Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
You do need something to actually respond to the serial port interrups--
ie a serial port interrupt driver. Linux includes on
I have a "gps" script in /etc/init.d
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>> years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said, ppscheck works fine and PPS seems to get processed if gpsd is run as an ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
On 07/28/21 20:22, Jim Pennino wrote:\
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing
the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and
xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over
the
years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
William Unruh <unruh@invalid.ca> wrote:
<snip old stuff>
You do need something to actually respond to the serial port interrups--
ie a serial port interrupt driver. Linux includes on
I have a "gps" script in /etc/init.d
gpsd does all that for you and simplifies configuration, which is why
I'm using gpsd at all.
I did a modprobe check to make sure all the pps stuff was loaded; it
was.
Next is a post to the gpsd support group.
If all else fails, gpsd gets dumped and ntpd gets configured for the
NMEA interface.
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>>> years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
chris <chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>>>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>>>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>>>> years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from
several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz
+/- 0.0002 Hz output has been verified.
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>>> years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
On 2021-07-28, chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>>>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>>>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>>>> years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from
several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
Nice theory. But AFAIK almost noone builds RS232 chips/booards with
follow the RS232 guidelines. I do not say noone because you may well
come up with an example, but why should they follow the standards?
Nothing else does. That standard was set about 1950, (Oops I exagerate,
it was 1960)when they were
driven by tubes.
If the only thing changed is the serial card then it's most likely
something related...
It is possible. Just as it is possible that if your car does not start,
you have to insert a crank into a little hole under the radiator and
turn it.
But AFAIK almost noone builds RS232 chips/booards with
follow the RS232 guidelines.
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>>>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>>>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>>>> years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from
several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz
+/- 0.0002 Hz output has been verified.
On 07/28/21 22:18, chris wrote:
On 07/28/21 20:22, Jim Pennino wrote:\
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing >>>>> the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and
xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over >>>>> the
years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a
teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to
rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from
several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
Here's the relevant fragment from ntp.conf, dcd line as pps input:
server 127.127.22.0 minpoll 2 maxpoll 2
fudge 127.127.22.0 flag2 1 flag3 1 flag4 1
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-28, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>>>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial >>>>>> post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you >>>>>> can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said, >>>>> ppscheck works fine and PPS seems to get processed if gpsd is run as an >>>>> ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD. >>>> The time tools pps is ttl, as are all the other gps do boxes here, from >>>> several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels >>>> are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is >>>> only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the
nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz
+/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
True, but it is just fine and dandy for calibrating frequency counters without having to send them somewhere and it allows testing of radio
dial accuracy no matter the state of the ionosphere.
Also that is one part in 10^11, and GPS is incapable of giving that.
I presume that the PPS output is once per second, and that the pulse
width is not too small (milliseconds) , or gpsd might well have trouble recogizing it as
a pulse.
This is a commercial box sold specifically as a time and frequency
standard and purchased because it's specifications are only a bit worse
than a rubidium standard and about $600 cheaper.
I have already verified that gpsd CAN recognize the PPS on CTS, but NOT running with the permissions that it needs to feed ntp.
On 2021-07-28, Jim Pennino <jimp@gonzo.specsol.net> wrote:
chris <chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the >>>>>> PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps >>>>>> did show PPS.
Guessing it was some sort of permissions issue I had introduced over the >>>>>> years, I did an apt purge gpsd, apt install gpsd and started fresh with >>>>>> a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial >>>>> post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm
using here. Single supply line, 5 or 12 v, don't remember, that you
can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said, >>>> ppscheck works fine and PPS seems to get processed if gpsd is run as an >>>> ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from
several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels
are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the
nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz
+/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
Also that is one part in 10^11, and GPS is incapable of giving that.
I presume that the PPS output is once per second, and that the pulse
width is not too small (milliseconds) , or gpsd might well have trouble recogizing it as
a pulse.
On 2021-07-28, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
<snip old stuff>
You do need something to actually respond to the serial port interrups-- >>> ie a serial port interrupt driver. Linux includes on
I have a "gps" script in /etc/init.d
gpsd does all that for you and simplifies configuration, which is why
I'm using gpsd at all.
Except it does not work!
I did a modprobe check to make sure all the pps stuff was loaded; it
was.
Do
cat /sys/devices/virtual/pps/pps0/assert and
cat sys/devices/virtual/pps/pps0/clear
to see if either is giving you signals once a second when the gpsd is supposed to be servicing the interrupts.
(run each successively)
It is usually assert you want. Your gpsd might be trying to read the
clear, and it never changes.
Next is a post to the gpsd support group.
If all else fails, gpsd gets dumped and ntpd gets configured for the
NMEA interface.
If all you want is the time to the nearest second then that is fine. If
you want it to better than a few tens of ms, then it is not.
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
I've used ex telco gps do for frequency standard work for years and they
typically have a pps offset in the 10nS range. You can even adjust the
antenna cable length to compensate. Have a look at the time nuts
mailing list for more info...
I bought this thing because I was interested in the state of the art of affordable devices.
I am well past the point in my life when I would haunt surplus stores
looking for treasure. I might violate that if I find a good source of
really cheap wave guide.
I've used ex telco gps do for frequency standard work for years and they typically have a pps offset in the 10nS range. You can even adjust the antenna cable length to compensate. Have a look at the time nuts
mailing list for more info...
On 07/29/21 00:24, Jim Pennino wrote:
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-28, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device,
reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not
processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon >>>>>>>> shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 >>>>>>>> and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced >>>>>>>> over the
years, I did an apt purge gpsd, apt install gpsd and started
fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a
serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm >>>>>>> using here. Single supply line, 5 or 12 v, don't remember, that you >>>>>>> can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I
said,
ppscheck works fine and PPS seems to get processed if gpsd is run
as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running
FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here,
from
several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels >>>>> are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive
chip is
only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the
nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz
+/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
True, but it is just fine and dandy for calibrating frequency counters
without having to send them somewhere and it allows testing of radio
dial accuracy no matter the state of the ionosphere.
Also that is one part in 10^11, and GPS is incapable of giving that.
I presume that the PPS output is once per second, and that the pulse
width is not too small (milliseconds) , or gpsd might well have
trouble recogizing it as
a pulse.
This is a commercial box sold specifically as a time and frequency
standard and purchased because it's specifications are only a bit worse
than a rubidium standard and about $600 cheaper.
I have already verified that gpsd CAN recognize the PPS on CTS, but NOT
running with the permissions that it needs to feed ntp.
Interesting, as some serial device chips have internal hardware locking between some of the control functions. I used the dcd line for the
system here, which was fine, but haven't tested with the cts line...
On 07/29/21 00:32, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
I've used ex telco gps do for frequency standard work for years and they >>> typically have a pps offset in the 10nS range. You can even adjust the
antenna cable length to compensate. Have a look at the time nuts
mailing list for more info...
I bought this thing because I was interested in the state of the art of
affordable devices.
I am well past the point in my life when I would haunt surplus stores
looking for treasure. I might violate that if I find a good source of
really cheap wave guide.
Depends on the item, but have more or less equipped the lab here with
older high end test gear, even some parts / repair. Trade time against
money. Never get tired of just making stuff work...
On 07/29/21 00:24, Jim Pennino wrote:
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-28, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured >>>>>>>> and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows >>>>>>>> only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial >>>>>>> post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm >>>>>>> using here. Single supply line, 5 or 12 v, don't remember, that you >>>>>>> can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said, >>>>>> ppscheck works fine and PPS seems to get processed if gpsd is run as an >>>>>> ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD. >>>>> The time tools pps is ttl, as are all the other gps do boxes here, from >>>>> several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels >>>>> are incorrect. Even if it does work, there may be excessive jitter,
which doesn't help performance. Propagation delay using a drive chip is >>>>> only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely
something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the
nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz
+/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
True, but it is just fine and dandy for calibrating frequency counters
without having to send them somewhere and it allows testing of radio
dial accuracy no matter the state of the ionosphere.
Also that is one part in 10^11, and GPS is incapable of giving that.
I presume that the PPS output is once per second, and that the pulse
width is not too small (milliseconds) , or gpsd might well have trouble recogizing it as
a pulse.
This is a commercial box sold specifically as a time and frequency
standard and purchased because it's specifications are only a bit worse
than a rubidium standard and about $600 cheaper.
I have already verified that gpsd CAN recognize the PPS on CTS, but NOT
running with the permissions that it needs to feed ntp.
Interesting, as some serial device chips have internal hardware locking between some of the control functions. I used the dcd line for the
system here, which was fine, but haven't tested with the cts line...
Ignore above but how about permissions on the serial device itself ?...
chris <chris-nospam@tridac.net> wrote:
On 07/29/21 00:24, Jim Pennino wrote:
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-28, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial >>>>>>>> post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm >>>>>>>> using here. Single supply line, 5 or 12 v, don't remember, that you >>>>>>>> can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said, >>>>>>> ppscheck works fine and PPS seems to get processed if gpsd is run as an >>>>>>> ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD. >>>>>> The time tools pps is ttl, as are all the other gps do boxes here, from >>>>>> several vendors collected over the years. While such mismatch may
work with some interfaces, there's no guarantee, as the voltage levels >>>>>> are incorrect. Even if it does work, there may be excessive jitter, >>>>>> which doesn't help performance. Propagation delay using a drive chip is >>>>>> only 10's of nS, havng measure it on a scope, so can be ignored.
If the only thing changed is the serial card then it's most likely >>>>>> something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO
box which has an RS-232 interface and a specified PPS accuracy in the >>>>> nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz >>>>> +/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
True, but it is just fine and dandy for calibrating frequency counters
without having to send them somewhere and it allows testing of radio
dial accuracy no matter the state of the ionosphere.
Also that is one part in 10^11, and GPS is incapable of giving that.
I presume that the PPS output is once per second, and that the pulse
width is not too small (milliseconds) , or gpsd might well have trouble recogizing it as
a pulse.
This is a commercial box sold specifically as a time and frequency
standard and purchased because it's specifications are only a bit worse
than a rubidium standard and about $600 cheaper.
I have already verified that gpsd CAN recognize the PPS on CTS, but NOT
running with the permissions that it needs to feed ntp.
Interesting, as some serial device chips have internal hardware locking
between some of the control functions. I used the dcd line for the
system here, which was fine, but haven't tested with the cts line...
There are no "serial device chips", there is a commercial PCI interface
board specified to work as RS-232 ports with full hardware handshake.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
I've used ex telco gps do for frequency standard work for years and they typically have a pps offset in the 10nS range
On 2021-07-29, Jim Pennino <jimp@gonzo.specsol.net> wrote:
chris <chris-nospam@tridac.net> wrote:
On 07/29/21 00:24, Jim Pennino wrote:
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-28, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial >>>>>>>>> post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>>>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm >>>>>>>>> using here. Single supply line, 5 or 12 v, don't remember, that you >>>>>>>>> can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u
time tools gps ntp server, and a micro atx motherboard running FreeBSD. >>>>>>> The time tools pps is ttl, as are all the other gps do boxes here, from >>>>>>> several vendors collected over the years. While such mismatch may >>>>>>> work with some interfaces, there's no guarantee, as the voltage levels >>>>>>> are incorrect. Even if it does work, there may be excessive jitter, >>>>>>> which doesn't help performance. Propagation delay using a drive chip is >>>>>>> only 10's of nS, havng measure it on a scope, so can be ignored. >>>>>>>
If the only thing changed is the serial card then it's most likely >>>>>>> something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO >>>>>> box which has an RS-232 interface and a specified PPS accuracy in the >>>>>> nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz >>>>>> +/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
True, but it is just fine and dandy for calibrating frequency counters >>>> without having to send them somewhere and it allows testing of radio
dial accuracy no matter the state of the ionosphere.
Also that is one part in 10^11, and GPS is incapable of giving that. >>>>> I presume that the PPS output is once per second, and that the pulse >>>>> width is not too small (milliseconds) , or gpsd might well have trouble recogizing it as
a pulse.
This is a commercial box sold specifically as a time and frequency
standard and purchased because it's specifications are only a bit worse >>>> than a rubidium standard and about $600 cheaper.
I have already verified that gpsd CAN recognize the PPS on CTS, but NOT >>>> running with the permissions that it needs to feed ntp.
Interesting, as some serial device chips have internal hardware locking
between some of the control functions. I used the dcd line for the
system here, which was fine, but haven't tested with the cts line...
There are no "serial device chips", there is a commercial PCI interface
board specified to work as RS-232 ports with full hardware handshake.
But those "PCI interface board" contains "serial device chips"
On 2021-07-28, Jim Pennino <jimp@gonzo.specsol.net> wrote:
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it
looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
How exactly is gpsd started? With the serial device specified on its
command line, or is it added only later by gpsdctl (via udev+systemd)
when gpsd already dropped the root privileges?
So a ps shows:
gpsd 1483 1 0 Jul28 ? 00:00:52 /usr/sbin/gpsd -n -b -s9600 /dev/ttyS4 /dev/pps0
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-29, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 00:24, Jim Pennino wrote:
William Unruh<unruh@invalid.ca> wrote:
On 2021-07-28, Jim Pennino<jimp@gonzo.specsol.net> wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 20:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/28/21 17:56, Jim Pennino wrote:
Using Ubuntu 20.04.2 and gpsd 3.20.
I replaced the long running USB GPS with a serial device, reconfigured
and rebooted.
NMEA data is just fine and ppscheck works but gpsd is not processing the
PPS signal on CTS.
ppstest and ppswatch fail, xgps does not show PPS, and ntpshmmon shows
only NTP0, i.e. not NTP1 and NTP2.
I ran gpsd as a regular user with -D3 -N to look for errors but it >>>>>>>>>>> looked OK and was accepting PPS. ntpshmmon showed NTP2 and NTP3 and xgps
did show PPS.
Guessing it was some sort of permissions issue I had introduced over the
years, I did an apt purge gpsd, apt install gpsd and started fresh with
a reboot.
No change.
Thinking maybe apparmor was doing something, I stopped it and did a >>>>>>>>>>> teardown.
No change.
Now I have no clue...
Any suggestions?
One possibility might be that the pps is at ttl level, while a serial
post needs rs3232 +/- levels to trigger properly. You can buy ttl to >>>>>>>>>> rs232 board level adapter for a few $ on Ebay, which is what i'm >>>>>>>>>> using here. Single supply line, 5 or 12 v, don't remember, that you >>>>>>>>>> can tap off from the host b/board...
Nope, this is a RS-232 device connected to a RS-232 port and as I said,
ppscheck works fine and PPS seems to get processed if gpsd is run as an
ordinary user, but not if started by root and run as gpsd.
This is NOT a hobby device.
Worth a try. It was the problem on the setup here, a s/hand 1u >>>>>>>> time tools gps ntp server, and a micro atx motherboard running FreeBSD.
