[LinuxPPS] DCF77-PPS

[Remco den Besten]

On my blog (remco.org) I published the idea to use DCF77 pulses as PPS 
source. I did some experiments,
found out that PHK already figured this out, and was a bit disappointed by 
the performance then.

However, a few months ago I gave a friend of mine a DCF77-receiver with the 
instructions 'plug it in' , direct it 'so and so', and I remotely installed 
and configured his machine to add him to the ntp.pool. Although the machine 
performed well,
I was intrigued with the posting(s) of Hal concerning the timer options. I 
found out that the machine of my friend has a 'pit' timer, and I remotely 
compiled and upgraded a kernel (2.6.8 to 2.6.27.5 <- ouch! ;-) with LinuxPPS 
and compiled
ntpd with STA_NANO. Although DCF77 misses the 59th second, I think the whole 
contrapsion behaves relatively
well. The offset and jitter remain within the 1 ms range. Results can be 
seen at remco.org/ntp.

Helium = LinuxPPS (kernel 2.6.27.4) with Oncore UT+ and acpi_pm timer (very 
temp dependent!)
Lithium = normal kernel with DCF77 via radioclkd2
FreeBSD = eh.. FreeBSD 7.0 (and sorry... behaves better here than LinuxPPS + 
nano)
ntp2.remco.org = DCF77-LinuxPPS (kernel 2.6.27.5) + STA_NANO

Be quick, this weekend I have a party at this home, and I will connect an 
Oncore VP GPS receiver ;-)

What is also very visible, is that helium and freebsd are located in the 
same room (20 cm seperated ;-) but react totally different on the (same) 
temperature variations. The outcome (i.e. offset and jitter) of PLL 
(in)stabilities of LinuxPPS vs
FreeBSD is striking.