frequency_nominal / channel mismatch in compute_MVBS

[LOCEANlloydizard]

While testing compute_MVBS on an ONC AZFP dataset (device ASLAZFP55196), a mismatch was observed (thanks @slonimer!) between channel labels, frequencies, and the resulting echograms. The 38, 120, and 200 kHz channels appear to be shifted in the output MVBS dataset.

Dataset information:

• Source: ONC
• Device: ASLAZFP55196
• Data product: AAPTS
• Time span: 2026-06-14T00:00:02Z to 2026-06-14T23:59:57Z

Input Sv dataset:

channel = ['55196-38-1', '55196-120-2', '55196-200-3']
frequency_nominal = [38000., 120000., 200000.]

Output MVBS dataset:

channel = ['55196-38-1', '55196-120-2', '55196-200-3']
frequency_nominal = [120000., 200000., 38000.]

The input ds_Sv appears correct, suggesting the mismatch may occur during MVBS computation or when frequency_nominal is propagated to the output dataset. In this case, frequency_nominal is stored as a (ping_time, channel) variable in the input dataset, which may be related to the issue.

TODO:

• fix
• add test in gthub assets
• add test with this file

Also check notebooks

compute_MVBS is used in several example notebooks. Once this issue is fixed, verify that these notebooks still behave correctly and that channel / frequency_nominal metadata remain consistent:

• Getting started with Echopype
• Watching a solar eclipse using a moored, upward-looking OOI echosounder
• Label-aware subselection of echo data based on ship tracks
• Perform frequency-differencing and masking to compute NASC attributed to krill from ship echosounder data
• Incorporate an external mask to compute NASC attributed to Pacific hake from ship echosounder data
• Seafloor detection