Re: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
Hi Esa, Sorry, I haven't tracked this group for a while. Too much work :-) You are intuitive. I expect that the TS21 is triggering on the wrong edge. Way, way back when the TS21 was invented, we were using an external Acutime antenna that had an open collector pps, hence falling edge trigger. When we added support for an onboard gps, they muxed the two 1pps sources together and we had an incompatible trigger logic problem. The way I fixed it at the time was to empirically measure the pulse width and use the antenna delay compensation field to "correct" the pps. This value was stored in e2 at mfg time based on gps receiver module. At some point, your unit must have not been set or been set to incorrect value or had its gps receiver replaced. Is there an antenna prop delay command available you can use? Greg Dowd | Symmetricom®, Inc. Staff Scientist 2300 Orchard Parkway, San Jose, CA 95131 Direct: 408.964.7643 gd...@symmetricom.com | www.symmetricom.com Symmetricom. Leading the world in precise time solutions. -Original Message- From: time-nuts-boun...@febo.com [mailto:time-nuts-boun...@febo.com] On Behalf Of Esa Heikkinen Sent: Friday, May 10, 2013 11:06 AM To: time-nuts@febo.com Subject: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset Hello! I bought Symmetricom Tymserve TS2100-GPS some time ago. When comparing it's 1PPS with Thunderbolt they are not in sync: http://www.amigazone.fi/files/gpsdo/ts2100-pps-offset.png Symmetricom seems to be always about 10 uSec ahead. If I turn of the power and then on again, it will be in sync very short moment after the tracking led is turned in. But then it will always roll itself at 10 uSec ahead and lock there, when the locked led turns on. It seems that it has somehow wrong GPS PPS offset or maybe it's folloring the 1PPS signal from it's internal GPS from the wrong edge (this is just my guess; Thunderbolt has 10 uSec long PPS signal, maybe the internal GPS as well?) I understand that TS2100 is not as accurate PPS source as Thunderbolt. But it's manual claims that 1PPS accuracy should be 1 uS. So this is clearly out of spec. This is strange because the Symmetricom seems to be high quality product. Obviously not? Is there any way to fix this by adjusting some offset with telnet etc? If not, maybe I try to invert the 1PPS from internal GPS. Or if someone knows how to use Thunderbolt as a time source, I could use that and remove the internal GPS... -- 73s! Esa OH4KJU ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
On Sun, May 12, 2013 at 1:34 AM, Esa Heikkinen wrote: > Jason Rabel kirjoitti: > >> I think nobody ever really bothered to fix anything since the NTP time was > >> "good enough". > > Yes - definitely... If it's only used for NTP then the 10 usec error is > totally insignificiant. A startum one NTP server typically runs with abot 2 uSec error fro UTC. ButA stratum two server, that means one that gets time from the strum one over the network NTP solution seems to be quite inaccurate, if I poll > my server from http://support.ntp.org/ntpq.php the results can vary couple > of milliseconds every time. Most Statum One NTP server are better then 10 uSec. 2 uSec is about what you should shoot for. It shouldn't work that way. What NTP does is discipline a local clock. It take hours or days before it settles and a lot depends on how good the local oscillator is. Mostly your local oscillator is just a TTL can that cost about 50 cents solder to a PC mother board. The the error in one "poll" does not matter to much because hundreds of them will be done and NTP wil make tiny adjustments to the locals clock's RATE. NTP is not adjusting the time, it adjusts the rate. (OK there is one time exception if the time is found to tbe far off) It does not do an adverse of servers. It compares them to find the best subset. This stops a broken server from polluting the average. Yes there might be a 10ms error in a polled time Bt what NTP is doing is noteing the local time, letting the local clock run for say 20 minutes then comparing to the server(s) to see if the local clock is fast or slow. So that 10ms error is only parts per million. It is not a 10ms absolute error in UTC time. If your local oscillator were better then NTP would open up that 20 minutes window and if the results aren't good it closes it down. It tries to keep several of those 20 minute intervals open at once. On the other hand SMTP, to thing that Windows uses simply gets the time and jumps the local clock to match and that's it. It does not discipline the rate. SNTP is a one-time thing -- Chris Albertson Redondo Beach, California ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
Jason Rabel kirjoitti: I think nobody ever really bothered to fix anything since the NTP time was > "good enough". Yes - definitely... If it's only used for NTP then the 10 usec error is totally insignificiant. NTP solution seems to be quite inaccurate, if I poll my server from http://support.ntp.org/ntpq.php the results can vary couple of milliseconds every time. However the offset shown by test page is always less than 10 milliseconds. Maybe this is because the route is too long: delay is more than 230 ms - the test site is far side of the world for me... It' the nature of the TCP/IP and Internet that the delays are quite random. Same result if I set up "software" NTP server which polls many NTP servers from the net. The results are always vayring many milliseconds and the server will constantly change the "best" server to follow. I think the multitasking operating system will also add some errors to the ntp server operation, since the internal system delays will always change as well. The hidden "eng" menu does have some interesting stuff. I've tweaked > the offset to bring time in sync with my other NTP servers, usually around -4000 does the trick since the TS2100 doesn't use a true NTP implementation (I think it is more SNTP). According to manual it should support both protocols, but maybe this NTP vs SNTP is more like client side difference? I understood that NTP polls many servers to get kind of "averaged" timing, when SNTP uses only one server. For example Windows has SNTP only, you can define only one time server to use at time. Well of course TS2100 doesn't poll any servers and uses only one time source, GPS. I have seen different offset settings in units, so I would think there would be *some* way to save the setting, but like you have not found out a way to do it yet. :( Yes! That's now the only question... If this will remain unknown, then maybe I will setup a microcontroller connected to the console port and set it to follow somehow, when TS2100 is restarted. Then it could send the "eng tim offset..." command via serial port to restore the offset setting automatically every time when needed. This is quite stupid way to "store" the setting, but it might be the only way if the offical way to store the settings will remain unknown. -- 73s! Esa OH4KJU ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
