TOPICS
Search

Circular Convolution


Circular convolution, also called cyclic convolution, is the convolution of finite sequences with their indices interpreted modulo a common length N. For x_0, ..., x_(N-1) and y_0, ..., y_(N-1), it is defined by

 z_n=sum_(m=0)^(N-1)x_my_((n-m)modN),

for n=0, ..., N-1. Here mod denotes the mod operation. Circular convolution is commutative and associative.

With the discrete Fourier transform convention X_k=sum_(n=0)^(N-1)x_ne^(-2piikn/N), where i is the imaginary unit, and similarly for Y_k and Z_k, the convolution theorem gives

 Z_k=X_kY_k.

Thus circular convolution can be computed using a fast Fourier transform. Ordinary convolution of two finite sequences of lengths L and M is obtained by appending zeros to both up to a length N>=L+M-1 before taking their circular convolution. This prevents the end of the result from wrapping around to its beginning.


See also

Convolution, Convolution Theorem, Discrete Fourier Transform, Fast Fourier Transform

Explore with Wolfram|Alpha

References

Smith, J. O. III. "Fourier Theorems for the DFT." Ch. 7 in Mathematics of the Discrete Fourier Transform (DFT) with Audio Applications, 2nd ed. Stanford, CA: W3K Publishing, 2007. https://www.dsprelated.com/freebooks/mdft/Fourier_Theorems_DFT.html.

Cite this as:

Weisstein, Eric W. "Circular Convolution." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/CircularConvolution.html

Subject classifications