ntpMerlin Drift ppm slow

[figorr]

Hi,

I am starting using ntpMerlin with Chrony. By default the Chrony.conf uses "ntp.pool.org" but i have been seeing some peaks in the Offset, so I decided to use another ntp servers that are closer to my location. I edited the crhony.conf and now I am using es.pool.ntp.org, europe.pool.ntp.org and other severs closer to my location.

As a result ... I can see the there is no more peaks and the Offset is more stable around +/- 0.1 values. While when I was using ntp.pool.org the values were around +/-0.5 and some peaks of +/-1.0. So now the Offset Graph is looking like a straight line.

But the Drift is always around 8.000ppm. There was no change by using the new ntp servers.

When I am typing the command "chronyc tracking" ... it shows that 8000ppm is "slow".

And using the command "chronyc -n sources" I can see that they are stratum 1 servers.

There is anything I can do to reduce the "Drift" in order to achieve values closer to 1.000ppm, fast?

Regards,

 

[heysoundude]

Hi,

I am starting using ntpMerlin with Chrony. By default the Chrony.conf uses "ntp.pool.org" but i have been seeing some peaks in the Offset, so I decided to use another ntp servers that are closer to my location. I edited the crhony.conf and now I am using es.pool.ntp.org, europe.pool.ntp.org and other severs closer to my location.

As a result ... I can see the there is no more peaks and the Offset is more stable around +/- 0.1 values. While when I was using ntp.pool.org the values were around +/-0.5 and some peaks of +/-1.0. So now the Offset Graph is looking like a straight line.

But the Drift is always around 8.000ppm. There was no change by using the new ntp servers.

When I am typing the command "chronyc tracking" ... it shows that 8000ppm is "slow".

And using the command "chronyc -n sources" I can see that they are stratum 1 servers.

There is anything I can do to reduce the "Drift" in order to achieve values closer to 1.000ppm, fast?

Regards,

Hi-

Good news: you're misinterpreting the data-

Look at the averages in the GUI - the closer to an avg of 0.0 on Offset and drift the better, but you also have to take into account the time unit on the offset number:
ms > us > ns <- each step here is a factor of 1000 in terms of accuracy (0.000usec is a 1000x better than 0.000ms.)

8ppm drift...ppm means pulses or parts per MILLION, and slow means your router's internal clocking crystal is running slower than the reference...by 8.0ns or 8 millionths of a second.

impressed and relieved? you should be. Especially if your offset average is in the usec range or lower

want to help the timing crystal in the router? I can't remember (or am not familiar) with how electrical mains grounding (earthing) is handled in your part of the world, but if you can improve it, you should - everything electrical/electronic in your house will benefit in terms of performance and lifespan. If you want to be super sexy and top-level about it, the next step is getting a pure-sine wave UPS that outputs "perfect" mains voltage/frequency, and connect both your router AND modem to it.

(I just checked my chrony sources and some of them have ms numbers in the last column...2 of the 4 have usec listed there...I need to look at the chrony manpage to grok what that means)

 

[figorr]

Thank you for the explanation. Now I am a bit less worried about drift,

The Offset Average is 0,023ms (so only 23us). I think It could be Ok then.

I changed the servers this morning (around 8.00 am) and there was only one big peak at midday when I made a change in order to test time.cloudflare.com with nts. But this seems not working well for me.

The modem and the router (RT-AX88U) are connected to an UPS, Riello iPlug 800. This is not a pure-sine wave UPS, but for household use it is enough for me.

 

[Make WiFi Great Again]

With all due respect why would a sinewave ups matter? Most electronic devices use switching power supplies which dont care about voltage or frequency as long as they are within perhaps 50% of designed unput voltage and freq. For exMple, my router power supply can run on anything from 70volts to 290v and and can do this on anything from straight dc input all the way to probably hundreds of hz. it puts out a perfect 18v dc. The router runs on dc not ac. And its switching power supply isnt finicky about input voltage or frequency. Remember out routers run on dc and do not have any sort of timing signal conected to the mains.

If this was an old clock radio or vintage plug in wall clock or something i coud perhaps agree with you. But modern electronics running on switching powrr supplies dont care about clean or dirty power or voltage and freq thats iff spec .

 

[Jack Yaz]

With all due respect why would a sinewave ups matter? Most electronic devices use switching power supplies which dont care about voltage or frequency as long as they are within perhaps 50% of designed unput voltage and freq. For exMple, my router power supply can run on anything from 70volts to 290v and and can do this on anything from straight dc input all the way to probably hundreds of hz. it puts out a perfect 18v dc. The router runs on dc not ac. And its switching power supply isnt finicky about input voltage or frequency. Remember out routers run on dc and do not have any sort of timing signal conected to the mains.