Here's some good reading you might want to check out about the PPS and different GPS modules: http://www.febo.com/pipermail/time-nuts/2008-September/033475.html I think nobody ever really bothered to fix anything since the NTP time was "good enough". Plus when all those companies got bought out and merged managing / tweaking legacy products just wasn't a priority I guess. The hidden "eng" menu does have some interesting stuff. I've tweaked the offset to bring time in sync with my other NTP servers, usually around -4000 does the trick since the TS2100 doesn't use a true NTP implementation (I think it is more SNTP). I have seen different offset settings in units, so I would think there would be *some* way to save the setting, but like you have not found out a way to do it yet. :( Jason > There's hidden "eng" menu with many options. No documentation is > available for this menu, so I had to find things out by trial and error. > > In the "timing" menu there's offset_time option. This shows -100 at > default and one step is 100ns --> 10 uS. Setting this to zero the PPS > moves to sync with Thunderbolt! > > But the setting is not permament, it's lost with reboot. Unable to find > out how to "save" this permamently. And hard to understood why this is > even set to -10 uS! > > There's also commands to read and write the eeprom contents (it's good > idea to read it and copy-paste the contenst to the safe place). However > the offset_time is not stored to eeprom, even with JP4 in place. Too sad! ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
Additional info: There's hidden "eng" menu with many options. No documentation is available for this menu, so I had to find things out by trial and error. In the "timing" menu there's offset_time option. This shows -100 at default and one step is 100ns --> 10 uS. Setting this to zero the PPS moves to sync with Thunderbolt! But the setting is not permament, it's lost with reboot. Unable to find out how to "save" this permamently. And hard to understood why this is even set to -10 uS! There's also commands to read and write the eeprom contents (it's good idea to read it and copy-paste the contenst to the safe place). However the offset_time is not stored to eeprom, even with JP4 in place. Too sad! -- 73s! Esa OH4KJU ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
If one of the two units has a 10 uS wide pulse, I bet you have one of the two pluses inverted. Make sure one of the two is not flipped On Fri, May 10, 2013 at 11:05 AM, Esa Heikkinen wrote: > Hello! > > I bought Symmetricom Tymserve TS2100-GPS some time ago. When comparing it's > 1PPS with Thunderbolt they are not in sync: > http://www.amigazone.fi/files/gpsdo/ts2100-pps-offset.png > > Symmetricom seems to be always about 10 uSec ahead. If I turn of the power > and then on again, it will be in sync very short moment after the tracking > led is turned in. But then it will always roll itself at 10 uSec ahead and > lock there, when the locked led turns on. > > It seems that it has somehow wrong GPS PPS offset or maybe it's folloring > the 1PPS signal from it's internal GPS from the wrong edge (this is just my > guess; Thunderbolt has 10 uSec long PPS signal, maybe the internal GPS as > well?) > > I understand that TS2100 is not as accurate PPS source as Thunderbolt. But > it's manual claims that 1PPS accuracy should be 1 uS. So this is clearly out > of spec. This is strange because the Symmetricom seems to be high quality > product. Obviously not? > > Is there any way to fix this by adjusting some offset with telnet etc? If > not, maybe I try to invert the 1PPS from internal GPS. Or if someone knows > how to use Thunderbolt as a time source, I could use that and remove the > internal GPS... > > -- > 73s! > Esa > OH4KJU > ___ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. -- Chris Albertson Redondo Beach, California ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
[time-nuts] Symmetricom TS2100-GPS - 1PPS has 10uS offset
Hello! I bought Symmetricom Tymserve TS2100-GPS some time ago. When comparing it's 1PPS with Thunderbolt they are not in sync: http://www.amigazone.fi/files/gpsdo/ts2100-pps-offset.png Symmetricom seems to be always about 10 uSec ahead. If I turn of the power and then on again, it will be in sync very short moment after the tracking led is turned in. But then it will always roll itself at 10 uSec ahead and lock there, when the locked led turns on. It seems that it has somehow wrong GPS PPS offset or maybe it's folloring the 1PPS signal from it's internal GPS from the wrong edge (this is just my guess; Thunderbolt has 10 uSec long PPS signal, maybe the internal GPS as well?) I understand that TS2100 is not as accurate PPS source as Thunderbolt. But it's manual claims that 1PPS accuracy should be 1 uS. So this is clearly out of spec. This is strange because the Symmetricom seems to be high quality product. Obviously not? Is there any way to fix this by adjusting some offset with telnet etc? If not, maybe I try to invert the 1PPS from internal GPS. Or if someone knows how to use Thunderbolt as a time source, I could use that and remove the internal GPS... -- 73s! Esa OH4KJU ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