The time tools pps is ttl, as are all the other gps do boxes here, from
several vendors collected over the years. While such mismatch may >>>>>>>> work with some interfaces, there's no guarantee, as the voltage levels >>>>>>>> are incorrect. Even if it does work, there may be excessive jitter, >>>>>>>> which doesn't help performance. Propagation delay using a drive chip is
only 10's of nS, havng measure it on a scope, so can be ignored. >>>>>>>>
If the only thing changed is the serial card then it's most likely >>>>>>>> something related...
What changed is I replaced the USB GPS with a GNSS disciplined OCXO >>>>>>> box which has an RS-232 interface and a specified PPS accuracy in the >>>>>>> nanoseconds.
It remains to be seen if it actually does nanoseconds, but the 10 MHz >>>>>>> +/- 0.0002 Hz output has been verified.
Which is pretty useless for timing purposes.
True, but it is just fine and dandy for calibrating frequency counters >>>>> without having to send them somewhere and it allows testing of radio >>>>> dial accuracy no matter the state of the ionosphere.
Also that is one part in 10^11, and GPS is incapable of giving that. >>>>>> I presume that the PPS output is once per second, and that the pulse >>>>>> width is not too small (milliseconds) , or gpsd might well have trouble recogizing it as
a pulse.
This is a commercial box sold specifically as a time and frequency
standard and purchased because it's specifications are only a bit worse >>>>> than a rubidium standard and about $600 cheaper.
I have already verified that gpsd CAN recognize the PPS on CTS, but NOT >>>>> running with the permissions that it needs to feed ntp.
Interesting, as some serial device chips have internal hardware locking >>>> between some of the control functions. I used the dcd line for the
system here, which was fine, but haven't tested with the cts line...
There are no "serial device chips", there is a commercial PCI interface
board specified to work as RS-232 ports with full hardware handshake.
But those "PCI interface board" contains "serial device chips"
Yes, they also contain capacitors, resistors, crystals, etc.
An interface board is a commercial item with a specification, i.e. the
board is specified to work in a PCI slot, provide RS-232 ports that have standard connectors with full hardware handshaking and work with all
native OS serial drivers.
Since the board obviously works to specification, obsessing to what may
or may not be happening at the chip level seems to me to be pointless.
On 2021-07-29, Jim Pennino <jimp@gonzo.specsol.net> wrote:
So a ps shows:
gpsd 1483 1 0 Jul28 ? 00:00:52 /usr/sbin/gpsd -n -b -s9600 /dev/ttyS4 /dev/pps0
Is that /dev/pps0 necessary? IIRC that is needed only when gpsd cannot
create the PPS device on its own (similarly to ldattach). If it is
needed, is gpsd started when it already exists?
On 29/07/2021 00:03, chris wrote:
I've used ex telco gps do for frequency standard work for years and they
typically have a pps offset in the 10nS range
Offset relative to what? Do you have an atomic clock that is traceable
to UTC?
The official GPS time transfer accuracy is <= 40ns, so whilst 10ns may
be achievable most of the time, it isn't guaranteed.
Depending on the chip id, some functions may be inhibited if the control lines are in the wrong state. So yes, chip hardware architecture does
matter and not always something that can be abstracted away within the
device driver. Devil is in the detail, as usual...
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 00:32, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
I've used ex telco gps do for frequency standard work for years and they >>>> typically have a pps offset in the 10nS range. You can even adjust the >>>> antenna cable length to compensate. Have a look at the time nuts
mailing list for more info...
I bought this thing because I was interested in the state of the art of
affordable devices.
I am well past the point in my life when I would haunt surplus stores
looking for treasure. I might violate that if I find a good source of
really cheap wave guide.
Depends on the item, but have more or less equipped the lab here with
older high end test gear, even some parts / repair. Trade time against
money. Never get tired of just making stuff work...
The inducement here is that while GPS disciplined oscillators are very
old technology which used to live in 19" 5U rack mount boxes and cost
tens of thousands of dollars, they now consist of some chip sets on a
circuit board in a 4"X4"X2" high box and costs about $180 with shipping.
On 07/29/21 02:11, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 00:32, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
I've used ex telco gps do for frequency standard work for years and they >>>>> typically have a pps offset in the 10nS range. You can even adjust the >>>>> antenna cable length to compensate. Have a look at the time nuts
mailing list for more info...
I bought this thing because I was interested in the state of the art of >>>> affordable devices.
I am well past the point in my life when I would haunt surplus stores
looking for treasure. I might violate that if I find a good source of
really cheap wave guide.
Depends on the item, but have more or less equipped the lab here with
older high end test gear, even some parts / repair. Trade time against
money. Never get tired of just making stuff work...
The inducement here is that while GPS disciplined oscillators are very
old technology which used to live in 19" 5U rack mount boxes and cost
tens of thousands of dollars, they now consist of some chip sets on a
circuit board in a 4"X4"X2" high box and costs about $180 with shipping.
Yes, they do, but they are unlikely to have the phase noise figures of
a better quality unit, nor short term stability. In fact some gpsdo
are using cheap dil vcxo's that that are not even ovenised and wander
all over the place.
Lower cost units should be good enough for ntp use, even if not good
enough for a primary frequency standard...
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
Depending on the chip id, some functions may be inhibited if the control
lines are in the wrong state. So yes, chip hardware architecture does
matter and not always something that can be abstracted away within the
device driver. Devil is in the detail, as usual...
Yeah, so?
The card obviously works to specification, i.e. RS-232 and it does see CTS.
If it did not work to specification, I would return the card to the
vendor and buy a different one.
On 07/29/21 17:45, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
Depending on the chip id, some functions may be inhibited if the control >>> lines are in the wrong state. So yes, chip hardware architecture does
matter and not always something that can be abstracted away within the
device driver. Devil is in the detail, as usual...
Yeah, so?
The card obviously works to specification, i.e. RS-232 and it does see CTS. >>
If it did not work to specification, I would return the card to the
vendor and buy a different one.
This is hard work, obviously never designed any hardware, nor written
code for it. While the card may very well meet spec, the spec can't
cover all the myriad cases of operation or use.The spec is often
generalised, to allow some slack in component selection and to take
account of advances in the art, but it's the corner cases that can
catch out the unwary...
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 17:45, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
Depending on the chip id, some functions may be inhibited if the control >>>> lines are in the wrong state. So yes, chip hardware architecture does
matter and not always something that can be abstracted away within the >>>> device driver. Devil is in the detail, as usual...
Yeah, so?
The card obviously works to specification, i.e. RS-232 and it does see CTS. >>>
If it did not work to specification, I would return the card to the
vendor and buy a different one.
This is hard work, obviously never designed any hardware, nor written
code for it. While the card may very well meet spec, the spec can't
cover all the myriad cases of operation or use.The spec is often
generalised, to allow some slack in component selection and to take
account of advances in the art, but it's the corner cases that can
catch out the unwary...
Do you understand the difference between DIY and purchasing a commercial product sold to thousands upon thousands if not millions of purchasers?
Do you understand that RS-232 is many decades old and is specified by TIA-232-F Interface Between Data Terminal Equipment and Data Circuit-Terminating Equipment Employing Serial Binary Data Interchange, issued in 1997?
The box and the card are both RS-232.
The card and box combination obviously works.
None of this has ANYTHING to do with the issue of gpsd.
chris <chris-nospam@tridac.net> wrote:
On 07/29/21 18:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 17:45, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
Depending on the chip id, some functions may be inhibited if the control >>>>>> lines are in the wrong state. So yes, chip hardware architecture does >>>>>> matter and not always something that can be abstracted away within the >>>>>> device driver. Devil is in the detail, as usual...
Yeah, so?
The card obviously works to specification, i.e. RS-232 and it does see CTS.
If it did not work to specification, I would return the card to the
vendor and buy a different one.
This is hard work, obviously never designed any hardware, nor written
code for it. While the card may very well meet spec, the spec can't
cover all the myriad cases of operation or use.The spec is often
generalised, to allow some slack in component selection and to take
account of advances in the art, but it's the corner cases that can
catch out the unwary...
Do you understand the difference between DIY and purchasing a commercial >>> product sold to thousands upon thousands if not millions of purchasers?
Do you understand that RS-232 is many decades old and is specified by
TIA-232-F Interface Between Data Terminal Equipment and Data
Circuit-Terminating Equipment Employing Serial Binary Data Interchange,
issued in 1997?
The box and the card are both RS-232.
The card and box combination obviously works.
None of this has ANYTHING to do with the issue of gpsd.
That's just the spec for the signals on the wires, nothing to do
with how that is implemented inside devices, or on your pc card,
which may be implemented in many different ways, from bit banging
to fully integrated devices with buffering and even some
intelligence. It only affects your low cost gpsd if you can't
make it work, and need to dig into why it doesn't work...
1) I no more care how the manufacturer inplemented what is inside the
box or on the PC card than I care what kind of steel was used to make
the exhaust manifold on my car.
2) The box is low cost compared to the tens of thousands of dollars such things used to cost.
Thank you for your suggestions. I always notice the host name part of the prompt string, which is different from each other for the test case discussed here, when I'm on different hosts.
3) If the box doesn't work, it goes back for a refund. I have no
interest in fixing things with a warrenty.
4) It is blazingly obvious to anyone paying the slightest bit of
attention that the problem is in software, i.e. gpsd, and not the box.
On 07/29/21 18:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 17:45, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
Depending on the chip id, some functions may be inhibited if the control >>>>> lines are in the wrong state. So yes, chip hardware architecture does >>>>> matter and not always something that can be abstracted away within the >>>>> device driver. Devil is in the detail, as usual...
Yeah, so?
The card obviously works to specification, i.e. RS-232 and it does see CTS.
If it did not work to specification, I would return the card to the
vendor and buy a different one.
This is hard work, obviously never designed any hardware, nor written
code for it. While the card may very well meet spec, the spec can't
cover all the myriad cases of operation or use.The spec is often
generalised, to allow some slack in component selection and to take
account of advances in the art, but it's the corner cases that can
catch out the unwary...
Do you understand the difference between DIY and purchasing a commercial
product sold to thousands upon thousands if not millions of purchasers?
Do you understand that RS-232 is many decades old and is specified by
TIA-232-F Interface Between Data Terminal Equipment and Data
Circuit-Terminating Equipment Employing Serial Binary Data Interchange,
issued in 1997?
The box and the card are both RS-232.
The card and box combination obviously works.
None of this has ANYTHING to do with the issue of gpsd.
That's just the spec for the signals on the wires, nothing to do
with how that is implemented inside devices, or on your pc card,
which may be implemented in many different ways, from bit banging
to fully integrated devices with buffering and even some
intelligence. It only affects your low cost gpsd if you can't
make it work, and need to dig into why it doesn't work...
On 2021-07-29, Jim Pennino <jimp@gonzo.specsol.net> wrote:
chris <chris-nospam@tridac.net> wrote:
On 07/29/21 18:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 07/29/21 17:45, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
<snip old stuff>
Depending on the chip id, some functions may be inhibited if the control
lines are in the wrong state. So yes, chip hardware architecture does >>>>>>> matter and not always something that can be abstracted away within the >>>>>>> device driver. Devil is in the detail, as usual...
Yeah, so?
The card obviously works to specification, i.e. RS-232 and it does see CTS.
If it did not work to specification, I would return the card to the >>>>>> vendor and buy a different one.
This is hard work, obviously never designed any hardware, nor written >>>>> code for it. While the card may very well meet spec, the spec can't
cover all the myriad cases of operation or use.The spec is often
generalised, to allow some slack in component selection and to take
account of advances in the art, but it's the corner cases that can
catch out the unwary...
Do you understand the difference between DIY and purchasing a commercial >>>> product sold to thousands upon thousands if not millions of purchasers? >>>>
Do you understand that RS-232 is many decades old and is specified by
TIA-232-F Interface Between Data Terminal Equipment and Data
Circuit-Terminating Equipment Employing Serial Binary Data Interchange, >>>> issued in 1997?
The box and the card are both RS-232.
The card and box combination obviously works.
None of this has ANYTHING to do with the issue of gpsd.
That's just the spec for the signals on the wires, nothing to do
with how that is implemented inside devices, or on your pc card,
which may be implemented in many different ways, from bit banging
to fully integrated devices with buffering and even some
intelligence. It only affects your low cost gpsd if you can't
make it work, and need to dig into why it doesn't work...
1) I no more care how the manufacturer inplemented what is inside the
box or on the PC card than I care what kind of steel was used to make
the exhaust manifold on my car.
2) The box is low cost compared to the tens of thousands of dollars such
things used to cost.
That is of course completely irrelevant. The same could be said of an aluminum
spoon (cost mcuch more than a gold spoon), and yet it would still not be most people's first choice for
tableware.
Thank you for your suggestions. I always notice the host name part of the prompt string, which is different from each other for the test case discussed here, when I'm on different hosts.
3) If the box doesn't work, it goes back for a refund. I have no
interest in fixing things with a warrenty.
But then as point 4 says, you do not know if it is the box or some misconfiguration (the wetware behind the keyboard) or a bug in gpsd.
4) It is blazingly obvious to anyone paying the slightest bit of
attention that the problem is in software, i.e. gpsd, and not the box.
And yet, you also have been cogitating about the possibility that the
serial port (either on the box or on the computer) does not satisfy the RS232 standards, which is hardware
How about trying the direct use of ldattach, instead of using gpsd, to
see if you can get that to work. If it does, and gpsd does not then your hypothesis that gpsd is at fault is born out. Then you can decide
whether it is better to suffer the slings and arrows of debugging gpsd,
or with one fell swoop use ldattach and have a working system.
William Unruh <unruh@invalid.ca> wrote:
2) The box is low cost compared to the tens of thousands of dollars such >>> things used to cost.
That is of course completely irrelevant. The same could be said of an aluminum
spoon (cost mcuch more than a gold spoon), and yet it would still not be most people's first choice for
tableware.
At one time the cost of aluminum was much higher than gold...
But then as point 4 says, you do not know if it is the box or some
misconfiguration (the wetware behind the keyboard) or a bug in gpsd.
Point 4 says I DO know the problem is in gpsd.
4) It is blazingly obvious to anyone paying the slightest bit of
attention that the problem is in software, i.e. gpsd, and not the box.
And yet, you also have been cogitating about the possibility that the
serial port (either on the box or on the computer) does not satisfy the RS232 standards, which is hardware
No, I have not, that is someone else fixating on that. That was settled
as far as I am concerned before I made the first post.
How about trying the direct use of ldattach, instead of using gpsd, to
see if you can get that to work. If it does, and gpsd does not then your
hypothesis that gpsd is at fault is born out. Then you can decide
whether it is better to suffer the slings and arrows of debugging gpsd,
or with one fell swoop use ldattach and have a working system.
