Spectrogram (SSID 24, L4) timing information issues

[samaloney]

While investigating some potential FSW bugs it was discovered that the spectrogram bin edges derived from the fits files are not contiguous, there are small under and overlaps.

• Confirmed this is present in the raw TM so not ground processing issue
• Investigation found the reported closing time offset (NIX00029) is not accurate between science substructures
• Simple fix was to stack all start times and only use the closing time offset (NIX00029) for the last science substructures
• Unclear if closing time offset (NIX00029) should be valid for all substructures or only last but don't need to know for this fix

[samaloney]

I've reviewed some recent and older spectrogram data and there seems to be an unexpected shifts of ~0.04s +/- 0.001/2. There are other larger shifts which seem to be related to non-data gaps and non-overlapping requests.

Need to check source of offset is in:

• the requests (IORs)
• the TM
• the fits

q = Fido.search(a.Time('2025-05-01T00:00', '2025-06-01T00:00'), 
                a.Instrument.stix, a.stix.DataProduct.sci_xray_spec)
spec = Product(q[0]['url'].tolist())

diffs = []
for i in range(len(spec)-1):
    last_match = np.argmin(np.abs(spec[i].times[-1] - spec[i+1].times))
    diff = (spec[i].times[-1] - spec[i+1].times[last_match]).to('s')
    diffs.append(diff)
    print(i, last_match, diff)

plt.plot([s.time_range.start.datetime for s in spec[:-1]], [d.value for d in diffs])