Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Convolution with a gaussian G(t)

  1. Jan 6, 2005 #1
    Hi. Suppose I have a function T(x,t), units are in Kelvins. I then do a convolution with a gaussian G(t), and the result is also in Kelvins. What are the units of the gaussian G(t)? Thanks.
  2. jcsd
  3. Jan 6, 2005 #2


    User Avatar
    Science Advisor
    Homework Helper

    [tex] T(x,t)\ast G(t)=:\int_{0}^{t} T(x,\tau)G(t-\tau) d\tau [/tex]
    ,and if u want the convolution to have the same units as the T,this means that the 'G' must be dimensionless (a genuine exponential (Gauss-bell) is dimensionless).

  4. Jan 6, 2005 #3
    Thanks, that helps a lot.

    Now, the Gaussian that I have has a defined width (duration). But what about its height (amplitude, maximum etc.)?

    I would expect the result to have a defined maximum T in Kelvins, that I can use for further physical calculations. How can such a known maximum exist? How must I define my gaussian amplitude? I suspect normalization is involved but I'm not sure how to do it so I have a meaningful maximum T in the end.
  5. Jan 6, 2005 #4
    Actually, doesn't the dtau have units (say s)? So that if G has no units, the integral would have K.s as units?
  6. Jan 6, 2005 #5


    User Avatar
    Science Advisor
    Homework Helper

    That [itex] \tau [/tex] is viewed as a variable of integration and,for obvious reasons,it has the same dimension as "t".
    This function
    [tex] G(\tau)=\exp(-A\tau^{2}) [/tex]
    is an example of dimensionless function defined everywhere.For obvious reasons,the constant 'A' is dimensional and it has the dimensions of
    [tex] <A>_{SI} =(<\tau>_{SI})^{-2} [/tex]

    Yes,the integral will have dimensions of K.s,and that's because the parameter is dimensional.If it wasn't,it would have been K.

  7. Jan 8, 2005 #6
    Thanks for the help. I figured out what units my gaussian was in, everything came into place. Reading back the thread, everything now seems so trivial. How typicial.
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook

Have something to add?

Similar Discussions: Convolution with a gaussian G(t)