ADF4360-3 leakage into AD8349 inputs

I'm observing LO leakage into the inputs (pins 1 & 2 and 15 & 16) of an AD8349 modulator in a direct conversion transmitter (tuned to 942.8 MHz).  The analog inputs to the modulator originate from an AD9763 DAC with a low-pass anti-aliasing filter with f(3db)= 3.5 MHz.   The LO signal is provided by an  ADF4360-3.  The RF outputs are configured using a broadband match as on p. 22 of the data sheet.   There is a large amount of RF leakage into the 3.3 V DC power rail thorough the 51 ohm resistors.   RF energy is also leaking into the 5V DC power rail that powers both the DAC and the modulator.   The layout guy suspects that the power net attached to pin 7 of the AD8349 may be picking up some noise from adjacent LOIP and LOIN pins 5 and 6. The net result is a large LO feedthrough signal at the output of the modulator that's subsequently amplified.  I replaced the ADF4360-3 RF output broadband match with choke inductors and tombstoned an extra inductor in series with the DC bypass capacitor and it helps a bit.   Does ADI have any other PCB layout or other advice for combating this unfortunate behavior?

  • 0
    •  Analog Employees 
    on Apr 11, 2011 2:27 PM

    I guess we don't have any specific recommendations beyond using appropriate decoupling as close as possible to the power supply pins. If you are driving differentially it may be a good idea to put in some filtering for LO harmonics as these can sometimes degrade the sideband performance.

    I agree that the inductive match will result in less RF contamination of the DC supply rails.

  • 0
    •  Analog Employees 
    on Apr 11, 2011 8:24 PM

    What is the output power level of the ADF4360-9. The AD8349 works well for LO levels down to -10 dBm so if you have more power than this coming out of the ADF4360-9, you can safely reduce the level. 

    You mentioned that the coupling is onto the I and Q baseband pins. If that is indeed the coupling path? Those inputs have limited bandwidth.

    Check that the LO Leakage is not being caused by external dc offsets. If you can turn the DACs off so that the four baseband lines are set to 0.5 V, try doing this. With these lines set to 0.5 V, are you still seeing the LO Leakage. You might also want to put a DVM on the four inputs when they are in this mode. If IP is not equal to IN to within around 1 mV (and if QP is not equal to QN), then you will get LO leakage.

    Is it possible to power the Modulator externally (from a bench lab supply) to verify if that is indeed the coupling path?

  • When the DAC is off (TX_SLEEP on AD9763 is high), there is almost no feedthrough.  However, the analog outputs of the DAC are at 0V when the DAC is asleep-- it looks like the FSADJ1 line is at 0V and only goes to 1.2V when TX_DAC_SLEEP is low.  There are no observable DC offsets between the I and Q analog outputs of the DAC.  Unfortunately, it's not really practical to power the DAC and Modulator from a lab supply.   Probing around, I definitely see RF noise from the ADF4360-3 on the +5V power line supplying both the DAC & and the modulator and it also appears to be leaking through the decoupling capacitors to analog ground.  Decreasing the LO output power decreases the RF leakage into the DC power lines a bit, but the feedthrough level at the output of the modulator does not got down appreciably.

  • 0
    •  Analog Employees 
    on Apr 13, 2011 10:32 PM

    How big the LO feedthrough is it at Baseband inputs and Modulator output?  It should be around -45dBm Carrier feedthrough performance at modulator output.

    It sounded that the problem is high LO feedthrough level at Modulator output and suspect LO leakage from ADF4360-3 into AD8349 through power line.

    RF signal can go around through power line and ground. Usually, it is recommended to use different power supply and ground for analog and digital circuit. And somecase it is also recommended to use different power supply for PLL.  It sounded power supply pin take power directly from power plane.If we assume that the leakage at baseband inputs of AD8349 is from ADF4360-3 through AD8349 chip or through power line and gnd plane, there should be similar amount of leakage whatever DAC is ON and OFF. But you said there is almost no feedthrough at DAC off state. Then I am not sure if we think ADF4360-3 leakage is coupled through DAC power line. So if there is a way to take a power externally or to put something like ferrite bead between power plane and power supply pins of the chip, I suggest to try that.

    And let's think about LO feedthrough at Modulator RFout. In an IQ modulator, there are multiple LO Leakage paths. As shown at the below figure , most familiar mechanism is the product with offset voltages on the I and Q inputs multiplying with the LO in the mixers and cause LO Leakage(1 case at below figure). And some of the LO power that is applied at the LO input will leak across the die or couple between bond wires and reach the output.

    When the LO is applied to the LO input, it goes through a series of gain stages and the peak-to-peak voltage of the LO signal that drives the mixers will be much bigger than the original LO signal. Some of this signal will also leak across the die to the output(case 2 at below figure).

    Finally, we have also observed on some not well-designed PCB which can also provide a leakage from the LO the Output.

    Even if there is LO leakage signal at baseband inputs, I think it will not be shown as LO feedthrough at RFout due to RF frequency rejection at baseband circuit and high isolation between BB and RFout . Instead of that, it will make another issue, DC or even order harmonics of LO.

  • I realized that I never posted about how this issue was resolved.  Thank you Ian, Eamon, and Hunsung for the helpful suggestions.  The board has separate power supplies for analog and digital components (in fact the synthesizers each have dedicated LDO DC supplies), there are ferrite beads pretty much everywhere, and separate analog and digital ground planes that are connected right under the main power supply for the board.  But unfortunately, the PCB trace that is feeding the LO input to the AD8349 was routed way too close to the modulator output, so the leakage was in fact not into the inputs from the DAC, rather directly from the LO into the output.   Unfortunately, the only way to resolve this problem once and for all was to respin the PCB, which is in progress now (although we're now using an ADL4350 and an ADL5375).  The PCB designer referred to the ADL5375 eval board layout as an example this time. Live and learn...