[ntp:questions] DCF/SHM calibration problems

Helmut Wollmersdorfer helmut at wollmersdorfer.at
Fri Oct 8 10:20:45 UTC 2004


For calibration and test of two DCF77 receivers I had them attached to 
my workstation (CPU: AMD Athlon(tm) XP 2500+, bogomips: 3643.80, Kernel 
2.4.26-1-386) and got the following values

# fudge to time1 after calibration (peer.awk was -41.933)
fudge 127.127.28.0 time1 0.041933
# fudge to time1 after calibration (peer.awk was -46.114)
fudge 127.127.28.1 time1 0.046114

Now I have them connected to the nodes of my cluster, consisting of two 
machines with identical hardware
(CPU: VIA Samuel 2, bogomips: 1183.74, Kernel 2.6.8-1-386). NTP now runs 
for some days against the same two stratum 1 servers I used on my 
workstation before for calibration, but I get very different values:

via1:~# ntpq
ntpq> peers
      remote           refid      st t when poll reach   delay   offset 
  jitter
=============================================================================
  mail.happy-fish .INIT.          16 u  14h 1024    0    0.000    0.000 
4000.00
xntp1.ien.it     .IEN.            1 u  573 1024  377  332.787  -94.915 
16.721
*212-82-32-15.ip .PPS.            1 u  588 1024  377   84.259   29.927 
  5.178
+SHM(0)          SHM(0)           5 l  247 1024  377    0.000  -19.142 
  0.460
+nav.metrologie. 71.80.83.0      14 u   16   64  377   14.884   37.766 
19.735
  LOCAL(0)        LOCAL(0)        13 l   45   64  377    0.000    0.000 
   0.004

via2:~# ntpq
ntpq> peers
      remote           refid      st t when poll reach   delay   offset 
  jitter
==============================================================================
+ruou.ifi.unizh. 129.132.2.21     2 u  841 1024  377   46.083   26.967 
  1.719
xntp1.ien.it     .IEN.            1 u  856 1024  373  385.893  -17.204 
70.084
*212-82-32-15.ip .PPS.            1 u  897 1024  377  143.339   55.047 
83.569
+SHM(0)          SHM(0)           5 l   46   64  377    0.000   -4.704 
  0.397
-nav.metrologie. 71.80.83.0      14 u   52   64  377   14.264   35.692 
21.256
  LOCAL(0)        LOCAL(0)        13 l   26   64  377    0.000    0.000 
   0.004

First, the difference to the workstation is hardware, kernel and ttyS1 
instead of ttyS0. Maybe kernel 2.6 produces shorter offsets as 2.4?

Second, signal lengths of the two receivers differ systematically around 
  ~5-9 ms. This maybe explains a part of the differences.

See here for the same moments:

via2:~# radioclkd -t ttyS1
[...]
DCD:  51    1    195246
DCD:  52    0     95524
DCD:  53    0     98436
DCD:  54    1    195326

via1:~# radioclkd -t ttyS1
[...]
DCD:  51    1    189162
DCD:  52    0     90625
DCD:  53    0     90749
DCD:  54    1    187530

Can somebody give me examples, how large typical offsets and signal 
lengths should be for radio clocks? The part of the radio signal delay 
for my distance of 600 km should theoretically be 1.8 ms.

TIA

Helmut Wollmersdorfer



More information about the questions mailing list