To try and reply in tandem here...
David Taylor and Dave Hart:
David Taylor wrote:
I don't know whether this is relevant, but on my system I needed to
add "flag3 1" to tell NTP to use the kernel-mode PPS:
I've tried 'flag3 1' before on my PPS refclock with the Oncore and it's
really flaky with it. At least with these Oncore receivers, I've always
controlled the PPS in the ntp.oncore.<n> config file and setting the PPS
on-second preference to either
server 127.127.22.0 minpoll 4 maxpoll 4
fudge 127.127.22.0 flag3 1 refid PPS
Also, for PPS clocks, I thought that minpoll and maxpoll should be
set to 4, but perhaps that is now an out-of-date recommendation.
I left them at 6 with the testing I have been doing. Polling more
frequently may be an idea to get more samples to make the algorithm sway
the in favor of the 1PPS refclock. I thought extending the polling out
would be better, and give me time time to have the GPS/PPS refclock
stabalize instead of throwing really high samples into the mix.
*shrugs* Just a guess.
However, thanks for the tip to convert that PPS CMOS TTL logic level
over to RS-232. I did put a MAX232 in between that signal and PPSAPI
still picked it up as a la 'good' PPS source, but at least I'm keeping
sane logic levels now between mediums now:
---------------
trying PPS source "/dev/pps0"
found PPS source "/dev/pps0"
ok, found 1 source(s), now start fetching data...
source 0 - assert 1352670921.396327477, sequence: 108709 - clear
1352670922.197455262, sequence: 108293
source 0 - assert 1352670922.396377713, sequence: 108710 - clear
1352670922.197455262, sequence: 108293
source 0 - assert 1352670922.396377713, sequence: 108710 - clear
1352670923.197455145, sequence: 108294
source 0 - assert 1352670923.396431911, sequence: 108711 - clear
1352670923.197455145, sequence: 108294
source 0 - assert 1352670923.396431911, sequence: 108711 - clear
1352670924.197453704, sequence: 108295
source 0 - assert 1352670924.395480092, sequence: 108712 - clear
1352670924.197453704, sequence: 108295
source 0 - assert 1352670924.395480092, sequence: 108712 - clear
1352670925.197454189, sequence: 108296
source 0 - assert 1352670925.395532695, sequence: 108713 - clear
1352670925.197454189, sequence: 108296
----------------
...but I still kept getting a falseticker.
Dave Hart wrote:
I would suggest removing the external S1 servers from the
configuration to see if their relatively wide range of offsets is a
contributing factor. If the PPS is still excluded as a falseticker.
Ok. I did try this and it seemed to help. So them I started playing
with 'tos mindist' some more without much luck.
### Kernel PPS Selection
enable pps
pps /dev/oncore.pps.0 hardpps
I believe both of the above lines are ignored by your recent 4.2.7
ntpd. There should be syntax errors in syslog at each ntpd startup
confirming that.
Ya, I jumped from ntpd 4.2.4-stable, ntp-dev-4.2.5p<n> and now at
ntp-4.2.7. I know it's obsolete. I'll purge it once I get a stable new
config I can live with.
Also, this isn't the best list for the discussion. If the above
suggestions don't help, please consider posting to
[email protected] a similar message describing the modified
configuration as a starting point (no external servers, minpoll 4 on
the refclocks).
Right on. My apologies for that, Mr Hart. So putting my last 'plug'
onto this problem before I continue this on the correct list, I did come
across a post you made, Mr. Hart, on the google comp.protocols.time.ntp
board
(https://groups.google.com/forum/?fromgroups=#!topic/comp.protocols.time.ntp/9lmJo66BSLU)
and feel as though I'm battling this problem.
Before reading that post, I had almost opted to to use the 'true'
option to just force it to be truechimer status and screw it. After
reading this and adding 'prefer' to the PPS refclock, I have assumed
pps.peer status with PPS now.
However, having other external server peers listed does seem to make my
PPS a falseticker, it does show up as peer once in awhile with or
without the external servers listed as well.
-Adam
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool
