• State TI Thinks Resolved
  • Locked Locked
  • Replies 2 replies
  • Answers 2 answers
  • Subscribers 21 subscribers
  • Views 89 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

LMK04832: Minimising Phase Difference

Athuljith R
Intellectual 601 points
Part Number: LMK04832
Other Parts Discussed in Thread: LMX1204, LMX1214, LMX1205

Tool/software:

Hi, I am working on a new project were minimising phase difference is very critical. On my board we are having 8 Analog Channels ie, 2 numbers of 4 channel ADCs (14 bit,125MSPS) and 4 numbers of 2 Channel DACs (16 bit 800 MSPS), in between FPGA for the processing. The constrain we have here is the phase difference between Each channel should be a maximum of 3ps. So I was planning to use LMK04832, for clocking all the ADCs and DAC and keeping them in synch, But the skew of LMK04832 is 60ps. Is there any way I can achieve my phase difference requirement. Or any other device i can use to achieve this. The inputs of ADC and Outputs of DAC will be coming from an RF card, which will up/down convert the signal respect to the direction.

  • 0
    TI__Mastermind 28912 points

    Hello Athuljith, 
    Yes even after sync, LMK04832 output-to-output skew can be around 60ps but the prop delays will definitely have an effect over temperature. 

    You can use the delays (digital delays as small as half the VCO period + analog delays) to try and minimize this. 

    Another option I recommend a bit more is using a device such as LMX1204 and cascading them as the tempco is lower. 

    Best regards, 

    Vicente 

  • 0
    TI__Mastermind 33430 points

    Adding a bit more depth:

    LMK04832 lists 60ps typical skew between channels, but the majority of that skew is constant due to on-die layout of the channels, or due to supply voltage variation affecting LVPECL or LCPECL (where the slew rate is directly dependent on the supply voltage). LVDS and HSDS are relatively close across PVT, within about 10ps. I'd also strongly advise you to avoid CMOS, as CMOS varies substantially across temperature.

    The graph below shows 250MHz skew between different output formats in different channels across voltage and temperature. The absolute skew for non-PECL ouputs changes by a few picoseconds across each condition, and appears to be quite predictable at fixed temperature and voltage.

    Still, 3ps max skew is pushing the limits of feasibility just due to process variation. If you need a single system, it is plausible to fix the supply voltage and the temperature, calibrate out the process delay for one device at the given temperature and voltage, and design a PCB to offset the channel-to-channel skew. If you need multiple systems, or if you expect to use multiple clocking devices, the process variation for individual devices is about the same order of magnitude as your entire skew budget.

    LMX1204 suggested by Vicente is a better choice, though the output-to-output skew can be as high as 5ps. I'll suggest two other devices:

    • LMX1214 is a cost-reduced version of LMX1204 without the SYSREF features or the multiplier. It offers similar skew performance. It might be a good choice for fanout, though there is an unfortunate flipping of the outputs that complicates the routing.
    • LMX1205 is layout-compatible with LMX1204 (some output polarities are flipped) and provides nearly the same features, along with both 60ps of adjustable delay at the input and 50ps of independently adjustable delay at each output (in about 1ps steps). Provided you have a way to calibrate the delays, the LMX1205 could potentially provide a knob to turn to match the output-to-output and device-to-device skew to within 1ps.

    All that said, the minimum input frequency to the LMX devices above is 300MHz, and although I don't see a frequency or frequency range mentioned, 125MSPS ADCs suggests this might pose a problem. We don't have a comparable device at lower frequency with the kind of skew requested, so for <300MHz needs, LMK04832 may be the best bet.