I am fairly new to the topic of DSP and FFT and I have hit a mental block. For a project, I am trying to pick out frequency's with amplitude peaks in a 47-53 MHz signal. This signal is being put through a 24 bit ADC with a sample rate of 96000 Hz. In order to successfully measure a sample, don't you have to have a sample rate >= to twice the desired frequency (Nyquist rate)? How can you get an accurate representation of a 47-53 MHz signal with such a low sample rate?

The second block I have hit involves the FFT of this signal. From my limited knowledge of FFT, the first "bin" (in the real part) represents 0 Hz and the last "bin" (in the real part) represents: $$ F=i\times \frac{Fs}{N} $$ Where $i$=Bin Index, $Fs$=Sample Rate, $N$=FFT Size

In my case with size 1024 and above sample rate, $511 \times 96000/1024 = 47906.25$ Hz, right? How can this range ($0$ Hz-$47906.25$ Hz) possibly represent my $47-53$ MHz signal? I get the feeling that I am missing something really crucial but I just can't figure it out.

Sorry if this is a really basic question and thanks for the help!

  • $\begingroup$ You need a high frequency ADC to capture such signal. $\endgroup$
    – jojeck
    Feb 13, 2015 at 17:39
  • 1
    $\begingroup$ Is the bandwidth 6MHz or does the range 47MHz to 53MHz just describe the range of possible carrier frequencies? The ability to reconstruct a signal depends on the bandwidth of the signal, not on the absolute frequency. $\endgroup$
    – Jazzmaniac
    Feb 13, 2015 at 19:26
  • $\begingroup$ @Jazzmaniac The bandwidth is 6Mhz $\endgroup$
    – Alex Hamm
    Feb 13, 2015 at 20:53

1 Answer 1


In short, what you're trying to do is impossible. The technique is known as band-pass sampling, or undersampling, but for it to work, you need that the signal's bandwidth be less than half your sampling frequency.

When you sample a signal with spectrum $S(f)$ and bandwidth $B$ at sampling frequency $f_s$, the sampled signal has spectrum $$S_s(f)=k\sum_i S(f-if_s)$$ where $k$ is a constant. If $B>f_s/2$ then the replicas $S(f-if_s)$ overlap and you end up with a distorted spectrum.

You are correct that the highest frequency bin in your FFT is 47906.25 Hz.

  • $\begingroup$ Thanks for the answer, I was going about the problem wrong. $\endgroup$
    – Alex Hamm
    Feb 16, 2015 at 4:55

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.