SparkFun GNSS Disciplined Oscillator
Hi Time Nuts,
I'm Sean and I work at SparkFun Electronics. Historically, my company has
made dev boards and primarily been in the embedded space, but our founder
has recently gotten really into the world of high-precision GNSS, which has
also led us into the wacky world of timing.
I'm not really here to sell you anything (but don't let that stop you, hah)
but more curious what your thoughts are on a new product we just came out
with (https://www.sparkfun.com/sparkpnt-gnss-disciplined-oscillator.html).
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps. It uses the mosaic-T
from Septentrio which has anti jamming and spoofing capabilities as well.
On the product page you can find links to the documentation for the
product, as well as the oscillator and GNSS module inside.
We're curious what your thoughts are on the product. Namely, is this a good
accuracy/price ratio compared to the methods you can DIY? Also, what
applications might you be using sub-nano accuracy for? We aren't really
sure how people will use this product, we just thought it would be cool to
make (while I'm on the subject, are there any features you would
want/expect at this price point that we aren't offering?). Also, what are
your thoughts on Atomichron's subscription service?
Looking forward to having a discussion about this technology with you all.
I really appreciate niche communities like this!
Happy hacking,
Sean
--
Sean Hollister (He/Him)
Marketing Amplification and Outreach Manager
E: sean.hollister@sparkfun.com
P: (303) 443-0048 M: (303) 000-1234
A: 6333 Dry Creek Pkwy, Niwot, CO 80503
Hi
Since you brought it up …. :) :)
Just how in the world do you get an Atomicron subscription? I’ve asked about this a number of times.
I have never been able to get a “click here” sort of answer.
Bob
On Mar 17, 2025, at 2:41 PM, Sean Hollister via time-nuts time-nuts@lists.febo.com wrote:
Hi Time Nuts,
I'm Sean and I work at SparkFun Electronics. Historically, my company has
made dev boards and primarily been in the embedded space, but our founder
has recently gotten really into the world of high-precision GNSS, which has
also led us into the wacky world of timing.
I'm not really here to sell you anything (but don't let that stop you, hah)
but more curious what your thoughts are on a new product we just came out
with (https://www.sparkfun.com/sparkpnt-gnss-disciplined-oscillator.html).
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps. It uses the mosaic-T
from Septentrio which has anti jamming and spoofing capabilities as well.
On the product page you can find links to the documentation for the
product, as well as the oscillator and GNSS module inside.
We're curious what your thoughts are on the product. Namely, is this a good
accuracy/price ratio compared to the methods you can DIY? Also, what
applications might you be using sub-nano accuracy for? We aren't really
sure how people will use this product, we just thought it would be cool to
make (while I'm on the subject, are there any features you would
want/expect at this price point that we aren't offering?). Also, what are
your thoughts on Atomichron's subscription service?
Looking forward to having a discussion about this technology with you all.
I really appreciate niche communities like this!
Happy hacking,
Sean
--
Sean Hollister (He/Him)
Marketing Amplification and Outreach Manager
E: sean.hollister@sparkfun.com
P: (303) 443-0048 M: (303) 000-1234
A: 6333 Dry Creek Pkwy, Niwot, CO 80503
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Had to dig to find the phase noise spec - one reason you discipline an oscillator is to deal with the long term drift/Allan dev, but often you need good phase noise (e.g. if you're multiplying it up to microwave frequencies, or if you're using it as a sampling clock in a SDR, particularly one that's undersampling).
The SiT5358 isn't super high performance in that area. Comparing to a fairly vanilla inexpensive Wenzel Streamline OCXO (it was $100-200 when I got it 15 years ago):
Offset Wenzel SiT5358
10 -130 -108
100 -155 -127
1k -165 -148
10k -165 -154
100k -165 -154
If you're multiplying 10 MHz up to, say, 10 GHz, for a narrow band (<1 MHz BW) communication link, that close in noise is a real problem. 10-20 dB is a big hit (when you're obsessing about driving the NF down to a fraction of a dB).
