[opensuse] NTP precision in openSUSE
One thing I've wondered about is the precision of the time displayed in openSUSE. I connect a stratum 1 NTP server and also have a clock that receives time from WWVB. Sometimes there seems to be a slight difference in the displayed time between the two. I suspect the clock may be less precise, as I bought it from Radio Shack several years ago and it wasn't that expensive. Has anyone looked into the display precision on Linux? Of course, the accuracy should be identical on both, as they are both traceable back to International Atomic Time. Please note, precision refers to how accurately the clock can be read, not how accurate the clock is. https://en.wikipedia.org/wiki/WWVB https://en.wikipedia.org/wiki/International_Atomic_Time -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
James Knott wrote:
One thing I've wondered about is the precision of the time displayed in openSUSE.
It will probably depend on where it is displayed, but if you do a gettimeofday(), the result is returned with precision in microseconds.
I connect a stratum 1 NTP server and also have a clock that receives time from WWVB. Sometimes there seems to be a slight difference in the displayed time between the two. I suspect the clock may be less precise, as I bought it from Radio Shack several years ago and it wasn't that expensive.
As long as it receives the time signal, the precision will be that of the WWVB time code. -- Per Jessen, Zürich (19.1°C) http://www.dns24.ch/ - your free DNS host, made in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On Wed, 08 Jul 2020 09:05:03 +0200 Per Jessen <per@computer.org> wrote:
James Knott wrote:
One thing I've wondered about is the precision of the time displayed in openSUSE.
It will probably depend on where it is displayed, but if you do a gettimeofday(), the result is returned with precision in microseconds.
From https://en.wikipedia.org/wiki/Network_Time_Protocol "The 64-bit timestamps used by NTP consist of a 32-bit part for seconds and a 32-bit part for fractional second, giving a time scale that rolls over every 232 seconds (136 years) and a theoretical resolution of 2−32 seconds (233 picoseconds)." So the precision is 233 picoseconds, but the same article says: "NTP can usually maintain time to within tens of milliseconds over the public Internet, and can achieve better than one millisecond accuracy in local area networks under ideal conditions." According to https://en.wikipedia.org/wiki/WWVB WWVB broadcasts once every second so its precision is 1 second. It broadcasts 60 bits per second, so that's the limit of resolution. So I'm not clear why anybody would care about the precision of either. But clearly a well-installed NTP reference can be more accurate.
I connect a stratum 1 NTP server and also have a clock that receives time from WWVB. Sometimes there seems to be a slight difference in the displayed time between the two. I suspect the clock may be less precise, as I bought it from Radio Shack several years ago and it wasn't that expensive.
As long as it receives the time signal, the precision will be that of the WWVB time code.
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
Le 08/07/2020 à 14:36, Dave Howorth a écrit :
From https://en.wikipedia.org/wiki/Network_Time_Protocol "The 64-bit timestamps used by NTP consist of a 32-bit part for seconds and a 32-bit part for fractional second, giving a time scale that rolls over every 232 seconds (136 years) and a theoretical resolution of 2−32 seconds (233 picoseconds)."
for sure 2^32 fore each
public Internet, and can achieve better than one millisecond accuracy
that have little meaning. With the present (and for what I know also future :-) light speed limit, the fiber light spend around 13ms to go to/from my next internet link (minimum ping). so the accuracy of the atomic clock have to be modified by the time between it's position and mine. How can one achieve this? May be a GPS knowing the respective position of the satellite and the receiver can make a fix, but on internet? it can only give legal time not the precise time I have here on an unknown position
According to https://en.wikipedia.org/wiki/WWVB WWVB broadcasts once every second so its precision is 1 second. It broadcasts 60 bits per second, so that's the limit of resolution.
if the time is casted every second, it can anyway be as precise as you want if the local clock is reasonably good, it wont derive so much in one second jdd -- http://dodin.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 08:46 AM, jdd@dodin.org wrote:
that have little meaning. With the present (and for what I know also future :-) light speed limit, the fiber light spend around 13ms to go to/from my next internet link (minimum ping).
so the accuracy of the atomic clock have to be modified by the time between it's position and mine. How can one achieve this? May be a GPS knowing the respective position of the satellite and the receiver can make a fix, but on internet? it can only give legal time not the precise time I have here on an unknown position
Actually, that's compensated for in NTP. https://en.wikipedia.org/wiki/Network_time_protocol#Clock_synchronization_al... Also, the GPS satellites have their own atomic clocks, which are synced with the U.S. Navy observatory. They wouldn't be using NTP to do that. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
jdd@dodin.org wrote:
that have little meaning. With the present (and for what I know also future :-) light speed limit, the fiber light spend around 13ms to go to/from my next internet link (minimum ping).
so the accuracy of the atomic clock have to be modified by the time between it's position and mine. How can one achieve this?
