diff --git a/s2fft/sampling/so3_samples.py b/s2fft/sampling/so3_samples.py
index 194c2c4d..1731606c 100644
--- a/s2fft/sampling/so3_samples.py
+++ b/s2fft/sampling/so3_samples.py
@@ -13,8 +13,8 @@ def f_shape(
     :math:`SO(3)`.
 
     Note:
-        Importantly, the convention adopted for storage of f is :math:`[\beta, \alpha,
-        \gamma]`, for Euler angles :math:`(\alpha, \beta, \gamma)` following the
+        Importantly, the convention adopted for storage of :math:`f` is :math:`[\gamma,
+        \beta, \alpha]`, for Euler angles :math:`(\alpha, \beta, \gamma)` following the
         :math:`zyz` Euler convention, in order to simplify indexing for internal use.
         For a given :math:`\gamma` we thus recover a signal on the sphere indexed by
         :math:`[\theta, \phi]`, i.e. we associate :math:`\beta` with :math:`\theta` and
diff --git a/s2fft/utils/signal_generator.py b/s2fft/utils/signal_generator.py
index 407e9614..799de9b9 100644
--- a/s2fft/utils/signal_generator.py
+++ b/s2fft/utils/signal_generator.py
@@ -39,11 +39,11 @@ def complex_el_and_m_indices(L: int, min_el: int) -> tuple[np.ndarray, np.ndarra
 
     Equivalent to nested list-comprehension based implementation
 
-    ```
-    el_indices, m_indices = np.array(
-        [(el, m) for el in range(min_el, L) for m in range(1, el + 1)]
-    ).T
-    ```
+    .. code-block:: python
+
+        el_indices, m_indices = np.array(
+            [(el, m) for el in range(min_el, L) for m in range(1, el + 1)]
+        ).T
 
     For `L, min_el = 1024, 0`, this implementation is around 80x quicker in
     benchmarks compared to list-comprehension implementation.