4 ms·

Very interesting stuff. One of the fun bits of being a time-nut is that there are somewhere around 15 orders of magnitude available to play with. Getting clocks

by aftbit 3y ago

Very interesting stuff. One of the fun bits of being a time-nut is that there are somewhere around 15 orders of magnitude available to play with. Getting clocks synced to within a few milliseconds is child's play, even for an amateur with $15 of hardware. Getting clocks synced within a few dozen microseconds is possible with GPS and some minor effort. Getting clocks synced within a few nanoseconds requires a lot more effort. I have never attempted anything below the ns range, but this process seems to produce clocks in sync within hundreds of attoseconds! This is within around 10^-16 seconds. Quite amazing if you ask me.

comboy 3y ago
I cannot comprehend. Napkin math says if you move like 50 nanometers within one second this is enough acceleration over time to make these clocks out of sync because relativity. I mean, do points on earth do not move that much by themselves if they are far enough?
- aftbit 3y ago
  It turns out that cheap chip-scale optical atomic clocks would be a sensor revolution. They are sensitive enough to tell the difference between the first and second floor of a building by gravitational effects alone.
  - failbuffer 3y ago
    This would require a device to compare the on-board time with an external reference clock of some sort, right?
dogline 3y ago
I didn't realize that a "time-nut" was a thing. Never thought about people doing this as a sort of hobby. Congrats!
- hotpotamus 3y ago
  https://en.wikipedia.org/wiki/Clock_of_the_Long_Now https://en.wikipedia.org/wiki/Clock_of_the_Long_Now Jeff Bezos might be among them since he has funded this project.
- fanf2 3y ago
  You can find a community of time nuts via http://www.leapsecond.com/time-nuts.htm http://www.leapsecond.com/time-nuts.htm
- ramenbytes 3y ago
  In addition to the link in the sibling comment, there are also some timenuts over here: https://www.eevblog.com/forum/metrology/ https://www.eevblog.com/forum/metrology/ It's not dedicated to just time nuts though, so you'll see voltnuts and other species too.
stagger87 3y ago
You're numbers are off ~5 orders of magnitude? A 15$ GPS module can get you 40ns of timing accuracy.
- foota 3y ago
  Maybe hobbyist vs commercial?
- aftbit 3y ago
  Quite possibly yes. I have always been trying to synchronize computer clocks. I was imagining the ms level solution to be plain jane NTP over the internet (which is really quite impressive in its own right). The ns level solution requires a GPS module, but in my lab I am also using a 10MHz OXCO for hold-over as my GPS reception is marginal and I lose it several times a day. This adds a bit to the cost and complexity. Getting ns level timing distributed throughout an ethernet network using PTP is on my list, although my current Raspberry Pi pps solution does not support PTP. I am hoping to find a single board computer with a decent GPIO and an ethernet controller that supports PTP. Regardless, this is a very deep rabbit hole to go down. For the right kind of nerd, it can be a fun and rewarding experience. :)
adolph 3y ago
How long does it take to get clocks synchronized within a few nanoseconds?
kimburgess 3y ago
Nanosecond level sync is fairly achievable with Precision Time Protocol [1]. This is common in most media systems where small amount of drift will cause phase issues with audio. Not my area, but I believe it's seeing a few using in the DC space too [2] where they may be strong requirements around event ordering for distributed systems. The White Rabbit Project [3] also extends it to the picosecond realm for research control and instrumentation systems. Not a scratch on this project though. [1]: https://en.wikipedia.org/wiki/Precision_Time_Protocol https://en.wikipedia.org/wiki/Precision_Time_Protocol [2]: https://engineering.fb.com/2022/11/21/production-engineering/precision-time-protocol-at-meta/ https://engineering.fb.com/2022/11/21/production-engineering... [3]: https://ohwr.org/project/white-rabbit/wikis/home https://ohwr.org/project/white-rabbit/wikis/home

by aftbit 3y ago

comboy 3y ago

I cannot comprehend. Napkin math says if you move like 50 nanometers within one second this is enough acceleration over time to make these clocks out of sync because relativity. I mean, do points on earth do not move that much by themselves if they are far enough?

aftbit 3y ago
It turns out that cheap chip-scale optical atomic clocks would be a sensor revolution. They are sensitive enough to tell the difference between the first and second floor of a building by gravitational effects alone.
- failbuffer 3y ago
  This would require a device to compare the on-board time with an external reference clock of some sort, right?

dogline 3y ago

I didn't realize that a "time-nut" was a thing. Never thought about people doing this as a sort of hobby. Congrats!

hotpotamus 3y ago
https://en.wikipedia.org/wiki/Clock_of_the_Long_Now https://en.wikipedia.org/wiki/Clock_of_the_Long_Now Jeff Bezos might be among them since he has funded this project.
fanf2 3y ago
You can find a community of time nuts via http://www.leapsecond.com/time-nuts.htm http://www.leapsecond.com/time-nuts.htm
ramenbytes 3y ago
In addition to the link in the sibling comment, there are also some timenuts over here: https://www.eevblog.com/forum/metrology/ https://www.eevblog.com/forum/metrology/ It's not dedicated to just time nuts though, so you'll see voltnuts and other species too.

stagger87 3y ago

You're numbers are off ~5 orders of magnitude? A 15$ GPS module can get you 40ns of timing accuracy.

foota 3y ago
Maybe hobbyist vs commercial?
aftbit 3y ago
Quite possibly yes. I have always been trying to synchronize computer clocks. I was imagining the ms level solution to be plain jane NTP over the internet (which is really quite impressive in its own right). The ns level solution requires a GPS module, but in my lab I am also using a 10MHz OXCO for hold-over as my GPS reception is marginal and I lose it several times a day. This adds a bit to the cost and complexity. Getting ns level timing distributed throughout an ethernet network using PTP is on my list, although my current Raspberry Pi pps solution does not support PTP. I am hoping to find a single board computer with a decent GPIO and an ethernet controller that supports PTP. Regardless, this is a very deep rabbit hole to go down. For the right kind of nerd, it can be a fun and rewarding experience. :)

adolph 3y ago

How long does it take to get clocks synchronized within a few nanoseconds?

kimburgess 3y ago

Nanosecond level sync is fairly achievable with Precision Time Protocol [1]. This is common in most media systems where small amount of drift will cause phase issues with audio. Not my area, but I believe it's seeing a few using in the DC space too [2] where they may be strong requirements around event ordering for distributed systems. The White Rabbit Project [3] also extends it to the picosecond realm for research control and instrumentation systems. Not a scratch on this project though. [1]: https://en.wikipedia.org/wiki/Precision_Time_Protocol https://en.wikipedia.org/wiki/Precision_Time_Protocol [2]: https://engineering.fb.com/2022/11/21/production-engineering/precision-time-protocol-at-meta/ https://engineering.fb.com/2022/11/21/production-engineering... [3]: https://ohwr.org/project/white-rabbit/wikis/home https://ohwr.org/project/white-rabbit/wikis/home