That gpsd has some sort of issue is well established at this point.
Any further debugging of gpsd will be done with the gpsd support list.
On 2021-07-30, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
2) The box is low cost compared to the tens of thousands of dollars such >>>> things used to cost.
That is of course completely irrelevant. The same could be said of an aluminum
spoon (cost mcuch more than a gold spoon), and yet it would still not be most people's first choice for
tableware.
At one time the cost of aluminum was much higher than gold...
Yes. Precisely. You are the one claiming that because prices are cheaper
now, it means something about the quality of the object for the purpose.
But then as point 4 says, you do not know if it is the box or some
misconfiguration (the wetware behind the keyboard) or a bug in gpsd.
Point 4 says I DO know the problem is in gpsd.
No.
Why do you feel that you know that the problem is gpsd?
You have tested
the interrupt handling with other programs (lets say like ldattach)? Not
that you have given any indication of. You simply say "it is obvious" Unfortunately I have seen far far to many instances in wich someone says
"it is obvious" and they turned out to be entirely wrong.
4) It is blazingly obvious to anyone paying the slightest bit of
attention that the problem is in software, i.e. gpsd, and not the box.
And yet, you also have been cogitating about the possibility that the
serial port (either on the box or on the computer) does not satisfy the RS232 standards, which is hardware
No, I have not, that is someone else fixating on that. That was settled
as far as I am concerned before I made the first post.
How about trying the direct use of ldattach, instead of using gpsd, to
see if you can get that to work. If it does, and gpsd does not then your >>> hypothesis that gpsd is at fault is born out. Then you can decide
whether it is better to suffer the slings and arrows of debugging gpsd,
or with one fell swoop use ldattach and have a working system.
That gpsd has some sort of issue is well established at this point.
Proof by blatant assertion is always a very poor argument.
Since ldattach is a stadard part of linux, and since the use of it is trivial, your blind refusal to test is pretty suspicious.
Any further debugging of gpsd will be done with the gpsd support list.
And here I thought that the purpose was to get your time working, not to debug gpsd. Oh well, sorry.
Jim Pennino <jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have
no issues and will see CTS being asserted but anything that looks at /dev/ppsx will fail.
According to Focal, a kernel update is imminent...
Do you have a reference for this-- eg a URL?
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Jim Pennino <jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have
no issues and will see CTS being asserted but anything that looks at
/dev/ppsx will fail.
According to Focal, a kernel update is imminent...
William Unruh <unruh@invalid.ca> wrote:
Do you have a reference for this-- eg a URL?
A reference for the kernel bug?
No, this was determined by downloading source for various GPS utility programs and running them under a debugger noting where the failure
occured and seeing that it always happened in a ppsapi routine.
And yes, the kernel says all the pps related modules are loaded.
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Jim Pennino <jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have >>> no issues and will see CTS being asserted but anything that looks at
/dev/ppsx will fail.
According to Focal, a kernel update is imminent...
William Unruh <unruh@invalid.ca> wrote:
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
Do you have a reference for this-- eg a URL?
A reference for the kernel bug?
No, this was determined by downloading source for various GPS utility
programs and running them under a debugger noting where the failure
occured and seeing that it always happened in a ppsapi routine.
So you reported the bug? What is the bug report then.
As there is an immenent kernel update, I didn't bother.
If the update doesn't fix the issue, then I will consider a bug report
for that kernel.
What was the "failure"?
Note that the pps driver is highly time sensative, and a debugger could
upset that, and report bug which is due to the interaction with the
debugger.
I have been writiting and debugging software since mid 1970 and do have
some idea of how it is done, but thank you for your comments.
Quiz of the day: What does "RT" in RT-11 stand for?
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
Do you have a reference for this-- eg a URL?
A reference for the kernel bug?
No, this was determined by downloading source for various GPS utility
programs and running them under a debugger noting where the failure
occured and seeing that it always happened in a ppsapi routine.
So you reported the bug? What is the bug report then.
What was the "failure"?
Note that the pps driver is highly time sensative, and a debugger could
upset that, and report bug which is due to the interaction with the
debugger.
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
Do you have a reference for this-- eg a URL?
A reference for the kernel bug?
No, this was determined by downloading source for various GPS utility
programs and running them under a debugger noting where the failure
occured and seeing that it always happened in a ppsapi routine.
So you reported the bug? What is the bug report then.
As there is an immenent kernel update, I didn't bother.
Kernels are updated for many reasons, and there seems to be no reason
why it would fix this particular bug if noone had ever reported it.
And are you sure that the evidence for this bug is not there in many of
the kernel versions? Ie, it could be a peculiarity of your own system
rather than being generic.
You were originally completely sure that this was a bug in gpsd. Now it
is a kernel bug.
If the update doesn't fix the issue, then I will consider a bug report
for that kernel.
What was the "failure"?
Note that the pps driver is highly time sensative, and a debugger could
upset that, and report bug which is due to the interaction with the
debugger.
I have been writiting and debugging software since mid 1970 and do have
some idea of how it is done, 5.4.0-80-genericbut thank you for your comments.
Quiz of the day: What does "RT" in RT-11 stand for?
Quiz of the day: What does "RT" in RT-11 stand for?
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have
no issues and will see CTS being asserted but anything that looks at /dev/ppsx will fail.
On 2021-07-31, Jim Pennino<jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have
no issues and will see CTS being asserted but anything that looks at
/dev/ppsx will fail.
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
gpsd supports the kernel PPS and also (a much less precise) user-space
PPS if the kernel PPS does not work.
However, I'm not sure why it worked for you only when started directly
from command line. It might help if we could compare the gpsd debug
output or strace output in both cases.
On 2021-07-31, Jim Pennino <jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have
no issues and will see CTS being asserted but anything that looks at
/dev/ppsx will fail.
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
gpsd supports the kernel PPS and also (a much less precise) user-space
PPS if the kernel PPS does not work.
However, I'm not sure why it worked for you only when started directly
from command line. It might help if we could compare the gpsd debug
output or strace output in both cases.
On 08/02/21 11:23, Miroslav Lichvar wrote:
On 2021-07-31, Jim Pennino<jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have >>> no issues and will see CTS being asserted but anything that looks at
/dev/ppsx will fail.
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
gpsd supports the kernel PPS and also (a much less precise) user-space
PPS if the kernel PPS does not work.
However, I'm not sure why it worked for you only when started directly
from command line. It might help if we could compare the gpsd debug
output or strace output in both cases.
The dcd line is what I used, as that is what was specified for
ntpd under FreeBSD. It provides a local microsecond sync, as the
following shows:
root@ntp-host:~ # ntpq -pn
remote refid st t when poll reach delay offset jitter ==================================================================== o127.127.22.0 .PPS. 0 l 5 8 377 0.000 0.002 0.001 +192.9.200.168 .GPS. 1 u 23 64 377 0.168 0.004 0.059 *192.9.200.169 .GPS. 1 u 34 64 377 0.365 -0.002 0.042
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
All the real devices I've used on Windows and multiple variants of Linux
use DCD.
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
All the real devices I've used on Windows and multiple variants of Linux use DCD.
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:I have to agree with David here:
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
All the real devices I've used on Windows and multiple variants of Linux
use DCD.
I have been a member of the NTP Hackers team for more than 25 years, and
I have yet to see a PPS device on RS232 which doesn't use DCD.
I have soldered together several (Motorola, Garmin, Sure) GPS reference clocks, all of those used DCD.
Terje
chris<chris-nospam@tridac.net> wrote:
On 08/02/21 11:23, Miroslav Lichvar wrote:
On 2021-07-31, Jim Pennino<jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have >>>> no issues and will see CTS being asserted but anything that looks at
/dev/ppsx will fail.
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
gpsd supports the kernel PPS and also (a much less precise) user-space
PPS if the kernel PPS does not work.
However, I'm not sure why it worked for you only when started directly
from command line. It might help if we could compare the gpsd debug
output or strace output in both cases.
The dcd line is what I used, as that is what was specified for
ntpd under FreeBSD. It provides a local microsecond sync, as the
following shows:
root@ntp-host:~ # ntpq -pn
remote refid st t when poll reach delay offset jitter
====================================================================
o127.127.22.0 .PPS. 0 l 5 8 377 0.000 0.002 0.001
+192.9.200.168 .GPS. 1 u 23 64 377 0.168 0.004 0.059
*192.9.200.169 .GPS. 1 u 34 64 377 0.365 -0.002 0.042
"In NTP versions after 4.0.99k23 is NMEA driver atomized what means that
for PPS processing we don't need neither ATOM driver nor PPS command in ntp.conf."
http://www.ntp.org/ntpfaq/NTP-s-config-adv.htm 6.2.4.3.3
Which means, assuming your PPS device outputs NMEA sentences, all you
need to get both data and PPS is:
server 127.127.20.0 mode (for your device) minpoll 4 maxpoll 4
fudge 127.127.20.0 flag1 1
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
Terje Mathisen<terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:I have to agree with David here:
I think that is expected if you have the PPS signal connected to CTS. >>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
All the real devices I've used on Windows and multiple variants of Linux >>> use DCD.
I have been a member of the NTP Hackers team for more than 25 years, and
I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
I have soldered together several (Motorola, Garmin, Sure) GPS reference
clocks, all of those used DCD.
Terje
I have a commercial GNSSDO, which contains an OCXO as the rubidium
option was a bit too pricey for me, and both models use CTS.
There is no soldering involved as this is, again, a commercial device
with a DB-9 on the panel with pin 8, i.e. CTS on the DTE end, labeled
1PPS.
If one runs gpsmon against this device it shows PPS being asserted on the data display line, so whoever wrote gpsmon has seen GPS devices with PPS
on CTS.
If one runs ppstest against this device it shows PPS being asserted, so whoever wrote ppstest has seen GPS devices with PPS on CTS.
FYI the OCXO model is about $180 and the rubidium model is about $800.
Arent't advances in technology great?
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS. >>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
I have to agree with David here:
All the real devices I've used on Windows and multiple variants of Linux >>> use DCD.
I have been a member of the NTP Hackers team for more than 25 years, and
I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
I have soldered together several (Motorola, Garmin, Sure) GPS reference
clocks, all of those used DCD.
Terje
I have a commercial GNSSDO, which contains an OCXO as the rubidium
option was a bit too pricey for me, and both models use CTS.
There is no soldering involved as this is, again, a commercial device
with a DB-9 on the panel with pin 8, i.e. CTS on the DTE end, labeled
1PPS.
If one runs gpsmon against this device it shows PPS being asserted on the data display line, so whoever wrote gpsmon has seen GPS devices with PPS
on CTS.
If one runs ppstest against this device it shows PPS being asserted, so whoever wrote ppstest has seen GPS devices with PPS on CTS.
FYI the OCXO model is about $180 and the rubidium model is about $800.
Arent't advances in technology great?
On 08/02/21 15:33, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/02/21 11:23, Miroslav Lichvar wrote:
On 2021-07-31, Jim Pennino<jimp@gonzo.specsol.net> wrote:
It turns out there is a kernel bug in Ubuntu 5.4.0-80-generic which
breaks PPS processing.
The net result is that an application that looks at /dev/ttySx will have >>>>> no issues and will see CTS being asserted but anything that looks at >>>>> /dev/ppsx will fail.
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
gpsd supports the kernel PPS and also (a much less precise) user-space >>>> PPS if the kernel PPS does not work.
However, I'm not sure why it worked for you only when started directly >>>> from command line. It might help if we could compare the gpsd debug
output or strace output in both cases.
The dcd line is what I used, as that is what was specified for
ntpd under FreeBSD. It provides a local microsecond sync, as the
following shows:
root@ntp-host:~ # ntpq -pn
remote refid st t when poll reach delay offset jitter
====================================================================
o127.127.22.0 .PPS. 0 l 5 8 377 0.000 0.002 0.001
+192.9.200.168 .GPS. 1 u 23 64 377 0.168 0.004 0.059
*192.9.200.169 .GPS. 1 u 34 64 377 0.365 -0.002 0.042
"In NTP versions after 4.0.99k23 is NMEA driver atomized what means that
for PPS processing we don't need neither ATOM driver nor PPS command in
ntp.conf."
http://www.ntp.org/ntpfaq/NTP-s-config-adv.htm 6.2.4.3.3
Which means, assuming your PPS device outputs NMEA sentences, all you
need to get both data and PPS is:
server 127.127.20.0 mode (for your device) minpoll 4 maxpoll 4
fudge 127.127.20.0 flag1 1
The pps line has nothing to do with nmea sentances directly, but is
a ttl level pulse that marks the beginning of each second,
typically with microsecond accuracy. It's a sync pulse for the current second.
The nmea sentences provide the overall time of day, month, year,
but fine resolution within the second can only be provided by the
subsecond accuracy of the pps pulse. Obvious really, in that
differing times of flight of the nmea sentences and decoding
can never provide the required accuracy alone...
On 02/08/2021 17:26, chris wrote:
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
ntpd doesn't provide that information. It only provides the offset
between the timing source and the PC's internal clock.
On 08/02/21 16:39, Jim Pennino wrote:
Terje Mathisen<terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:I have to agree with David here:
I think that is expected if you have the PPS signal connected to CTS. >>>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
All the real devices I've used on Windows and multiple variants of Linux >>>> use DCD.
I have been a member of the NTP Hackers team for more than 25 years, and >>> I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
I have soldered together several (Motorola, Garmin, Sure) GPS reference
clocks, all of those used DCD.
Terje
I have a commercial GNSSDO, which contains an OCXO as the rubidium
option was a bit too pricey for me, and both models use CTS.
There is no soldering involved as this is, again, a commercial device
with a DB-9 on the panel with pin 8, i.e. CTS on the DTE end, labeled
1PPS.
If one runs gpsmon against this device it shows PPS being asserted on the
data display line, so whoever wrote gpsmon has seen GPS devices with PPS
on CTS.
If one runs ppstest against this device it shows PPS being asserted, so
whoever wrote ppstest has seen GPS devices with PPS on CTS.
FYI the OCXO model is about $180 and the rubidium model is about $800.
Arent't advances in technology great?
They are, but if it's all now working, what output do you get
from the ntpq utility. or equivalent, in terms of accuracy ?.
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
milliseconds, 3 orders of magnitude error...
What part of the kernel PPS routines in the OS on that machine are broken
and do NOT process PPS did you fail to understand?
I expect the ntpq output for this device, when PPS is working, to be in
the low nanoseconds range as that is what the manufactures specifications say. Right now the offset and and error are already in the microsecond
range.
You need to look at both the offset and estimated error fields ofThat may give you the random errors, not the systematic (eg from the
loopstats and calculate the standard deviation to get any real idea of
the accuracy of ntpd.
You do do numerical analysis and graphing of loopstats, don't you?