As to potential uses
Radio interferometry is what springs to my mind (since that's what I'm doing at work) - for low frequency interferometry (<30 MHz) you want timing uncertainty <1 ns, with post processing. And that's sort of the key - most interferometers do a fair amount of post processing of the raw GNSS observables to solve for the clock bias.
Some kinds of Synthetic Aperture Radar (particularly bistatic) might use something like this (but the close in phase noise might be a problem, especially if you're looking at Doppler).
Other possible uses are synchronizing distributed sources where it's impractical to run a cable or radio link - I'm not sure how good that has to be (for instance, TV broadcasts that are "simulcast" from multiple transmitters- I know there's people on the list that know this, though).
Also, it's not super clear what that Atomichron service is doing..I went to the link, but it just wants me to subscribe. Some browsing found:
Timing accuracy: <5 ns UTC (95% of the time)
Accuracy: <1 ns to Fugro AtomiChron® timescale (independent of world-wide location
. I'm not sure what the Fugro timescale is and how it relates to, say, TAI. 5ns is something you might be able to get with a disciplined OXCO without the service.
And "Proprietary Global Reference Network" is kind of scary in a "research" application - might be fine in a commercial application. Seems it provides high availability, which might be justified if you have something like time tagging financial transactions.
On Mon, 17 Mar 2025 12:41:34 -0600, Sean Hollister via time-nuts time-nuts@lists.febo.com wrote:
Hi Time Nuts,
I'm Sean and I work at SparkFun Electronics. Historically, my company has
made dev boards and primarily been in the embedded space, but our founder
has recently gotten really into the world of high-precision GNSS, which has
also led us into the wacky world of timing.
I'm not really here to sell you anything (but don't let that stop you, hah)
but more curious what your thoughts are on a new product we just came out
with (https://www.sparkfun.com/sparkpnt-gnss-disciplined-oscillator.html).
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps. It uses the mosaic-T
from Septentrio which has anti jamming and spoofing capabilities as well.
On the product page you can find links to the documentation for the
product, as well as the oscillator and GNSS module inside.
We're curious what your thoughts are on the product. Namely, is this a good
accuracy/price ratio compared to the methods you can DIY? Also, what
applications might you be using sub-nano accuracy for? We aren't really
sure how people will use this product, we just thought it would be cool to
make (while I'm on the subject, are there any features you would
want/expect at this price point that we aren't offering?). Also, what are
your thoughts on Atomichron's subscription service?
Looking forward to having a discussion about this technology with you all.
I really appreciate niche communities like this!
Happy hacking,
Sean
--
Sean Hollister (He/Him)
Marketing Amplification and Outreach Manager
E: sean.hollister@sparkfun.com
P: (303) 443-0048 M: (303) 000-1234
A: 6333 Dry Creek Pkwy, Niwot, CO 80503
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
I've been really excited by the sparkfun timing (and RTK) products.
I'm also a fan of the Septentrio receivers, though I've not used the
-T parts yet. This looks like an interesting product.
I've thought about playing with the SI digitally controlled
oscillators before, particularly because of the digital interface
likely eliminates a bunch of sources of problems for anyone nervous
about making their own GPSDO like noise on the control lines trashing
the performance.
I have a bit of an impression that the price point might be somewhat
mismatched with the cost of the Atomichron subscription, in the sense
that the additional cost of a MAC-SA53 or PRS10 would be insubstantial
compared to the atomicchron cost and would presumably at least improve
the close in adev. Obviously it's easy to backseat quarterback
someone elses design when I lack info like how much the close in adev
is really from PLL vs the underlying oscillator-- so I don't mean this
as a criticism but I might guess that anyone splurging on the
atomichron would probably appreciate some alternative on the
oscillator.
That said, I think for a lot of time-nuts sort of audience (any
myself, for that matter) would probably PLL a surplus PRS10 against
the 1PPS output and then not really care about the <100s adev or the
phase noise of the 10MHz output. The sparkpnt shouldn't, if I
understand it, have any sawtooth on the 1PPS which is usually one of
the biggest irritations for just PLLing an external oscillator against
a GPS output.
