• State Resolved
  • Locked Locked
  • Replies 17 replies
  • Answers 5 answers
  • Subscribers 21 subscribers
  • Views 989 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.

Original question:

LMX2571: About in-band spurious by TCXO freqency

LMX2571: In band spurious 100KHz and 150K off

Mochi
Genius 4165 points
Part Number: LMX2571
Other Parts Discussed in Thread: LMX2582,

Hello,414.25MHz_RX_Local_150kHzoff_spurious_20211119.zip

Our customer is verifying performance of LMX2571EVM for 470MHz TX/RX.

Their finding is 100KHz and 150KHz off spurious issue.

There is combination condition, TCXO 16.8MHz with PLL output 414.25MHz is one of significant tough condition. We have observed 100kHz -39dBc and 150kHz -53dBc.

In the case of 100KHz spurious, if use setting of Multiplier is x1 and Post-R is 1/, then spurious level is improved to below noise floor.

About 150KHz spurious, it is large level as -53dBc when Fractional setting is 1st order. But change it to 4th order, spurious is disappeared as attached file.

 

What is root cause of phenomena? How can we optimize registers setting?

 

Regards,

Mochizuki

  • 0
    TI__Guru** 107060 points

    Hi Mochizuki-san,

    According to your configuration, the spurs frequency should be multiples of 900kHz. I think the 100kHz and 150kHz (multiples of 50kHz) spurs appear only when DEN = 16800 and 3rd order modulator. If you change the MULT=1 and Post-R=1, the spurs will appear at multiples of 150kHz, that is why the 100kHz spur disappear. If 4th order modulator is used, then the spur will appear at multiples of 75kHz. 

    We can calculate the spur frequency, details can be found in LMX2582 datasheet section 8.1.1.

  • 0 in reply to Noel Fung
    Genius 4165 points

    Hello Noel,

    Thank you very much for your support.

    We have read LMX27582 datasheet and related document SNAA289 Using a Programmable Input Multiplier to Minimize Integer Boundary Spurs.

    As quick re-solution, should our customer push "Spur-b-Gone" button in TICSPro to get optimized PLL parameter setting?

     

    Regards,

    Mochizuki

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochizuki-san,

    I just tried, spur-b-gone suggests MULT=4 and Post-R=3 Slight smile

    Can you check the spurs performance if DEN = 16777215?

  • 0 in reply to Noel Fung
    Genius 4165 points

    Hello Noel,

    Thank you for confirming the result of “Spur-b-gone” output in TICSpro during your holiday week.

    As we have sent captured FFT in previous zip file which setting is actually MULT=4, Post-R=3 and DEN=16777215 with 4th order, it is same as“Spur-b-gone” suggestion. Resulted in only 150KHz spurious is exist -74dB. And 100KHz spurious is disappeared by 4th order modulator setting. Is it enough answer for your request?

    Here are remaining questions from us.

    1. As you commented “According to your configuration, the spurs frequency should be multiples of 900kHz.” However our FFT result is not include 900KHz spurs, there is only 150KHz spurs existed. What is reason of this difference?
    2. “If 4th order modulator is used, then the spur will appear at multiples of 75kHz” We did not see 75KHz spurs at 4th order modulator in our FFT capture, do you see reason why?
    3. “If you change the MULT=1 and Post-R=1, the spurs will appear at multiples of 150kHz” Can you recommend to use MULT=1 and Post-R=1 with 3rd order modulator configuration, even it is not follow“Spur-b-gone” suggestion? Because this setting gave us better balance of spurs.

    Regards,

    Mochizuki

     

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochizuki-san,

    let's do some tests to verify:

    Output frequency = 414.25 MHz
    Config MULT Post-R DEN Order Expect spurs frequency Actual spur frequency
    1 4 3 16777215 3rd order Multiples of 900kHz ?    e.g 900kHz, 1.8MHz
    1 4 3 16777215 4th order Multiples of 900kHz ?
    1 4 3 16800 3rd order Multiples of 50kHz ?
    1 4 3 16800 4th order Multiples of 25kHz ?
    2 1 1 16777215 3rd order Multiples of 900kHz ?
    2 1 1 16777215 4th order Multiples of 900kHz ?
    2 1 1 16800 3rd order Multiples of 150kHz ?
    2 1 1 16800 4th order Multiples of 75kHz ?

    Attached are the .tcs files of Config1 and Config2, please ask the customer to change DEN and Order and fill in the last column.

    2571Config1.tcs2571Config2.tcs

  • 0 in reply to Noel Fung
    Genius 4165 points

    Hi Noel,

     

    Thank you for your proposal.

    Our customer will run those trials on Monday. We hope to see good alignment between EVM and Simulation.

     

    Regards,

    Mochizuki

  • 0 in reply to Mochi
    Genius 4165 points

    Hello Noel,

     

    Attached is measurement result.

