@selene: "what do we do with it?" is really more of a class of questions that depends on the application. Since identifying and addressing it in any form requires some IMO tedious attention to detail, I'm inclined to first ask, "is it causing a problem that needs solving for your application? i.e. if you don't address it, does it change your results/interpretation?" Only if the answer is "yes", then is it worth delving into potential solutions. Otherwise probably not worth addressing (e.g. if it's just to satisfy a compulsion for things to line up nicely/perfectly).

This mismatch between clock times and sample rates (and nominal vs actual sample rate) is a common issue for most acoustic recorders, not just SoundTraps. See related discussion on the forum at bioacoustics.stackexchange.com/questions/1632/…

A bit of a stretch answer with my limited experience with unsupervised classification methods: I've found non-linear multi-dimensional scaling useful to cluster features measured from small data (n=10^4). T-SNE may also be used to cluster spectrograms directly (see for example Jamie Macaulay's SoundSort: wildlabs.net/discussion/…). UMAP is another similar concept with a different underlying implementation. And of course transformers are all the rage these days (e.g. arxiv.org/html/2406.01253v1) so maybe worth checking out animal2vec etc.

Yes, thank you this was indeed an error -- good catch. I have replaced wavFilt with wavFiltData, which I believe should now be correct.

Can you elaborate on your application - specifically why you want to build your own hydrophone instead of buying a tried and tested existing product? Is this a hobby? personal learning? engineering, art or science project? Have you read the existing threads here on building your own hydrophones with some references to and discussion of sensors (bioacoustics.stackexchange.com/questions/1269/…; bioacoustics.stackexchange.com/questions/1313/…)?

Can you share what you've already tried? From the seewave documentation, it seems that bwfilter allows for the specification of channel. Might be informative if you try passing a list of channels (e.g. channels = list(1,2) or similar) and share the results. Even if bwfilter can't handle multiple channels in a single call, this suggests at the very least that one could call bwfilter for each channel individually.

This seems like a question that is either best posed to the authors of the respective software, or directly answered yourself. One could answer it themself by either inspecting the code (all packages appear to have source code available), and/or running the same dataset through each software package and comparing the results.

I think the questions posed here might be a bit too general and open-ended for the StackExchange format. The question of "what else should I take into consideration?" sounds like it might be better answered via a formal course on- or literature review of ecoacoustics. Perhaps if you could clarify why you think acoustic diversity and spectral diversity should agree (e.g. provide a reference, provide some data, explain what and where you are recording, or post some data and results) you might be more likely to get a response.