I am working on ultrasound signals, which in general are narrowband and short duration, however, my aim is to use spread spectrum signals that give wider bandwidth.

The problem with such is that upon specific duration i.e. 50 µs signals overlaps

The overlapping problem that appears is that the first reflection of transducer is overlapped with the second reflection in case of sample (I am using air-coupled transducers).

One idea could be to use blind source separation, however, I am not an expert, but as far as I understood, it performs individual component analysis, which is not applicable in this case, as source is same but second reflection is only the delayed version of first reflection.

  • $\begingroup$ If by overlap you mean that returns are overlapping in time as part of the return, then you might have a resolution problem and need more bandwidth in your waveform. If you're running into the issue where you need to receive a signal before you can transmit the next pulse, that's a different problem. $\endgroup$
    – Envidia
    Commented May 26 at 18:09

1 Answer 1


The nice thing about spectrum spreading is that it gives you freedom in choosing on how to do that. A common method is using DSSS (direct-sequence spread spectrum), which essentially means you're transmitting a pseudorandom sequence, instead of just a pulse.

You can choose this sequence! And when you do so, you'd choose something that has good autocorrelation properties: when correlated with a delayed version of itself, there's exactly one high correlation "peak".

If you use that for your signals, then it doesn't matter that the reflections overlap in time - by correlating with the same sequence in the receiver, you get one peak per reflection, well-defined.

There's a lot such sequences to choose from; classical choices are Hadamard sequences if you are free to transmit pulses of inverse phase (i.e., +1 or -1, modulated onto a carrier). If you have a transmitter that can only continuously adjust the frequency of a tone, you'd go for Zadoff-Chu sequences. There's many more solutions (including just picking random sequences!), but which you use depend on the technical capabilities of your transmitter system, which you haven't told us.

  • $\begingroup$ I really appreciate your explanation. I am already using (nonlinear frequency modulation) chirp signal which are broadband, however, question arises when long duration signal are produced they overlapp. Now correlation is good answer and I had already tried it. Not very satisfactory results, moreover, I am confused as i need to window (gate) the reflection based on assumtion and then autocorelation with this gated version. The problem is where should the window end, we know the start, however, there is no end as start of second and end of first are overlapped. What should we do in such case? $\endgroup$ Commented Apr 25 at 16:24
  • $\begingroup$ well, if the results are not good after correlation, thent your chirps are not good in terms of autocorrelation, as explained above. You need different chirps, then. See the Zadoff-Chu sequences, I referred to these for a reason ;) $\endgroup$ Commented Apr 25 at 16:33
  • $\begingroup$ I always feel the best way to start is to first do a basic level simulation. In that you can adjust the first reflection to the worst case scenario and see the output of your correlator. You do not need to adjust the window but cancel the effect of the first reflection. $\endgroup$
    – learner
    Commented May 25 at 18:08

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.