BUG - CoDICE L3a Lo line plots (epoch deltas?) (Phase D)

[jtniehof]

Description of the issue

Line plotting the CoDICE L3a Lo products in CAVA leaves many data points unconnected to their neighbors.

The epoch_delta appears to create time bins that are noncontiguous by a few milliseconds. I don't know if that's enough to trigger CAVA's detection of missing data. This might be a CAVA problem if this accurately represents the CoDICE sampling of the data. Or there might be a rounding or something in the L2 that needs to be looked into.

Affects partial densities and ratios most obviously. The charge state distributions might be affected but don't manifest in the same way on a spectrogram.

Steps to reproduce the issue

Plot imap_codice_l3a_lo-sw-ratios_20260121_v003.cdf in CAVA; I was using o7_to_o6_ratio. Turn off log plot to remove the "goes to zero" problem for diagnosis.

with spacepy.pycdf.CDF('imap_codice_l3a_lo-sw-ratios_20260121_v003.cdf') as f:
    e = f.raw_var('epoch')[...]
    de = f['epoch_delta'][...]
expected = e + de
err = numpy.abs(expected[:-1] - e[1:])
print(numpy.max(err) / 1e6)
print(numpy.median(err) / 1e6)

Expected vs Actual behavior

No response

Code Snippet (If applicable)

Code

Additional notes, affected areas, and suggested fixes

No response

[pleasant-menlo]

CAVA does have a couple places where we (arbitrarily) chose 1ms as the limit for connecting lines:

https://github.com/pleasant-menlo/cava/blob/910184c82109afcd98a3bfcd61bb564221f1aea2/cava/science/visualization/timeseries/timeseries.py#L127

https://github.com/pleasant-menlo/cava/blob/910184c82109afcd98a3bfcd61bb564221f1aea2/cava/science/visualization/formatted_cava_data_segment.py#L109

[pleasant-menlo]

@jtniehof - we have done an initial investigation. What's happening is that in the L2 SW Species, the epoch_delta_plus and epoch_delta_minus values are symmetrical at each distinct epoch, but differ when comparing them from epoch to epoch. We propagate the epoch and epoch_deltas directly from the L2 data. In the instance where we are plotting the L3a ratios, the gaps/overlaps are exceeding the 1ms buffer (see above) and are not drawing lines between points.

Is this the expectation for what the epoch_deltas will look like at L2? If so, we could accommodate this in CAVA by increasing the buffer period, to 5 or 10ms for example. Otherwise, we may be able to handle it during data processing, although we are generally hesitant to modify the epoch/epoch_delta values coming in from the L2 data.

Let us know what you think,
Logan & Ethan

[jtniehof]

How complicated would it be to make the tolerance relative (e.g. 1% of the mean epoch_delta) instead of absolute?

For most of our measurements, 10ms would be perfectly reasonable--a 1s cadence is very fast and that's 1%. But I want to allow for the possibility of some bursty stuff (e.g. magnetometer) so a %age would be better.

Agreed we don't want to mess with the timestamps coming from L2.

[pleasant-menlo]

@jtniehof It wouldn't be that complicated - we estimate 4 hours. Should we get started on this?

Is this an issue that the SDC should be aware of for CoDICE L2 processing?

Logan & Ethan

[jtniehof]

Check in with Jamey for authorization to proceed on CAVA, but I'm good with it.

I'll take a look at the L2. I suspect there is a reason for it.

[pleasant-menlo]

@jtniehof We've made this change in CAVA. Here is a screenshot of what that data looks like now:

~ Logan and Ethan