Reading mseed file with bad timestamp

cpaitor · January 21, 2026, 8:49am

Hi

Unfortunately there seems to be a bug in the software of one of our digitizer that resulted in a timestamp 2025-366 (year-julday) for the the first day of this year. Trying to read this file using obspy fails with the error

...
 File "/usr/local/lib/python3.9/dist-packages/obspy/core/utcdatetime.py", line 432, in __init__
    raise ValueError(msg)
ValueError: 'julday' out of bounds for year 2025: 366

Which of course is correct behaviour. Question though is if it is possible somehow then, using obspy, to read in the file and correct the date (we have no reasons to believe that the timing is bad, only the date seems erroneous) so that the corrected file can be written.

If, not, info on what other software to use will be appreciated.

ThomasLecocq · January 21, 2026, 3:42pm

you could probably edit the io.mseed code to intercept that julian day and modifiy it, e.g. somewhere like:

github.com/obspy/obspy

obspy/io/mseed/util.py

a1953b9ea


      
          info["network"] = _decode_header_field("network", data[10:12].strip())
          
          # Use the date to figure out the byte order.
          file_object.seek(record_start + 20, 0)
          # Capital letters indicate unsigned quantities.
          data = file_object.read(28)
          
          def fmt(s):
              return '%sHHBBBxHHhhBBBxlxxH' % s
          
          def _parse_time(values):
              if not (1 <= values[1] <= 366):
                  msg = 'julday out of bounds (wrong endian?): {!s}'.format(
                      values[1])
                  raise InternalMSEEDParseTimeError(msg)
              # The spec says values[5] (.0001 seconds) must be between 0-9999 but
              # we've  encountered files which have a value of 10000. We interpret
              # this as an additional second. The approach here is general enough
              # to work for any value of values[5].
              msec = values[5] * 100
              offset = msec // 1000000

cpaitor · January 21, 2026, 6:30pm

That’s an alternative but I was sort of hopping that there was a quick way out of this.

But looking at the peace of code you cite, isn’t that a bug we are looking at (the if condition at line 672) since for leap years 366 is a valid day number every leap year?

ThomasLecocq · January 21, 2026, 6:43pm

It looks pretty quick to me, just add a condition there and do -1 for your wrong data.

Regarding your question, I’m not getting what you mean?

cpaitor · January 21, 2026, 6:52pm

In line 672:

if not (1 <= values[1] <= 366):

Would trigger e.g. for december 31 2024 which has day number 366, wouldn’t it?

ThomasLecocq · January 21, 2026, 7:12pm

euh ? no ? if value[1] = 366 will make the check be True and it’ll pass ?

cpaitor · January 21, 2026, 7:24pm

Hmm, yes, just forget about this walk astray, think the ski-trip earlier today consumed most oxygen in my grey sludge…

megies · January 23, 2026, 3:18pm

@cpaitor it is relatively simple to fix fixed headers in MSEED files, so you could “repair” your file manually.

e.g.

from struct import unpack
data = open("my_file.mseed", "rb"), read()
print(unpack('>2H', data[20:24]))

Now what you can do is go through your file and put back together the byte data as is and just replace those 2x2 bytes that encode year and day-of-year with the fixed ones and write back to a repaired file.

Usually MSEED files will be quite well behaved, so just hopping around with record length (512, 2048, …) should be fine and for every record you could add a quick check by just doing a obspy.read(io.BytesIO(data[x:x+record_length]), 'MSEED') and see if it makes sense.