I had the impression that one common reason people who want precise
timing is to sync radio receivers at different locations for
beamforming/positioning/etc. So an interesting figure of merit might
be adev between two static units 10km apart or something. Since many
of the error sources should be common between the two receivers
perhaps the sync performance would be better implied by the
performance charts.
I think (???) the mosiac-t has some time transfer affordances which
might be interesting to time-nuts residents. To e.g. transfer some
list members 5071a or maser time to other people with GNSS satellite
in common view. But I don't believe the sparkpnt exposes enough for
this (I think you'd want to switch the 10mhz source to their external
clock). In terms of potentially 'missing' time nuts relevant
functionality I'd probably put time transfer at the top of the list.
If the specs have the power usage, I missed it-- it's a useful figure
for planning since no clock has holdover longer than its power can
stay on.
On Mon, Mar 17, 2025 at 10:05 PM Sean Hollister via time-nuts
time-nuts@lists.febo.com wrote:
Hi Time Nuts,
I'm Sean and I work at SparkFun Electronics. Historically, my company has
made dev boards and primarily been in the embedded space, but our founder
has recently gotten really into the world of high-precision GNSS, which has
also led us into the wacky world of timing.
I'm not really here to sell you anything (but don't let that stop you, hah)
but more curious what your thoughts are on a new product we just came out
with (https://www.sparkfun.com/sparkpnt-gnss-disciplined-oscillator.html).
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps. It uses the mosaic-T
from Septentrio which has anti jamming and spoofing capabilities as well.
On the product page you can find links to the documentation for the
product, as well as the oscillator and GNSS module inside.
We're curious what your thoughts are on the product. Namely, is this a good
accuracy/price ratio compared to the methods you can DIY? Also, what
applications might you be using sub-nano accuracy for? We aren't really
sure how people will use this product, we just thought it would be cool to
make (while I'm on the subject, are there any features you would
want/expect at this price point that we aren't offering?). Also, what are
your thoughts on Atomichron's subscription service?
Looking forward to having a discussion about this technology with you all.
I really appreciate niche communities like this!
Happy hacking,
Sean
--
Sean Hollister (He/Him)
Marketing Amplification and Outreach Manager
E: sean.hollister@sparkfun.com
P: (303) 443-0048 M: (303) 000-1234
A: 6333 Dry Creek Pkwy, Niwot, CO 80503
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
When digging into the SiT5358 datasheet, I noticed that the tuning input goes to an ADC and that the step size is 5 parts per trillion.
Compared to the PRS10 being documented as having a 22 bit ADC, with tuning steps of 2 parts per trillion, better phase noise and the possibility of loop time constants up to 18 hours, that is a deal killer.
My use case would be more demanding that most, in that I'd probably be using it as a fixed reference to drive all of the 10MHz reference inputs on a dedicated E5504 phase noise system. For that I need extremely low phase noise. While that doesn't require extreme frequency accuracy, it would be desirable for other purposes. Ideally one source would serve both needs.
Doing something like slaving a 10811 to something like a PRS10 with a time constant of hours in principle is close to ideal, because a PHM or a cesium isn't really in the budget.
From: Jim Lux via time-nuts time-nuts@lists.febo.com
Sent: Monday, March 17, 2025 7:44:12 PM
To: time-nuts@lists.febo.com time-nuts@lists.febo.com
Cc: sean.hollister@sparkfun.com sean.hollister@sparkfun.com; Jim Lux jim@luxfamily.com
Subject: [time-nuts] Re: SparkFun GNSS Disciplined Oscillator
Had to dig to find the phase noise spec - one reason you discipline an oscillator is to deal with the long term drift/Allan dev, but often you need good phase noise (e.g. if you're multiplying it up to microwave frequencies, or if you're using it as a sampling clock in a SDR, particularly one that's undersampling).
The SiT5358 isn't super high performance in that area. Comparing to a fairly vanilla inexpensive Wenzel Streamline OCXO (it was $100-200 when I got it 15 years ago):
Offset Wenzel SiT5358
10 -130 -108
100 -155 -127
1k -165 -148
10k -165 -154
100k -165 -154
If you're multiplying 10 MHz up to, say, 10 GHz, for a narrow band (<1 MHz BW) communication link, that close in noise is a real problem. 10-20 dB is a big hit (when you're obsessing about driving the NF down to a fraction of a dB).
