Time Syncing solved without NTP software

[JayStar]

I use the TheSkyX Pro as a planetarium and for telescope and equipment control.  One of the things that bugged me a lot was the inability for Windows to keep accurate time, which meant that my software would always be slightly out in terms of the time and therefore pointing and alignment was not spot on.

Windows syncs its time every week - and at the end of the week it can be quite a bit out of sync.  I tried using software (like Meinberg NTP) to regularly sync my time with NTP servers but kept running into issue with Windows blocking 3rd party tools trying to control system processes.

I eventually found the linked tip on setting the existing windows process to sync every 5 minutes instead of every week.  Works perfectly and no 3rd party software... my PC is now synced every 5 minutes and I am always less than 1 second out from internet NTP servers recorded times. (in fact usually just milliseconds out)

Anyway here is the link for anyone have similar issues trying to get their PC to use 3rd party time sync tools... or just don't want to use 3rd party tools.  I spent ages trying to get a decent syncing solution in place before landing on this  so hopefully it saves some other people some time if they don't have a solution yet.

http://www.pretentiousname.com/timesync/

If anyone has some of the solutions then let me know!  One limitation is that windows can only check one NTP server... I would like to improve that but probably means other software.

[perfrej]

I have spent aeons of time on this subject and the answer is definitely not the Windows Time Service (w32time) which is what sets the windows clock. W32Time is not full NTP, but rather an odd version of SNTP coupled with internal Windows-specific time operations applicable in a Windows Active Directory domain. Real NTP does not sync time instantly, except for the initial "burst sync", and needs a good 24 hours to set your clock accurately. It uses a complex procedure to measure network latencies and response times of all parties involved, and needs to gradually adjust your clock. W32Time does none of that.

I use NTP in my remote observatory in Provence. For me it is needed as I am constantly dealing in the arc-second game, and in low single digits at that.

The Meinberg NTP service is very easy to get going and will be able to sync your PC time to within 0.997 milliseconds. The strange number comes from the way windows handles network packages that NTP needs to measure, and is actually the jitter in the package timing precision. There is a registry change that one can apply to remove that jitter, and at that time you could be synced to within microseconds of UTC. Of all the implementations of NTP that exist for WIndows, Meinberg is the obvious choice as it does not involve Cygwin or any other such "linux emulation", but is written in Windows native code.

For the normal user, I really recommend the Meinberg NTP. Even with a French Orange internet connection (no offence intended) like the one we have in Provence, Meinberg keeps my clock within a few milliseconds of UTC. I have a stratum 1 NTP server with GPS PPS ready that I will install in the observatory, and with that on the local network we will achieve either 0.997 ms or a few micrososeconds depending on registry settings in the windows machines. 

Why is this so important? Well, three of the mounts in the observatory are very accurate and use in-mount modeling, so if the clock in the mount or PC (or both) are off by half a second, the resulting model point is off by 7.5 arc-seconds, which is way off for these instalaltions which usually puts a target within 0 to 5 pixels of the center and tracks 0.1" RMS without guiding.

Anyway, Meinberg it is. If you have any problem getting it to run, you should know that within the 10Micron user community it is the standard for time keeping, and very few have had problems with it. I have it on my seven Windows 7 machines, none of which exhibit any problems. Give it a shot again

/per

Hrere's my remote observatory computer right now:

[JayStar]

Yes I would like to use Meinberg but it seems that it will just not work properly with Windows 8.1.  The service will only start under a system account and once it does, many of the settings do not work.

Just for clarity on the Windows solution I am using pool.ntp.org.....The server you sync with can be set to whatever you want.  I have been checking against the NIST atomic clock and can see I am within milliseconds of the time.

The big advantage of Meinberg is that it can be synced more regularly than 5 minutes and several servers can be checked as a pool.  I will try it again but the windows solution is for those (liek me!) that cant get Meinberg to work.

There is also ClockWatch Pro from Beagle Software

[JamesF]

Oh, Per!  You have such a downer on open source software normally saying it can't meet the standards of quality of commercial software and there you are using it all along for something you deem performance critical!

James

[perfrej]

Haha! Yes, NTP and the other core components of TCP/IP will not be subject to forks and peoply geting tired of supporting it, so it is safe!

As for NTP and how often it is updated... It doesn't work that way and there is no set interval. The algorithm measures drift of the clock and latency in speaking to the time servers. This goes on all the time and the clock is adjusted by small amounts until the drift is minimized. 

The windows time service does a quick roundtrip judgement and a single sync, so it simply canot get it right, simlpe as that.

I do not have any computers running Windows 8.1 as I have chosen to skip it entirely. I do, however, have a couple of laptops (3 to more precise) that run Windows 10, which should be about as difficult as WIndows 8.1 to get running. Install commencing... Will check in in a while.

/per

[JayStar]

Haha! Yes, NTP and the other core components of TCP/IP will not be subject to forks and peoply geting tired of supporting it, so it is safe!

