locked Re: Clock Sync #Timesync


Jim Pennino
 

Re time from a network: Unless you are on the local network of USNO, no network time will be of equal accuracy to a $15 GPS puck.

Overkill: Yes, ntp is overkill for amateur digital modes. However, it the only method that costs nothing, you only set it up once and forget it, and if you operate remote/portable, $15 makes it work in the middle of nowhere. That it is far better than required by WSJT is just frosting on the cake.

minpoll: I don't find setting minpoll to less than 4 has any advantage and just takes up more CPU cycles. That is also the opinion of the developers of ntp which is why for most distributions the minimum accepted is 4. The setting of minpoll has nothing to do with how often ntp evaluates GPS data, that happens once per second. The minpoll for a GPS source is the time between querying that internal, once per second, evaluation for a result.

gpsd: Yes, the raspberry versions of LOTS of things are a bit behind the Linux versions. However the utilities provided with gpsd were never intended to monitor the system clock and IMHO useless for this task. Once you have verified you have a good view of the sky with xgps or cgps, there is no reason, other than idle curiosity to ever run it again.

Monitoring ntp: For windows, the simplest thing to do is to copy the quick ntp status icon to the desktop and click on it once in a while.

I currently have 5 *nix GPS servers to keep track of. I use simple awk scripts which get executed daily and the results mailed to me. I can execute all of them remotely from the command line at any time. This also a gnuplot script to generate graphs of system offset and error that I can execute remotely. All these use the data from the ntp loopstats and peerstats log files. None of this was done for anything related to amateur radio but because I have a personal interest in the state of the art of ntp. If anyone is interested in the scripts, mail me directly and I will send them to you with some comments about them.

cesium clock: Cesium, rubidium, etc. clocks are not real clocks in that they have no idea of what the current time is as they are just high precision oscillators. If you buy something called a rubidium or cesium clock, what you get is a rubidium or cesium oscillator with a GPS receiver. The device uses the GPS information in essentially what is a phase locked loop configuration to keep the oscillator stable, divides the oscillator output down to 1 Hz, and provides the raw GPS data with the oscillator derived high accuracy 1 Hz signal to replace the GPS PPS signal. You then connect this device to a computer running ntp to get actual time. Brand new rubidium clocks go for about $800 on ebay and about $180 for an OCXO clock.

Windows native time keeping: Yes, recent Windows releases allow specifying several ntp servers, but not an attached GPS. The windows timekeeping does not run continuously and the system time occasionally jumps around and how badly is dependent on your hardware. It is also subject to the latest Windows update. And since it can't use GPS, will not work for remote/portable operation. Raise your hand if you have never had a Windows update screw something up for you.

ntp with GPS and other servers: I have found that in setting up ntp with GPS that having network ntp servers is usefully in calibrating the GPS offset. However once that is done network servers cause "noise" in the clock accuracy and should be commented out of the ntp configuration.

Jim WB6DKH

Join main@WSJTX.groups.io to automatically receive all group messages.