As for "can never provide the required accuracy alone", that depends on
what your required accuracy is. I have 3 USB GPS ntp machines that are
all CONSISTENTLY accurate to about 2 milliseconds, which is better than
you will ever get with network sources.
And to state the blazingly obvious, the whole point of running ntpd at
all is to, hopefully, correct the system internal clock.
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS. >>>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
The above is simply wrong.
All the real devices I've used on Windows and multiple variants of Linux >>>> use DCD.I have to agree with David here:
I have been a member of the NTP Hackers team for more than 25 years, and >>> I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
Not really.
I have soldered together several (Motorola, Garmin, Sure) GPS reference
clocks, all of those used DCD.
Terje
I have a commercial GNSSDO, which contains an OCXO as the rubidium
option was a bit too pricey for me, and both models use CTS.
Jim, I did not state that CTS was invalid, just that all I've personally
seen have used DCD!
In reality you can have a PPS signal on any bit line which have
interrupt capability, this includes using a classic "Centronics"
parallel port along with the serial port for NMEA or some other protocol
to number the seconds.
There is no soldering involved as this is, again, a commercial device
with a DB-9 on the panel with pin 8, i.e. CTS on the DTE end, labeled
1PPS.
If one runs gpsmon against this device it shows PPS being asserted on the
data display line, so whoever wrote gpsmon has seen GPS devices with PPS
on CTS.
If one runs ppstest against this device it shows PPS being asserted, so
whoever wrote ppstest has seen GPS devices with PPS on CTS.
FYI the OCXO model is about $180 and the rubidium model is about $800.
Arent't advances in technology great?
Oh, absolutely! :-)
I wrote the software for Meinberg's super NTP server, i.e. a custom ntpd version which can run on all cores and share the processing load across
them. This 1U rack server have either a TCXO or Rb internal frequency standard which provides excellent stability, much better than the
short-term GPS stability.
Terje
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
....
As for "can never provide the required accuracy alone", that depends on
what your required accuracy is. I have 3 USB GPS ntp machines that are
all CONSISTENTLY accurate to about 2 milliseconds, which is better than
you will ever get with network sources.
Uh, depends on your network sources. If it is nearby (eg in your
building) then you can certainly get sub-ms accuracy from a network
source (ie better than NMEA). If it is half way around the world, then
you may well be right.
On 02/08/2021 18:52, Jim Pennino wrote:
And to state the blazingly obvious, the whole point of running ntpd at
all is to, hopefully, correct the system internal clock.
But not against UTC, but rather against various secondary standards, and
also not on the basis that the secondary standard is always right (in
which case offset would always be exactly zero, immediately after a poll).
People treat offset as too much of a holy grail when it is actually
measured against something which is wrong, and doesn't have a stable
error. Jitter is probably a better measure of the quality of the source
(Incidentally, it is impossible to achieve zero offset from UTC in real
time, as UTC isn't known until days, or maybe weeks, later.)
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
What part of the kernel PPS routines in the OS on that machine are broken
and do NOT process PPS did you fail to understand?
I expect the ntpq output for this device, when PPS is working, to be in
the low nanoseconds range as that is what the manufactures specifications
say. Right now the offset and and error are already in the microsecond
range.
Well, no. The interrupt handling routines in Linux and even the line
from the gps to the machine are much slower than
"low nanoseconds"
That may give you the random errors, not the systematic (eg from the interrupt time, signal propagation time, etc)
You need to look at both the offset and estimated error fields of
loopstats and calculate the standard deviation to get any real idea of
the accuracy of ntpd.
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS. >>>>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
The above is simply wrong.
OK, let me restate that.
Most current, commercial devices not intended for the experimenter or
hobbist market use CTS.
FWIW, I did some research of such devices with a DB-9 connector on them.
All the real devices I've used on Windows and multiple variants of Linux >>>>> use DCD.I have to agree with David here:
I have been a member of the NTP Hackers team for more than 25 years, and >>>> I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
Not really.
Yes, really, e.g. GNSS and relatively cheap rubidium oscillators.
With GNSS and a $15 USB GPS, one can obtain consistent time accuracy in
the single digit milliseconds these days.
David Woolley <david@ex.djwhome.demon.invalid> wrote:
On 02/08/2021 18:52, Jim Pennino wrote:
And to state the blazingly obvious, the whole point of running ntpd at
all is to, hopefully, correct the system internal clock.
But not against UTC, but rather against various secondary standards, and
also not on the basis that the secondary standard is always right (in
which case offset would always be exactly zero, immediately after a poll).
Assuming such a standard existed, which it doesn't.
People treat offset as too much of a holy grail when it is actually
measured against something which is wrong, and doesn't have a stable
error. Jitter is probably a better measure of the quality of the source
Which is why I analyze and graph MUCH more than just offset.
(Incidentally, it is impossible to achieve zero offset from UTC in real
time, as UTC isn't known until days, or maybe weeks, later.)
The people that run GPS claim the GPS offset from USNO UTC to be less
than 40 nanoseconds worst case. That is good enough for me.
William Unruh <unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
What part of the kernel PPS routines in the OS on that machine are broken >>> and do NOT process PPS did you fail to understand?
I expect the ntpq output for this device, when PPS is working, to be in
the low nanoseconds range as that is what the manufactures specifications >>> say. Right now the offset and and error are already in the microsecond
range.
Well, no. The interrupt handling routines in Linux and even the line
from the gps to the machine are much slower than
"low nanoseconds"
We shall see...
However, the absolute speed of everything is far less important than
the consistency of PPS processing.
That is, if one has an absolutely accurate PPS signal that is exactly
1Hz, it doesn't matter how long, within reason, it takes to process the
PPS signal. What matters is how much jitter the hardware and software
has in processing the signal. Since the jitter in modern digital
electronic is in the picosecond, that is not a factor. That means the determining factor will be the jitter in the interrupt handling.
Which with 1 Ghz+ multiple core CPUs is hopefully in the low
nanoseconds.
That may give you the random errors, not the systematic (eg from the
You need to look at both the offset and estimated error fields of
loopstats and calculate the standard deviation to get any real idea of
the accuracy of ntpd.
interrupt time, signal propagation time, etc)
Such will show up in the graphing of the various ntpd statistics files,
and if it doesn't, that means the people that wrote the documentation
for the ntpd statistics files were lying, which I highly doubt.
But not against UTC, but rather against various secondary standards, and
also not on the basis that the secondary standard is always right (in
which case offset would always be exactly zero, immediately after a poll).
Assuming such a standard existed, which it doesn't.
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS. >>>>>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
The above is simply wrong.
OK, let me restate that.
Most current, commercial devices not intended for the experimenter or
hobbist market use CTS.
FWIW, I did some research of such devices with a DB-9 connector on them.
All the real devices I've used on Windows and multiple variants of Linux >>>>>> use DCD.I have to agree with David here:
I have been a member of the NTP Hackers team for more than 25 years, and >>>>> I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
Not really.
Yes, really, e.g. GNSS and relatively cheap rubidium oscillators.
With GNSS and a $15 USB GPS, one can obtain consistent time accuracy in
the single digit milliseconds these days.
With PPS. Without PPS, even the length of the nmea sentences vary more
than that.
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:That may be true for your purposes. However, accuracy is Computer time -
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
What part of the kernel PPS routines in the OS on that machine are broken >>>> and do NOT process PPS did you fail to understand?
I expect the ntpq output for this device, when PPS is working, to be in >>>> the low nanoseconds range as that is what the manufactures specifications >>>> say. Right now the offset and and error are already in the microsecond >>>> range.
Well, no. The interrupt handling routines in Linux and even the line
from the gps to the machine are much slower than
"low nanoseconds"
We shall see...
However, the absolute speed of everything is far less important than
the consistency of PPS processing.
That is, if one has an absolutely accurate PPS signal that is exactly
1Hz, it doesn't matter how long, within reason, it takes to process the
PPS signal. What matters is how much jitter the hardware and software
has in processing the signal. Since the jitter in modern digital
electronic is in the picosecond, that is not a factor. That means the
determining factor will be the jitter in the interrupt handling.
UTC. So clearly if to 1 ns the difference is always 3 days, one would
not call that accurate. The interrupt handling/propagation delays are
one way. They always make the computer time later than UTC. My
measurements a few years ago showed that the interrupt delay was about 1 microsecond.
Which with 1 Ghz+ multiple core CPUs is hopefully in the low
nanoseconds.
There is no way that the interrupt processing is that fast. Even the
reading of the local clock is not that fast.
That may give you the random errors, not the systematic (eg from the
You need to look at both the offset and estimated error fields of
loopstats and calculate the standard deviation to get any real idea of >>>> the accuracy of ntpd.
interrupt time, signal propagation time, etc)
Such will show up in the graphing of the various ntpd statistics files,
and if it doesn't, that means the people that wrote the documentation
for the ntpd statistics files were lying, which I highly doubt.
There is no way that ntpd can measure the delay. It has no idea what UTC actually is. All it has is reading of the computer system time when the interrupt comes in-- which takes a while to process.
You could get a measure if, as soon as the interrupt process measured
the system time, it toggled some output port up, and then you attached a oscilliscope to that port, and the gps input and looked at the delay
between the two.
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote:
On 02/08/2021 18:52, Jim Pennino wrote:Assuming such a standard existed, which it doesn't.
And to state the blazingly obvious, the whole point of running ntpd at >>>> all is to, hopefully, correct the system internal clock.
But not against UTC, but rather against various secondary standards, and >>> also not on the basis that the secondary standard is always right (in
which case offset would always be exactly zero, immediately after a poll). >>
People treat offset as too much of a holy grail when it is actually
measured against something which is wrong, and doesn't have a stable
error. Jitter is probably a better measure of the quality of the source
Which is why I analyze and graph MUCH more than just offset.
(Incidentally, it is impossible to achieve zero offset from UTC in real
time, as UTC isn't known until days, or maybe weeks, later.)
The people that run GPS claim the GPS offset from USNO UTC to be less
than 40 nanoseconds worst case. That is good enough for me.
And apparently usually a sawtooth error. But the fact that the leading
edge of the pulse is within 40ns does not mean that the system time is
within 40ns. As I said, the computer receives the signal. At some point
in the rise time of the pulse (smeared out by propagation delays and impedance mismatches in the cable connecting the gps to the computer) it triggers and interrupt. The computer has to stop what it is doing,
branch out to the interrupt handler (assuming that some non-maskable interrupt is not running which delays that) start running the interrupt handler program, send out a command to another subroutine in the kernel
to read the system clock and translate that into time, and deliver that
time back to the interrupt handler. All of that take time. As I said, my measurements gave something like 1 microsecond in best case (2 or 3 or
even 10 in worst cases) in delay. There is nothing that ntp can do to to
know what that delay is. You can, with diffuculty, measure it, and you
could tell ntpd to subtract that from the time, but it is not
necessarily consistant, and will depend on the state of use of the cpus.
William Unruh <unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote:
On 02/08/2021 18:52, Jim Pennino wrote:Assuming such a standard existed, which it doesn't.
And to state the blazingly obvious, the whole point of running ntpd at >>>>> all is to, hopefully, correct the system internal clock.
But not against UTC, but rather against various secondary standards, and >>>> also not on the basis that the secondary standard is always right (in
which case offset would always be exactly zero, immediately after a poll). >>>
People treat offset as too much of a holy grail when it is actuallyWhich is why I analyze and graph MUCH more than just offset.
measured against something which is wrong, and doesn't have a stable
error. Jitter is probably a better measure of the quality of the source >>>
(Incidentally, it is impossible to achieve zero offset from UTC in real >>>> time, as UTC isn't known until days, or maybe weeks, later.)
The people that run GPS claim the GPS offset from USNO UTC to be less
than 40 nanoseconds worst case. That is good enough for me.
And apparently usually a sawtooth error. But the fact that the leading
edge of the pulse is within 40ns does not mean that the system time is
within 40ns. As I said, the computer receives the signal. At some point
in the rise time of the pulse (smeared out by propagation delays and
impedance mismatches in the cable connecting the gps to the computer) it
triggers and interrupt. The computer has to stop what it is doing,
branch out to the interrupt handler (assuming that some non-maskable
interrupt is not running which delays that) start running the interrupt
handler program, send out a command to another subroutine in the kernel
to read the system clock and translate that into time, and deliver that
time back to the interrupt handler. All of that take time. As I said, my
measurements gave something like 1 microsecond in best case (2 or 3 or
even 10 in worst cases) in delay. There is nothing that ntp can do to to
know what that delay is. You can, with diffuculty, measure it, and you
could tell ntpd to subtract that from the time, but it is not
necessarily consistant, and will depend on the state of use of the cpus.
Absolute and consistent delay is irrelevaent.
The processes of reading the PPS edge and getting the system time are independent processes.
If things were as you say, rubidium clocks would not work.
William Unruh <unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS. >>>>>>>>> AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
The above is simply wrong.
OK, let me restate that.
Most current, commercial devices not intended for the experimenter or
hobbist market use CTS.
FWIW, I did some research of such devices with a DB-9 connector on them. >>>
All the real devices I've used on Windows and multiple variants of LinuxI have to agree with David here:
use DCD.
I have been a member of the NTP Hackers team for more than 25 years, and >>>>>> I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
Not really.
Yes, really, e.g. GNSS and relatively cheap rubidium oscillators.
With GNSS and a $15 USB GPS, one can obtain consistent time accuracy in
the single digit milliseconds these days.
With PPS. Without PPS, even the length of the nmea sentences vary more
than that.
No, without PPS.
How much the NMEA sentences vary depends on the speed of the electronics
in the receiver and how many satellites the receiver can process.
Some old USB GPS pucks have jitter at the hundred millisecond level andIt is RS232 that is the problem.
a very few at the tens of millisecond level, and what you said used to
be true.
Modern USB GNSS pucks have jitter at the millsecond level and a standard deviation of about 1 millisecond.
So, in summary, CURRENT, MODERN GNSS receivers have better sensitity,
can see and process more satellites, and have faster electronics inside
them.
How do I know this you say?
As I have said many times before, I am doing a study of CURRENT, MODERN
GNSS devices and how well they work with ntpd.
On 08/02/21 18:05, David Woolley wrote:
On 02/08/2021 17:26, chris wrote:
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
ntpd doesn't provide that information. It only provides the offset
between the timing source and the PC's internal clock.
Correct me if i'm wrong, but pps is traceable to UTC, so that would be
used within ntpd to set the system clock. In the case of several
ntp server sources, ntpd will weed out the outliers, keep the best and
use that to set local system time. ntpq then provides a list showing
what it thinks are the timing relationships between them, with pps utc
as the primary reference ?...
On 02/08/2021 17:26, chris wrote:
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
ntpd doesn't provide that information. It only provides the offset
between the timing source and the PC's internal clock.
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Absolute and consistent delay is irrelevaent.