As for NTP and how often it is updated... It doesn't work that way and there is no set interval. The algorithm measures drift of the clock and latency in speaking to the time servers. This goes on all the time and the clock is adjusted by small amounts until the drift is minimized. 

The windows time service does a quick roundtrip judgement and a single sync, so it simply canot get it right, simlpe as that.

I do not have any computers running Windows 8.1 as I have chosen to skip it entirely. I do, however, have a couple of laptops (3 to more precise) that run Windows 10, which should be about as difficult as WIndows 8.1 to get running. Install commencing... Will check in in a while.

/per

Yes I have been reading up a lot on Meinberg's site - think I can grasp what it is doing now.  

I have managed to get it working in Windows 8.1 (to a point).  The NTP Time Server Monitor will not allow the stopping of the service.  I need to actually kill the process in Windows services to stop it.  Starting or restarting the service only works that way also.

I also have a Windows 10 machine so might try it there too.

Thanks

John

[perfrej]

Yes indeed. Wasn't a total success and I had to run the installer twice, but works it does

This is on WIndows 10:

/per

[Earl]

Per is there any advantage to users in general doing AP to do this?

[perfrej]

Just the fact that you get proper time. One minute off in time, 15 arc-minutes off in pointing...

/per

[JayStar]

Well apart from not being able to start or stop the service within the Server Monitor tool (have to use windows service screens for both) I seem to have Meinberg working now and have left it to work on my system for the last few hours.

Will keep it running and measure it against the internal windows syncing process.  Looking good though.

[JayStar]

Per is there any advantage to users in general doing AP to do this?

For me it was just that I am bit obsessive about getting the best pointing possible. I had noticed that my PC clock was drifting by 10-20 seconds a week and therefore TheSkyX Pro software was not using the correct time for pointing and telescope control.

[perfrej]

It is a service that starts at boot time, so it is ALWAYS running, and that is the way it should be. The longer it has been running, the better it is at keeping your PC time accurately synced to the world standard.

I think that you may have to run the tool as administrator in order for it to be able to start and stop the service (which you always o after a config file change but never otherwise). You can also start and stop it from the "Services" administrative tool, which they have tried to hide on 8.1 but is available nonetheless.

You will find this to be something completely different than Windows Time...

/per

[james_screech]

I have spent aeons of time on this subject and the answer is definitely not the Windows Time Service (w32time) which is what sets the windows clock. W32Time is not full NTP, but rather an odd version of SNTP coupled with internal Windows-specific time operations applicable in a Windows Active Directory domain. Real NTP does not sync time instantly, except for the initial "burst sync", and needs a good 24 hours to set your clock accurately. It uses a complex procedure to measure network latencies and response times of all parties involved, and needs to gradually adjust your clock. W32Time does none of that.

I use NTP in my remote observatory in Provence. For me it is needed as I am constantly dealing in the arc-second game, and in low single digits at that.

The Meinberg NTP service is very easy to get going and will be able to sync your PC time to within 0.997 milliseconds. The strange number comes from the way windows handles network packages that NTP needs to measure, and is actually the jitter in the package timing precision. There is a registry change that one can apply to remove that jitter, and at that time you could be synced to within microseconds of UTC. Of all the implementations of NTP that exist for WIndows, Meinberg is the obvious choice as it does not involve Cygwin or any other such "linux emulation", but is written in Windows native code.

For the normal user, I really recommend the Meinberg NTP. Even with a French Orange internet connection (no offence intended) like the one we have in Provence, Meinberg keeps my clock within a few milliseconds of UTC. I have a stratum 1 NTP server with GPS PPS ready that I will install in the observatory, and with that on the local network we will achieve either 0.997 ms or a few micrososeconds depending on registry settings in the windows machines. 

Why is this so important? Well, three of the mounts in the observatory are very accurate and use in-mount modeling, so if the clock in the mount or PC (or both) are off by half a second, the resulting model point is off by 7.5 arc-seconds, which is way off for these instalaltions which usually puts a target within 0 to 5 pixels of the center and tracks 0.1" RMS without guiding.

Anyway, Meinberg it is. If you have any problem getting it to run, you should know that within the 10Micron user community it is the standard for time keeping, and very few have had problems with it. I have it on my seven Windows 7 machines, none of which exhibit any problems. Give it a shot again

/per

Hrere's my remote observatory computer right now:

Mein.PNG

Where did this windows user interface come from? I've just installed Meinberg and the status app just shows a flickering dos box.

James

[steppenwolf]

Where did this windows user interface come from?

You can download this monitor application.

[Earl]

fom what i see this will improve initial slews fro me, I plate solve and use that sync so ill probably not gain much from this?

[JayStar]

Well I think following Per's advice on this has changed this thread - rightly so.  I have had Meinberg running now for a day and can already see it at work improving my time accuracy. The windows syncing is a solution to improving the basic windows syncing perfromance, but it doesn't match Meinberg.

Happy to have it running... my offset time is running as low as 0.9 micro seconds.

For pointing this is just another dimension to improve accuracy I think.  As you say I dont think that it will improve on plate solving - maybe Per can advise on that though as I am only just getting into plate solving.