Electrical Engineering Stack Exchange is a question and answer site for electronics and electrical engineering professionals, students, and enthusiasts. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I want to check the spectrum of a DDS signal ("pure" sinewave) for undesired frequency components. The way I am doing it is by storing the waveform from a 2GS/s oscilloscope (Agilent DSO81204) and then applying FFT (or Chirp Z) in MATLAB. I want to compare the results with the spectrum I see on the Spectrum Analyzer. Well, long story short, after the signal processing of the time domain signal in MATLAB I see two peaks on both sides of my desired frequency i.e. similar to AM.

fc = 15.625 MHz

When I feed the same signal to the Spectrum analyzer the peaks are not there anymore (even when using the smallest possible Bandwidth). So my question is - which spectrum is the right one? In MATLAB I tried using different windows including Rectangular (i.e. no window) but the peaks are still there. I filtered the signal with a lowpass filter before measuring with the oscilloscope and the peaks became smaller but were still there. The only reason that came to my mind is that it is some higher order mixing product in the oscilloscope. Anyone any ideas?

P.S. I tried another oscilloscope but it is some low end model that can store only 1600 points which is far from enough for an accurate FFT.

share|improve this question
Is your desired frequency zero kHz because that is what your picture indicates. – Andy aka Jul 23 '14 at 19:15

Those sidelobes are the result of the limited resolution of the data from the DSO, along with the fact that the sampling in the DSO is not precisely synchronized to the DDS output waveform, which results in a tiny amount of apparent (not actual) amplitude modulation of the actual waveform.

If you oversample and apply a proper sin(x)/x reconstruction filter to your DSO data, those sidelobes will disappear.

share|improve this answer

This is one of the reasons why (digital sampling + FFT) != spectrum analyzer. The true anwser is the spectrum analyzer. While the DSO + FFT is fine for large scale trends it can't give you detailed answers. For example looking at the plot I would guess that it is a 8 bit - 10 bit digitizer.

The SA (Spectrum Analyzer) is an almost ideal low noise instrument and, aside from taking spectrums, it is very useful in analyzing noise as the instrument basically has a noise floor of a 50 Ohm termination. It's hard to imagine a high speed DSO doing that.

share|improve this answer
Yes, the issue is probably the sampler. A modern spectrum analyzer typically (ultimately) uses digital sampling and an FFT too; but a painstakingly careful implementation thereof. – Chris Stratton Jul 23 '14 at 16:54

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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