diff --git a/janus_core/helpers/post_process.py b/janus_core/helpers/post_process.py
index ebcca5d2..a750ad32 100644
--- a/janus_core/helpers/post_process.py
+++ b/janus_core/helpers/post_process.py
@@ -189,7 +189,7 @@ def compute_vaf(
     use_velocities: bool = False,
     fft: bool = False,
     index: SliceLike = (0, None, 1),
-    filter_atoms: MaybeSequence[MaybeSequence[int]] = ((),),
+    filter_atoms: MaybeSequence[MaybeSequence[Optional[int]]] = ((None),),
 ) -> NDArray[float64]:
     """
     Compute the velocity autocorrelation function (VAF) of `data`.
@@ -209,7 +209,7 @@ def compute_vaf(
     index : SliceLike
         Images to analyze as `start`, `stop`, `step`.
         Default is all images.
-    filter_atoms : MaybeSequence[MaybeSequence[int]]
+    filter_atoms : MaybeSequence[MaybeSequence[Optional[int]]]
         Compute the VAF averaged over subsets of the system.
         Default is all atoms.
 
@@ -238,9 +238,7 @@ def compute_vaf(
     data = data[slice(*index)]
 
     if use_velocities:
-        momenta = np.asarray(
-            [datum.get_momenta() / datum.get_masses() for datum in data]
-        )
+        momenta = np.asarray([datum.get_velocities() for datum in data])
     else:
         momenta = np.asarray([datum.get_momenta() for datum in data])
 
@@ -249,7 +247,7 @@ def compute_vaf(
 
     # If filter_atoms not specified use all atoms
     filter_atoms = [
-        atom if atom and atom[0] else range(n_atoms) for atom in filter_atoms
+        atom if atom[0] is not None else range(n_atoms) for atom in filter_atoms
     ]
 
     used_atoms = {atom for atoms in filter_atoms for atom in atoms}
@@ -259,7 +257,9 @@ def compute_vaf(
         np.asarray(
             [
                 [
-                    np.correlate(momenta[:, j, i], momenta[:, j, i], "same")
+                    np.correlate(momenta[:, j, i], momenta[:, j, i], "full")[
+                        n_steps - 1 :
+                    ]
                     for i in range(3)
                 ]
                 for j in used_atoms
diff --git a/tests/data/vaf-lj-1-2-3.dat b/tests/data/vaf-lj-1-2-3.dat
new file mode 100644
index 00000000..81aaba52
--- /dev/null
+++ b/tests/data/vaf-lj-1-2-3.dat
@@ -0,0 +1,11 @@
+1.0066870750448105
+0.6418190953234829
+0.3563911120413356
+0.3051070730422549
+0.23009251840934272
+0.26906890724754023
+0.24509942703971466
+0.15062229045327677
+-0.23061920240560663
+-0.42160528271693426
+-0.20524096931167723
diff --git a/tests/data/vaf-lj-3-4.dat b/tests/data/vaf-lj-3-4.dat
new file mode 100644
index 00000000..03d62c99
--- /dev/null
+++ b/tests/data/vaf-lj-3-4.dat
@@ -0,0 +1,11 @@
+1.1263716900873755
+0.8820158202526532
+0.784347065477728
+0.6806780688236092
+0.5265986467643831
+0.5295303067488114
+0.5150663396964206
+0.5188617483216265
+0.3239983242603667
+0.19859295298597962
+0.058499204536948346
diff --git a/tests/data/vaf-lj.dat b/tests/data/vaf-lj.dat
new file mode 100644
index 00000000..154db564
--- /dev/null
+++ b/tests/data/vaf-lj.dat
@@ -0,0 +1,11 @@
+0.940392953490707
+0.581286162528787
+0.3474623957116591
+0.19574532179574347
+0.08404080467605056
+0.02723301143121412
+0.03320823810966539
+-0.0004815372548760061
+-0.05545244283596119
+-0.06580258912853633
+-0.09743930723331182
diff --git a/tests/test_post_process.py b/tests/test_post_process.py
index bba91146..f394d448 100644
--- a/tests/test_post_process.py
+++ b/tests/test_post_process.py
@@ -4,6 +4,7 @@
 
 from ase.io import read
 import numpy as np
+from pytest import approx
 from typer.testing import CliRunner
 
 from janus_core.calculations.md import NVE
@@ -172,14 +173,19 @@ def test_rdf_by_elements():
         assert (np.isclose(expected_peaks[element], rdf[0][peaks])).all()
 
 
-def test_vaf():
+def test_vaf(tmp_path):
     """Test vaf will run."""
+    vaf_names = ("vaf-lj-3-4.dat", "vaf-lj-1-2-3.dat")
+    vaf_filter = ((3, 4), (1, 2, 3))
+
     data = read(DATA_PATH / "lj-traj.xyz", index=":")
     vaf = post_process.compute_vaf(data)
+    expected = np.loadtxt(DATA_PATH / "vaf-lj.dat")
 
     assert isinstance(vaf, list)
     assert len(vaf) == 1
     assert isinstance(vaf[0], np.ndarray)
+    assert vaf[0] == approx(expected, rel=1e-9)
 
     vaf = post_process.compute_vaf(data, fft=True)
 
@@ -187,8 +193,17 @@ def test_vaf():
     assert len(vaf) == 1
     assert isinstance(vaf[0], np.ndarray)
 
-    vaf = post_process.compute_vaf(data, filter_atoms=((3, 4), (1, 2, 3)))
+    vaf = post_process.compute_vaf(
+        data, filter_atoms=vaf_filter, filenames=[tmp_path / name for name in vaf_names]
+    )
 
     assert isinstance(vaf, list)
     assert len(vaf) == 2
     assert isinstance(vaf[0], np.ndarray)
+
+    for i, name in enumerate(vaf_names):
+        assert (tmp_path / name).exists()
+        expected = np.loadtxt(DATA_PATH / name)
+        written = np.loadtxt(tmp_path / name)
+        assert vaf[i] == approx(expected, rel=1e-9)
+        assert vaf[i] == approx(written, rel=1e-9)