You don't need pandas, and a cleaner check to ensure Unicode support etc is to use numpy.
I built this little function which I use instead of the standard versions that ship natively
def save_txt(a, name="arr.txt", sep=", ", dt_hdr=True):
"""Save a NumPy structured, recarray to text.
Parameters
----------
a : array
input array
fname : filename
output filename and path otherwise save to script folder
sep : separator
column separater, include a space if needed
dt_hdr : boolean
if True, add dtype names to the header of the file
"""
a_names = ", ".join(i for i in a.dtype.names)
hdr = ["", a_names][dt_hdr]
s = np.array(a.tolist(), dtype=np.unicode_)
widths = [max([len(i) for i in s[:, j]])
for j in range(s.shape[1])]
frmt = sep.join(["%{}s".format(i) for i in widths])
np.savetxt(name, a, fmt=frmt, header=hdr, comments="")
print("\nFile saved...")
since you already have your array, you can use the above with any separator you want, not just comma-delimited.
save_txt(ar, "c:/temp/arr.txt", sep=", ", dt_hdr=True)
