Post Go back to editing

ADRF6720-27 DC bias does not lead to datasheet results


I'm working with the ADRF6720-27 eval board. In the datasheet a DC bias of 2.68 Vdc for the baseband inputs (I+, I-, Q+, Q-) with 1 Vpp is recommended. At the moment, I'm only using both the I inputs. The signal source is a function generator where I apply the settings of 1 Vpp and 2.68 Vdc for a 1 MHz sine wave. The signals were checked with an oscilloscope and look fine (min ~2.18 V, max ~3.18 V). When applying these signals to the eval board I get the results below. The current consumption of the board is at 385 mA. I'm using an external reference at 153.6 MHz (DIV4 for PFD). My LO frequency is 2300 MHz.

However, when dividing the offset in half to 1.34 V, the RF output spectrum looks good and the output power peak Pout is more or less at 3.5 dBm. The values for second and third harmonic are not matching at all. Current is now at 350 mA. The new spectrum is shown in the next figure.

Is there something I misunderstood? I'd really like to get the same results as shown in the datasheet on p. 4 for 2300 MHz.

Thank you in advance for your help! :)

  • Hello,

    ADRF6720-27 suggests definitely around 2.68V common-mode voltage at all baseband inputs and all performance data in the datasheet is at 2.68V. Please find the performance data vs. common-mode voltage at Figure 55 in the datasheet.  Not sure how it works at 1.34V Common-mode voltage as expected when it runs at 2.68V. 

    Few things to check:

    • baseband Input : you said you use both I+ / I- , but not use Q+/Q-. And  confirmed 1Vp_p biased 2.68V with the oscilloscope

             : Please connect the function generator to the input connectors on the evaluation board. And double-check DC bias and signal swing level with the oscilloscope.

             : there would be 6dB lower output power by driving I only, not both I and Q. So there might be around -2.5dBm Pout @ 1Vp_p diff at your condition. and there should be a double-sideband signal as you saw here since one mixer core in a modulator works.

              1. DC Bias at 2.68V : about -23dBm LO Feedthrough @ 2300MHz says that there is dc-offset between I + and I- ( or between Q+ and Q-). So make sure the dc bias level should be same between I+(Q+) and I-(Q-). Difference between makes a high LO feedthrough. it should be around -40dBm.

              2. swing level on each I+ , I- : should swing in 2.43V~ 2.93V (0.5Vpp on each I + and I -,  1Vpp differential between I+ and I-)

              3. phase relationship between I + and I- : make sure it is at 180 degree phase difference.

              4. unused baseband inputs : need to dc-bias the unused baseband inputs, Q+, Q- to 2.68V.

    If you want to get the same results as shown in the datasheet, you need to drive both I and Q baseband inputs at the specified condition. 



  • Hello Tony,

    Thank you very much for your answer, I'm sorry for my late response, but here we go. I'm using a function generator with an inverse coupled output on both channels (as inputs for I+ and I-). I checked the signal conditions once again on the oscilloscope, see the figure below. As you said, I'm now applying a 0.5 Vpp signal between ~2.43 V and ~2.93 V on each pin I+ and I-. The offset between I+ and I- is very close to 0 as it should be. The phase is 180°. 

    After applying this signal to the ADRF6720-27 and adding a DC bias of 2.68 Vdc for each Q+ and Q-, the following spectrum was measured:

    I understand the 6 dB losses due to only driving one input pair and not all four inputs. Still, the result above is not as it should be, right? Do you or someone else has any more ideas? Thank you once again!

  • Marked as answered from lack of activity