Open j9ac9k opened 4 years ago
The assumption is that folks are ok with loosing the additional precision that would be necessary to capture the details of the resampling. I'm ok with making this assumption. Maybe we should have explicit bit-depth converters between 16 (int16), 32 (float32), and 64 (float64) bits.
Description
When calling tracking.wave.resample(some_freq), I get a new track back, but the dtype of the new_track was floating point (when the original track dtype was int16). This of course causes all sorts of havok with wave.min and wave.max, and thus plotting
Here is the code for resampling:
https://github.com/TimeViewers/signalworks/blob/master/signalworks/tracking/wave.py#L189-L202
In the end, it calls the Wave constructor, with the new values from
resample_poly
.The thing to do here would be to convert the returned values into the dtype of the original array, something along the lines of