Steven Murray

Nice... well I tinkered around with the code and made it much faster (see [this gist](https://gist.github.com/steven-murray/d334cd3fa31292fffbbbd9ce065157b0)). It is now vectorized and this means the most intense part of the calculation...

> Regarding the direction='inverse' versus inverse=True, I think that I prefer the direction='inverse', since it places the two directions on even footing, and makes the user consider which transform they...

With the "Curve A", we could attempt to determine the solution to the Fourier transform and then the Hankel transform (analytically) to see where things are going astray.

For Curve A, is the curve zero beyond ``r=1``? If so, the solution is (I think): 

Thanks @stggh , this looks promising -- certainly using the zeros of the Bessel function is more akin to the way that ``hankel`` works. Making this a proper DHT will...

> Neat! Amazing ``sympy``? ;-) Actually, this time I just used Wolfram|Alpha on each term separately :-)

I've used autosummary a bit. It's nice, but it can be a little finnicky, depending on how the code is laid out. I can see if I can help with...

@bhazelton and @mkolopanis : this is essentially working (I will write some tests soon). My non-unit-test way of testing this was to run the following script: ```python import os import...

@mkolopanis and @bhazelton I believe this PR is ready.

Thanks @florpi for the detailed issue. This is confusing indeed. Can you make sure that using a power spectrum of say `lambda k: k**-2` gives the correct normalization for you?