    Regards,

    Kia Rahbar

  • 0 in reply to Kia Rahbar
    Prodigy 40 points

    Thanks for the reply!

    I assume those values come in addition to the +/-50 ps skew between outputs of equal type, as the uncertainty of the latter is higher than the uncertainty  of the [20..60] ps range. But it's still a big improvement compared to the worst-case differences between the propagation delays.

    Regards,

    Torstein

  • 0 in reply to Torstein Sunde
    TI__Expert 8585 points

    Hello Torstein,

    The values I presented do not come in addition to the +/- 50 ps skew between outputs of equal type. Output skew is the propagation delay difference between any two outputs, so a subtraction of the LVPECL propagation delay from the LVDS propagation delay will provide you the min, typ, and max output skews. I have provided the LVDS and LVPECL propagation values below so you can see the values I used to find a min output skew of 20 ps, a typical output skew of 40 ps, and a max output skew of 60 ps.

    Regards,

    Kia Rahbar

  • 0 in reply to Kia Rahbar
    Prodigy 40 points

    Hello Kia,

    I'm sorry, but I have a problem with the reasoning.

    First, with two outputs of the same type, the skew is +/-50 ps, i.e. one output has an uncertainty range of 100 ps when the other is the reference. But an LVDS output is supposed to be 20 to 60 ps delayed compared to an LVPECL output, i.e. an uncertainty range of only 40 ps, significantly less than the former. I find it hard to believe that different types of outputs would give less uncertainty than two equal types.

    Second, if the approach is valid, it should also be applicable to two outputs of the same type, but this yields a skew of zero.

    Regards,

    Torstein

  • 0 in reply to Torstein Sunde
    TI__Expert 8585 points

    Hello Torstein,

    We will be obtaining measurements for the output skew between an LVDS and LVPECL output and we will get back to you once we have the results.

    Regards,

    Kia Rahbar

  • 0 in reply to Torstein Sunde
    TI__Expert 8585 points

    Hello Torstein,

    After performing measurements in our lab, we determined that the typical output skew between an LVDS output and LVPECL output on our LMK00304EVM board is approximately 30 ps (as shown in the image below). The average value we measured is shown to the right of the mu in the delay row (bottom of image) and the standard deviation is shown next to the sigma.

    For the LMK00304EVM board we used in our lab, the output skew between these two formats appeared to vary anywhere from 5 ps to 60 ps. Since this is a small sample size, it is difficult to come to a conclusion on the actual min, typical, and max values, but it is safe to assume that the values I presented in my first post are reasonable output skew values between a LVDS and LVPECL output. From this small sample size it appears the min = 0 ps, the typical = 30 ps, and the max = 60 ps.

    Below is also an image of the typical output skew between two LVPECL output formats. As you can see the typical value is around 20 ps, which aligns with our data sheet values. The average value we measured is shown to the right of the mu in the delay row (bottom of image).

    Regards,

    Kia Rahbar

  • 0 in reply to Kia Rahbar
    Prodigy 40 points

    Hello Kia,

    Thanks for doing the measurements!

    As we are using these values to ensure we meet setup and hold times, it's the minimum and maximum values that are used in the calculations. From your measurements, it appears that the LVPECL to LVDS skew is centered around 30 ps. As the max skew from the datasheet (+/- 50 ps) is valid over the temperature range and isn't limited to just a few samples, it may be reasonable to use that as a base, but with an offset of those 30 ps, i.e. [-20..+80] ps, perhaps with an additional 10 or 20 ps margin just to be safe.

    By the way, a 20 ps delay corresponds to a PCB trace length of approximately 3 mm, whereas the package size of the LMK00304 is 5x5 mm. So one might assume that much of the skew between outputs of the same type can be attributed to different signal path lengths within the device. Cf. the measurement between two LVPECL outputs.

    Regards,

    Torstein

  • 0 in reply to Torstein Sunde
    TI__Expert 8585 points

    Hello Torstein,

    Agreed, please carry forward with a min of -20 ps, a typical of 30 ps, and a max of 80 ps as rough approximations for the output skew between a LVPECL and LVDS output.

    Regards,

    Kia Rahbar