convbin bug - SBF MeasEpoch (Meas2) → RINEX: Doppler missing in v2.5.0

[miguel-ribot-skydio]

When converting Septentrio SBF files that use the classic MeasEpoch/MeasExtra (“Meas2”) blocks to RINEX with v2.5.0, Doppler observations are not written (or entire epochs appear to be missing, depending on how the data is inspected). The same SBF file converts correctly with b34i.

The root cause seems to be that decode_measepoch() in septentrio.c buffers observations while waiting for an EndOfMeas / ID_MEASEPOCH_END (block 5922) marker, but that marker is not present in Meas2 data streams. As a result, the observation buffer is never flushed for those epochs, so Doppler is never output for Meas2 logs.

Proposed fix (high level):

The approach I tested locally (which fixes the issue for my Meas2 data) is:
1. Flush immediately after decoding MeasEpoch (Meas2)
In decode_measepoch(raw_t *raw), after populating raw->obuf / raw->obuf.n, call flushobuf(raw) before returning for MeasEpoch-style data.
2. Do not flush in Meas3 extension blocks
In decode_meas3Doppler() and decode_meas3CN(), remove or avoid the time-based flush (timediff(...) != 0 -> flushobuf(raw)).

I’m not completely sure this is the best way to handle the issue, so I’d really appreciate feedback. I’m happy to open a PR with this patch if it seems reasonable.

[JensReimann]

Thank you for reporting the issue.
The idea behind this logic is to allow for both, "Meas2" an "Meas3" data blocks. If one flushes already at the end of decode_measepoch(), there is no chance to add Doppler data through decode_measepochextra() or data from meas3 data blocks.

The main question is, why is flushobuf() not triggered in decode_measepoch(). It should detect a new time-tag and than call flushobuf() to commit the data including the Doppler information. I saw, that this is the only time the code tests for the time difference to be greater than zero while at other locations it tests for inequality to zero. But this shouldn't cause the problem.

Can you figure out, why time difference test in decode_measepoch() (septentrio.c:705) is not triggered as it is supposed to do?

Best regards,
Jens

[ourairquality]

The following branch attempted to address some issues with the septentrio decoder. It uses fabs() to compare these times, and perhaps that helps. https://github.com/ourairquality/RTKLIB/blob/oaq/src/rcv/septentrio.c#L857 Even if the EndOfMeas blocks are ignored it appears to make little difference to the output. If you control the receiver configuration then might the EndOfMeas blocks be enabled to address this - it does not appear to be specific to Meas3 and appears to be usable with Meas2 and will also help flush the data through with less latency. Otherwise perhaps a hack flush after MeasExtra would suit your data stream.