AnsweredAssumed Answered

AD9364 RX baseband analog filter passband response is not flat

Question asked by pharbour on Apr 16, 2015
Latest reply on Apr 16, 2015 by tlili

Hi,

 

I am using the AD9364 and running the no-OS code.  Specifically I am observing the spectral response of the RX baseband I/Q output signals.  Let me first describe my setup of the AD9364 RX signal chain.

 

- ADC sample rate: 192Msps

- HB3 off

- HB2 off

- HB1 on

- RX FIR off

- TRX RX clk chain {768000000, 192000000, 192000000, 192000000, 96000000, 96000000}

- rf-rx-bandwidth-hz = 26 MHz

- RX RF gain = 70dB

- no RF input signal (input terminated)

 

Thus the RX DSP chain is set to 2x decimation with half-band filter on in HB1, all other DSP filters are off.  And the RX baseband analog anti-alias low pass filters are programming to a bandwidth that is ~13.5% of the ADC nyquist bandwidth.  The goal here is to measure the gain response of the front-end LNA (white noise plot).

Given this configuration I would expect to see a flat spectral response in the passband.  But that is not observed.  Please see the attached plot 'AD9364_RX_baseband_complex_FFT_noise_input.png'.  There are a few discussions listed in this forum showing plots of the IO-oscilloscope under the same conditions with similar baseband I/Q signal spectra.  An example is

 

https://ez.analog.com/thread/43114?start=0&tstart=0

(see pic 4 in this post)

 

Is this the expected response of the DSP chain / baseband analog anti-alias filter?  (I think the passband should have a flat spectrum response)

 

Do I have a config error in my no-OS input structure?

 

Also observing the time domain I/Q output (see 'AD9364_RX_baseband_IQ_Lissajous_noise_input.png') suggests some portion of the DSP chain is clipping, but the signal level is roughly -14 dBFS so the "partially clipped edge" of the Lissajous plot is unexpected.

 

Please let me know if this RX chain response is expected.

 

 

Thanks,

 

Pete

 

Outcomes