diff --git a/docs/synphot/filter_par.rst b/docs/synphot/filter_par.rst
index 5883ca69..313d1837 100644
--- a/docs/synphot/filter_par.rst
+++ b/docs/synphot/filter_par.rst
@@ -59,7 +59,7 @@ By default, 10 FFT parameters are returned as complex numbers::
      (38.635087381362396-13.02803811279449j)]
 
 .. TODO: Only skipping the fft_pars comparison above because output is very
-   different for Numpy 1.16. Unskip it and replace with +FLOAT_CMP when
+   different for NUMPY_LT_1_17. Unskip it and replace with +FLOAT_CMP when
    Numpy minversion is 1.17.
 
 It is up to you to decide how to store this data, though storing it in a
@@ -70,15 +70,19 @@ will store the results in a table for you::
     >>> from synphot.filter_parameterization import filters_to_fft_table
     >>> mapping = {'HST/ACS/HRC/F555W': (bp, None)}
     >>> filter_pars_table = filters_to_fft_table(mapping)
-    >>> filter_pars_table  # doctest: +FLOAT_CMP +ELLIPSIS
+    >>> filter_pars_table  # doctest: +SKIP
     <Table length=1>
           filter      n_lambda ...                  fft_9
                                ...
-          str17        int32   ...                complex128
+          str17        int...  ...                complex128
     ----------------- -------- ... ---------------------------------------
     HST/ACS/HRC/F555W    10000 ... (38.635087381362396-13.02803811279449j)
     >>> filter_pars_table.write('my_filter_pars.fits')  # doctest: +SKIP
 
+.. TODO: Only skipping the filter_pars_table comparison above because output
+   is slightly different for NUMPY_LT_1_17. Unskip it and replace with
+   +FLOAT_CMP +ELLIPSIS when Numpy minversion is 1.17.
+
 .. _filter_fft_construction:
 
 Reconstructing Filter from FFT
diff --git a/synphot/compat.py b/synphot/compat.py
index c0a23d1e..6c5b2818 100644
--- a/synphot/compat.py
+++ b/synphot/compat.py
@@ -2,6 +2,7 @@
 """Module to handle backward-compatibility."""
 
 import astropy
+import numpy
 from astropy.utils.introspection import minversion
 
 try:
@@ -19,8 +20,9 @@
     HAS_DUST_EXTINCTION = True
 
 
-__all__ = ['ASTROPY_LT_4_1', 'ASTROPY_LT_4_0', 'HAS_SPECUTILS',
-           'HAS_DUST_EXTINCTION']
+__all__ = ['ASTROPY_LT_4_1', 'ASTROPY_LT_4_0', 'NUMPY_LT_1_17',
+           'HAS_SPECUTILS', 'HAS_DUST_EXTINCTION']
 
 ASTROPY_LT_4_1 = not minversion(astropy, '4.1')
 ASTROPY_LT_4_0 = not minversion(astropy, '4.0')
+NUMPY_LT_1_17 = not minversion(numpy, '1.17')
diff --git a/synphot/filter_parameterization/filter_fft.py b/synphot/filter_parameterization/filter_fft.py
index 9e7a73af..909ec889 100644
--- a/synphot/filter_parameterization/filter_fft.py
+++ b/synphot/filter_parameterization/filter_fft.py
@@ -5,6 +5,7 @@
 from astropy.modeling.models import custom_model, Sine1D
 from astropy.table import Table
 
+from synphot.compat import NUMPY_LT_1_17
 from synphot.models import Empirical1D
 from synphot.spectrum import SpectralElement
 from synphot.units import validate_quantity
@@ -210,18 +211,23 @@ def filters_to_fft_table(filters_mapping, n_terms=10):
 
     """  # noqa
     wave_unit = SpectralElement._internal_wave_unit
-
-    fft_table = Table(names=['filter', 'n_lambda', 'lambda_0', 'delta_lambda',
-                             'tr_max'] + [f'fft_{i}' for i in range(n_terms)],
-                      dtype=[np.str, np.int32, np.float32, np.float32,
-                             np.float32] + [np.complex] * n_terms)
-    fft_table['lambda_0'].unit = wave_unit
-    fft_table['delta_lambda'].unit = wave_unit
+    colnames = ['filter', 'n_lambda', 'lambda_0', 'delta_lambda',
+                'tr_max'] + [f'fft_{i}' for i in range(n_terms)]
+    rows = []
 
     for key, (bp, wavelengths) in filters_mapping.items():
         n_lambda, lambda_0, delta_lambda, tr_max, fft_pars = filter_to_fft(
             bp, wavelengths=wavelengths, n_terms=n_terms)
-        fft_table.add_row(tuple(
-            [key, n_lambda, lambda_0, delta_lambda, tr_max] + fft_pars))
+        if not NUMPY_LT_1_17:
+            rows.append(tuple(
+                [key, n_lambda, lambda_0, delta_lambda, tr_max] + fft_pars))
+        else:  # Numpy 1.16 cannot handle unit here
+            rows.append(tuple(
+                [key, n_lambda, lambda_0.value, delta_lambda.value,
+                 tr_max.value] + fft_pars))
+
+    fft_table = Table(rows=rows, names=colnames)
+    fft_table['lambda_0'].unit = wave_unit
+    fft_table['delta_lambda'].unit = wave_unit
 
     return fft_table