    Spurious frequency itself is not changed in all of configurations but peak level is 10dB difference.

     

    Regards,

    Mochizuki

    414R25MHz spuri.xlsx

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochi-san,

    Looks like the 150kHz spurs is created from the input path. 

    Could you try the following configuration? This will make the fpd same as Config 2 but with the MULT and Post-R active. If the result is same as Config 2, most likely the 150kHz spurs is generated from the Post-R divider.

    Output frequency = 414.25 MHz
    Config MULT Post-R DEN Order
    1 4 4 16777215 3rd order
  • 0 in reply to Noel Fung
    Genius 4165 points

    Hello Noel,

    Thank you for the suggestion, we will tryout this configuration.

    This time we are using SE output TCXO and LMX2751 setting is differential input mode with pseudo differential connection.

    So, do you feel GND noise between EVM and TCXO module would be another suspected root cause of 150KHz spurious?

     

    Regards,

    Mochizuki

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochizuki-san,

    Probably. Spurs are very sensitive to the OSCin signal, you may get different spurs performance when the TCXO is on-board or from an external module. 

    You can set OSCin as SE buffer, the min. input oscin signal can be 0.5V so there is no issue working with a TCXO.

  • 0 in reply to Noel Fung
    Genius 4165 points

    414R25MHz spuri2.pdfHi Noel,

    Please look into attached data which is your proposed configuration.

    150KHz spuri level is better than previous one.

    Also our customer tried SE mode setting, it gave us a few dB lower spuri. However if enable 50ohms input load, spuri is increase +6dB.

    Is there any workaround to reduce spuri at post R divider output.

     

    Regards,

    Mochizuki

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochizuki-san,

    I looked through the register maps again, there is no post R-divider related configuration bits, not even with hidden registers. I am afraid there is no workaround for this.

    TCXO cannot drive a 50Ω load, don't enable the internal termination with TCXO.

  • 0 in reply to Noel Fung
    Genius 4165 points

    Hi Noel,

    We appreciate very much for your support so far, our customer has proceeded spurious measurement at all of LO frequency channels.

    Please find attached data.

    Target level is 76dBc, when look at overall performance "1/1/1 and 4th order" setting show us better result than "1/4/3 and 3rd order", however we still could not find suitable configuration at 403.21250/403.22500/403.23750,,,MHz even we applied "Spur-b-gone".

    Would you give us suggestion of better configuration at those frequencies ? Especially nearby carrier spurious is tough one.

     

    Regards,

    Mochizuki

    20211208.pdf

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochizuki-san,

    Spur-b-gone is used to suggest a different fpd in order to avoid IBS. However, IBS is a kind of crosstalk spurs, we cannot eliminate it completely. Worst of all, with channel spacing of 6.25kHz, IBS spurs frequency could be at 6.25kHz offset. When using LMX2571 in synthesizer mode, loop bandwidth is usually wide. As such, most of the IBS spurs will be in-band so there is no attenuation from the loop filter. We have run some spurs tests before, with a 19.2MHz reference clock and 12.5kHz channel spacing, output frequency between 400MHz and 520MHz, spur failure rate (test limit is -75dBc) is around 20%. That means, out of 9600 frequency channels, 20% of them will fail with spurs higher than -75dBc. There is no way to resolve this problem with registers or configuration. The only effective way to reduce failure rate is use a different reference clock at these spurs channels. For example, at 420.00625MHz, with fosc = fpd = 16.8MHz, you have a bunch of low frequency spurs. If you change the reference clock to 27MHz, then all the spurs will be gone away. To implement this solution in the system, use two reference clocks and switch to use one of them at a time. If this approach is not acceptable, then we will have to use external VCO with small loop bandwidth so the the spurs could be attenuated by the loop filter.

  • 0 in reply to Noel Fung
    Genius 4165 points

    Hi Noel,

    Our customer had been used discrete VCO and achieved -76dBc IBS in existing products.

    This time they were trying to move VCO integrated PLL device but two reference clock configuration will be last choice . Do you have another recommended solution like DDS?

    Regards,

    Mochizuki

  • 0 in reply to Mochi
    TI__Guru** 107060 points

    Hi Mochi-san,

    Unfortunately all integrated VCO synthesizers, even from our competitors, have similar spurs behavior. To use synthesizer in this application, tradeoffs have to be made. For example, restricted usage of some of the spur channels; scarify phase noise in return of better spurs. Our customers have been doing this. The migration from discrete VCO+PLL to synthesizer solution is not straight forward. I have supported many customers for the migration, I understand how this customer feel about this, the migration isn't easy at all.

    In TI, we don't have DDS devices. 

  • 0 in reply to Noel Fung
    Genius 4165 points

    Hello Noel,

     

    We appreciate very much for your dedicated support.

    Our customer well understood existing PLL device capability, and got useful information of workaround for IBS optimization.

     

    Regards,

    Mochizuki