cdrnet
commented
5 years ago It seems there are three common choices for these constants. To my understanding, our coefficients minimize peak side-lobe levels, but it indeed deviates slightly from the standard definition. In similar cases in the past we've let the user decide which definition is needed (e.g. quantiles where there are a dozen of definitions).
It seems we could provide:
- Hamming (25/46)
- Approximative Hamming (0.54), like Matlab -> lower side-lobes
- Equiripple Hamming (0.53836), i.e. the existing implementation -> minimal max side-lobe
Any thoughts on that?
Andi11449
while replacing the Hamming function used in an existing project with the MathNet version in the Window class I noticed that the constant values are different to the ones used by Matlab and Python leading to failing unit tests. Is there a reason why a, b are defined as 0.53836 and 0.46164 instead of 0.54 and 0.46 (which seems to be the implementation in most libraries)? The used values are even farther away from 25/46 and 21/46 which seems to be the most common interpretation of the window. I would suggest to change them to 0.54 and 0.46 to be consistent with other common libraries like Matlab and Python.