As to potential uses
Radio interferometry is what springs to my mind (since that's what I'm doing at work) - for low frequency interferometry (<30 MHz) you want timing uncertainty <1 ns, with post processing. And that's sort of the key - most interferometers do a fair amount of post processing of the raw GNSS observables to solve for the clock bias.
Some kinds of Synthetic Aperture Radar (particularly bistatic) might use something like this (but the close in phase noise might be a problem, especially if you're looking at Doppler).
Other possible uses are synchronizing distributed sources where it's impractical to run a cable or radio link - I'm not sure how good that has to be (for instance, TV broadcasts that are "simulcast" from multiple transmitters- I know there's people on the list that know this, though).
Also, it's not super clear what that Atomichron service is doing..I went to the link, but it just wants me to subscribe. Some browsing found:
Timing accuracy: <5 ns UTC (95% of the time)
Accuracy: <1 ns to Fugro AtomiChron® timescale (independent of world-wide location
. I'm not sure what the Fugro timescale is and how it relates to, say, TAI. 5ns is something you might be able to get with a disciplined OXCO without the service.
And "Proprietary Global Reference Network" is kind of scary in a "research" application - might be fine in a commercial application. Seems it provides high availability, which might be justified if you have something like time tagging financial transactions.
On Mon, 17 Mar 2025 12:41:34 -0600, Sean Hollister via time-nuts time-nuts@lists.febo.com wrote:
Hi Time Nuts,
I'm Sean and I work at SparkFun Electronics. Historically, my company has
made dev boards and primarily been in the embedded space, but our founder
has recently gotten really into the world of high-precision GNSS, which has
also led us into the wacky world of timing.
I'm not really here to sell you anything (but don't let that stop you, hah)
but more curious what your thoughts are on a new product we just came out
with (https://urldefense.proofpoint.com/v2/url?u=https-3A__www.sparkfun.com_sparkpnt-2Dgnss-2Ddisciplined-2Doscillator.html&d=DwIGaQ&c=euGZstcaTDllvimEN8b7jXrwqOf-v5A_CdpgnVfiiMM&r=JsDsKeR7cZC8wbZhIlxxBQ&m=ESHpC_BtCWK4zOfoW4gWbyNR3zxpSmvim5C8OcXniGgrldqLvdg5bOTHZMl0gnGp&s=PfyD7LbH2lGG5qen7VzOvZkVmgfyiQYhyE-hQwg6enM&e=).
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps. It uses the mosaic-T
from Septentrio which has anti jamming and spoofing capabilities as well.
On the product page you can find links to the documentation for the
product, as well as the oscillator and GNSS module inside.
We're curious what your thoughts are on the product. Namely, is this a good
accuracy/price ratio compared to the methods you can DIY? Also, what
applications might you be using sub-nano accuracy for? We aren't really
sure how people will use this product, we just thought it would be cool to
make (while I'm on the subject, are there any features you would
want/expect at this price point that we aren't offering?). Also, what are
your thoughts on Atomichron's subscription service?
Looking forward to having a discussion about this technology with you all.
I really appreciate niche communities like this!
Happy hacking,
Sean
--
Sean Hollister (He/Him)
Marketing Amplification and Outreach Manager
E: sean.hollister@sparkfun.com
P: (303) 443-0048 M: (303) 000-1234
A: 6333 Dry Creek Pkwy, Niwot, CO 80503
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Hi,
I really like that SparkFun is getting into timing! I’ve been following you guys for a little while now, it’s been cool to see!
From a very different perspective, I am using a GPSDO (and some custom built servers & boards) to serve NTP - and hopefully PTP soon - to my servers in a rack. I didn’t see whether or not this device serves NTP, but it would be immensely useful to me for it to do so. The price is a little high for a hobbyist, but I recognize that the device is not really aimed at hobbyists.
Other than that, it looks awesome!