The processes of reading the PPS edge and getting the system time are
independent processes.
If things were as you say, rubidium clocks would not work.
Rubidium clocks are not general purpose computers. Their electronics are
all specially built to reduce the errors as much as possible.
Absolute and consistant delay is NOT irrelevant. It is part of the error budget that determines the accuracy of the clock. Or would you not care
if your clock was 3 days out in its time? If not, that is fine, but most people would not call it accurate.
On 08/02/21 18:05, David Woolley wrote:
On 02/08/2021 17:26, chris wrote:
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
ntpd doesn't provide that information. It only provides the offset
between the timing source and the PC's internal clock.
Correct me if i'm wrong, but pps is traceable to UTC, so that would be
used within ntpd to set the system clock. In the case of several
ntp server sources, ntpd will weed out the outliers, keep the best and
use that to set local system time. ntpq then provides a list showing
what it thinks are the timing relationships between them, with pps utc
as the primary reference ?...
On 2021-08-02, chris<chris-nospam@tridac.net> wrote:
On 08/02/21 18:05, David Woolley wrote:
On 02/08/2021 17:26, chris wrote:
If it's working properly and has been running for a few hours,
offset against UTC should be in the microsecond range, not
ntpd doesn't provide that information. It only provides the offset
between the timing source and the PC's internal clock.
Correct me if i'm wrong, but pps is traceable to UTC, so that would be
used within ntpd to set the system clock. In the case of several
If one wants sub-microsecond accuracy, then how it is used within ntpd becomes important. Lets assume that the gps module itself is good. The hardware inside the receiver delivers a pps pulse at the UTC second (it
does not, it fluctuates a few 10s of nanoseconds from that). That pulse
then travels down the cable to the computer. If the impedance matching between the receiver and the computer is not good, then that pulse gets smeared out over many travel times down the cable. At a certain voltage level, a trigger in the driver flips, and that raises and interrupt
voltage. The hardware watches for that flip, and tells the cpu to stop (controllably) what it is doing, caches the currently running program,
loads the interrupt driver program, and branches to that program. If
that program is well written, it then calls a kernel subroutine, loading
it into memory if need be, and calls a program to convert the counter,
which is the system clock, reads out the count, and converts that count
to a time. It then returns that time to the interrupt driver routine.
All of that takes time. All of that takes a variable amount of time. For example, if another interrupt service routine is running ( eg due to
disk reads) this one will wait for it to finish (in the sense of
signalling that it is OK for it to be interrupted )
ntp server sources, ntpd will weed out the outliers, keep the best and
use that to set local system time. ntpq then provides a list showing
what it thinks are the timing relationships between them, with pps utc
as the primary reference ?...
No. The PPS timing is simply one amongst the lot. It may well be the
chosen one, and its low stratum will mean that ntpd will be more likely
to choose it. But not necessarily.
William Unruh<unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino<jimp@gonzo.specsol.net> wrote:
<snip old crap>
Absolute and consistent delay is irrelevaent.
The processes of reading the PPS edge and getting the system time are
independent processes.
If things were as you say, rubidium clocks would not work.
Rubidium clocks are not general purpose computers. Their electronics are
all specially built to reduce the errors as much as possible.
Absolute and consistant delay is NOT irrelevant. It is part of the error
budget that determines the accuracy of the clock. Or would you not care
if your clock was 3 days out in its time? If not, that is fine, but most
people would not call it accurate.
Rubidium clocks are NOT clocks and do not tell time.
A modern rubidium clock is a box containing a rubidium oscillator that is disciplined by an included GNSS receiver. The electronics is off the shelf chips.
And there are no vacuum tubes in there either.
The GNSS receiver provides the GNSS data output and the precision
rubidium oscillator provides the PPS signal.
A rubidium clock has PPS jitter of around 1 nanosecond and to a computer looks like an extremely low jitter GNSS receiver.
Most also provide an output as a frequency standard.
You keep arm waving about "3 days", why?
Why would you say something as blazingly obvious as "Rubidium clocks are
not general purpose computers"?
Are you purposely trying to be condescending?
William Unruh <unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
<snip old crap>
Absolute and consistent delay is irrelevaent.
The processes of reading the PPS edge and getting the system time are
independent processes.
If things were as you say, rubidium clocks would not work.
Rubidium clocks are not general purpose computers. Their electronics are
all specially built to reduce the errors as much as possible.
Absolute and consistant delay is NOT irrelevant. It is part of the error
budget that determines the accuracy of the clock. Or would you not care
if your clock was 3 days out in its time? If not, that is fine, but most
people would not call it accurate.
Rubidium clocks are NOT clocks and do not tell time.
A modern rubidium clock is a box containing a rubidium oscillator that is disciplined by an included GNSS receiver. The electronics is off the shelf chips.
And there are no vacuum tubes in there either.
The GNSS receiver provides the GNSS data output and the precision
rubidium oscillator provides the PPS signal.
A rubidium clock has PPS jitter of around 1 nanosecond and to a computer looks like an extremely low jitter GNSS receiver.
Most also provide an output as a frequency standard.
You keep arm waving about "3 days", why?
Why would you say something as blazingly obvious as "Rubidium clocks are
not general purpose computers"?
Are you purposely trying to be condescending?
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
David Taylor wrote:
On 02/08/2021 14:01, Jim Pennino wrote:
I think that is expected if you have the PPS signal connected to CTS.
AFAIK the kernel PPS serial driver works only with DCD.
Then that is a bug because most real devices use CTS.
The above is simply wrong.
OK, let me restate that.
Most current, commercial devices not intended for the experimenter or
hobbist market use CTS.
FWIW, I did some research of such devices with a DB-9 connector on them. >>>>
All the real devices I've used on Windows and multiple variants of LinuxI have to agree with David here:
use DCD.
I have been a member of the NTP Hackers team for more than 25 years, and
I have yet to see a PPS device on RS232 which doesn't use DCD.
It is now 2021 and there have been changes.
Not really.
Yes, really, e.g. GNSS and relatively cheap rubidium oscillators.
With GNSS and a $15 USB GPS, one can obtain consistent time accuracy in >>>> the single digit milliseconds these days.
With PPS. Without PPS, even the length of the nmea sentences vary more
than that.
No, without PPS.
How much the NMEA sentences vary depends on the speed of the electronics
in the receiver and how many satellites the receiver can process.
No. The NMEA sentences themselves are variable length. And the RS22
standard means that the data goes up the line not very fast. Each
character takes milliseconds to be transmitted.
It is RS232 that is the problem.
Some old USB GPS pucks have jitter at the hundred millisecond level and
a very few at the tens of millisecond level, and what you said used to
be true.
Modern USB GNSS pucks have jitter at the millsecond level and a standard
deviation of about 1 millisecond.
So, in summary, CURRENT, MODERN GNSS receivers have better sensitity,
can see and process more satellites, and have faster electronics inside
them.
Of course.
How do I know this you say?
As I have said many times before, I am doing a study of CURRENT, MODERN
GNSS devices and how well they work with ntpd.
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
William Unruh <unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
<snip old crap>
Absolute and consistent delay is irrelevaent.
The processes of reading the PPS edge and getting the system time are
independent processes.
If things were as you say, rubidium clocks would not work.
Rubidium clocks are not general purpose computers. Their electronics are >>> all specially built to reduce the errors as much as possible.
Absolute and consistant delay is NOT irrelevant. It is part of the error >>> budget that determines the accuracy of the clock. Or would you not care
if your clock was 3 days out in its time? If not, that is fine, but most >>> people would not call it accurate.
Rubidium clocks are NOT clocks and do not tell time.
Neither is your computer. It also just has a crystal and a counter. You computer converts that counting to a time after calibrating that counter (ticks per second).
A modern rubidium clock is a box containing a rubidium oscillator that is
disciplined by an included GNSS receiver. The electronics is off the shelf >> chips.
And there are no vacuum tubes in there either.
The GNSS receiver provides the GNSS data output and the precision
rubidium oscillator provides the PPS signal.
A rubidium clock has PPS jitter of around 1 nanosecond and to a computer
looks like an extremely low jitter GNSS receiver.
Most also provide an output as a frequency standard.
You keep arm waving about "3 days", why?
As an obviously huge error which noone would regard as accurate, even if
that offset has vary small fluctuation.
Why would you say something as blazingly obvious as "Rubidium clocks are
not general purpose computers"?
Because you compared a Rubidium clock to a general purpose computer.
Are you purposely trying to be condescending?
Obvious perhaps, condescending no.
On 08/02/21 23:59, Jim Pennino wrote:
William Unruh<unruh@invalid.ca> wrote:
On 2021-08-02, Jim Pennino<jimp@gonzo.specsol.net> wrote:
<snip old crap>
Absolute and consistent delay is irrelevaent.
The processes of reading the PPS edge and getting the system time are
independent processes.
If things were as you say, rubidium clocks would not work.
Rubidium clocks are not general purpose computers. Their electronics are >>> all specially built to reduce the errors as much as possible.
Absolute and consistant delay is NOT irrelevant. It is part of the error >>> budget that determines the accuracy of the clock. Or would you not care
if your clock was 3 days out in its time? If not, that is fine, but most >>> people would not call it accurate.
Rubidium clocks are NOT clocks and do not tell time.
A modern rubidium clock is a box containing a rubidium oscillator that is
disciplined by an included GNSS receiver. The electronics is off the shelf >> chips.
And there are no vacuum tubes in there either.
The GNSS receiver provides the GNSS data output and the precision
rubidium oscillator provides the PPS signal.
Not really the full story. The rb source or vcxo is locked to
gps 1pps, also providing the external pps signal to ntpd. The
rb is typically used as a low phase noise frequency standard
for external equipment such as signal generators, counters etc.
They are not separate entities, but are intimately connected.
Suggest swot up on gps disciplined oscillators, for more
background on how the systems work...
That is, if one has an absolutely accurate PPS signal that is exactly
1Hz, it doesn't matter how long, within reason, it takes to process the
PPS signal. What matters is how much jitter the hardware and software
has in processing the signal. Since the jitter in modern digital
electronic is in the picosecond, that is not a factor. That means the determining factor will be the jitter in the interrupt handling.
Which with 1 Ghz+ multiple core CPUs is hopefully in the low
nanoseconds.
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
Correct me if i'm wrong, but pps is traceable to UTC
And there are no vacuum tubes in there either.
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second transition most probably lies. You pretty much acknowledge that later
in your description.
On 03/08/2021 01:13, Jim Pennino wrote:
And there are no vacuum tubes in there either.
I'm sure the reference to vacuum tubes was rhetorical, as I don't think
GPS was started until well into the solid state age.
However, some quick research on how rubidium clocks work suggests that a vacuum tube would be a good description of the rubidium containing part
of the system, as they are low pressure discharge lamps. They have to
be low pressure to prevent atoms interacting with each other.
Also rubidium oscillator is something of a misnomer, as they are really rubidium controlled crystal oscillators.
On 2021-08-02, Jim Pennino <jimp@gonzo.specsol.net> wrote:
That is, if one has an absolutely accurate PPS signal that is exactly
1Hz, it doesn't matter how long, within reason, it takes to process the
PPS signal. What matters is how much jitter the hardware and software
has in processing the signal. Since the jitter in modern digital
electronic is in the picosecond, that is not a factor. That means the
determining factor will be the jitter in the interrupt handling.
Which with 1 Ghz+ multiple core CPUs is hopefully in the low
nanoseconds.
No, a typical jitter of the interrupt-based PPS timestamping on a
computer serial port is around 1 microsecond and there is a delay of at
least few microseconds. If you didn't disable the CPU power-saving
features, the delay can easily get to tens of microseconds and a large
offset can be observed when the CPU is loaded on and off.
It does not matter if your PPS signal is stable to 1 or 50 nanoseconds.
It makes no difference in that noise. A $10 GPS from ebay would likely perform the same as long as it can get a good signal.
If you were serious about accuracy and stability, you wouldn't be
connecting PPS to a serial port. You need something that has hardware timestamping, like a networking card (e.g. the I210). That would get you
to the low nanosecond range, at least for the hardware clock.
Synchronization of the system clock to the hardware clock is the weakest
link of the chain. It is typically accurate only to few hundreds of nanoseconds. The problem is large and asymmetric latency of the PCIe
bus, which cannot be easily measured and compensated. Still, that's at
least an order of magnitude better that the serial port.
A new feature that some NICs have is PTM, which is basically a hardware implementation of NTP on PCIe. It should improve the accuracy
significantly. I have not had a chance to play with it yet.
The cheapest, current, commercial GPS with a serial connector and PPS
that I can find is over $100.
On 08/03/21 15:22, Jim Pennino wrote:
The cheapest, current, commercial GPS with a serial connector and PPS
that I can find is over $100.
Then you are not looking very hard. Bare board gps modules,
with pps and serial are on Ebay in droves at around the
< $10 upwards. Of course, if you are not happy doing a bit
of simple soldering and interfacing, then you are at the
mercy of those that can...
Usb is the last thing anyone would use if accuracy is the goal.
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible. >>>
PPS provides information on where within the cycle the true UTC second
transition most probably lies. You pretty much acknowledge that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible. >>>>
PPS provides information on where within the cycle the true UTC second
transition most probably lies. You pretty much acknowledge that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:22, Jim Pennino wrote:
The cheapest, current, commercial GPS with a serial connector and PPS
that I can find is over $100.
Then you are not looking very hard. Bare board gps modules,
with pps and serial are on Ebay in droves at around the
< $10 upwards. Of course, if you are not happy doing a bit
of simple soldering and interfacing, then you are at the
mercy of those that can...
What part of the words "commercial" and "serial connector" is it that
you did not understand?
FYI I have been soldering since about 1963.
Usb is the last thing anyone would use if accuracy is the goal.
That depends on the accuracy required.
As I said, there are thousands of people who have no serial or parallel
port that have a required accuracy of about 100 milliseconds.
For them, USB would be the FIRST thing they would use as it is the
cheapest and simplest solution and easily meets the requirement.
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible. >>>>>
PPS provides information on where within the cycle the true UTC second >>>> transition most probably lies. You pretty much acknowledge that later >>>> in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC.
If you can not, you are just pulling this from your butt.
On 08/03/21 16:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:22, Jim Pennino wrote:
The cheapest, current, commercial GPS with a serial connector and PPS
that I can find is over $100.
Then you are not looking very hard. Bare board gps modules,
with pps and serial are on Ebay in droves at around the
< $10 upwards. Of course, if you are not happy doing a bit
of simple soldering and interfacing, then you are at the
mercy of those that can...
What part of the words "commercial" and "serial connector" is it that
you did not understand?
FYI I have been soldering since about 1963.
Then you should have no problem taking a $10- gps module, soldering a
few wires to it and a 9 pin D and you are good to go.
Usb is the last thing anyone would use if accuracy is the goal.
