Use channel offsets, not gains, when scaling amplitudes to uV

src/spikeinterface/postprocessing/amplitude_scalings.py

@@ -179,7 +179,7 @@ def __init__(
         if return_in_uV and recording.has_scaleable_traces():
             self._dtype = np.float32
             self._gains = recording.get_channel_gains()
-            self._offsets = recording.get_channel_gains()
+            self._offsets = recording.get_channel_offsets()
         else:
             self._dtype = recording.get_dtype()
             self._gains = None

src/spikeinterface/postprocessing/spike_amplitudes.py

@@ -69,7 +69,7 @@ def __init__(
         if return_in_uV and recording.has_scaleable_traces():
             self._dtype = np.float32
             self._gains = recording.get_channel_gains()
-            self._offsets = recording.get_channel_gains()
+            self._offsets = recording.get_channel_offsets()
         else:
             self._dtype = recording.get_dtype()
             self._gains = None

src/spikeinterface/postprocessing/tests/test_spike_amplitudes.py

@@ -1,8 +1,35 @@
+import numpy as np
+
+from spikeinterface.core import NumpyRecording, NumpySorting
+from spikeinterface.core.node_pipeline import SpikeRetriever
 from spikeinterface.postprocessing import ComputeSpikeAmplitudes
 from spikeinterface.postprocessing.tests.common_extension_tests import AnalyzerExtensionCommonTestSuite
+from spikeinterface.postprocessing.spike_amplitudes import SpikeAmplitudeNode
 
 
 class TestComputeSpikeAmplitudes(AnalyzerExtensionCommonTestSuite):
 
     def test_extension(self):
         self.run_extension_tests(ComputeSpikeAmplitudes, params=dict())
+
+    def test_scaled_amplitude_uses_channel_offset(self):
+        traces = np.zeros((200, 2), dtype=np.int16)
+        traces[100, 0] = 10
+        recording = NumpyRecording([traces], sampling_frequency=1000)
+        recording.set_property("gain_to_uV", [2.0, 3.0])
+        recording.set_property("offset_to_uV", [-100.0, -200.0])
+        sorting = NumpySorting.from_times_and_labels(
+            [np.array([100], dtype=np.int64)],
+            [np.array([1], dtype=np.int64)],
+            sampling_frequency=1000,
+            unit_ids=[1],
+        )
+        sorting.set_property("main_channel_id", [0])
+        spike_retriever = SpikeRetriever(sorting, recording, channel_from_template=True)
+        node = SpikeAmplitudeNode(recording, parents=[spike_retriever], peak_shifts={1: 0}, return_in_uV=True)
+        peaks = np.zeros(1, dtype=[("sample_index", "int64"), ("unit_index", "int64"), ("channel_index", "int64")])
+        peaks[0] = (100, 0, 0)
+
+        amplitudes = node.compute(traces, peaks)
+
+        np.testing.assert_array_equal(amplitudes, [-80.0])