diff --git a/examples/tutorials/audio_data_augmentation_tutorial.py b/examples/tutorials/audio_data_augmentation_tutorial.py
index d191b079595..f11a2471a16 100644
--- a/examples/tutorials/audio_data_augmentation_tutorial.py
+++ b/examples/tutorials/audio_data_augmentation_tutorial.py
@@ -334,7 +334,7 @@ def get_noise_sample(*, resample=None):
 noise_power = noise.norm(p=2)
 
 for snr_db in [20, 10, 3]:
-    snr = math.exp(snr_db / 10)
+    snr = 10 ** (snr_db / 20)
     scale = snr * noise_power / speech_power
     noisy_speech = (scale * speech + noise) / 2