That depends on the accuracy required.
As I said, there are thousands of people who have no serial or parallel
port that have a required accuracy of about 100 milliseconds.
They do ?, who are these thousands of people ?...
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible. >>>>>>
PPS provides information on where within the cycle the true UTC second >>>>> transition most probably lies. You pretty much acknowledge that later >>>>> in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC.
Typical gps modules will claim a few 10's of ns offset limits
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:22, Jim Pennino wrote:
The cheapest, current, commercial GPS with a serial connector and PPS >>>>> that I can find is over $100.
Then you are not looking very hard. Bare board gps modules,
with pps and serial are on Ebay in droves at around the
< $10 upwards. Of course, if you are not happy doing a bit
of simple soldering and interfacing, then you are at the
mercy of those that can...
What part of the words "commercial" and "serial connector" is it that
you did not understand?
FYI I have been soldering since about 1963.
Then you should have no problem taking a $10- gps module, soldering a
few wires to it and a 9 pin D and you are good to go.
There are no $10 modules with RS-232 outputs.
Do you understand the difference between serial output and RS-232?
Usb is the last thing anyone would use if accuracy is the goal.
That depends on the accuracy required.
As I said, there are thousands of people who have no serial or parallel
port that have a required accuracy of about 100 milliseconds.
They do ?, who are these thousands of people ?...
Probably the biggest group is amateur radio operators running any of the modern modes such as FST4, FST4W, FT4, FT8, JT4, JT9, JT65, Q65, MSK144,
and WSPR, which is the majority of the about 3 million world wide.
They also have the issue of operating from a remote location with no
internet or cell access.
This is typically done with a Windows laptop that has no device I/O other than USB.
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second >>>>>> transition most probably lies. You pretty much acknowledge that later >>>>>> in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC.
Typical gps modules will claim a few 10's of ns offset limits
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to UTC.
You claim PPS is synced to UTC.
Prove it.
On 08/03/21 17:34, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second >>>>>>> transition most probably lies. You pretty much acknowledge that later >>>>>>> in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC.
Typical gps modules will claim a few 10's of ns offset limits
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to UTC.
You claim PPS is synced to UTC.
Prove it.
Millions of people depend on it every day, so we know it works. PPS
sync is directly traceable to the satellite master clocks and thence
to utc maintenance at the national standards labs worldwide. Yes Jim,
it's major worldwide collaboration, with everything from GSM mobile, broadcasting, industry,. medicine and more depending on it. ...and you
have doubts ?.
Not a specialist in the subject,
Please provide the GPS specification that defines the PPS sync to UTC.
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 17:34, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:Typical gps modules will claim a few 10's of ns offset limits
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>>>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second >>>>>>>> transition most probably lies. You pretty much acknowledge that later >>>>>>>> in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC. >>>>
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to UTC.
You claim PPS is synced to UTC.
Prove it.
Millions of people depend on it every day, so we know it works. PPS
sync is directly traceable to the satellite master clocks and thence
to utc maintenance at the national standards labs worldwide. Yes Jim,
it's major worldwide collaboration, with everything from GSM mobile,
broadcasting, industry,. medicine and more depending on it. ...and you
have doubts ?.
You claim the PPS pulse is synced to UTC.
I say it is not.
Prove your statement with a link to a GPS specification, not arm waving.
Not a specialist in the subject,
To say the least...
On 08/03/21 17:18, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:22, Jim Pennino wrote:
The cheapest, current, commercial GPS with a serial connector and PPS >>>>>> that I can find is over $100.
Then you are not looking very hard. Bare board gps modules,
with pps and serial are on Ebay in droves at around the
< $10 upwards. Of course, if you are not happy doing a bit
of simple soldering and interfacing, then you are at the
mercy of those that can...
What part of the words "commercial" and "serial connector" is it that
you did not understand?
FYI I have been soldering since about 1963.
Then you should have no problem taking a $10- gps module, soldering a
few wires to it and a 9 pin D and you are good to go.
There are no $10 modules with RS-232 outputs.
Do you understand the difference between serial output and RS-232?
We already had that discussion, where you seemed unaware of the
need or not for level conversion. Sure, the gps module will have
serial output and pps at ttl levels. 2400 or 4800 baud ascii
format being typical.
Usb is the last thing anyone would use if accuracy is the goal.
That depends on the accuracy required.
As I said, there are thousands of people who have no serial or parallel >>>> port that have a required accuracy of about 100 milliseconds.
They do ?, who are these thousands of people ?...
Probably the biggest group is amateur radio operators running any of the
modern modes such as FST4, FST4W, FT4, FT8, JT4, JT9, JT65, Q65, MSK144,
and WSPR, which is the majority of the about 3 million world wide.
They also have the issue of operating from a remote location with no
internet or cell access.
This is typically done with a Windows laptop that has no device I/O other
than USB.
Fair comment, but I would try to find a more accurate solution than
using a usb interface. Bluetooth gps engines are quite popular now
fwics, and not expensive either...
On 08/03/21 18:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 17:34, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:Typical gps modules will claim a few 10's of ns offset limits
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>>>>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second
transition most probably lies. You pretty much acknowledge that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally >>>>>>>> independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC. >>>>>
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to UTC. >>>>
You claim PPS is synced to UTC.
Prove it.
Millions of people depend on it every day, so we know it works. PPS
sync is directly traceable to the satellite master clocks and thence
to utc maintenance at the national standards labs worldwide. Yes Jim,
it's major worldwide collaboration, with everything from GSM mobile,
broadcasting, industry,. medicine and more depending on it. ...and you
have doubts ?.
You claim the PPS pulse is synced to UTC.
I say it is not.
Prove your statement with a link to a GPS specification, not arm waving.
Not a specialist in the subject,
To say the least...
You really are beyond help, but how do you think gsm cell handover, or
high frequency stock trading works ?. Are you even curious about how
things like that work and make some effort to find out ?.
The thing about gps time and pps is that you can have a gps rx in London
and one on NY and both pps will be in sync to within 10's
of nS typically. How amazing is that ?. nS accuracy just about
anywhere on the planet where you can see satellites.
Do your own homework and stop being so lazy :-)...
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 17:18, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:28, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:22, Jim Pennino wrote:
The cheapest, current, commercial GPS with a serial connector and PPS >>>>>>> that I can find is over $100.
Then you are not looking very hard. Bare board gps modules,
with pps and serial are on Ebay in droves at around the
< $10 upwards. Of course, if you are not happy doing a bit
of simple soldering and interfacing, then you are at the
mercy of those that can...
What part of the words "commercial" and "serial connector" is it that >>>>> you did not understand?
FYI I have been soldering since about 1963.
Then you should have no problem taking a $10- gps module, soldering a
few wires to it and a 9 pin D and you are good to go.
There are no $10 modules with RS-232 outputs.
Do you understand the difference between serial output and RS-232?
We already had that discussion, where you seemed unaware of the
need or not for level conversion. Sure, the gps module will have
serial output and pps at ttl levels. 2400 or 4800 baud ascii
format being typical.
Sigh.
I was NEVER unaware of the need or not for level conversion for anything.
Some GPS modules have TTL output, but most have CMOS levels these days.
4800 and 9600 are the most common bauds.
GNSS modules are all at least 9600 because a GNSS receiver has more NMEA sentences and they will not fit into one second window at 4800 baud.
Usb is the last thing anyone would use if accuracy is the goal.
That depends on the accuracy required.
As I said, there are thousands of people who have no serial or parallel >>>>> port that have a required accuracy of about 100 milliseconds.
They do ?, who are these thousands of people ?...
Probably the biggest group is amateur radio operators running any of the >>> modern modes such as FST4, FST4W, FT4, FT8, JT4, JT9, JT65, Q65, MSK144, >>> and WSPR, which is the majority of the about 3 million world wide.
They also have the issue of operating from a remote location with no
internet or cell access.
This is typically done with a Windows laptop that has no device I/O other >>> than USB.
Fair comment, but I would try to find a more accurate solution than
using a usb interface. Bluetooth gps engines are quite popular now
fwics, and not expensive either...
Sigh.
Bluetooth has no quality that would make it more accurate than USB for timekeeping applications.
COMMERCIAL bluetooth GPS receivers are more expensive than USB.
Why do you continually babble about DIY projects when I have repeatedly
said I have no interest in them?
Why would you "try to find a more accurate solution" when the solution offered is two orders of magnitude better than the requirement?
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 18:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 17:34, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:Typical gps modules will claim a few 10's of ns offset limits
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second
transition most probably lies. You pretty much acknowledge that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally >>>>>>>>> independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification. >>>>>>>>>
It may generate a pps signal if divided down, but it would not >>>>>>>> be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC. >>>>>>
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to UTC. >>>>>
You claim PPS is synced to UTC.
Prove it.
Millions of people depend on it every day, so we know it works. PPS
sync is directly traceable to the satellite master clocks and thence
to utc maintenance at the national standards labs worldwide. Yes Jim,
it's major worldwide collaboration, with everything from GSM mobile,
broadcasting, industry,. medicine and more depending on it. ...and you >>>> have doubts ?.
You claim the PPS pulse is synced to UTC.
I say it is not.
Prove your statement with a link to a GPS specification, not arm waving. >>>
Not a specialist in the subject,
To say the least...
You really are beyond help, but how do you think gsm cell handover, or
high frequency stock trading works ?. Are you even curious about how
things like that work and make some effort to find out ?.
The thing about gps time and pps is that you can have a gps rx in London
and one on NY and both pps will be in sync to within 10's
of nS typically. How amazing is that ?. nS accuracy just about
anywhere on the planet where you can see satellites.
Do your own homework and stop being so lazy :-)...
Yet again, nothing but arm waving.
You have claimed the GPS PPS pulse is synced to UTC.
I say it is not and the GPS PPS pulse is only specified as an interval.
Prove your statement with a link to a GPS specification, not arm waving.
On 08/03/21 18:57, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 18:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 17:34, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a >>>>>>>>>>>> reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, >>>>>>>>>>>> as possible.
PPS provides information on where within the cycle the true >>>>>>>>>>> UTC second
transition most probably lies. You pretty much acknowledge >>>>>>>>>>> that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally >>>>>>>>>> independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification. >>>>>>>>>>
It may generate a pps signal if divided down, but it would not >>>>>>>>> be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync >>>>>>>> to UTC.
Typical gps modules will claim a few 10's of ns offset limits
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to
UTC.
You claim PPS is synced to UTC.
Prove it.
Millions of people depend on it every day, so we know it works. PPS
sync is directly traceable to the satellite master clocks and thence >>>>> to utc maintenance at the national standards labs worldwide. Yes Jim, >>>>> it's major worldwide collaboration, with everything from GSM mobile, >>>>> broadcasting, industry,. medicine and more depending on it. ...and you >>>>> have doubts ?.
You claim the PPS pulse is synced to UTC.
I say it is not.
Prove your statement with a link to a GPS specification, not arm
waving.
Not a specialist in the subject,
To say the least...
You really are beyond help, but how do you think gsm cell handover, or
high frequency stock trading works ?. Are you even curious about how
things like that work and make some effort to find out ?.
The thing about gps time and pps is that you can have a gps rx in London >>> and one on NY and both pps will be in sync to within 10's
of nS typically. How amazing is that ?. nS accuracy just about
anywhere on the planet where you can see satellites.
Do your own homework and stop being so lazy :-)...
Yet again, nothing but arm waving.
You have claimed the GPS PPS pulse is synced to UTC.
I say it is not and the GPS PPS pulse is only specified as an interval.
Prove your statement with a link to a GPS specification, not arm waving.
Rofl, i'm quite happy to accept that pps is deterministically
related to UTC, nS offset and all, but you believe what you like.
Millions of users worldwide would disagree with you, but hey,
perhaps you are the only man in step ?...
On 03/08/2021 17:34, Jim Pennino wrote:
Please provide the GPS specification that defines the PPS sync to UTC.
I can't think of any good reason why the GPS specification should even mention PPS.
Although there may be specialist equipment that only provides a 1Hz
frequency standard (although I'd think 10MHz more likely), I think we
would have seen lots of people reporting problems if real world PPS
outputs weren't aligned with GPS seconds.
As for a typical product specification, the Garmin 18x LVC series has
this specification:
"The highly accurate one-pulse-per-second (PPS) output is provided for applications requiring precise timing measurements. After the initial position fix has been calculated, the PPS signal is generated and
continues until the unit is powered down. The rising edge of the signal
is aligned to the start of each GPS second within 1 μs for all
conditions in which the receiver has reported a valid and accurate
position for at least the previous 4 seconds."
<https://static.garmin.com/pumac/GPS_18x_Tech_Specs.pdf>
Which beg the question, why are you here if you already have
all the answers ?. T learn perhaps, as you seem to have no understanding
of the relationship between utc and pps, even to
the point of denying there is one ?...
Just as a final question, how do you think UTC is generated and
maintained these days ?. Clue: What source is used as a reference ?...
On 08/03/21 18:57, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 18:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 17:34, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/03/21 16:22, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:Typical gps modules will claim a few 10's of ns offset limits
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid> wrote: >>>>>>>>>>> On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as possible.
PPS provides information on where within the cycle the true UTC second
transition most probably lies. You pretty much acknowledge that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally >>>>>>>>>> independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification. >>>>>>>>>>
It may generate a pps signal if divided down, but it would not >>>>>>>>> be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
Please provide the GPS specification that defines the PPS sync to UTC. >>>>>>>
for local pps in relation to UTC, which is the final arbiter.
GPS have clocks traceable to national standards and UTC. They
are the primary reference, to which any local clocks or
oscillators, vcxo or rb, etc will be locked in sync with.
If you need any more info, download the data sheet for a
typical sub $10 gps module, which should have all the info
you need...
I did not ask for spec sheets for GPS modules.
Please provide the GPS specification that defines the PPS sync to UTC. >>>>>>
You claim PPS is synced to UTC.
Prove it.
Millions of people depend on it every day, so we know it works. PPS
sync is directly traceable to the satellite master clocks and thence >>>>> to utc maintenance at the national standards labs worldwide. Yes Jim, >>>>> it's major worldwide collaboration, with everything from GSM mobile, >>>>> broadcasting, industry,. medicine and more depending on it. ...and you >>>>> have doubts ?.
You claim the PPS pulse is synced to UTC.
I say it is not.
Prove your statement with a link to a GPS specification, not arm waving. >>>>
Not a specialist in the subject,
To say the least...
You really are beyond help, but how do you think gsm cell handover, or
high frequency stock trading works ?. Are you even curious about how
things like that work and make some effort to find out ?.
The thing about gps time and pps is that you can have a gps rx in London >>> and one on NY and both pps will be in sync to within 10's
of nS typically. How amazing is that ?. nS accuracy just about
anywhere on the planet where you can see satellites.
