• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 1 answer
  • Subscribers 21 subscribers
  • Views 413 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.

LMX2592: PLLatinum Sim Question

Tanveer Raina
Prodigy 50 points
Part Number: LMX2592

Hi, I want an integrated phase noise plot for a range of frequencies. How I can do that?

Thanks

  • 0
    TI__Mastermind 33580 points

    Hi Tanveer,

    Since LMX2592 operates with a single frequency at a time, you would take a separate plot for every frequency in your range. We don't really have any automation support built-in to make taking lots of data easy.

    What are you trying to do? Do you need many plots (10s, 100s, 1000s)? Do you need a specific element from the plot (jitter, spurs, etc)? There may be a way to perform some kind of 3rd party software automation e.g. AutoHotKey to simplify the task for large quantities of data.

    Regards,

    Derek Payne

  • 0
    Prodigy 50 points

    Hi Derek, 

    Thanks for the reply. 

    I am trying to plot phase noise. I am new to synth design and circuit and trying to learn how to get better phase noise performance with LMX2592. 

    For each frequency, there is a separate phase noise plot at offset frequencies (1,10 and 1000KHz)? If yes, then to look at phase noise of at 10 frequencies, I will have 10 plots?

    Thanks

    Tanveer  

  • 0 in reply to Tanveer Raina
    TI__Mastermind 33580 points

    Tanveer,

    Yes, there will be ten separate plots and phase noise values at desired offsets for ten separate frequencies. However, you can make some reasonable predictions about the performance at different frequencies, which hopefully helps cut down the number of plots you need:

    • Correlated noise (reference input, in-band PLL, VCO) will change by a factor of 20*log(Fnew/Fold)
    • Uncorrelated noise (noise floor) will change by a factor of 10*log(Fnew/Fold)
    • As long as the VCO frequencies are close, the gain of the VCO and the loop characteristics will remain very similar. But if the VCO gain changes a lot between frequencies, there could be differences that are not accurately modeled. As a general rule, if the loop gain changes by less than 40% (which appears to be true across the VCO range of the LMX2592), the results will be very similar to theoretical predictions at different frequencies.
    • Engaging the divider or the doubler will increase the noise floor above the value achievable with the VCO, so this would be a step increase in noise floor.

    You can verify this yourself in PLLatinum Sim.

    In PLLatinum Sim, on the phase noise tab for LMX2592, there's a button you can click titled "LMX2592 Phase Noise Tips" which provides some suggestions for how to improve the phase noise.

    Regards,

    Derek Payne

  • 0
    Prodigy 50 points

    Thanks Derek, May I ask you one more question. What is MASH order? Does it need to be of high value? What will happen if I select it to be 1?

    Best Regards,

    Tanveer

  • +1 in reply to Tanveer Raina
    TI__Mastermind 33580 points

    Tanveer,

    Take a look at Chapter 10 for the basics, and Chapter 20/21 for spurious effects, in Dean Banerjee's book (https://www.ti.com/lit/ml/snaa106c/snaa106c.pdf). The full scope of explaining the MASH engine (Multi-stAge SHaping, a very tortured backronym) is beyond an E2E question, but you can think of it as a pseudorandomization technique overlayed on top of a delta-sigma block inside of the N-divider.

    In the simplest case, you could imagine a divide of 20.5 implemented with an N-divider that switches back and forth between divide-by-20 and divide-by-21 at some rate. If the divider value switches periodically, it introduces a spur on the output, which is a function of the period of the switching and the pattern of the switching. The MASH adds some randomization to the switching function's pattern, which on average results in the correct fraction and spreads the divide-switching energy out across multiple higher frequencies in a pseudorandom fashion to reduce the amplitude of the switching spurs. By shaping the spurious energy so that more of the energy is concentrated at higher frequencies, the loop filter can roll off more of the spurious noise and the effect of the fractional spurs can be mitigated to some extent. As the MASH order increases, the length and factors of the pseudorandom feedback function change, resulting in different noise-shaping profiles and different spurious interactions. Generally increasing the MASH order pushes in-band noise lower, but there are rules and caveats: certain fractions will be periodic with some MASH orders and will not be effectively shaped; N-divides below a certain size cannot effectively take advantage of higher MASH orders, because the instantaneous error of a higher order MASH fraction may be too high and introduce unacceptable clock distortion at the feedback port of the phase detector; additional randomization effects such as using a larger (irreducible) denominator value, or dithering the MASH values so that they are only periodic over a much longer duration, could be helpful or harmful depending on where the spur energy is distributed.

    The book I linked includes several tables near the end of Chapter 21 that provide suggestions about when a given MASH order will be helpful or harmful.

    Regards,

    Derek Payne

  • 0 in reply to Derek Payne
    Prodigy 50 points

    Thanks much Derek.

    Have a great day,

    Tanveer