Thanks,
Eamonn Nugent
On Mar 17, 2025, at 6:06 PM, Sean Hollister via time-nuts time-nuts@lists.febo.com wrote:
Hi Time Nuts,
I'm Sean and I work at SparkFun Electronics. Historically, my company has
made dev boards and primarily been in the embedded space, but our founder
has recently gotten really into the world of high-precision GNSS, which has
also led us into the wacky world of timing.
I'm not really here to sell you anything (but don't let that stop you, hah)
but more curious what your thoughts are on a new product we just came out
with (https://www.sparkfun.com/sparkpnt-gnss-disciplined-oscillator.html).
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps. It uses the mosaic-T
from Septentrio which has anti jamming and spoofing capabilities as well.
On the product page you can find links to the documentation for the
product, as well as the oscillator and GNSS module inside.
We're curious what your thoughts are on the product. Namely, is this a good
accuracy/price ratio compared to the methods you can DIY? Also, what
applications might you be using sub-nano accuracy for? We aren't really
sure how people will use this product, we just thought it would be cool to
make (while I'm on the subject, are there any features you would
want/expect at this price point that we aren't offering?). Also, what are
your thoughts on Atomichron's subscription service?
Looking forward to having a discussion about this technology with you all.
I really appreciate niche communities like this!
Happy hacking,
Sean
--
Sean Hollister (He/Him)
Marketing Amplification and Outreach Manager
E: sean.hollister@sparkfun.com
P: (303) 443-0048 M: (303) 000-1234
A: 6333 Dry Creek Pkwy, Niwot, CO 80503
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Sean Hollister via time-nuts writes:
Without the Atomichron subscription ($1k/month), it can get accuracy around
5ns. With the subscription, it gets to within 100 ps.
Are those numbers based on vendor specifications, or did you hike up to NIST to measure it ?
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Time-nuts!
Without the Atomichron subscription ($1k/month), it can get
accuracy around 5ns. With the subscription, it gets to within 100
ps.
UTC(NIST) can vary from UTC(USNO) by more than 5 ns.
https://www.nist.gov/pml/time-and-frequency-division/time-services/nist-usno
So waht does "within 100 ps" mean?
RGDS
GARY
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
gem@rellim.com Tel:+1 541 382 8588
Veritas liberabit vos. -- Quid est veritas?
"If you can't measure it, you can't improve it." - Lord Kelvin
On Tue, Mar 18, 2025 at 8:50 PM Eamonn Nugent via time-nuts <
time-nuts@lists.febo.com> wrote:
From a very different perspective, I am using a GPSDO (and some custom
built servers & boards) to serve NTP - and hopefully PTP soon - to my
servers in a rack. I didn’t see whether or not this device serves NTP, but
it would be immensely useful to me for it to do so. The price is a little
high for a hobbyist, but I recognize that the device is not really aimed at
hobbyists.
So the septentrio mosaic receivers do NTP at least, though so far I've
found their performance isn't particularly impressive compared to a
laureline.
I have a septentrio mosaic-x5 receiver (adrusimple board) connected to the
same antenna and same ethernet switch as a laureline gps. Logging is
enabled on the mosaic-x5 (since what I use it for is rtk postprocessing).
Chrony sourcestats on a host monitoring both reports 414ns std_dev for the
laureline and 122us for the mosaic. There are two switches between the host
and the receiver and the host is doing some other stuff, so its the
relative figures that are interesting. I see roughly the same numbers when
the laureline is noselect and the mosaic is prefer, so it's not due to
chrony preferring the laureline.
I've not yet tested with sparkpnt or a mosaic-t board, I'll post again when
I do.
In any case the performance is fine for typical computer usage and better
than NTP over the internet. My WAG is that NTP support is just some random
userspace process on the receiver and even with a better clock and receiver
it will underperform a super optimized ntp device since on an NTP device
the biggest jitter sources are the ones that come from handling the
ethernet. But maybe the story is different on mosaic-t.
(And obviously hold-over is going to be better with a source with a better
oscillator... and offset to UTC is going to be more trustworthy from a
multi-frequency device and/or one with a PPP resolved antenna location...).
If you want casual PTP, the Mosaic-T is being wasted.
