Limit on order in makefilter to ensure positivity of the filter in physical space

[glwagner]

@kburns suggests it may be so.

FourierFlows.jl/src/domains.jl

Line 343 in 0802dc3

decay = 15*log(10) / (outerK-innerK)^order # decay rate for filtering function

[navidcy]

This is a wavenumber filter. When you say "physical space" you mean when you take, e.g., ifft(filter)?

Why does the wavenumber filter have to be positive in physical space...? Am I missing something @glwagner?

[navidcy]

OK, I see..
You are making the point that changing the order can make the filter transition from something like a smooth gaussian filter to a "top-hat" filter in wavenumber space. And this would induce ringing artifacts.