The NUFFTadjoint result is very different from that of irt by jeff fessler in scale

[Michaelsqj]

Dear sigpy developers,

I had some NUFFT code written using irt by jeff fessler, and I was hoping to migrate to sigpy. However I found the results are quite different, even in absolute scales.

% ktraj: [Npts,3]
% kd: [Npts]
st   =   nufft_init(ktraj,...
                          im_size,...
                          [6,6,6],...
                          im_size*2,...
                          im_size/2);
dd = nufft_adj(kd, st);

and the python script

coords=ktraj/np.pi/2*im_size
dd = sp.nufft_adjoint(kd, coords, oshape=im_size)

the sum of square for dd reconstructed using irt nufft_adj was 5.5726e+06, whereas that reconstructed by sigpy was 1.0250025.

May I ask do you have idea what causes the scale difference?

Yours,
Qijia