BIPM only goes up to 2021, and may not be fully correct

[dlakaplan]

@vhaasteren noted:

When running the update_clock_corrections.py script, I got a warning that tai2tt_bipm2022.clk was missing. Looking into it, it seems that tai2tt_bipm2021.clk andtai2tt_bipm2020.clk are actually copies of tai2tt_bipm2019.clk. I don't think that's supposed to happen? Or is it?

This may need a little deeper dive, since there is some confusion about those files. But I think this can be fixed within this repo.

[dlakaplan]

@jorisverbiest

[vhaasteren]

The workaround I used in the Dockerfile that I was building at the time consisted of just copying tai2tt_bipm2019.clk to upcoming years, and letting pulsar-clock-corrections update it properly. It seems to work.

WORKDIR $SOFTWARE_DIR/pulsar-clock-corrections
RUN mkdir -p gh-pages/.this_is_gh_pages && \
    for year in $(seq 2022 $(( $(date +%Y) - 1 ))); do \
        cp T2runtime/clock/tai2tt_bipm2019.clk T2runtime/clock/tai2tt_bipm${year}.clk; \
    done

[dlakaplan]

Is https://www.bipm.org/en/time-ftp/tt-bipm- the correct source location?

[vhaasteren]

It looks like it's in these locations now: https://webtai.bipm.org/ftp/pub/tai/ttbipm/TTBIPM.2022

[dlakaplan]

I am also a little confused because:

• the final dates in e.g., BIPM2021 is 2024-09-26 MJD 60579.0, but https://webtai.bipm.org/ftp/pub/tai/ttbipm/TTBIPM.2021 only goes up to 2021-12-31. Probably this is because it adds prediction data to smooth out newer TOAs. I think this follows lines like:

  Until the next realization, users can extend TT(BIPM22) after MJD 59939 as
  TT(BIPM22) = TAI + 32.184 s + 27670.6 ns - 0.01x(MJD-59939) ns 

  and the header does say this, so maybe we just need to understand how these are computed.

• Separately, https://github.com/ipta/pulsar-clock-corrections/blob/main/bipm.py#L289 looks like it does try to get newer files but hasn't: maybe just because this hasn't been run since April?

[dlakaplan]

If I run it locally I do pull down BIPM2022 but other things fail because the file doesn't already exist and neither does a relevant log file. If I copy in a stub then things run but I'm not sure all of the appropriate files are actually updated.

[dlakaplan]

I couldn't convince this to automatically update the file, but I was able to create the appropriate file using:

import bipm

bipm2022 = bipm.get_ttbipmxy_corrections_clock(2022)
bipm2022.write_tempo2_clock_file('tai2tt_bipm2022.clk')

Along with that there was a bug(?) in the code that made the top comment in BIPM 2020/2021/2022 all read BIPM 2019.

[dlakaplan]

And I also agree that BIPM2020 seemed to be the same as BIPM2021 (which looks OK).

[dlakaplan]

Closed by #11