Post Go back to editing

ad9209 Clocking Scheme

Category: Hardware
System shall have 5 DRx modules. Each module shall have its synthesizer (LTC6952 and ADF4368) to generate the JESD204B clocks and sampling clocks and 2 AD9209 chips, accounts to 8 Rx Channels per module. So, all these five DRx modules (40 Rx channels) have to be phase aligned.
  1. Clock- Distribution Module shall have on-Board OCXO to generate 120MHz which is then passed via power Divider(1:8) to provide 120 MHz clock reference to all the six Modules (5 DRx Modules (10Chips) + One controller Clock Modules) and unused ports are terminated.
  2. controller Clock Modules shall generate EZS-SRQ (SYNC or Sysref) to reset the dividers of LTC6952 and ADF4368 of all five DRx modules
  3. Each of the five DRx modules shall generate clocks required for ADC, JESD204B interface and 120MHz system clock for FPGA
  4. Kindly verify the power level once for a single module.
  5. LTC6952 shall have HD2 harmonic, does it have to be suppressed before feeding to AD9209 clock inputs?
  6. Kindly suggest VCXO part of frequency 600MHz. I am planning to use CVCO55CL-0575-0675. Is it fine?
 
Limitations:
      1. All 120 MHz routed via RF Cable to five modules may have phase mismatch of max 3 degree
      2. All sync signals routed via ethernet cables may have phase mismatch of max 5 degree across the five modules
      3. 120MHz is from Clock-Distribution Module and SYNC is from controller Clock Modules, it may result in fixed phase-offset between SYNC from controller Clock module and 120MHz from Clock -distribution module
      4. All these phase offset has to be compensated at ADF4368 and LTC6952 to meet the timing specifications to do divider reset
      5. As per UG-1578, DC-Coupled sysref operation is supported as per Page. No: 29. In this design, we are planning to use one-shot sysref to reset the dividers. Does the configuration in Page.No:29 is applicable only for two-shot, four-shot sysref pulses or even it applies to single-shot sysref pulses
      6. What is the minimum number of sysref pulses required to reset the dividers of AD9209 including NCO, LTC6952 and ADF4368. (i.e. Pulse width should be four times of Clock input of AD9209)
      7.What is the use of NCO-subordinate sync in Multi-chip synchronization, even though one-shot sync shall be used to reset NCO
      8. How to interface SYNC from LTC6952 to ADF4368 as the common -mode voltage in different (2.3V to 1.3V). As the SYNC signal is a one-shot waveform
      9. Kindly share some application note to delay the clock output with reference to reference-signal phase mismatch. (If I have a 5-degree phase mismatch in reference clock, how much delay should I add in output clock)
      
All EZS-SRQ shall be dc-coupled from controller Clock Module to DRx Modules. Board-1 to Board-5 in clock scheme pdf is DRx Modules.
 
 
PDFs:
clock_scheme With the list of components over the clock path
clock_scheme_divider settings Dividers settings of synthesizer to generate onboard clock.
 


Added PDFs
[edited by: Sagar_BN at 5:43 PM (GMT -5) on 19 Jan 2024]
  • Hi  ,

    Thanks for the detailed explanation and block diagrams. It is a very well-built diagram for clock distribution. 

    I am an application engineer for ADF4368 and LTC6952 products. Therefore, my comments will be limited to the clock distribution scheme. Please see my comments and questions below. 

    • What is the reason for not using other JESD204B/C clock distributor parts such as LTC6953 to distribute both references and sync signals to DRX boards? Buffer or Clock distribution circuits can have additive noise contribution however it can simplify your design. LTC6953 has 11 outputs and can be configured for 5 continuous reference outputs and 5 SYNC outputs.
      If you desire to provide SYNC separately from the 120MHz reference signal, I recommend placing a flip-flop retimer circuit to align SYNC signals to reference signals to avoid setup and hold time violations.
      (Page 10 of the AN-161 App Note)

    • Did you check ADF4377 instead of ADF4368? If fractional PLL/VCO is not needed, ADF4377 or ADF4378 can be a more suitable option for clocking data converters. If ADF4368’s SYNC input only resets the fractional engine to achieve multiple ADF4368 synchronization. If ADF4368 is integer mode, or ADF4377 is used, there is no need to route a SYNC signal. Aligning the reference signals is sufficient to synchronize our new generation integer PLL/VCO product such as ADF4377/ADF4378.
    • RFOUTx outputs of ADF4368 are aligned. You can use each output to provide a clock to individual AD9209 chips. I assume this balun/Coupler/Switch Combination is for debugging options. However, data converters perform better with high slew rate signals and this structure might harm the slew of the ADF4368 output signal.
    • 6 dBm input level is suitable for ADF4368. However, higher slew rate signals have a positive effect on reducing the inband noise of the PLL/VCO product. Please see page 16 of the ADF4368 datasheet for the effect of the slew rate and page 22 of the ADF4368 datasheet for proper reference input network.
    • The application engineer for AD9209 can answer this question.
    • Did you consider the clock distribution part with integrated VCO such as HMC7044? HMC7044 internal VCO can cover your frequency range for SYSREF and management clocks. You can disable the PLL1, and you can use PLL2 and Clock distributor by supplying a 120 MHz reference signal to OSCIN input.
      CVCO55CL-0575-0675 seems a suitable candidate for a 600MHz signal. You can model VCO parameters and design a loop filter using LTC6952Wizard software.

