2
$\begingroup$

I am doing the spectral analysis of some antenna recorded signals by computing PSD (logarithmic absolute value of the FFT) and the spectrogram of the signal. The problem is that I get some unexpected spikes in the signal that appear through all the signal, as it can be seen in the next figure:

enter image description here

The two peaks that appear are the spikes I am speaking about, and they are the same in all the signal. The center spike is not located in $0\textrm{ Hz}$, so it is not a DC offset of the signal. This is giving me some problems, so I would like to know what can they be or how can they be removed (maybe I am not computing the spectral analysis correctly).

I have also performed Welch PSD estimations, and the peaks still appear. I have been told that it can be some internal interference form the electronics recording the data or some issue with ADCs.

Improve this question
$\endgroup$
6
  • $\begingroup$ Can you please add information about the origin of the signal? How was the signal created, at what distance from you, in what band, etc? If you see the signals in your analysis, that means that the signals are real. The main question is, are they present in the air, or are they produced by non-idealities in your receiver? $\endgroup$
    – MBaz
    Commented Nov 16, 2016 at 14:54
  • $\begingroup$ The signals are recordings of GNSS signals received from an antenna through the air. My main goal is to get signals that have higher power than an specific threshold. However, this spikes look not to come through the air, and I am wondering if they are a consequence of non-idealities in the receiver or if the way I am computing the psd is not correct, as when plotting in the spectrogram, most of those peaks power is much lower or they do not even appear. $\endgroup$
    – Josu Etxezarreta Martinez
    Commented Nov 16, 2016 at 15:03
  • $\begingroup$ I'd put my money on external interference (I see that happening all the time). BTW, the large peak to the right is there in the spectrogram too. There are several things you can try. For instance, connect the antenna to a good spectrum analyzer and look for the interference. Also, try changing your center frequency slightly and see if the two peaks shift the same amount. That would point to external interference. If possible, try to narrow your front end's bandpass filter to be as close as possible as your desired signal's BW. $\endgroup$
    – MBaz
    Commented Nov 16, 2016 at 16:53
  • $\begingroup$ I am not sure that they may be external interference, as when I change the length of the FFT the power of the peaks decrease sgnificantly compared to the signals that I assume to be external interference, sometimes completely disappearing. That's why I am confused about the nature of these peaks. I would appreciate to know some other possible things in order that I can research them deeper. Thanks. $\endgroup$
    – Josu Etxezarreta Martinez
    Commented Nov 22, 2016 at 9:44
  • $\begingroup$ I'm a bit confused by your PSD plot and wondered about your frequency axis. I'd expect GNSS signals to be higher frequency signals (GPS L1 & L2 carrier frequencies are about 1.2 and 1.5 GHz...). Am I missing something? Is it expected to have a PSD that's non-symmetric around DC frequency? $\endgroup$
    – snowflake
    Commented Dec 9, 2016 at 16:12

1 Answer 1

Reset to default
-1
$\begingroup$

The large spike at ~9MHz could be some form of interference from electronics near the receiver or internal to the receiver (maybe from using a poorly conditioned power source). My best guess for the spike at DC is a DC bias from quantization. If the digitizer truncates instead rounds, the resulting digital signal will have a non-zero mean and therefore a DC offset. We had this happen in an FPGA downcoverter implementation at a previous job. Once they changed the FPGA to round, the DC offset went away. If you have ever used an Ettus USRP radio, you see the same DC offset in many of those boxes.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.