There are a few SKUs on Mellanox(now nvidia) ConnectX-6 DX cards that have PPS and a 10 MHz input. They’re rare enough to be hard to even find new ones with published prices. Used last I looked they were in the 600 USD price range.
Then there’s the timebeat board to go to GPIO header in certain Intel multigig and higher Ethernet cards. While that’s inexpensive, it’s a bit of a hack with some limitations.
Then there’s two Intel X810 cards with timing versions that can be bought with or without the embedded uBlox ZED-F9T. One card has four 1/10/25G SFP28 ports. The other has two 10/25/40/50/100G ports that can be configured with fanout cables to give 8x10G,2x40G,2x50G or other combinations. Due to the lack of 1G rate on the QSFP optics, 10G is slowest it supports. Both of the Intel timing cards support synchronous Ethernet, and Intel has a pdf that goes over the particulars for how to configure it. It claims to support one step or two step PTP, as well as telecom profiles. Those can be had for somewhere in the $300-700 range on eBay for what’s probably new old stock even though it’s a current product. The Intel X810 NICs claim I think around 1us holdover over four hours without GNSS PPS signal. The seem to be made to serve the cellular market for coherent processing of 4G/5G and E911 without GNSS on a user terminal.
I suppose the other option would be white rabbit switch hardware. I haven’t a clue what those cost.
Finally, the last and probably cheapest option would be certain Cisco routers that support synchronous Ethernet, and have PPS and 10MHz input. Those might be interesting to mount at a tower to provide a 10MHz reference over fiber. I did confirm frequency transfer by slewing 10MHz from 3335A to input and watching the outside it frequency lock and follow. I didn’t characterize phase noise (yet).
I spent probably a month worth of evenings studying the hardware choices, and caveats and I’d be happy to share.
I’m personally really interested in comparing the performance of a rubidium disciplined by a LEA-M8F, M8T, ZED-F9T, and the Mosaic-T and a 10811 slaved to each of the above with a loop filter with milliertz bandwidth to hide the rubidium frequency steps from internal DAC on my PRS-10.
From: Greg Maxwell via time-nuts time-nuts@lists.febo.com
Sent: Tuesday, March 18, 2025 5:56:45 PM
To: Discussion of precise time and frequency measurement time-nuts@lists.febo.com
Cc: Greg Maxwell gmaxwell@gmail.com
Subject: [time-nuts] Re: SparkFun GNSS Disciplined Oscillator
On Tue, Mar 18, 2025 at 8:50 PM Eamonn Nugent via time-nuts <
time-nuts@lists.febo.com> wrote:
From a very different perspective, I am using a GPSDO (and some custom
built servers & boards) to serve NTP - and hopefully PTP soon - to my
servers in a rack. I didn’t see whether or not this device serves NTP, but
it would be immensely useful to me for it to do so. The price is a little
high for a hobbyist, but I recognize that the device is not really aimed at
hobbyists.
So the septentrio mosaic receivers do NTP at least, though so far I've
found their performance isn't particularly impressive compared to a
laureline.
I have a septentrio mosaic-x5 receiver (adrusimple board) connected to the
same antenna and same ethernet switch as a laureline gps. Logging is
enabled on the mosaic-x5 (since what I use it for is rtk postprocessing).
Chrony sourcestats on a host monitoring both reports 414ns std_dev for the
laureline and 122us for the mosaic. There are two switches between the host
and the receiver and the host is doing some other stuff, so its the
relative figures that are interesting. I see roughly the same numbers when
the laureline is noselect and the mosaic is prefer, so it's not due to
chrony preferring the laureline.
I've not yet tested with sparkpnt or a mosaic-t board, I'll post again when
I do.
In any case the performance is fine for typical computer usage and better
than NTP over the internet. My WAG is that NTP support is just some random
userspace process on the receiver and even with a better clock and receiver
it will underperform a super optimized ntp device since on an NTP device
the biggest jitter sources are the ones that come from handling the
ethernet. But maybe the story is different on mosaic-t.
(And obviously hold-over is going to be better with a source with a better
oscillator... and offset to UTC is going to be more trustworthy from a
multi-frequency device and/or one with a PPP resolved antenna location...).
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com