diff --git a/qupulse/utils/types.py b/qupulse/utils/types.py
index 2d2267e97..0aeabd06a 100644
--- a/qupulse/utils/types.py
+++ b/qupulse/utils/types.py
@@ -219,7 +219,7 @@ def from_float(cls, value: float, absolute_error: typing.Optional[float] = None)
         elif absolute_error > 1:
             raise ValueError('absolute_error needs to be <= 1')
         else:
-            return cls(qupulse_numeric.approximate_double(value, absolute_error, fraction_type=cls._InternalType))
+            return cls(qupulse_numeric.approximate_double(value, absolute_error, fraction_type=cls._to_internal))
 
     @classmethod
     def from_fraction(cls, numerator: int, denominator: int) -> 'TimeType':
diff --git a/tests/utils/time_type_tests.py b/tests/utils/time_type_tests.py
index 87d201c28..6273d501c 100644
--- a/tests/utils/time_type_tests.py
+++ b/tests/utils/time_type_tests.py
@@ -97,6 +97,8 @@ def assert_fraction_time_from_float_with_precision_works(self, time_type):
                          fractions.Fraction(1))
         self.assertEqual(time_type.from_float(2.50000000000008, absolute_error=1e-10),
                          time_type.from_fraction(5, 2))
+        self.assertEqual(time_type.from_float(9926.666666667, absolute_error=1e-9),
+                         time_type.from_fraction(29780, 3))
 
     def test_fraction_time_from_float_with_precision(self):
         self.assert_fraction_time_from_float_with_precision_works(qutypes.TimeType)