I have several analog signals with constant amplitude in time. What I want is a way to convert these analog signals to an impulse signal i.e. a signal that consists of impulses occurring at different instants of time.

The conversion should be such that it should be possible to differentiate between different analog signals just by looking at the impulse signals that they have been converted into? Also the impulse signals should be of finite duration.

I don't come from a signal processing background so my question might not have a lot of info required.

Please let me know if there is any other info that's needed.

Thank you.

  • $\begingroup$ So your continuous signals are simply $x(t)=c$ for all $t$ with a given constant $c$? I don't understand what kind of information should be conveyed by the location of the impulses of the converted signal. Or do you mean to convert the number $c$ (i.e., the amplitude) to a binary format consisting of 1s and 0s? $\endgroup$
    – Matt L.
    Jun 25, 2013 at 14:25
  • $\begingroup$ All the analog signals don't have a constant amplitude. Lets say I have certain numbers between 0 & 1. There can be infinite numbers between 0 & 1. I need a way to convert these numbers to an impulse signal such that different numbers would have different impulse pattern. $\endgroup$
    – Shaun
    Jun 25, 2013 at 14:35
  • 1
    $\begingroup$ From what you're saying I understand that you actually want to quantize analog signal values to binary bit streams. Of course you can only distinguish a finite number of values, otherwise your binary bit streams become infinitely long. Have you had a look at pulse-code modulation? $\endgroup$
    – Matt L.
    Jun 25, 2013 at 19:52
  • $\begingroup$ You mean there is no way I can work with finite number of bits if I have infinite analog values... $\endgroup$
    – Shaun
    Jun 26, 2013 at 3:46
  • 1
    $\begingroup$ Well, a finite number of bits offers a finite number of different bit combinations, so you can only represent a finite number of different values. What you get is called quantization, i.e. an (arbitrarily small) interval of analog values is mapped to the same bit pattern. $\endgroup$
    – Matt L.
    Jun 26, 2013 at 7:12

1 Answer 1


An approach for converting an analog signal into a bitstream is to use delta-sigma modulation. Intuitively, the idea is to build a sequence of binary pulses such that, when it is low-pass filtered, you recover the input signal.

It is not clear what you mean by "differentiate". There are many conversion processes for which distinct inputs will yield distinct outputs. But I you suspect that you expect something like "similar-looking inputs should yield similar-looking outputs". The question is what kind of definition of "similar-looking" you have in mind...

  • $\begingroup$ I will use my example of infinite numbers between 0 & 1. Lets say that I have the numbers .01356, .01349 & .01692. I am trying to convert then to an impulse signal. The impulse signals should be such that they are different for all the three numbers but the signals for first 2 numbers are closer to each other than that of the third number. $\endgroup$
    – Shaun
    Jun 25, 2013 at 14:46
  • $\begingroup$ I guess I should have asked this question along with dsp.stackexchange.com/questions/9721/… But now I can't combine them into one, still you can refer that question and it might answer your question. $\endgroup$
    – Shaun
    Jun 25, 2013 at 14:47
  • $\begingroup$ "Closer to each other" according to each distance? $\endgroup$ Jun 25, 2013 at 14:47
  • $\begingroup$ What if I answer you like this: represent the value 0.01356 as a 32-bit floating point in the IEEE 754 format. You get a sequence of 32 0s or 1s. To compare these sequences, interpret them as IEEE 754 32-bit floats and compute the distance. $\endgroup$ Jun 25, 2013 at 14:49
  • $\begingroup$ Another solution: represent 0.01356 as a sequence of 1356 1s and 98644 zeros. To compare sequences, count the number of 1s. $\endgroup$ Jun 25, 2013 at 14:51

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.