diff --git a/newsfragments/759.bugfix b/newsfragments/759.bugfix
new file mode 100644
index 000000000..9dc343a64
--- /dev/null
+++ b/newsfragments/759.bugfix
@@ -0,0 +1 @@
+``dxtbx.dlsnxs2cbf``: add bit_depth_image explicitly to work around issues with data recorded at 32 bit
diff --git a/src/dxtbx/util/dlsnxs2cbf.py b/src/dxtbx/util/dlsnxs2cbf.py
index 2d45fcf85..cdb6df8cb 100644
--- a/src/dxtbx/util/dlsnxs2cbf.py
+++ b/src/dxtbx/util/dlsnxs2cbf.py
@@ -185,7 +185,7 @@ def make_cbf(
         print(f"Writing images to {template}{'#' * num_digits}.cbf:")
         for j in tqdm(range(num_images), unit=" images"):
             header = compute_cbf_header(nxmx_obj, j)
-            (data,) = dxtbx.nexus.get_raw_data(nxdata, nxdetector, j)
+            (data,) = dxtbx.nexus.get_raw_data(nxdata, nxdetector, j, bit_depth_readout)
             if bit_depth_readout:
                 # if 32 bit then it is a signed int, I think if 8, 16 then it is
                 # unsigned with the highest two values assigned as masking values