Post Go back to editing

Quad MxFE (ADQUADMXFE2EBZ) analog Rx response

Category: Choose a category
Product Number: ADQUADMXFE2EBZ

Hi,

Can I ask what is the expected Rx analog response of the ADQUADMXFE2EBZ quad MxFe variant board?

The hardware details page on the wiki states "Rx Analog Input Frequency Range: Up to ~1.8GHz" but I cannot find any further information. Judging from the ADC front end in the schematic, it should be fairly flat up to 1.8 GHz with some high pass filtering (cutoff?) from the dc blocking caps.

I ask because during testing we observed a dramatic high pass filtering effect. We are operating 8 ADC at 4GSPS with a decimation of 2, leading to 2GSPS complex output. Sweeping a sinewave input across the range 0 to 1.8 GHz (using rohde and schwarz SMA100B sig gen), we see the following frequency response with an Rx NCO frequency of 1GHz:

WIth low analog frequencies (approx < 300 MHz) strongly suppressed.  Is this an analog issue? Or is there likely something wrong with our design? I see the same roll off regardless of NCO frequency, with low frequency input signals always suppressed.

  • Hi callum123,

    Thanks for reaching out.

    You are correct in that the blocking caps are mostly causing this high-pass response behavior.

    The active components in the ADC-side signal chain operate:

    • HMC8411LP2FE: 0.01GHz-10GHz
    • HMC540SLP3E: 0.1-8GHz

    The passive components in the ADC-side signal chain operate:

    • TCM1-83X+: 0.01-8GHz
    • LFCN-1700D+: DC-1.7GHz

    The limiting components in the ADC-side signal chain on the low-end of the frequency range are mainly the 0.9pF capacitors on the differential legs prior to entering the ADCs. At 100MHz those caps offer a high series impedance which attenuate the signal. This, in tandem with the series 100pF caps, causes a high-pass response which you show above. Swapping those capacitors to a lower impedance alternative for the lower frequency ranges will likely improve the lower frequencies at the expense of potentially sacrificing performance at the higher frequencies within your band of interest.

    If this rolloff is prohibitive for your application you could consider flattening the response using the hardened DSP pFIR filters on MxFE. An example on how to do this can be found at the links below and is available via an example MATLAB script.

    Quad-MxFE Quick Start Guide [Analog Devices Wiki]

    HighSpeedConverterToolbox/hsx_examples/qmxfe at master · analogdevicesinc/HighSpeedConverterToolbox · GitHub

  • Thanks for the quick response. Flattening the response digitally is not really an option for us as we need to maintain high SNR across the ADC bandwidth.

    Can I ask the reason behind the design choices here? It seems quite strange especially considering the single AD9081 eval board (AD9081-FMCA-EBZ) uses a completely different balun/cap combination that is able to achieve a broadband frequency response.

    Indeed, it seems quite misleading of ADI to state that "Rx Analog Input Frequency Range: Up to ~1.8GHz" for the ADQUADMXFE2EBZ variant, as in link below. This seems to defy any reasonable definition of bandwidth given the roll off shown above.

    https://wiki.analog.com/resources/eval/user-guides/quadmxfe/boardhardwaredetails

    Thanks.

  • Hi Callum123,

    Thank you for your comments.

    The TCM1-83X balun was selected as its footprint was small enough to fit within the lattice spacing targeted for the platform. The single-chip evaluation board did not have a lattice spacing requirement and used larger footprint baluns as you mention.

    The filtering was a design trade-off where we wanted to provide modest filtering of any 2nd Nyquist content on-board at the smallest footprint and price target we could achieve.