PPS is not aligned for a 50km link
Peter Jansweijer wrote to the white-rabbit-dev:
Currently I'm testing with a fiber of 50 Km between a switch (v5.0) and
a SPEC (v4.1). The PPS offset between switch and SPEC is ~196 usHuge
I didn't have this kind of offset when using 8 Km of fiber. It should be
noted that I didn't calibrate the link at all (no delta tx, rx nor
alpha) so I expect numbers in the order of a few tenth of ns.
I downloaded older binaries from the web
(https://www.ohwr.org/project/wr-cores/wikis/Previous_releases) to see
how previous versions behave. Here is the outcome:
8Km
50Km
50+8Km
pps [ns] crtt [ps] pps [ns] crtt [ps] pps [ns] crtt [ps]
wrpc-v3.0 -20 82529330 -48 491643739 -54 573758947
wrpc-v4.0 -20 82529367 196560 491643595 262090 573758868
wrpc-v4.1 -20 82529508 196559 491644095 262090 573759025
Conclusions:
- wrpc v3.0 looks fine! (different pps offsets are due to alpha being set to 0)
- all crtt times look fine and are in the range you would expect for
all versions and all fiber lengths.
- in version v4.0 and v4.1 there is a weird pps offset jump when the
fiber length exceeds some number between 8 Km and 50 Km.
My gut feeling:
Since all crtt's look fine there is probably an arithmetic overflow or
round that spoils the pps offset calculation in v4.
Maybe someone can at least verify my observation in an independent setup.
Cheers,
Peter