Merge EDI input work

1 files changed, 30 insertions, 0 deletions

diff --git a/doc/TIMESTAMPS.rst b/doc/TIMESTAMPS.rst
new file mode 100644
index 0000000..458f462
--- /dev/null
+++ b/doc/TIMESTAMPS.rst
@@ -0,0 +1,30 @@
+Some knowledge accumulated about timestamping
+=============================================
+
+The meaning of the timestamps changed between v2.3.1 and v3.0.0, this document gives some guidance about the interaction between different settings.
+
+The following table tries to summarise the differences.
+
++-----------------------------+----------------------------------------------+-------------------------------------+-----------------------------------------------+
+| ODR-DabMux version          | Meaning of timestamp inside EDI              | ODR-ZMQ2EDI wait time w             | Offset that should be set in the mod          |
++=============================+==============================================+=====================================+===============================================+
+| Up to and including v2.3.1  | t_frame = t_mux (No offset in mux available) | positive, meaning delay after t_mux | Something larger than w + mod processing time |
+| Later than v2.3.1           | t_frame = t_tx = t_mux + tist_offset         | negative, meaning delay before t_tx | Something larger than mod processing time     |
++-----------------------------+----------------------------------------------+-------------------------------------+-----------------------------------------------+
+
+For historical reasons, ODR-DabMod decodes absolute timestamp from MNSC, not from “EDI seconds”.
+The edilib tool decodes both EDI timestamp and MNSC, and can be used to verify both are identical.
+
+Issues in ODR-DabMux v2.3.1
+---------------------------
+
+Running ODR-DabMux against the absolute timestamp firmware has uncovered a few issues:
+
+ * At startup, the UTCO was not properly applied to the EDI seconds. This offset was 5 seconds (TAI-UTC offset - 32s, see EDI spec);
+ * odr-zmq2edi did not compensate for UTCO, hiding the above issue;
+ * ODR-DabMux needs a configurable offset;
+ * (minor) MNSC and EDI timestamps did not use the same internal representation, making it difficult to prove that they encode the same value;
+ * (minor) odr-zmq2edi swapped endianness when regenerating EDI from ETI (minor because only ODR-DabMod considers MNSC, and usually isn't used with EDI);
+
+**Important** Do not combine odr-zmq2edi with odr-dabmux of a different version!
+