Post Go back to editing

AD9174-FMC-EBZ output gets corrupted

Category: Hardware
Product Number: ZCU102, AD9174-FMC-EBZ

I used following HW & SW setups and found that DAC output was corrupted by maybe DAC and/or circuit (balun, RF choke, etc.).

ZCU102 & AD9174-FMC-EBZ as HW.

JESD204B mode 18 + Wideband mode real dual (no NCO, interpolation used)

DAC sampling rate : 3Gsps

Input digital data : sinusoidal wave (10/20/30/40/50 MHz), sinusoidal wave (10/20/30/40/50 MHz) + some modulation

Case 1.

When I input simple sinusoidal wave data with different frequency and with "the same" amplitude from FPGA to DAC, the output amplitude measured by oscilloscope varies according to the frequency of the input sinusoidal wave.

The output amplitude becomes larger with higher input frequency. In this case, the amplitude of the 50MHz sinusoidal wave was twice as large as that of 10MHz sinusoidal wave. Looks like some HPF is applied.

Case 2.

When I input modulated sinusoidal wave data, the output waveform was corrupted. In this case, I applied some kind of phase modulation where 180 degree phase rotation is applied per 2 cycle of sinusoid.

From the above, I think the output of AD9174-FMC-EBZ has some frequency characteristics around 10~50 MHz or something I should case.

Could someone point this out?

  • Hello, the output impedance of the dac is complex, so the matching with the balun varies slightly over frequency. For example the green line on this graph shows the measured output power on a different eval board but with the same balun that is used on the 917x evb (TCM-83X+) and the same dac output architecture.

    Can you elaborate on the corruption you see in your output waveform for case 2?

  • Hello.

    Thank you for providing with an example of the frequency response with TCM-83X+.

    For case 1, output "voltage" in my environment is as follows.

    10MHz: 600mVp-p

    20MHz: 800mVp-p

    30MHz: 900mVp-p

    40MHz: 1200mVp-p

    50MHz: 1200mVp-p

    The voltage in 40MHz is twice as large as that in 10MHz (6dB difference in "power"), which seems too large compared with Figure 88...

    For case 2, the digital input and the corresponding output measured by oscilloscope are as follows.



  • Hi, can you try removing inductors L6 and L8 on the eval board upstream of the balun? These are not required since the balun has a DC bias path to ground. At low frequencies they drop the overall output impedance and output power on the board. Removing them may also improve your output signal, since they can't react quickly to the discontinuity in current at the inversion points in the digital data.