Do your own homework and stop being so lazy :-)...
Yet again, nothing but arm waving.
You have claimed the GPS PPS pulse is synced to UTC.
I say it is not and the GPS PPS pulse is only specified as an interval.
Prove your statement with a link to a GPS specification, not arm waving.
Rofl, i'm quite happy to accept that pps is deterministically
related to UTC, nS offset and all, but you believe what you like.
Millions of users worldwide would disagree with you, but hey,
perhaps you are the only man in step ?...
It is still my position, and will be until someone comes up with a GPS specification that says otherwise, that the only specification for the
GPS PPS signal is interval accuracy, i.e. that it is 1 Hz +/- some specification.
It also seems the ntpd algorithms fall apart if you have multiple high accuracy reference clocks attached. I doubt any of the designers ever envisioned the day would come when someone could afford more than one as
at that time such things would set you back many tens of thousands of dollars.
Now they cost less than $50.
This is irrelevant as if you have a high accuracy reference clock, you
only need one and ntpd will sort itself out just fine.
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid>Â wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference
that is as close to 1 Hz, i.e. an elapsed time of 1 second, as
possible.
PPS provides information on where within the cycle the true UTC second
transition most probably lies. You pretty much acknowledge that later
in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
chris wrote:
On 08/03/21 15:15, Jim Pennino wrote:
David Woolley<david@ex.djwhome.demon.invalid>Â wrote:
On 03/08/2021 00:37, Jim Pennino wrote:
PPS has no information. The purpose of PPS is to provide a reference >>>>> that is as close to 1 Hz, i.e. an elapsed time of 1 second, as
possible.
PPS provides information on where within the cycle the true UTC second >>>> transition most probably lies. You pretty much acknowledge that later >>>> in your description.
Nope.
In fact if one were to build a highly stable oscillator totally
independent of GPS, it would still work as a PPS signal.
If you think otherwise, please provide the GPS specification.
It may generate a pps signal if divided down, but it would not
be synchronised to UTC or anything else. Only a gps 1 pps
can provide that signal, or another clock locked to UTC.
I should stop digging mate, ever more ridiculous...
A PPS signal without some additional signal (serial or otherwise) which names/numbers the seconds will be dropped very quickly by ntpd.
I.e. even if those pulses are delivered by an HP atomic clock, they are worthless without the secondary info.
Terje
On 2021-08-03, Jim Pennino <jimp@gonzo.specsol.net> wrote:
It is still my position, and will be until someone comes up with a GPS
specification that says otherwise, that the only specification for the
GPS PPS signal is interval accuracy, i.e. that it is 1 Hz +/- some
specification.
Maybe this will help https://www.usno.navy.mil/USNO/time/gps/gps-info
GPS SYSTEM TIME
GPS system time is given by its Composite Clock (CC). The CC or "paper"
clock consists of all operational Monitor Station and satellite
frequency standards. GPS system time, in turn, is referenced to the
Master Clock (MC) at the USNO and steered to UTC(USNO) from which system
time will not deviate by more than one microsecond. The exact difference
is contained in the navigation message in the form of two constants, A0
and A1, giving the time difference and rate of system time against UTC(USNO,MC). UTC(USNO) itself is kept very close to the international benchmark UTC as maintained by the BIPM, and the exact difference, USNO
vs. BIPM is available in near real time.
Jim Pennino wrote:
It also seems the ntpd algorithms fall apart if you have multiple high
accuracy reference clocks attached. I doubt any of the designers ever
envisioned the day would come when someone could afford more than one as
at that time such things would set you back many tens of thousands of
dollars.
Rather the opposite, this is how we compare various PPS devices, i.e. by setting 'prefer' on one of them and comparing how the others behave
relative to this, then optionally swtich the prefer around and retry.
Good refclocks, like the original Oncore UT+ allow you to both offset
the PPS signal from the top of the second (to avoid any extra interrupt latency due to multiple events firing at the same time), and to
determine the sawtooth error, which is the +/-50 ns you get when the PPS signal is generated from the 10 MHz reference: The serial protocol
includes information about how far off the current pulse is/will be.
ntpd is designed to handle _many_ sources and (with the pool)
Now they cost less than $50.
This is irrelevant as if you have a high accuracy reference clock, you
only need one and ntpd will sort itself out just fine.
automatically settle on a near-optimal set of up to 10 sources.
Terje
chris <chris-nospam@tridac.net> wrote:
You really are beyond help, but how do you think gsm cell handover, or
high frequency stock trading works ?. Are you even curious about how
things like that work and make some effort to find out ?.
The thing about gps time and pps is that you can have a gps rx in London
and one on NY and both pps will be in sync to within 10's
of nS typically. How amazing is that ?. nS accuracy just about
anywhere on the planet where you can see satellites.
Do your own homework and stop being so lazy :-)...
Yet again, nothing but arm waving.
You have claimed the GPS PPS pulse is synced to UTC.
I say it is not and the GPS PPS pulse is only specified as an interval.
Jim Pennino wrote:
chris <chris-nospam@tridac.net> wrote:
You really are beyond help, but how do you think gsm cell handover, or
high frequency stock trading works ?. Are you even curious about how
things like that work and make some effort to find out ?.
The thing about gps time and pps is that you can have a gps rx in London >>> and one on NY and both pps will be in sync to within 10's
of nS typically. How amazing is that ?. nS accuracy just about
anywhere on the planet where you can see satellites.
Do your own homework and stop being so lazy :-)...
Yet again, nothing but arm waving.
You have claimed the GPS PPS pulse is synced to UTC.
I say it is not and the GPS PPS pulse is only specified as an interval.
Jim, now you are just being a troll:
Yes, we all know that UTC is a paper clock, determined a month behind
after collecting data from the (order of magnitude) 100+ atomic clocks
in many national laboratories.
This means that nobody can claim to know "This is the exact PPS
instant", only that statistical data for their local reference clocks
(like USNO whoch provides the baseline for GPS) show that they have been consistently within single-digit ns of the consensus/paper calculated value.
However, for our purposes, USNO used to have an original goal of being
within (afair) 25 ns of the UTC paper clock, and to keep each GPS sat
within another 25 (or so) ns of UTC(USNO), with a total max offset of
~45 ns.
In reality they are doing far better these days, so we can safely claim
GPS time to be within 10 ns of UTC, simply because they have
consistently been so for a decade or two.
In order to provide 3m positional accuracy, we get a relative timing
accuracy of better than 10 ns, i.e. all the visible sats needs to agree
what TAI(GPS) is within a ns or two.
Similarly, starting with the timing GPS receiver: Since it knows where
it is (either statically or due to long-term averaging) with sub-meter precision, it will only introduce time errors in the same ballpark, i.e.
a couple of ns.
The end of all this is that we can all safely claim that a working
GPS-based PPS source will be in the 5-50 ns (RMS) range of true TAI/UTC.
If you need beter than this then you need your own atomic clock and
long-term comparisons with your friendly neigborhood national lab.
Terje
Terje Mathisen <terje.mathisen@tmsw.no> wrote:[big snip]
Similarly, starting with the timing GPS receiver: Since it knows where
it is (either statically or due to long-term averaging) with sub-meter
precision, it will only introduce time errors in the same ballpark, i.e.
a couple of ns.
The end of all this is that we can all safely claim that a working
GPS-based PPS source will be in the 5-50 ns (RMS) range of true TAI/UTC.
If you need beter than this then you need your own atomic clock and
long-term comparisons with your friendly neigborhood national lab.
Terje
Yeah, I know all that, but nowhere do you address the actual issue.
Addressing the actual issue would require citing the GPS specification
that says how PPS is generated and it's specified accuracy.
Quoting observed things is interesting but not a specification.
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:[big snip]
GPS never talks about PPS, that is simply not a part of what a position/timing constellation needs to care about.Similarly, starting with the timing GPS receiver: Since it knows where
it is (either statically or due to long-term averaging) with sub-meter
precision, it will only introduce time errors in the same ballpark, i.e. >>> a couple of ns.
The end of all this is that we can all safely claim that a working
GPS-based PPS source will be in the 5-50 ns (RMS) range of true TAI/UTC. >>>
If you need beter than this then you need your own atomic clock and
long-term comparisons with your friendly neigborhood national lab.
Terje
Yeah, I know all that, but nowhere do you address the actual issue.
Addressing the actual issue would require citing the GPS specification
that says how PPS is generated and it's specified accuracy.
Quoting observed things is interesting but not a specification.
A PPS signal is always locally generated, for most (all?) those I have
read the specifications, this PPS signal is generated by the GPS
receiver chipset using a local 10 MHz oscillator, so that the PPS signal
is a pulse that rises on the edge which it has calculated will be the
closest to the true UTC/TAI second start.
With this approach we automatically get a standard deviation of 25-50 ns
from UTC(USNO), even if the average is perfectly centered around the
true second tick.
As you have alluded to multiple times, GPS time which is very close to UTC(USNO) introduces two more sources of jitter, i.e. the USNO to GPS
sat time transfer and the UTC to UTC(USNO) delta, but in the current
worl we are living in, UTC(GPS) is effectively the only real source for
time sync in global use.
Terje
using a local 10 MHz oscillator
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:[big snip]
GPS never talks about PPS, that is simply not a part of what aSimilarly, starting with the timing GPS receiver: Since it knows where >>>> it is (either statically or due to long-term averaging) with sub-meter >>>> precision, it will only introduce time errors in the same ballpark, i.e. >>>> a couple of ns.
The end of all this is that we can all safely claim that a working
GPS-based PPS source will be in the 5-50 ns (RMS) range of true TAI/UTC. >>>>
If you need beter than this then you need your own atomic clock and
long-term comparisons with your friendly neigborhood national lab.
Terje
Yeah, I know all that, but nowhere do you address the actual issue.
Addressing the actual issue would require citing the GPS specification
that says how PPS is generated and it's specified accuracy.
Quoting observed things is interesting but not a specification.
position/timing constellation needs to care about.
A PPS signal is always locally generated, for most (all?) those I have
read the specifications, this PPS signal is generated by the GPS
receiver chipset using a local 10 MHz oscillator, so that the PPS signal
is a pulse that rises on the edge which it has calculated will be the
closest to the true UTC/TAI second start.
With this approach we automatically get a standard deviation of 25-50 ns
from UTC(USNO), even if the average is perfectly centered around the
true second tick.
As you have alluded to multiple times, GPS time which is very close to
UTC(USNO) introduces two more sources of jitter, i.e. the USNO to GPS
sat time transfer and the UTC to UTC(USNO) delta, but in the current
worl we are living in, UTC(GPS) is effectively the only real source for
time sync in global use.
Terje
So you are saying every GPS receiver has a 10 MHz oscillator ergo the accuracy of the PPS output signal is totally dependent on the accuracy
of the internal oscillator, ignoring what one MAY externally add?
If so, you have validated my contention that GPS does NOT specify PPS syncronization to UTC, but rather it is totally up to whoever designed
the receiver.
On 2021-08-04, Jim Pennino <jimp@gonzo.specsol.net> wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
Terje Mathisen <terje.mathisen@tmsw.no> wrote:[big snip]
GPS never talks about PPS, that is simply not a part of what aSimilarly, starting with the timing GPS receiver: Since it knows where >>>>> it is (either statically or due to long-term averaging) with sub-meter >>>>> precision, it will only introduce time errors in the same ballpark, i.e. >>>>> a couple of ns.
The end of all this is that we can all safely claim that a working
GPS-based PPS source will be in the 5-50 ns (RMS) range of true TAI/UTC. >>>>>
If you need beter than this then you need your own atomic clock and
long-term comparisons with your friendly neigborhood national lab.
Terje
Yeah, I know all that, but nowhere do you address the actual issue.
Addressing the actual issue would require citing the GPS specification >>>> that says how PPS is generated and it's specified accuracy.
Quoting observed things is interesting but not a specification.
position/timing constellation needs to care about.
A PPS signal is always locally generated, for most (all?) those I have
read the specifications, this PPS signal is generated by the GPS
receiver chipset using a local 10 MHz oscillator, so that the PPS signal >>> is a pulse that rises on the edge which it has calculated will be the
closest to the true UTC/TAI second start.
With this approach we automatically get a standard deviation of 25-50 ns >>> from UTC(USNO), even if the average is perfectly centered around the
true second tick.
As you have alluded to multiple times, GPS time which is very close to
UTC(USNO) introduces two more sources of jitter, i.e. the USNO to GPS
sat time transfer and the UTC to UTC(USNO) delta, but in the current
worl we are living in, UTC(GPS) is effectively the only real source for
time sync in global use.
Terje
So you are saying every GPS receiver has a 10 MHz oscillator ergo the
accuracy of the PPS output signal is totally dependent on the accuracy
of the internal oscillator, ignoring what one MAY externally add?
No. He is saying that the gps picks the edge of 10MHz signal that is
closest to GPS time. That does not depend on the accuracy of the 10MHz. Sometime the gps signal may pick the edge that is 997 cycles away from
the lastone, sometimes the 998, sometimes the 1001 -- whichever is
closest. The GPS sattelites send down a time signal. The PPS tends to
ramp up in a sawtooth.and when it gets to 50 flips to the next one. A
good timing gps will figure out what its 10MHz is with respect to GPS
time and send down a correction so you can, after the fact correct for
the sawtooth offset.
If so, you have validated my contention that GPS does NOT specify PPS
syncronization to UTC, but rather it is totally up to whoever designed
the receiver.
You did not read what he said.
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
I can't see why a GPS receiver would need a 10MHz oscillator. TheWhy?
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
PLL tech is a bit better than it was in 1971 when I first used one.
David Woolley <david@ex.djwhome.demon.invalid> wrote:
So, in what today we call a gps, there are at least three separate
devices, a radio receiver to obtain the the GPS data, a device to
generate PPS, and a device to convert GPS data to NMEA sentences.
BTW, NMEA stands for National Marine Electronics Association. Guess why commercial GPS devices output data in a marine standard format.
In circa 1996 those three devices would likely be in three 19 inch rack
mount boxes containing mostly discrete components.
Today everything is fabricated on chip sets.
In regards to PPS alignment with UTC, since it is up to the equiment
maker, it will be whatever the equipment maker decides it will be.
On 08/04/21 22:17, David Woolley wrote:
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
Just to disambiguate the comments about 10MHz. While the gps receiver
module will have an internal oscillator to suit the design, that
oscillator is not normally available external to the module. What the
gps receiver does provide, is a 1pps signal traceable to UTC and it is
to that, that an external high stability 10 MHz or other frequency
oscillator may be phase locked to.
There all kinds of applications for gps capabilities, with navigation
being just one...
On 04/08/2021 22:56, Jim Pennino wrote:
I can't see why a GPS receiver would need a 10MHz oscillator. TheWhy?
