• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 2 answers
  • Subscribers 20 subscribers
  • Views 1237 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.

LMX2594: LMX2594 Phase Noise Prediction vs Measurement

Ben Brown91
Prodigy 60 points
Part Number: LMX2594

Hi, I am having some trouble with the phase noise from the LMX2594 synth.

I am using a 10MHz reference source with Fpd set to 10MHz as well. The output frequency is set to 6GHz.

When I look at the PLLatinum sim predicted phase noise performance I get around -90dBc/Hz at 100Hz offset but once this is implemented on my PCB I only achieve around -70dBc/Hz. See plots below (dark blue trace on the NoiseXT measurement data). As you can see from the plots I have tried numerous phase detector and oscillator frequencies as well as different frequency sources but I have not been able to improve on this phase noise.

Can anybody help me work out why there might be a 20dB degradation in phase noise from the prediction to the measurement?

  • 0
    TI__Mastermind 18070 points

    Hi Ben,

    The way that PLLatinum sim works is that it assumes perfect reference source. So if you use a Wenzel oscillator as reference, you can get very similar behavior as predicted in PLLatinum sim. However, if the reference source is not clean, then phase noise can be greatly degraded.

    Please try the reference pro board shipped along with the LMX2594 EVM, or use the on-board 100MHz crystal oscillator. I believe that the phase noise will be greatly improved.

    You can also import the noise profile of your reference source into PLLatinum sim and run the simulation.

    Regards,

    Hao

  • 0 in reply to Hao Z
    Prodigy 60 points
    Thanks for the response, I should have explained this better though.
    I suspected something like this too so I have already entered measurement data for my reference source into the PLLatinum sim (see the lime green trace in PLL sim, original measurement of the source is the red trace on the NoiseXT plot). When I did this I saw a slight degradation in the phase noise (~3 or 4dB) but it is still quite far off the measured performance.
    Are there limitations in the PLLatinum sim when using lower frequency references perhaps? What is your experience with this, is PLL sim normally pretty accurate to real world performance?

    Regards
    Ben
  • 0 in reply to Ben Brown91
    TI__Mastermind 27641 points
    Ben,

    I think that the issue is related to the input reference noise. Although you entered them, they are not working as you suspect. For instance, the in-band input refernce contribution at 10 Hz is -125 + 20*log(12GHz/10MHz) = -63.4 dBc/Hz.

    Notice that the box for the 1/f^3 noise is yellow. This is telling you that when you use this three point model, it is not finding the 1/f^3 normalized noise; in other words, it says you can't use this 3 point model with these offsets. Note that the 1/f^3 line is missing from the PLL noise graph on the lower right. You need to actually import the data. For this 3 point model, there are restrictions on the offsets and the data and when there is an issue it highlights the box.

    Also, I see you have tried different oscillators, but if you look at phase noise from 10 Hz to 100 Hz, you will see that it follows a 20 dB/decade slope, not 10 dB/decade as the PLL would do. This also indicates the input reference. Even a 100 MHz Wenzel oscillator dominates over the PLL 1/f noise at 10 Hz offset.
  • 0 in reply to Dean Banerjee
    Prodigy 60 points
    Thanks for that, I didn't realise there would be this much difference between using the three point model and importing actual test data.
    The trace is still not exactly the same as measured though but it is only out by 2-3 dB which seems reasonable.