That is what we have NTP for. :-) -- Per Jessen, Zürich (25.6°C) http://www.hostsuisse.com/ - dedicated server rental in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
Le 08/07/2020 à 16:13, Per Jessen a écrit :
jdd@dodin.org wrote:
that have little meaning. With the present (and for what I know also future :-) light speed limit, the fiber light spend around 13ms to go to/from my next internet link (minimum ping).
so the accuracy of the atomic clock have to be modified by the time between it's position and mine. How can one achieve this?
That is what we have NTP for. :-)
be aware it's legal time, not local real time (only the GPS can know where I am) jdd -- http://dodin.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 10:16 AM, jdd@dodin.org wrote:
be aware it's legal time, not local real time (only the GPS can know where I am)
Actually, it's International Atomic Time, which has some differences from UTC, such as no leap seconds. When an NTP server, connected to a GPS receiver, receives the time, it has to adjust accordingly. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 10:13 AM, Per Jessen wrote:
That is what we have NTP for.:-)
NTP isn't good enough to sync the atomic clocks. That process is mentioned in the book I linked to. What it requires is some event that's observable from both locations. One method is to observe the phase of the LORAN C signals or a signal from a satellite. Another method it to carry a portable atomic clock from one site to the other, though that would be a bit difficult with GPS satellites. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 08:36 AM, Dave Howorth wrote:
"The 64-bit timestamps used by NTP consist of a 32-bit part for seconds and a 32-bit part for fractional second, giving a time scale that rolls over every 232 seconds (136 years) and a theoretical resolution of 2−32 seconds (233 picoseconds)."
So the precision is 233 picoseconds, but the same article says: "NTP can usually maintain time to within tens of milliseconds over the public Internet, and can achieve better than one millisecond accuracy in local area networks under ideal conditions."
That is clock accuracy, not display. Over the long term, the display will be just as accurate as the original source, but might fluctuate in the short term or perhaps have a small bias.
According tohttps://en.wikipedia.org/wiki/WWVB WWVB broadcasts once every second so its precision is 1 second. It broadcasts 60 bits per second, so that's the limit of resolution.
Again, you're confusing accuracy with precision. Over the long term, the accuracy will be as good as the source. The dispaly might not be.
So I'm not clear why anybody would care about the precision of either. But clearly a well-installed NTP reference can be more accurate.
It's more of a curiousity thing. Also, with the pandemic, I have way too much time on my hands. ;-) In a note to Per, I mentioned "rate", which is the difference between a clock's accuracy and actual time. It must be accounted for in calculations. What people might not be aware of is that you can have atomic clocks, that are part of the basis for International Atomic Time, at Boulder, Colorado and Greenwich, England and, because they're at different attitudes, will actually have different times. That difference has to be accounted for when the IAT is determined. https://en.wikipedia.org/wiki/International_Atomic_Time -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On Wed, 8 Jul 2020 09:21:00 -0400 James Knott <james.knott@jknott.net> wrote:
On 2020-07-08 08:36 AM, Dave Howorth wrote:
"The 64-bit timestamps used by NTP consist of a 32-bit part for seconds and a 32-bit part for fractional second, giving a time scale that rolls over every 232 seconds (136 years) and a theoretical resolution of 2−32 seconds (233 picoseconds)."
So the precision is 233 picoseconds, but the same article says: "NTP can usually maintain time to within tens of milliseconds over the public Internet, and can achieve better than one millisecond accuracy in local area networks under ideal conditions."
That is clock accuracy, not display. Over the long term, the display will be just as accurate as the original source, but might fluctuate in the short term or perhaps have a small bias.
According tohttps://en.wikipedia.org/wiki/WWVB WWVB broadcasts once every second so its precision is 1 second. It broadcasts 60 bits per second, so that's the limit of resolution.
