• State
  • Locked Locked
  • Replies 1 reply
  • Subscribers 9 subscribers
  • Views 765 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LO aliases in sampled data with 4DSP FMC30RF

Jeffrey Melville
Prodigy 10 points
Other Parts Discussed in Thread: CDCE62005, TRF3765, TRF371125, AFE7225

I am integrating a 4DSP FMC30RF card into my design. To my understanding, this board was designed for TI as a card for demonstrating multiple TI chips. We are having some technical problems that they suggested would be better answered by TI. I am copying posts that I previously posted in a 4DSP forum.

The board uses the following TI components:

Main Clock generation: CDCE62005

LO PLL: TRF3765

IQ Downconverter: TRF371125

AFE: AFE7225

The CDCE62005 is generating 122.88 MHz clocks using from an external 10 MHz atomic reference. The TRF3765 is configured for 1575.420 MHz for baseband downconversion of a 1575.420 MHz signal. The TRF371125 is configured with filter bypass off, but with maximum filter bandwidth. The AFE7225 is configured with an input clock divider of 1 but with internal decimation enabled for a final sampling rate of 61.44 MHz. I am executing automatic calibration of the AFE.


In this configuration, I see spurs at +/- 22.02 MHz at about 30 dB above the noise floor. This frequency is an alias of the LO at 61.44 MHz sampling. I believe that the signal is appearing AFTER the filter in the IQ downconverter because I am unable to suppress the tones by decreasing the filter bandwidth. I confirmed that our filter configuration is taking effect because I was able to configure the filter to suppress a tone at LO minus 20 MHz that appears adjacent to the aliases. To me, it looks like the LO is coupling into the analog I and Q channels prior to entering the AFE.

If I disable the AFE's decimation for an effective sampling rate of 122.88 MHz, I am able to observe aliases of at least the first 4 harmonics of the LO.  What could be going on here?

Even worse, if I apply a tone at 2 MHz above the LO, I get an image 2 MHz below the IF as well. Could this be due to an IQ imbalance or problems in calibration?


For what it's worth, I am mostly an embedded software engineer and not an RF expert. It's possible that I overlooked something basic. However, I have been able to replicate some of this behavior with 4DSP's reference design as well.

I have attached a plot of the spectrum of the baseband data with a 1575.420 MHz LO and 122.88 MHz sampling (no decimation in the AFE).

Let me know if you need any additional information.

Thanks,

Jeff

  • 0
    TI__Mastermind 18065 points
    I agree with your analysis. It looks like the spur is indeed an aliased down version of the LO signal. There is some LO leakage to the BB section but the filter bandwidth and natural amplifier bandwidth will attenuate that signal significantly. The spur that you report seems too high and, in addition, is not affected by filter bandwidth set-point. It seems that there is another mechanism at play. I have two thoughts: radiated coupling and power supply coupling. Try a couple of easy experiments to see if there is any impact.
    First, with your hand, cover the area around the TRF3765 LO generator. Your hand will act like a shield and if there is a radiated problem I would expect to see the spur drop significantly. Second, touch the bypass caps on the supply pins (after properly grounding yourself for ESD) of the device. This too uses the capacitance of your fingers to see if there is coupling on the supply lines of the BB circuitry that is allowing the LO to seep through. Activate the 3 dB attenuator of the TRF3711 and see if the spur stays the same or drops 3 dB.
    ***Even worse, if I apply a tone at 2 MHz above the LO, I get an image 2 MHz below the IF as well. Could this be due to an IQ imbalance or problems in calibration?***
    RJH>> An image is expected, but it should be attenuated by 40 dB or so compared to the desired sideband. If the image and fundamental are the same value, then there is a problem in one of the BB paths. This would be a complete failure in one of those paths so it should be easy to probe the connections and find the issue. Double check the voltage levels around the ’11 device to ensure that they are actually at the proper levels.
    --RJH