ntpMerlin ntpMerlin - NTP Daemon for AsusWRT Merlin

[L&LD]

Thanks all. I'll probably try Diversion as my first foray into add-ons. This NTP server would probably be second.

Follow the appropriate steps in the link here.

https://www.snbforums.com/threads/any-issues-with-asus-rt86u-2-4ghz-band.55835/#post-476606

 

[cmkelley]

Thanks all. I'll probably try Diversion as my first foray into add-ons. This NTP server would probably be second.

As L&LD's link says, your first foray should actually be amtm. Set up your system per his link, and then add Diversion from amtm.

 

[maxbraketorque]

I was wondering if amtm would be needed/helpful. I thought I might be able to handle all the necessary actions within the Diversion installer itself. I guess its helpful to work from amtm.

 

[L&LD]

I was wondering if amtm would be needed/helpful. I thought I might be able to handle all the necessary actions within the Diversion installer itself. I guess its helpful to work from amtm.

If you like to spend your nights getting the spaces and { in the right places for your scripts, don't use amtm.

But, if you're a mere mortal like myself? amtm all the way!

 

[L&LD]

@Jack Yaz and others, I think I broke ntpMerlin on my RT-AC68U.

I tracked down and found a Stratum 1 NTP server in my city that states it is powered by GPS.

I edited the ntpMerlin config and entered just that server and deleted the other standard servers. I saved this file and saw that ntpMerlin restarted.

Now, the graphs are showing me zero offset, and zero jitter. Can this be possible?

Do I need to re-add more servers? Or is there a deeper issue I should be aware of?

(Jack Yaz, just wanted to also mention that the first post needs updating to show the latest version is 1.1.1).

 

[cmkelley]

@Jack Yaz and others, I think I broke ntpMerlin on my RT-AC68U.

I tracked down and found a Stratum 1 NTP server in my city that states it is powered by GPS.

I edited the ntpMerlin config and entered just that server and deleted the other standard servers. I saved this file and saw that ntpMerlin restarted.

Now, the graphs are showing me zero offset, and zero jitter. Can this be possible?

Do I need to re-add more servers? Or is there a deeper issue I should be aware of?

(Jack Yaz, just wanted to also mention that the first post needs updating to show the latest version is 1.1.1).

I have a single server listed in ntp.conf, but it's my own gps time server on my network. and my offset is working (usually ~20 usec), but the jitter isn't working.

 

[L&LD]

I have a single server listed in ntp.conf, but it's my own gps time server on my network. and my offset is working (usually ~20 usec), but the jitter isn't working.

Thanks, all five graphs show zero for current except for the last one 'Drift' which still shows 37.54ppm.

Maybe, the internal hardware/kernel can't accurately show these very high precision values?

 

[L&LD]

A further question for the NTP users here: what effect in terms of accuracy have when I choose a GPS, CDMA or PPS based NTP server?

Again, these are all public Stratum 1 Network Time Protocol (NTP) servers, above.

 

[cmkelley]

Thanks, all five graphs show zero for current except for the last one 'Drift' which still shows 37.54ppm.

Maybe, the internal hardware/kernel can't accurately show these very high precision values?

No, the graphs are auto-scaling, and you're not gonna get better than a few microseconds (.000001 sec). There is a boatload of information on NTP at: http://www.ntp.org/documentation.html, and the first link "Official NTP Documentation" is worth spending half a day or so going through. This link: https://support.ntp.org/bin/view/Support/ChoosingReferenceClocks talks about the different clock types, an atomic PPS is the best followed by GPS, then CDMA.

 

[heysoundude]

If terrestrial atomic references are considered more correct/preferable, it’s by a very narrow margin. With each GPS/GLONASS bird having at least 2 atomic clocks, We’ve quite the pool of time reference surrounding us, especially since a few of those satellites (and their clocks) are needed to derive any position on the planet, save a small bit of error from propagation/calculation delay which likely isn’t easy to account for/factor in.
GPS can pinpoint you within a few feet in 3 dimensions anywhere on the planet, for all intents and purposes. If 5G cellular takes that to millimeter resolution (which is pico or femtosecond accuracy in terms of this conversation, the time units computers function in), it’s literally a quantum shift in accounting for time accurately; if time is money, and money is power, mastering time raises the bar by orders of magnitude.

I sense I may need to read Hawking and possibly Einstein in the near future...

 

[heysoundude]

Now, the graphs are showing me zero offset, and zero jitter. Can this be possible?

Yes, if you read my last post. It’s possible you’re getting results below what the daemon is set to round up to.
Awesome.

 

[Jack Yaz]

A further question for the NTP users here: what effect in terms of accuracy have when I choose a GPS, CDMA or PPS based NTP server?

Again, these are all public Stratum 1 Network Time Protocol (NTP) servers, above.

Can you check the size of ntpdstats_rrd.rrd please? I inadvertently introduced a bug for new installs last night, which I have rolled back. Existing installs should have remained intact.

 

[cmkelley]

Yes, if you read my last post. It’s possible you’re getting results below what the daemon is set to round up to.
Awesome.

No, sorry, it isn't. High accuracy, non-temperature-controlled or compensated clocks have a error of roughly 1 PPM, which is .000001 seconds (1 us), and that's just the clock itself without taking into account the inevitable degradation due to the network connection. For reference, an error of 1 PPM is 31.5 seconds/year. Typical accuracy over the internet is .001 seconds (1 ms / 1000 PPM), and I've seen around 10 us error with a reference clock connected directly to the router through a usb port. I get 50 us or so accuracy on my router with my Raspberry Pi GPS setup that includes Pulse Per Second clock discipline on my local network. I know for a fact that kvic's graphs autoscale at least down to 1 us steps, because I've seen it. You cannot possibly get results that won't display.

 

[bluzfanmr1]

Easy to see when the change kicked in for me.

 

[cmkelley]

Easy to see when the change kicked in for me.

What is the 24 hour graph showing? Should be in the microsecond range.

 

[bluzfanmr1]

What is the 24 hour graph showing? Should be in the microsecond range.

 

[cmkelley]

If it's convenient, can you capture it again in about 20 minutes when that 20 ms spike drops off the graph?

 

[John Fitzgerald]

Does this help?
on AC68U:

 

[cmkelley]

Yeah, that's what I expect to see for synchronizing over the internet. Hard to do much better than that without a local time source.

 

[bluzfanmr1]

If it's convenient, can you capture it again in about 20 minutes when that 20 ms spike drops off the graph?

20 minutes later...