Again, you're confusing accuracy with precision. Over the long term, the accuracy will be as good as the source. The dispaly might not be.
I don't think I am. I think you actually confused precision with resolution when you asked your original question: "Please note, precision refers to how accurately the clock can be read, not how accurate the clock is." What you describe as precision is actually the resolution. So you can reread my answers and substitute 'resolution' in some places, sure. Precision refers to random variations in the size of errors. There's no way to know that for 'openSUSE' in general; it will depend on what hardware you have, what program you're using to make the display, what else you're running on the machine, what kernel settings you have and perhaps other factors as well.
So I'm not clear why anybody would care about the precision of either. But clearly a well-installed NTP reference can be more accurate.
It's more of a curiousity thing. Also, with the pandemic, I have way too much time on my hands. ;-)
Indeed, as apparently have I. But as described so far your question doesn't have a sensible answer. Hence our disappearing up our collective a*ses.
In a note to Per, I mentioned "rate", which is the difference between a clock's accuracy and actual time.
Yes that's one of the easiest consequences of general relativity to demonstrate, these days.
It must be accounted for in calculations. What people might not be aware of is that you can have atomic clocks, that are part of the basis for International Atomic Time, at Boulder, Colorado and Greenwich, England and, because they're at different attitudes, will actually have different times. That difference has to be accounted for when the IAT is determined.
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 08:36 AM, Dave Howorth wrote:
Fromhttps://en.wikipedia.org/wiki/Network_Time_Protocol "The 64-bit timestamps used by NTP consist of a 32-bit part for seconds and a 32-bit part for fractional second, giving a time scale that rolls over every 232 seconds (136 years) and a theoretical resolution of 2−32 seconds (233 picoseconds)."
Actually, NTP 4 uses 128 bit timestamps. https://tools.ietf.org/html/rfc5905 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 03:05 AM, Per Jessen wrote:
As long as it receives the time signal, the precision will be that of the WWVB time code.
That would be the clock accuracy, not the display precision. Precision refers to how accurately you can read something. That doesn't mean the clock is accurate, as it could be off but you, but you can read it precisely. One example I recall from years ago would be 2 rulers. One has coarse divisions, but they are extremely accurate. The other has fine divisions, but the entire ruler is the wrong length. With the first, you could accurately measure things that are close to the coarse divisions, but with the other, you'd be wrong, no matter how many divisions there were, if you didn't know about the error. Astronomers considered something called "rate". It is the known error in a clock and they would include it in their calculations. This reminds me of something at work many years ago. I used to work for a telecom company. In the network mangement centre, there was a large digital clock. I asked one of the guys how accurate it was. He said it was very accurate because it was driven by the same clock as the entire network, which was derived from LORAN C. However, that only provided an extremely accurate time base, not time of day. The actual displayed time was set by hand. This was back in the days before the Internet became popular, so NTP wasn't an option. I know ntp can be accurate within milliseconds, but that's an entirely separate issue from how accurate the display is in relation to it.
It will probably depend on where it is displayed, but if you do a gettimeofday(), the result is returned with precision in microseconds.
You'd have to be connected to an appropriate stratum 0 source for that. Stratum 1 will only be within milliseconds. Incidentally, I read a book about time & frequency, etc. It's a free download from the NIST: https://nvlpubs.nist.gov/nistpubs/Legacy/MONO/nistmonograph155e1999.pdf https://en.wikipedia.org/wiki/Network_time_protocol https://en.wikipedia.org/wiki/Loran-C -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
James Knott wrote:
On 2020-07-08 03:05 AM, Per Jessen wrote:
As long as it receives the time signal, the precision will be that of the WWVB time code.
That would be the clock accuracy, not the display precision. Precision refers to how accurately you can read something.
Didn't we deal with that in the previous paragraph ?
That doesn't mean the clock is accurate, as it could be off but you, but you can read it precisely.
Exactly.
It will probably depend on where it is displayed, but if you do a gettimeofday(), the result is returned with precision in microseconds.
You'd have to be connected to an appropriate stratum 0 source for that. Stratum 1 will only be within milliseconds.
I thought you said it is only about the precision _read_, not how accurate it is. On a stand-alone PC running off the undisciplined local clock, you can still read the timestamp with microsecond precision. -- Per Jessen, Zürich (26.9°C) http://www.cloudsuisse.com/ - your owncloud, hosted in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 10:20 AM, Per Jessen wrote:
I thought you said it is only about the precision_read_, not how accurate it is. On a stand-alone PC running off the undisciplined local clock, you can still read the timestamp with microsecond precision.
