ADRV9009 RX SFDR performance at low frequencies

Hello

I am using a custom board with three ADRV9009 devices. I have started doing performance evaluation of the RX inputs of the ADRV9009 devices. Testing includes Single Tone SFDR testing. Basically a signal generator is used (followed by an appropriate low pass / bandpass filter) to inject a signal into one of the RX inputs of the ADRV9009. The sample data is captured and the FFT is plotted. The SFDR is measured in dBc as the signal level of the input signal minus the signal level of the highest spurious component.

I am getting pretty good results at 942.5MHz and 1842MHz, in the region of 80dBc. However, there is a noticeable reduction of performance as my input frequency goes DOWN. At 79.55MHz input signal (150MHz LO frequency), I am only seeing around 60dBc.

Below is the FFT plot of such a test. The input signal is at 79.55MHz and the LO frequency is 150MHz. The level of the input signal is roughly -1dBFS. The sample rate is 208MSPS and the bandwidth is approximately 169MHz.

 

The numbers above the spurs indicate how they move with frequency relative to the input signal. So for example, a 1:1 spur moves by 1MHz for each 1MHz that the input signal moves, a 2:1 spur moves by 2MHz for each 1MHz that the input signal moves and so on.

I thought initially that it may be related to my custom hardware so I repeated the tests on the ADRV9009 evaluation board. However, I am seeing similar performance there.

I am aware of the limitations mentioned elsewhere about tuning the LO too low. That is why I chose an LO frequency of 150MHz.

I am performing initial calibrations (with the input terminated in 50ohm terminator) and I have enabled continuous calibration correction.

1) I know this question has been asked elsewhere but can you please give an indication of the expected ADRV9009 performance at frequencies close to 75MHz. According to RFSim, the performance of the ADRV9009 should be pretty consistent across input frequency but that does not happen with real hardware (both my custom hardware and the ADRV9009 evaluation board).

2) Can you please provide guidelines on how I can go about improving the low frequency performance of the ADRV9009. Is there any optimisation that I can do to get better results at lower frequencies?

Thank you.

  • I have made some progress on the performance at lower frequencies by changing the input AC coupling capacitors for 18pF to 100nF. This has resulted in a 10dB improvement in SFDR performance.

    I am still only getting roughly 68dBc performance at 79.55MHz. Now the limiting factor is one particular spur that has a ratio of 1:1 (it moves by 1MHz for every 1MHz I move the input signal) and it's level drops by 1dB for every 1dB drop in the input signal.

    1) Do you have any suggestions as to what could cause this 1:1 spur? HD2 and HD3 I understand, but I am not sure what could be causing this 1:1 spur.

    Thank you.

  • Can Analog Devices please provide some guidance?

    Thank you.

  • 0
    •  Analog Employees 
    on Jun 21, 2021 2:37 PM in reply to gavint

      At low frequency the gain inside the chip is more and it might be saturating the Rx creating these spurs. For SFDR measurements at lower frequency, try setting the signal to a level that will produce an output of approximately -5dBFS to avoid clipping/saturation that could cause more harmonics and noise. Can you check the performance by feeding the signal at -5dBFs?

    You can select the coupling capacitor as per the frequency of operation. 

    On the 1:1 spur check the input signal for harmonics. If the signal generator harmonics are significant they can beat and produce spurs. Try using a BPF at input for the measurements. 

  • I have also noticed a significant improvement in the performance of the ADRV9009 as the input signal level is dropped. 

    The following is the single tone test results when testing at -5dBFS:

    The performance is still only about 60dBc to the largest spur (3:1 spur). We are hoping to achieve at least 70dBc but preferably 80dBc. 

    The signal generator output is filtered by an 80MHz LPF that has at least 80dB of stop band attenuation. I am not sure I follow your reasoning for the 1:1 spurs being related to harmonics of the input signal. I am struggling to understand how this can be the cause. For example, a third harmonic of 79.55MHz would move by 3MHz for every 1MHz move in the input signal (79.55MHz).

    Do you have any other advice to help us get closer to our goal of 70dBc (80dBc ideal) performance with the ADRV9009? Are there any documented results of the performance of the ADRV9009 at lower frequencies?

     Thank you again.

  • 0
    •  Analog Employees 
    on Jun 22, 2021 2:39 PM in reply to gavint

    The harmonic of internal LO can mix with any signal around 2x the LO and fall in band which can show a 1:1 relation.

    Since you are using a LPF this should be getting coupled to the Rx port externally or internally. 

    Try connecting the Rx port to spectrum and see the frequency components you see there. 

    The 1:1 spur are you seeing even without the tone?  

    At 150 MHz  ,you can try using reduced bandwidth and try.

    If not, Try reducing the input further and see at what input level the spurs meet your requirement.