[time-nuts] beaglebone black, debian, NTP client
David J Taylor
david-taylor at blueyonder.co.uk
Wed Jan 6 04:57:14 EST 2016
That's one thing that annoys me with those graphs. If you average jitter
it loses its meaning. What you then get is the mean deviation (aka offset).
Without an accompanying standard deviation (and a test that you actually
have a gausian distribution) this value is not worth much.
What I am talking about is
The before ("full" OS) and after ("console" OS) is strange by itself.
What kind of process is running that increases interrupt latency jitter
by a factor of 2-3? Why does the "console" OS still exhibit a jitter that
is a factor 2 to 3 higher than what i'd expect as interrupt jitter?
PS: could you please quote mails properly? It makes them much easier to
The jitter is not averaged, and the RMS value is given in:
The Console BBB OS was recommended here. Likely the two were:
The GPS has a jitter well under a a tenth of a microsecond, so someone else
will need to investigate why the jitter does not meet you expected levels,
as I have abandoned the BBB as it created too much RF interference, and has
far less support than the Raspberry Pi. Timekeeping on the RPi is quite
good enough for my purposes.
Unfortunately, my e-mail software makes it very difficult to do proper
quoting. I always do correct quoting with NNTP using Thunderbird, but I'm
unwilling to start putting chevrons at the start of every quoted line in
e-mail with my inferior e-mail software. Perhaps one day I'll change e-mail
software - it annoys me as well. At least it doesn't top-post, or use HTML!
SatSignal Software - Quality software written to your requirements
Email: david-taylor at blueyonder.co.uk
More information about the time-nuts