Yep, you could read it, but how accurate is that precision? I mentioned that clock I have. It doesn't sync continuously, instead it tries every few hours. In between successful syncs, the clock can drift. On the other hand, NTP syncs every couple of minutes or so. So, based on that, the display on my computer is likely to be more precise, though both would be equally accurate long term. Also, I use stratum 1 servers that are located at the Internet exchange that my ISP uses to connect to the rest of the world. (BTW, it's in the same building where I used to work for the telecom company.) I don't know how much variation there will be on that, compared to the 60 KHz signal my clock receives on the trip between Boulder and Toronto. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
James Knott wrote:
On 2020-07-08 10:20 AM, Per Jessen wrote:
I thought you said it is only about the precision_read_, not how accurate it is. On a stand-alone PC running off the undisciplined local clock, you can still read the timestamp with microsecond precision.
Yep, you could read it, but how accurate is that precision? I mentioned that clock I have. It doesn't sync continuously, instead it tries every few hours. In between successful syncs, the clock can drift. On the other hand, NTP syncs every couple of minutes or so.
I am certainly no expert on the workings of NTP, but I thought NTP worked with a "clock discipline" to achieve the best time and synch.
So, based on that, the display on my computer is likely to be more precise, though both would be equally accurate long term. Also, I use stratum 1 servers that are located at the Internet exchange that my ISP uses to connect to the rest of the world. (BTW, it's in the same building where I used to work for the telecom company.) I don't know how much variation there will be on that, compared to the 60 KHz signal my clock receives on the trip between Boulder and Toronto.
The 60kHz from WWVB (or MSF Rugby, now gone), just like 77kHz from DCF77 or the 243kHz longwave carrier from Kalundborg are highly accurate frequency standards. They are excellent for a PPS signal for your local reference clock. Is that what your Radio Shack device does, act as reference clock? I have a DCF receiver, but it only provides the time signal, not a PPS signal. GPS receivers are also popular as PPS providers. -- Per Jessen, Zürich (27.3°C) http://www.hostsuisse.com/ - dedicated server rental in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2020-07-08 11:03 AM, Per Jessen wrote:
I am certainly no expert on the workings of NTP, but I thought NTP worked with a "clock discipline" to achieve the best time and synch.
Yep. It compensates for the delay between server and client and then the error. I posted the link about it in another message.
The 60kHz from WWVB (or MSF Rugby, now gone), just like 77kHz from DCF77 or the 243kHz longwave carrier from Kalundborg are highly accurate frequency standards. They are excellent for a PPS signal for your local reference clock. Is that what your Radio Shack device does, act as reference clock? I have a DCF receiver, but it only provides the time signal, not a PPS signal. GPS receivers are also popular as PPS providers.
No, it's just a clock that sits on my window sill. It has no output other than the display. As I mentioned, it only syncs periodically throughout the day. The display shows when it's attempting to sync and whether it's synced in the past few hours. Many years ago, Heathkit had a kit that received WWV or WWVB and produced an output that could be read by a computer. http://www.schaffrath.net/mac.html -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
James Knott wrote:
On 2020-07-08 11:03 AM, Per Jessen wrote:
The 60kHz from WWVB (or MSF Rugby, now gone), just like 77kHz from DCF77 or the 243kHz longwave carrier from Kalundborg are highly accurate frequency standards. They are excellent for a PPS signal for your local reference clock. Is that what your Radio Shack device does, act as reference clock? I have a DCF receiver, but it only provides the time signal, not a PPS signal. GPS receivers are also popular as PPS providers.
No, it's just a clock that sits on my window sill. It has no output other than the display. As I mentioned, it only syncs periodically throughout the day.
Ah, I get it - got one of those hanging on my kitchen wall :-)
Many years ago, Heathkit had a kit that received WWV or WWVB and produced an output that could be read by a computer. http://www.schaffrath.net/mac.html
Yup, ELV offers similar. -- Per Jessen, Zürich (27.7°C) http://www.dns24.ch/ - your free DNS host, made in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
participants (4)
-
Dave Howorth -
James Knott -
jdd＠dodin.org -
Per Jessen