sievers/corrcal2

Python/C code for calibration of quasi-redundant arrays. Different (fixed) algorithm from original corrcal

corrcal2's Issues

Reordering source model

Looking at this code, it seems that you do a more than just reorder the visibility data. So we should not be doing the same to the source model?
Also the ii used to reorder the visibilities is not the same one that is returned; its modified in ii=numpy.argsort(myind).

PS: The data is at hippo:/home/sphe/Sources

def grid_data(vis,u,v,noise,ant1,ant2,tol=0.1,do_fof=True):
    """Re-order the data into redundant groups.  Inputs are (vis,u,v,noise,ant1,ant2,tol=0.1)
    where tol is the UV-space distance for points to be considered redundant.  Data will be
    reflected to have positive u, or positive v for u within tol of zero.  If pyfof is
    available, use that for group finding."""

    ii=(v<0)|((v<tol)&(u<0))
    tmp=ant1[ii]
    ant1[ii]=ant2[ii]
    ant2[ii]=tmp
    vis[ii]=numpy.conj(vis[ii])

    if (have_fof & do_fof):
        uv=numpy.vstack([u,v]).transpose()
        groups=pyfof.friends_of_friends(uv,tol)
        myind=numpy.zeros(len(u))
        #for i in numpy.arange(len(groups)):
        for jj,group in enumerate(groups):
            for i in range(len(group)):
                myind[group[i]]=jj
        ii=numpy.argsort(myind)
        u=u[ii]
        v=v[ii]
        ant1=ant1[ii]
        ant2=ant2[ii]
        noise=noise[ii]
        myind=myind[ii]
        edges=numpy.where(numpy.diff(myind)!=0)[0]+1
        edges=numpy.append(0,edges)
        edges=numpy.append(edges,len(ant1))
        return vis,u,v,noise,ant1,ant2,edges,ii

Seg. fault when testing on simulated data

at this line of the test_corrcal2_fof.py file:

asdf=fmin_cg(corrcal2.get_chisq,gvec*fac,corrcal2.get_gradient,(big_vis+500*src,mycov,ant1,ant2,fac))

This is the output from the debuger.

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff2df8429 in apply_gains_to_mat () from libcorrcal2_funs.so