Easy Time Sync

[Patrick M. Hausen]

Try a CenterClick device:

https://centerclick.com/ntp/

Satisfied customer, nothing else. Works splendidly.

I use that as my "preferred" peer with the official time source for Germany as a fallback. Then distribute to all devices from my OPNsense.

I really do not understand why you refuse to run ntpd on every 24x7 up device. That's at least how it was intended to be done when the software was designed. And how I run it in my data centres with approximatly 100 machines. Of course we run ntpd on every single one of them.

[pfry]

[...]
In US we do have NIST WWVB 60kHz time carrier, supposedly very accurate[...]

The source is good, but the signal varies - although it penetrates structures pretty well. You also have to correct for location. GPS tends to be easier to deal with, if you can get a signal.

I have an old Ultralink 325 - the very last one they had. It's only supported by ntpd (that I know of). I haven't had it hooked up in some years, but one of these days I'll set it back up.

[BrandyWine]

Try a CenterClick device:

https://centerclick.com/ntp/

Satisfied customer, nothing else. Works splendidly.

I use that as my "preferred" peer with the official time source for Germany as a fallback. Then distribute to all devices from my OPNsense.

I really do not understand why you refuse to run ntpd on every 24x7 up device. That's at least how it was intended to be done when the software was designed. And how I run it in my data centres with approximatly 100 machines. Of course we run ntpd on every single one of them.

That's one of the amzon devices that start at $250(us). For home use that mean the antenna has to be in good location. It's too bad they didnt include a USB-A port and support 1-2 common USB mini wifi adapters, then it could be somewhat more flexible between location and network access.

As for ntpd, I just don't see any reason to leave it memory resident, anything I have that does NTP will do it to ntp.org or nist. An hourly ntpd -q is enough to keep fw time accurate. It's not stepping, only slew adjusts.

Not having a 123 listener there for 59min50sec every 1hr means less exposure (even if restricted by conf), and, less use of system resources, and it still accomplishes the same thing as running ntpd resident.

I know drift file can change with ntpd running all the time, so that is something I am looking at, not much info on how (or if) ntpd -q does drift file update functions, I know it reads the drift file, just not sure it will update that file, but I can check and see if value has changed. I know for sure using -q will not create a drift file.

[Linwood]

If they are properly built.

I was looking at some GPS ntp units on amzon, they are built into small metal enclosure with long wired antenna, $300(us)+

No idea about those.  That's 5-10 times what some people spend doing it on the cheap, so definitely has more potential.

"You don't get what you don't pay for"   :)

[Kets_One]

2x GPS and 1x DCF77

Reconciliation of multiple sources is an interesting subject.

Are both GPS receivers using same GPS service?

GPS (receiver) is accurate to approx 100ns. Is the DCF77 broadcast any better?

Yes, both GPS sources are professional NTP appliances from Meinberg. DCF77 is not more accurate than GPS however, but I wanted some diversity in the time sources.
for example in case GPS reception is disturbed.
When bought these are quite expensive, but they are regularly offered on Ebay for a few hundred USD/EUR. Support from Meinberg is excellent. They still offer firmware updates for many of their NTP appliances, even many years after production. Highly recommended.

[cookiemonster]

but my reasoning is on the other side of yours

If one needs fw to be NTP server, then by all means leave "it" memory resident.
I only want fw device to keeps it's own clock updated, and I don't run any NTP server anywhere locally.

I am curious about chronyd on your setup, can you post a "ps -aux |grep chronyd"

Sure.

Code Select Expand

penguin@OPNsense:~ $ sudo ps -aux | grep chronyd
Password:
chronyd  15768   0.0  0.1   26820   3700  -  I     3Oct25     0:06.49 /usr/local/sbin/chronyd
root     15797   0.0  0.0   22080   3108  -  I     3Oct25     0:00.30 /usr/local/sbin/chronyd
penguin  26228   0.0  0.0   13748   2004  0  S+   09:54       0:00.00 grep chronydJust the forked process:
You cannot view this attachment.

[BrandyWine]

you have 16GB ram?

[cookiemonster]

The mini pc host has 32 GB but this is a VM on it. It only has been given 6 GiB and 2 vCPUs.

[BrandyWine]

Am I right, the ps shows 0.1% ram for chronyd, so 6MB ?

[cookiemonster]

If I read htop correctly the RES is 3700 so 3.7 megabytes

[pfry]

Try a CenterClick device:[...]

Done, and agreed. I'd looked into building one, but I couldn't find a parts combo that I liked. (What I'd really wanted was an integrated PCI-e x1 GPS, and nobody seems to make it.) The included wall-wart power supply is atrocious, though - I used one from a Netgear AP temporarily and grabbed a decent Mean Well for future use. (Not an issue for PoE.)

The last time I had a GPS unit up and running (for anything, but NTP specifically) was about 15 years ago, and it (and its predecessor, older Garmin units) wouldn't pick up diddly indoors. The devices have improved tremendously, in sync time and reception. (I still got an outdoor antenna, just in case.)

[Greg_E]

I have a cheap GNSS device that I've used for a few years, accuracy always matches my phone within human perception. You can get the version without display for $80 on ebay, or more for the version with display, and more again for the version with the 90 meter external antenna. In theory it's accurate, but I don't have a good PTP source to check it against.

TF-NTP-Lite

And all that said, the prices have doubled since I bought mine. I will say the external antenna made an impact on the number of satellites tracked and how quickly it locks. I have one of the shorter antenna versions in my lab with antenna positioned near a window, good enough with carrying around 10-12 satellites, the roof top antenna often pulls 22+ satellites.

[BrandyWine]

the roof top antenna often pulls 22+ satellites.

That's some serious time data...............

I always wondered, does the receiver have 1thread/1cpu per sat? If not how does receiver account for cpu cycles (time) to process the data and report it out?