Record sampling rate as part of the Electrical series

[pickles-bread-and-butter]

This attribute is really a feature of the voltage measurements themselves. Also, it is possible for a probe to collect data at varying sampling rates.

AllenSDK/allensdk/brain_observatory/ecephys/nwb/AIBS_ecephys_extension.yaml

Lines 16 to 18 in a040661

	- name: sampling_rate
	dtype: float64
	doc: the sampling rate at which data were acquired on this probe's channels

Split off from: #1330

[njmei]

Hi, @rly @bendichter,

I tried implementing this suggestion, but it looks like ElectricalSeries does not allow both rate and timestamps to be defined at the same time. It looks like the error is being raised by:

https://github.com/NeurodataWithoutBorders/pynwb/blob/78ded2eb46d47a8dc400172fbbf97e0be4dbf7aa/src/pynwb/base.py#L146

We would like to store information about the lfp sampling rate without having to re-calculate that sampling rate from the LFP ElectricalSeries timestamps.

[bendichter]

@njmei We recommend storing data with a constant sampling rate using the starting_time and rate. I agree that it is cumbersome to calculate timestamps on the fly. I maintain this helper function to do this when I need to. We have discussed adding this type of functionality to pynwb or hdmf here. Right now, nothing exists within the current APIs for computing this on the fly. If you really need to store timestamps that would be permissible by NWB- it just would not be what we recommend.

If you can apply compression, that would help somewhat by making the timestamps should be very small.