diff --git a/poppy/accel_math.py b/poppy/accel_math.py
index 5e9fadee..4e1a4f58 100644
--- a/poppy/accel_math.py
+++ b/poppy/accel_math.py
@@ -387,7 +387,7 @@ def benchmark_fft(npix=2048, iterations=20, double_precision=True):
         time_fftw = timer.timeit(number=iterations) / iterations
         print("  {:.3f} s".format(time_fftw))
     else:
-        time_fftw = np.NaN
+        time_fftw = np.nan
 
     if poppy.accel_math._NUMEXPR_AVAILABLE:
         print("Timing performance with Numexpr + FFTW:")
@@ -396,7 +396,7 @@ def benchmark_fft(npix=2048, iterations=20, double_precision=True):
         time_numexpr = timer.timeit(number=iterations) / iterations
         print("  {:.3f} s".format(time_numexpr))
     else:
-        time_numexpr = np.NaN
+        time_numexpr = np.nan
 
     if poppy.accel_math._MKLFFT_AVAILABLE:
         print("Timing performance with Numexpr + MKL:")
@@ -407,7 +407,7 @@ def benchmark_fft(npix=2048, iterations=20, double_precision=True):
         time_mkl = timer.timeit(number=iterations) / iterations
         print("  {:.3f} s".format(time_mkl))
     else:
-        time_mkl = np.NaN
+        time_mkl = np.nan
 
     if poppy.accel_math._OPENCL_AVAILABLE:
         print("Timing performance with OpenCL:")
@@ -416,7 +416,7 @@ def benchmark_fft(npix=2048, iterations=20, double_precision=True):
         time_opencl = timer.timeit(number=iterations) / iterations
         print("  {:.3f} s".format(time_opencl))
     else:
-        time_opencl = np.NaN
+        time_opencl = np.nan
 
 
     poppy.conf.use_mkl, poppy.conf.use_fftw, poppy.conf.use_numexpr, poppy.conf.use_cupy,\
diff --git a/poppy/optics.py b/poppy/optics.py
index 66ffd009..06771549 100644
--- a/poppy/optics.py
+++ b/poppy/optics.py
@@ -566,7 +566,7 @@ def get_transmission(self, wave):
             else:
                 raise NotImplemented("No defined NIRCam wedge BLC mask for that wavelength?  ")
 
-            sigmas = scipy.poly1d(polyfitcoeffs)(scalefact)
+            sigmas = numpy.poly1d(polyfitcoeffs)(scalefact)
 
             sigmar = sigmas * np.abs(y)
             sigmar.clip(np.finfo(sigmar.dtype).tiny, out=sigmar)  # avoid divide by zero -> NaNs