Steve Gibson

Q1: > How does this compare with the "Fourier Hankel" method? The selection of this algorithm was from the reference given in issue [#61](https://github.com/PyAbel/PyAbel/issues/61#issuecomment-282784537), a method that appeared to be...

From what I can tell extracting cosine amplitudes from the FFT spectrum requires some form of peak-finding, ok for a simple oscillating function, but not clear cut for a Gaussian...

You are right, on a frequency scale the coefficients are `A0 ... An`. I had missed that subtly.

Update: I have not made much progress on reconciling the cosine series coefficients from `scipy.fftpack.rfft` with those from the least-squares fit of the cosine series to the (forward Abel transformed)...

Oh, I now see. The `rfft` should be take of the forward transform in order to compare coefficients `An`.

Unfortunately, no. I see no common pattern between the coefficients of the two methods. I think this may be due to the phase factors/complex components of `rfft`, although the lack...

Update: I have removed superfluous row storage for `hbasis` and fixed the basis storage functions. The `fourier_expansion` method is now usable, the time comparison for `two_point` is now *only* `x200`...

Finally! I have decode the `fft` output to coefficients that agree with the least-squares fitting. The index for the real coefficients `An` appears to be `2*n - 1`, for whatever...

Thanks @DanHickstein! The problem is that `rfftfreq()` returns `n/n[-1]/d/2` and hence the maximum frequency is always `

Do we just normalize the length of our image to 2pi, so `d = 2*np.pi/cols`? Edit, no: `freq_max = 1/d/2` Edit2: Yes, `d = (r[1]-r[0])/np.pi/2` puts the frequencies at the...