Post Go back to editing

(Low power) 0 Hz tone showing up on RX FFT in 12-bit mode

Category: Hardware
Product Number: AD9081
Software Version: ACE 1.25.3217.1403

When I run the AD9081 in 12-bit mode, I see a 0 Hz tone on the FFT output that I don't see in 16-bit mode. It's not very high (~-70 dBm) but is on the order of or higher than other spurs and harmonics in the spectrum and makes the spectrum asymmetric. I've put more details below.

I don't think this is a result of post-processing and suspect something in the digital truncation may be responsible, but I was wondering if someone can comment if this is expected.

  • Here is the spectrum in 12-bit mode. There is a 0 Hz tone that makes the spectrum asymmetric
    • Settings
    • The spectrum shown has a CDDC NCO of 50 MHz applied to show the 0 Hz tone. The ACE output shows the tone. 
    •  I also exported the IQ waveform and took a FFT of the samples and saw the same 0 Hz tone. The y-axis isn't scaled here. Just in case there was a rounding error, I also took the mean at 65535 samples , 65534 samples, 65533 samples, and so on and it looks like the DC bias/0 Hz tone is always present.
  • Here is the corresponding spectrum in 16-bit mode. No 0 Hz tone is in the FFT and things look as expected.
    • Settings
      • The reference was tweaked to keep the ADC sampling rate the same as the 12-bit case (3 Gsps)
    • 100 MHz is going into ADC2. All other ADCs are terminated to 50 Ohms so aren't shown. The FFT calculated from the IQ waveform seems to roughly match (y-axis wasn't scaled)
  • I could only find this comment about the 16 -> 12 bit conversion in the user guide.
  • Hi, what exactly is your setup? I am not fully understanding the aim of this test. 

  • Hello,

    The cause  is a truncation spur Issue related to a register setting in the data formatter that is not being set properly by the API.   This register bit field will "round" to 12-bit resolution before truncation.  The latest ACE Plug-in should include this modified API such that this DC spur is not present.  Please download lastest plug-in and let us know if problem still exists. The affected register is shown below and is currently not being set by API to 12-bit resolution.

  • Hi PMH,

    Thanks! This was it. I updated ACE and it didn't apply the 12-bit resolution by default (was set to 16-bit resolution), but I rewrote the register in the memory map and saw the 0 Hz spur disappear after setting this correctly.

    1. I updated ACE to version 1.26.3240.1417
    2. I programmed the chip and had the NCO for this ADC channel set to some value. The ADC is terminated to 50 Ohms (no signal coming in). The 0 Hz tone is present
    3. I rewrote register 0x02A8 to the value for 12-bit mode (was 16-bit mode by default)
    4. Looking at the FFT again, the 0 Hz tone is gone. Woo!