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
PLL tech is a bit better than it was in 1971 when I first used one.
Do you mean frequency synthesizer tech?
It does seem that the internal crystal frequency is somewhat arbitrary, because the real working frequencies are generated by synthesizers. I
found products using 26MHz, as well as 10MHz for their primary oscillators.
chris<chris-nospam@tridac.net> wrote:
On 08/04/21 22:17, David Woolley wrote:
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
Just to disambiguate the comments about 10MHz. While the gps receiver
module will have an internal oscillator to suit the design, that
oscillator is not normally available external to the module. What the
gps receiver does provide, is a 1pps signal traceable to UTC and it is
to that, that an external high stability 10 MHz or other frequency
oscillator may be phase locked to.
No, the receiver does not provide a PPS signal unless you are calling
the entire gps device a receiver, which contains three separate devices,
i.e a radio receiver, a time standard (PPS) generator, and a NMEA
sentence generator.
There all kinds of applications for gps capabilities, with navigation
being just one...
No shit, Captain Obvious?
On 08/05/21 16:06, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/04/21 22:17, David Woolley wrote:
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
Just to disambiguate the comments about 10MHz. While the gps receiver
module will have an internal oscillator to suit the design, that
oscillator is not normally available external to the module. What the
gps receiver does provide, is a 1pps signal traceable to UTC and it is
to that, that an external high stability 10 MHz or other frequency
oscillator may be phase locked to.
No, the receiver does not provide a PPS signal unless you are calling
the entire gps device a receiver, which contains three separate devices,
i.e a radio receiver, a time standard (PPS) generator, and a NMEA
sentence generator.
I said gps receiver module, which most familiar with the technology
would take to mean the sort of thing you can buy for a few $ on Ebay,
but I guess with have to make allowances. So, enlighten us, just
how do the pps and nmea sentances appear on output pins ?. Some
sort of black magic perhaps ?...
Jim Pennino wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote:
So, in what today we call a gps, there are at least three separate
devices, a radio receiver to obtain the the GPS data, a device to
generate PPS, and a device to convert GPS data to NMEA sentences.
Except that in a modern chipset as used in a cell phone, all of those functions, plus wifi and possibly even more, is a single chip.
BTW, NMEA stands for National Marine Electronics Association. Guess why
commercial GPS devices output data in a marine standard format.
In circa 1996 those three devices would likely be in three 19 inch rack
mount boxes containing mostly discrete components.
Your timing is off by a substantial number of years: Garmin sold
(portable) GPSs for $105M already in 1995, I have track logs which are
older than this.
https://en.wikipedia.org/wiki/Garmin#Handheld_GPS
Today everything is fabricated on chip sets.
In regards to PPS alignment with UTC, since it is up to the equiment
maker, it will be whatever the equipment maker decides it will be.
Yes, and when the time nuts compare the PPS signal from those cheap
boards, against proper atomic reference clocks, like the sub-80 dollar
SURE which was popular about 8-10 years ago, they found about 25 ns RMS offset.
Terje
chris<chris-nospam@tridac.net> wrote:
On 08/05/21 16:06, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/04/21 22:17, David Woolley wrote:
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
Just to disambiguate the comments about 10MHz. While the gps receiver
module will have an internal oscillator to suit the design, that
oscillator is not normally available external to the module. What the
gps receiver does provide, is a 1pps signal traceable to UTC and it is >>>> to that, that an external high stability 10 MHz or other frequency
oscillator may be phase locked to.
No, the receiver does not provide a PPS signal unless you are calling
the entire gps device a receiver, which contains three separate devices, >>> i.e a radio receiver, a time standard (PPS) generator, and a NMEA
sentence generator.
I said gps receiver module, which most familiar with the technology
would take to mean the sort of thing you can buy for a few $ on Ebay,
but I guess with have to make allowances. So, enlighten us, just
how do the pps and nmea sentances appear on output pins ?. Some
sort of black magic perhaps ?...
Your question is answered at some length in another post.
I am not going to repost it just for you.
This is a technical discussion group, not a marketing group, and as such precision in language and the lack of arm waving matters.
The phrase "gps receiver module" is ambiguous in a discussion of GPS technology in 2021 absent a definition of what you mean.
On 08/05/21 18:00, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/05/21 16:06, Jim Pennino wrote:
chris<chris-nospam@tridac.net> wrote:
On 08/04/21 22:17, David Woolley wrote:
On 04/08/2021 17:59, Terje Mathisen wrote:
using a local 10 MHz oscillator
I can't see why a GPS receiver would need a 10MHz oscillator. The
chipping rates are 1.023Mhz, and 10.023MHz, and I would expect a
multiple of one of those
Just to disambiguate the comments about 10MHz. While the gps receiver >>>>> module will have an internal oscillator to suit the design, that
oscillator is not normally available external to the module. What the >>>>> gps receiver does provide, is a 1pps signal traceable to UTC and it is >>>>> to that, that an external high stability 10 MHz or other frequency
oscillator may be phase locked to.
No, the receiver does not provide a PPS signal unless you are calling
the entire gps device a receiver, which contains three separate devices, >>>> i.e a radio receiver, a time standard (PPS) generator, and a NMEA
sentence generator.
I said gps receiver module, which most familiar with the technology
would take to mean the sort of thing you can buy for a few $ on Ebay,
but I guess with have to make allowances. So, enlighten us, just
how do the pps and nmea sentances appear on output pins ?. Some
sort of black magic perhaps ?...
Your question is answered at some length in another post.
I am not going to repost it just for you.
This is a technical discussion group, not a marketing group, and as such
precision in language and the lack of arm waving matters.
The phrase "gps receiver module" is ambiguous in a discussion of GPS
technology in 2021 absent a definition of what you mean.
You talk about arm waving, but the fact is, you don't know, don't
understand and are too lazy to make the effort to find out...
Jim Pennino wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote: So, in
what today we call a gps, there are at least three separate
devices, a radio receiver to obtain the the GPS data, a device
to generate PPS, and a device to convert GPS data to NMEA
sentences.
Except that in a modern chipset as used in a cell phone, all of
those functions, plus wifi and possibly even more, is a single
chip.
BTW, NMEA stands for National Marine Electronics Association.
Guess why commercial GPS devices output data in a marine
standard format.
In circa 1996 those three devices would likely be in three 19
inch rack mount boxes containing mostly discrete components.
Your timing is off by a substantial number of years: Garmin
sold (portable) GPSs for $105M already in 1995, I have track
logs which are older than this.
https://en.wikipedia.org/wiki/Garmin#Handheld_GPS
Ted Heise <theise@panix.com> wrote:
On Thu, 5 Aug 2021 17:50:16 +0200,
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote: So, in
what today we call a gps, there are at least three separate
devices, a radio receiver to obtain the the GPS data, a device
to generate PPS, and a device to convert GPS data to NMEA
sentences.
Except that in a modern chipset as used in a cell phone, all of
those functions, plus wifi and possibly even more, is a single
chip.
BTW, NMEA stands for National Marine Electronics Association.
Guess why commercial GPS devices output data in a marine
standard format.
In circa 1996 those three devices would likely be in three 19
inch rack mount boxes containing mostly discrete components.
Your timing is off by a substantial number of years: Garmin
sold (portable) GPSs for $105M already in 1995, I have track
logs which are older than this.
https://en.wikipedia.org/wiki/Garmin#Handheld_GPS
I for one have been fascinated by this discussion. I ran my own
linux server (Slackware) for the better part of two decades, and
spent quite a lot of time getting ntp up and running to my
satisfaction.
The Garmin page reminds me I have an old etrex Vista HCx sitting
in a closet at home. I don't suppose there's any way to hack it
into use as a GPS clock? Guessing it may not even generate a PPS
signal, let alone have any way to get it out of the unit.
It was a terrific device for its time. It rode on the handlebars
of my bike for tracking speed, mileage, and routes.
I doubt it even has a PPS signal anywhere as a general purpose
navigation device would have no need for it. If it outputs serial NMEA sentences, which I would doubt, you could use it without PPS assuming
that an NTP accuracy in the millseconds is good enough for you.
Assuming you want to use something Linux PC based, then your options
are for minimal investment:
Finding something old on ebay or in a surplus store.
A DYI project using inexpensive board sets. This will cost in the $50
range and requires soldering, building cables, etc.
A Commercial Off The Shelf device you can just plug in. The cheapest
COTS devices I have found that provide both NMEA sentences and PPS are
about $180 new on ebay. They also contain a disciplined oven crystal
oscillator with both 10 MHz and 1 PPS singnals on BNC connectors. These
are handy if you also do RF or test equipment stuff.
There does not seem to be any current market for simple GPS receiver
devices with a serial output and PPS. There is a huge market for USB
GPS, actually GNSS, devices and if NTP accuracy in the millseconds is
good enough for you, they cost about $15 on amazon. Look for the VK-162 G-Mouse available from several vendors.
If you have a SBC such as a Arduino or Raspberry Pi, you have LOTS of
cheap options for building a NTP server.
And finally, if you have more money than common sense and want to be the
envy of your geek friends, you can get a new rubidium standard box on ebay for about $800 complete with NMEA and PPS on a DB-9 connector and 10 MHz
and PPS on BNC's.
On Thu, 5 Aug 2021 17:50:16 +0200,
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote: So, in
what today we call a gps, there are at least three separate
devices, a radio receiver to obtain the the GPS data, a device
to generate PPS, and a device to convert GPS data to NMEA
sentences.
Except that in a modern chipset as used in a cell phone, all of
those functions, plus wifi and possibly even more, is a single
chip.
BTW, NMEA stands for National Marine Electronics Association.
Guess why commercial GPS devices output data in a marine
standard format.
In circa 1996 those three devices would likely be in three 19
inch rack mount boxes containing mostly discrete components.
Your timing is off by a substantial number of years: Garmin
sold (portable) GPSs for $105M already in 1995, I have track
logs which are older than this.
https://en.wikipedia.org/wiki/Garmin#Handheld_GPS
I for one have been fascinated by this discussion. I ran my own
linux server (Slackware) for the better part of two decades, and
spent quite a lot of time getting ntp up and running to my
satisfaction.
The Garmin page reminds me I have an old etrex Vista HCx sitting
in a closet at home. I don't suppose there's any way to hack it
into use as a GPS clock? Guessing it may not even generate a PPS
signal, let alone have any way to get it out of the unit.
It was a terrific device for its time. It rode on the handlebars
of my bike for tracking speed, mileage, and routes.
Jim Pennino wrote:
Ted Heise <theise@panix.com> wrote:
On Thu, 5 Aug 2021 17:50:16 +0200,
Terje Mathisen <terje.mathisen@tmsw.no> wrote:
Jim Pennino wrote:
David Woolley <david@ex.djwhome.demon.invalid> wrote: So, in
what today we call a gps, there are at least three separate
devices, a radio receiver to obtain the the GPS data, a device
to generate PPS, and a device to convert GPS data to NMEA
sentences.
Except that in a modern chipset as used in a cell phone, all of
those functions, plus wifi and possibly even more, is a single
chip.
BTW, NMEA stands for National Marine Electronics Association.
Guess why commercial GPS devices output data in a marine
standard format.
In circa 1996 those three devices would likely be in three 19
inch rack mount boxes containing mostly discrete components.
Your timing is off by a substantial number of years: Garmin
sold (portable) GPSs for $105M already in 1995, I have track
logs which are older than this.
https://en.wikipedia.org/wiki/Garmin#Handheld_GPS
I for one have been fascinated by this discussion. I ran my own
linux server (Slackware) for the better part of two decades, and
spent quite a lot of time getting ntp up and running to my
satisfaction.
The Garmin page reminds me I have an old etrex Vista HCx sitting
in a closet at home. I don't suppose there's any way to hack it
into use as a GPS clock? Guessing it may not even generate a PPS
signal, let alone have any way to get it out of the unit.
It was a terrific device for its time. It rode on the handlebars
of my bike for tracking speed, mileage, and routes.
I doubt it even has a PPS signal anywhere as a general purpose
navigation device would have no need for it. If it outputs serial NMEA
sentences, which I would doubt, you could use it without PPS assuming
that an NTP accuracy in the millseconds is good enough for you.
The etrex was a late relative to the Garmin GPS II, II+, III etc, the
very first devices Garmin made used a two -channel receiver which they
had to multiplex across the tracked channels, but then they very early
made their own 12-channel chipset which was used all over the their many offerings.
One of those was the model 18 (afair), a puck antenna/receiver which had
a serial interface, and which included a PPS signal (although at TTL
levels, but that usually worked as is, otherwise the needed RS232
buffering chip was dirt cheap to add.
I used both this and the Motorola Oncore UT+ as the sole internal
reference for a multi-national corporation with 70-80K employees and
offices and/or factories in 130 countries.
Assuming you want to use something Linux PC based, then your options
are for minimal investment:
Finding something old on ebay or in a surplus store.
A DYI project using inexpensive board sets. This will cost in the $50
range and requires soldering, building cables, etc.
Right, except it is in fact easy to get even lower these days.
A Commercial Off The Shelf device you can just plug in. The cheapest
COTS devices I have found that provide both NMEA sentences and PPS are
about $180 new on ebay. They also contain a disciplined oven crystal
Almost certainly not: OCXOs were almost completely replaced by TCXOs due
to both faster startup times and much lower power usage, while getting
more or less exactly the same performance.
Have this changed back over the last 5+ years?
oscillator with both 10 MHz and 1 PPS singnals on BNC connectors. These
are handy if you also do RF or test equipment stuff.
There does not seem to be any current market for simple GPS receiver
devices with a serial output and PPS. There is a huge market for USB
GPS, actually GNSS, devices and if NTP accuracy in the millseconds is
good enough for you, they cost about $15 on amazon. Look for the VK-162
G-Mouse available from several vendors.
If you have a SBC such as a Arduino or Raspberry Pi, you have LOTS of
cheap options for building a NTP server.
And finally, if you have more money than common sense and want to be the
envy of your geek friends, you can get a new rubidium standard box on ebay >> for about $800 complete with NMEA and PPS on a DB-9 connector and 10 MHz
and PPS on BNC's.
What's funny is that you can in fact do far better even using USB, if
you use the same type of protocol as some of the early (non-GPS)
reference clocks did, i.e. a two-way protocol where the computer would
tell the reference clock what it thought the current time was, while the receiver measured the round-trip delay, and then sent back a timestamp adjusted to be as close as possible when the computer read it.
On RS232 this allowed sub-ms, with USB3 speeds you should be able to get
into the low microseconds range.
| Sysop: | Keyop |
|---|---|
| Location: | Huddersfield, West Yorkshire, UK |
| Users: | 112 |
| Nodes: | 8 (1 / 7) |
| Uptime: | 206:59:28 |
| Calls: | 2,465 |
| Files: | 8,603 |
| Messages: | 1,878,442 |