    Limitations:

    • If this phase mismatch is deterministic. You can use ADF4368’s phase delay capability to shift the output of the ADF4368. Also, ADF4377 can delay its outputs.
    • As long as Reference signals are aligned to LTC6952 and Sync signals is inside the setup/hold time. Synchronization is achievable. Please see the first question for SYNC to reference alignment.
    • Please see above.
    • Yes, the phase shifting capability of ADF4368 and the Digital and analog delay capability of LTC6952 can compensate for these differences.
    • The application engineer for AD9209 can answer this question.
    • The application engineer for AD9209 can answer this question.
    • The application engineer for AD9209 can answer this question.
    • LTC6952 outputs can directly drive the ADF4368 SYNC inputs. DC coupling is recommended. I added an example interface.
    • AN-165 and AN-161 can be a good starting point. These app notes are focused on older products but they can give you a general idea and still applicable to LTC695

    Thanks,

    Emrecan

  • Hi, Emrecan,

    1. Yes, Using LTC6953 will ease the job. Since the Controller card LTC6952 has a requirement of better phase noise, I have to add one more distributer IC for this case and the module is size-constraint. 120MHz and SYSREF shall have the fixed offset from power on to power on. I have to delay the SYSREF in LTC6952 to meet the setup and hold time at Controller module which will remove external retimer circuit

    2.ADF4368 is fixed by the customer.

    3&4. As you said, ADC performance is based on Clock input power, Noise Floor, Phase Noise etc., As per UG-1578, Page.No:16, PCLK input power vs Phase noise is plotted for 1.8GHz output. The phase noise floor degradation is less across the power level. I am planning to have a back-up option of 10db amplifier after coupler. But, I have a doubt since ADF4368 has even harmonics as well. Do, I have to eliminate it before it is fed to AD9209. Board has to be synchronised with 3.6GHz as well, as a trail and the clock is monitored to check the phase noise so add these components is mandatory.

    5. Regarding the SYSREF and NCO-subordinate SYNC, could you please tag an application engineer to this queries

    6. As I am not confident, in designing loop-filters for 3.6GHz External VCO for LTC6952, I am currently configuring the VCO to around 600MHz. In future, if this works, 600MHz VCO shall be replaced by a 3.6GHz VCO and fed to AD9209

    Limitations:

    1. So the Phase adjust is at Output clock(max -255 22 degree phase adjustment), Not at Reference side

    5 to 7. Could you please tag an application engineer to this thread

    8. Common mode voltage is different for ADF4368 and LTC6952. 1.85V and 2.3V

    9. How to dc couple CML output of LTC6952 to LVDS input of FPGA

    Newly added Queries:

    1.  4.3nH is recommended for ADF4368, But in user guide, 43nH is used

    2. Why External VCO of LTC6952 is connected to VCO- of LTC6952? Is that for Stability purpose(180 Phase shift)

  • Hi  

    For AD9209 questions,  is already tagged. He may know who can answer these questions. 

    You can check below thread for CML to LVDS interface. 

    (+) LTC6952, LTC6953: SYSREF schematics when there is a common mode mis-match - Q&A - Clock and Timing - EngineerZone (analog.com)

    For inductor value at the output of the ADF4368, there is typo on datasheet. It should be 34nH. 

    Evaluation board uses VCO- input because of the ease of layout. There is no technical reason for using VCO- or REF-. You can use any of them without problem. 

    1.8V is the common mode voltage when ADF4368 is AC coupled. ADF4368 can accept common mode values between 1.4V to 3.1V as long as there is no absolute maximum rating violation.

    Thanks,

    Emrecan

  • Hello,


    With regard to following questions, the response is posted below each of the quesitonms.

     5. As per UG-1578, DC-Coupled sysref operation is supported as per Page. No: 29. In this design, we are planning to use one-shot sysref to reset the dividers. Does the configuration in Page.No:29 is applicable only for two-shot, four-shot sysref pulses or even it applies to single-shot sysref pulses.

    -The circuit shown in figure 23 is used to drop the common-mode voltage seen at the SYSREF inputs to meet the datasheet Max Limit values and the voltage drop across the 6 kohm resistors is due to the common-mode bias currents (green) being servo-ed such that the common-mode voltage appearing at the internal clock receiver is around 0.65 V.






          6. What is the minimum number of sysref pulses required to reset the dividers of AD9209 including NCO, LTC6952 and ADF4368. (i.e. Pulse width should be four times of Clock input of AD9209).

    -While it should all work on the 1st SYSREF pulse (meeting pulse width specifications), it is quite common for customers to use 2nd or higher SYSREF ref when the host (LTC6952)  provides a SYSREF pulse train during the initial multi-chip synchronization phase.


          7.What is the use of NCO-subordinate sync in Multi-chip synchronization, even though one-shot sync shall be used to reset NCO

    -If the NCO's are only going to be configured once upon initialization and synchronization of the system....than using SYSREF to generate internal one-shot to align NCO phases should be sufficient.  If one was going to load via SPI different NCO settings at some point after initial initialization than one could consider the Master-Slave NCO implementation so that a full-system synchronization can be avoided.



          8. How to interface SYNC from LTC6952 to ADF4368 as the common -mode voltage in different (2.3V to 1.3V). As the SYNC signal is a one-shot.
    -See Figure 23.


  • 5. 

    Does both paragraph applies for AC coupled applications or the top one is for DC-coupled and the second is for AC-Coupled applications and Figure 23 applies for one-shot pulse as well. 

    8.   shared input common mode range of ADF4368 is 1.4V to 3.1V. Could you please share the source for this? I am unable to find this in datasheet

  • Hi ,

    Input buffer structure SYNC and REF inputs are similar for ADF4368/77/78. This common mode range comes from ESD diodes at the input and maximum swing levels. 

     is the primary Apps engineer for ADF4368. He can help with the documentation. 

    Thanks,

    Emrecan

  • Could you recommended d-flip flop for divider reset of ltc6952. Input frequency of the clock is 120MHz since wave and Sysref  is less than 5MHz