If this was an old clock radio or vintage plug in wall clock or something i coud perhaps agree with you. But modern electronics running on switching powrr supplies dont care about clean or dirty power or voltage and freq thats iff spec .

A PC PSU that uses active power factor correction (PFC) can run into problems with a simulated sine wave from a UPS.
Agreed it doesn't apply to router, though.

 

[zaxcom]

Would you say these look fairly normal?

 

[JT Strickland]

Would you say these look fairly normal?

I think you could benefit from some closer servers maybe.
here's mine:

 

[zaxcom]

I think you could benefit from some closer servers maybe.
here's mine:

What is the path to that file?

 

[JT Strickland]

What is the path to that file?

I believe it's /opt/share/ntpmerlin.d/chrony.conf on my RT-AC86U.
I think it will tell you in amtm on the NtpMerlin page where it is.

 

[zaxcom]

I believe it's /opt/share/ntpmerlin.d/chrony.conf on my RT-AC86U.
I think it will tell you in amtm on the NtpMerlin page where it is.

Ok, since I am in the US, I modified the servers to us.pool.ntp.org. Restarted chrony so it will be a while before I have meaningful graphs.

 

[heysoundude]

pools suck - go straight to the stratum 1 sources:

StratumOneTimeServers < Servers < Network Time Foundation's NTP Support Wiki

support.ntp.org

by selecting your own servers you're making chrony make you your own pool on your router. the closer they are to your location, the better. this "pool of pools" attitude will make the resulting average of averages too high; we can do better, because we have the technology

 

[JT Strickland]

Ok, since I am in the US, I modified the servers to us.pool.ntp.org. Restarted chrony so it will be a while before I have meaningful graphs.

I was just going to share mine, when I saw @heysoundude post, now they don't seem very important, but they are, right now:

pool 0.us.pool.ntp.org iburst
pool 1.us.pool.ntp.org iburst
pool time1.facebook.com iburst
pool 0.amazon.pool.ntp.org iburst

Remember to restart chronyd after editing the config file. Never mind, I see you already did that.

I'm going to look into the SratumOneTimeServers myself. How does one specify them properly in the config file?
Edit: After reading the rules of engagement, I don't think I qaulify.

 

[JT Strickland]

I don't want to reply to myself, but my drift wasn't right in that last screenshot, I hadn't updated to the latest.
It is more correctly displayed here:

 

[dave14305]

pools suck - go straight to the stratum 1 sources:

StratumOneTimeServers < Servers < Network Time Foundation's NTP Support Wiki

support.ntp.org

by selecting your own servers you're making chrony make you your own pool on your router. the closer they are to your location, the better. this "pool of pools" attitude will make the resulting average of averages too high; we can do better, because we have the technology

Personal foul, heysoundude, being a bad netizen, 15 yard penalty, loss of down.

 

[heysoundude]

Personal foul, heysoundude, being a bad netizen, 15 yard penalty, loss of down.

I'm well out of snowball range, and the border is closed anyways: Mahomes is gonna eat your boys for breakfast tonight

 

[zaxcom]

Well this looks amazingly better.

 

[heysoundude]

@JT Strickland is right, and I apologise to @dave14305
unless we're running 100+ client devices we should steer clear of stratum 1 servers...but there are plenty of stratum 2, and I'm (well, chrony is..) referring to 4 of them and is only using 3.



I'm good with a 3ns offset and ~3us drift...close enough for horseshoes and hand grenades ;-D

(my ISP rotated my WAN IP back to a much closer server early this morning that's colocated with a stratum 1 server - thats why offset stopped swinging and drift dropped and stayed low. I can tell from my ping and connection jitter dropping in another addon tab)

 

[zaxcom]

My DNS provider is cloudflare and cloudflare is also first in my chrony.conf file. I happen to live three blocks from a cloudflare data center. So I guess adding them has caused pool.ntp.org to fall to a lower priority.

 

[zaxcom]

Does this look like it’s running properly? Seems too good.

 

[heysoundude]

Does this look like it’s running properly? Seems too good.

While I'm not familiar with what version of asuswrt-merlin is the latest for your router, I'm questioning whether the version you're running is compatible with this script.
Have you checked the compato statement in the first post of the ntpMerlin 3.0 thread?
also - error in your first CLI screenshot: I don't think it's meant to be querying the server that relentlessly. have you read the chrony manpage to better config your conf file?