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: LMX2595

Part Number: LMX2595

I am using LKMX2595 on my custom board to generate 2.8 GHz. For programming, I am using eval board. The loop filter design is as per the eval board. I have attached the Vtune plot, spectrum plot and register settings. Instead of CW , I am getting modulated spectrum, Is some setting missing or any other issue.

Regards

Vikas

  • Vikas,

    So this looks like you see a smearing centered at 2.8 GHz that's about +/- 20 MHz wide.
    On the scope, you see a sinusoidal wave of about 40 mV P-P and 600 kHz

    From the TICSPro picture, the VCO gain is about 85 MHz/V, but it's divided down by a factor of 4, so that means 20 MHz/V on the output.

    So 40 mV * 20 MHz/V = 0.8 MHz, but your smearing is +/- 20 MHz, so maybe your timescale on the oscilloscope is too slow to see what is really going on.

    In any case, it looks like the VCO has calibrated correctly to the correct frequency band of 2.8 GHz, but there is instability. Realize that the VCO has multiplle bands (7 cores with 183 bands). At 11200, maybe this band is +/- 100 MHz wide, so divide by 4, that's about +/- 25 MHz, which is about what your smearing looks like.

    So in other words, if the loop filter were unstable, this would make sense because the loop filter would be disconnected during calibration.

    So I would try these things:

    1. Increase CAL_CLK_DIV to the highest setting of 3. This slows the calibration clock, so if it is calibration related, this should help. With the slow clock, toggle FCAL_EN and see if it fixes the issue.

    2. Reduce the charge pump gain. If it improves the smearing to less wide, then it suggests loop filter stability.

    3. After programming properly to 2800 MHz output, try programming the VCO frequency WITHOUT toggling FCAL_EN. Set way low (say 1000 MHz).
    This should slam the VCO to the lower frequency of the band. Then set the VCO frequency to 20000 MHz, this should slam the VCO to the upper rail. This gives you an idea of how wide the band really is where you are working. Also it ensures that the loop filter is connected to the VCO.

    Regards,
    Dean
  • Hi Dean

    Thanks for the quick response. With the 150 KHz loop filter, the PLL is able to lock.
    Now I am facing two issues.
    1) Phase noise is degraded by more than 10 dB at 1 KHz and
    2) Inductive pull up for output is affecting the output power.

    As per the datasheet, with inductive pull up the impedance mis match occurs. When I try with 50 ohm, as its mentioned in datasheet, the current consumption increased drastically because of which I could not use it.

    Instead of single ended, I am using transformer for output. Any suggestions.

    Regards
    Vikas
  • 2084.High Frequency Open Collector Outputs.pdfVikas,

    1)  If decreasing the loop bandwidth degrades close in phase noise, it implies the VCO noise is cropping in the loop bandwidth, but I would expect that 150 kHz sould filter out VCO noise at 1 kHz.  But as the reduced gain suggests better stability, maybe there is some issue with stability.    If phase noise at 1 kHz is always worse, ensure the refernece is clean and also it has a fast slew rate.

    2)  With inductor, matching is bad, but no DC drop accross it.

    With resistor pull-up, DC drop accross it limits output power.  Also, not in all cases can we assume output is high impedance.

    Ideally, I think it makes more sense to use inductor followed by resistive pad.   Because inductor has higher power, you can afford to sacrifice 3 dB.  Also, it makes it less sensitive to whatever you load it with.

    The attached document goes into great depth on these outputs and output configurations.

  • Hi Dean

                      Thanks a lot for the suggestion.

    1)I have tried with different clock source for PLL input and did the simulation with their phase noise. With 130.44 MHz source,  I could achieve a phase noise of 109 at 1 KHz which is very good.

    2)For the output coupling, I have faced such issues with DDS as well when the output is referred to VCC. In those boards also, I faced problem when I try to use differential output through transformer which is causing more than 7 dB of loss. For the matching issue, I had provided a 3 dB pad after transformer. If I get a problem in final chain, I will skip the transformer to use single ended output.

    Regards

    Vikas

  • Dean

                       Sorry to trouble you again.

    I was going through data sheet but I could find out why the output spectrum differs so much with OUTA_PWR command. For the same PLL, I am generating 2.8 GHz. When I have lower OUTA_PWR, I get harmonics will single tone power less, which is accepted. Now, I try to increase the OUTA_PWR, What I observe is, 1.4 GHz power is more compared to 2.8 GHz.


    I am attaching the plot at 2.8 GHz with OUTA_PWR 31. With setting 32, spectrum is quite clean

  • Hi Vikas,

    FYI, I just tried with an EVM, I did not see any modulation, phase noise is good and there is no sub-harmonics at 1.4GHz.
    I was using your configuration.
  • Noel

            Thanks for the response.

    Modulation issues got solved by simulation of loop filter of 150 KHz.

    Phase noise also achieved. In one of my card, where I am using the single ended output, I was getting such response. One end , I am using for output, other one , I have terminated with 50 ohm. But in other card, I have transformer, where it will give single ended output, I get a good response with OUTA_PWR 31 also. I will check what fault has happened in the first card

  • Vikas,

    In regards to the power, be aware that the power increases from 0-31. But 32 is actually LOWER output power than 31; itgives the same power as 16 and 47 gives the same power as 31. So it sounds like the lower output power level is actually better.

    Regards,
    Dean
  • Dean 

              Thank you so much for the support.

    I have one clarification. I am using reference 100 MHz. So for the PLL, I tried with the PFD of 100 MHz and 200 MHz . What I observe is in band spur is better with 100 MHz PFD . It comes around 68. What could be other solution to improve this apart from changing charge pump.

  • Vikas,

    Fractional spurs are complicated with multiple mechanisms that can cause them.  For spurs, the key is to figure out what is the dominant mechanism that is causing them.  

    If you increase the phase detector frequency without compensating by changing the charge pump current or loop filter, the the loop bandwidth will increase.  But also, increasing phase detector frequency, especially with the doubler can increase the Fvco%Fpd spurs (these are modeled in PLLatinum Sim).  So in other words, there might be some trial and error involved in optimizing spurs to figure out the dominant cause.

    For instance, if you double the phase detector and make the charge pump current half, then the loop bandwidth is the same.  Doing this excercise, if the spurs are the same in this condition, it implies spurs improve with narrower loop bandwidth.  So in this case, either reduce the charge pump current or re-design the loop filter with a narrower loop bandwidth.

    On the other hand, if doubling the phase detector frequency and using the charge pump current still has worse spurs in 200 MHz Fpd mode, it implies the higher phase detector frequency is the issue.  So for this, maybe there's less you can do.  maybe the PFD_DLY_SEL has some impact or the bypassing.   Also, this is decreasing the PLL N divider, which might make fractional spurs worse, so try changing MASH_ORDER or MASH_SEED to see if this spur is related to the fraction.  

    Regards,

    Dean

  • Noel

            Based on my custom board design, what I observed is. This issue is because of using single ended input and terminating the other input when the PCB is designed for transformer coupling. I removed the usage of single ended and moved to the transformer coupling, this sub set frequencies are eliminated.

  • Vikas,

    It looks like your post discusses instability, but this comment is about spurs. But if it is a resolution that's great.

    In general, if there is an issue with the input reference, then this could cause a whole host of issues. You can drive this single-ended, but the unused side needs to be properly terminated with an AC coupled 50 ohm.

    Regards,
    Dean