• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 3 answers
  • Subscribers 20 subscribers
  • Views 1739 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.

LMX2595: Phase Noise Much Worse than Expected on EVM Board

Kyle Weidmann
Prodigy 130 points
Part Number: LMX2595
Other Parts Discussed in Thread: LMX2594

We recently purchased the LMX2595EVM board and when doing a phase noise analysis we received results that were far worse than expected. The phase noise chart generated by our ESA was about 30dB worse across the nearly the entire range. Is there anything that we are doing wrong? I've attached the PLL settings we used from TICS Pro along with the chart generated. We were testing at 15GHz and using the default settings, which I believed the EVM's loop filter was optimized for. Thanks

  • 0
    TI__Mastermind 28011 points

    Kyle,

    Looking at your plot, it seems that for offsets of 20 kHz and higher, you results actually do match the EVM instructions.  However, for offsets below 20 kHz, something is grossly off.  I see three possibilities:

    1.Measurement issue with the PSA spectrum analyzer

    I see that signal tracking is on, but I'm not so convinced that it is tracking very well.  The steep drop off in noise, looks like something that measurement equipment would do.  The PLL is in integer mode and one would not expect fractional spurs at this lower frequency.  I see nothing that would cause the PLL to do this and I have seen strange artifacts of measurement equipment, especially when I see abrupt changes when it switches to a new decade;  I suspect this is the issue.

    2.  Issue with the Input Reference

    The input reference is pretty good, but if there is poor soldering or connection or it wanders around too much, it could cause issues.  If you are inputting a signal generator as your input reference, then this noise is very likely to be the signal generator noise as I have seen strange looking noise profiles from signal generators.  If using a signal generator and that's all thats available, you can try this trick:  Increase the frequency and divide down.  For instance, try 1000 MHz and divide down by 5 to get the same 200 MHz Fpd.  If phase noise improves, this is a telltale sign of signal generator noise.

    3.  Noisy power supply

    If you have a very noisy power supply, then maybe this could cause some issue if you are using the on board XO on the board.  I suspect that the issue is likely something else though.


    Regards,
    Dean

  • 0 in reply to Dean Banerjee
    Prodigy 130 points

    Dean,

    We've tested with different power supplies and are using the built in crystal oscillator on the board. The EVM has not been changed at all from its default configuration. Our PSA's tracking should be good enough for this application as we've tested other frequency synthesizers and they have matched their specified performance. Thanks,

    - Kyle 

  • 0 in reply to Kyle Weidmann
    TI__Mastermind 28011 points
    Kyle,

    OK, its probably not the power supply, but I'm not convinced its not the PSA as I have seen it do strange things like this before.

    Here's some steps to shed some light on this.


    1. Suppose PSA is the cause:

    a. Try engaging the output divider. Try 7.5 GHz = 15 GHz/2 or 3.75 GHz = 15 GHz/4. In this case, the close in phase noise should scale down. However, sometimes these noise profiles are carrier specific. If the PSA is the issue, the noise profile could be dependent on the carrier frequency. For instance, the E5052 typically doesn't show these, but if I use the E5053 at higher frequencies, I can get these big decade jumps in noise. I verified with the FSWP noise analyzer that these are not real. I have seen these decade jumps with the PSA.

    b. Change the lower offset to 1 kHz (instead of 100 Hz).
    Maybe if the instrument sweeps faster, this will go away.


    2. Suppose XO is the cause:
    Realize we do not test the XO in production, nor do we burn them in. If there is any way to bypass it, that would rule this out. If you do bypass it, I would remove the resistor that powers it and just drive the OSCinP with the input signal.

    3. Suppose the LMX2594 IC is the cause
    Really doubt it, but maybe if configured in some way. I have worked with this device since definition and never seen the device do this, so I highly doubt it is a defective LMX2594 IC. I would go to default mode in TICSPro. If you have the programmable input multipler MULT set to anything greater than 1, then this is a potential uncertaintly. Default mode uses the OSC_2X multiplier. This should be fine, but it's easy enough to disable this. Other than these experiments with the input path, there's not too many other things do do with LMX2594 settings.


    Regards,
    Dean
  • 0 in reply to Dean Banerjee
    Prodigy 130 points

    Dean,

    I just finished following your part 1 suggestions . The phase noise did not improve with an output divide of 2, and slightly improved with an output divide of 4. Reducing the upper offset down to 100kHz also did not improve anything. I've attached the two divided phase noises and the reduced output for comparison. Is there any way to have a phone conversation of other means of direct communication to resolve this issue. We would like to use this chip but as of now the phase noise we are seeing is not appropriate for our application. Thanks,

    - Kyle Weidmann

    Phase Noise /2

    Phase Noise /4

    Phase Noise with Reduced Max

  • 0 in reply to Kyle Weidmann
    TI__Mastermind 28011 points
    Kyle,

    I'm sorry that this issue is being so problematic. Send an email to clock_support@list.ti.com and lets try to set something up. I am on Pacific Time in the US.

    I have an E4440A spectrum analyzer and I played it with a little. I noted that if I go to phase noise mode and push "Signal Track" to On, then the equipment makes a clicking sound at every decade and 5x Value. In other words, it did it at 500Hz, 1 kHz, 5 kHz, 10 kHz, 50 kHz, 100 kHz, and so on. I couldn't reproduce your phenomenon, but I did get it to create a small false "cliff" at 500 Hz for our board with an XO on it. When I used the signal generator, there was no cliff.

    Regards,
    Dean