Does a higher GPS module output frequency have any advantages for usage in a GPSDO?
Many GPS modules can output either a PPS with a sawtooth problem (and
possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second
and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this
is where I'm not sure) the opportunity of providing more information to
the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1
PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10 lower
than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g.
locked phase) where the 10 kHz output seems to go down half a decade
per decade
Is this correctly interpreted?
Could an application that requires only frequency stability (such as a 3
cm repeater) benefit from a GPSDO that uses the higher frequency output
and thus eliminate the need for sawtooth correction data as the
correction data is already included in the fractional (jittering) output
frequency?
Erik.
Feeding the prs10 with gps 1pps I think has a problem that there is no
straightforward way to sawtooth correct so you can end up with
unexpectedly bad performance-- the time can be off by the duration of
the gps device clock for arbitrary lengths of time. I think you
kinda want a cleanup GPSDO first. :(
On Tue, Feb 4, 2025 at 10:06 PM Erik Kaashoek via time-nuts
time-nuts@lists.febo.com wrote:
Many GPS modules can output either a PPS with a sawtooth problem (and
possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second
and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this
is where I'm not sure) the opportunity of providing more information to
the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1
PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10 lower
than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g.
locked phase) where the 10 kHz output seems to go down half a decade
per decade
Is this correctly interpreted?
Could an application that requires only frequency stability (such as a 3
cm repeater) benefit from a GPSDO that uses the higher frequency output
and thus eliminate the need for sawtooth correction data as the
correction data is already included in the fractional (jittering) output
frequency?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Hi Erik
There is a little more to it.
There exist GPS/GNSS receivers that produces position+time at a higher
rate than once per second. You can see rates as 1, 2, 5, 10, 20 Hz for
instance.
Then you can produce higher pulse rates than PPS.
Then you have the sawtooth correction.
The sawtooth correction corrects difference between where it wanted to
set the pulse, and where it ended up due to the clock quanitzation. That
part is trivial, but if you now consider the number of solutions per
second, you can actually get new such information. This will be helpful.
If you can get 10 Hz on updates, then using a 10 Hz signal with each a
separate sawtooth correction would be very helpful. As you average on
that you improve your performance. This moves your starting point for
the slope to 0.1 s and thus gives you a good slope starting-point,
putting your 1 s performance down to a tenth.
Just operating on a higher frequency may not be significantly better,
but it may make the design around it simpler, which is not a bad thing.
Cheers,
Magnus
On 2/4/25 15:42, Erik Kaashoek via time-nuts wrote:
Many GPS modules can output either a PPS with a sawtooth problem (and
possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second
and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this
is where I'm not sure) the opportunity of providing more information
to the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1
PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10
lower than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g.
locked phase) where the 10 kHz output seems to go down half a decade
per decade
Is this correctly interpreted?
Could an application that requires only frequency stability (such as a
3 cm repeater) benefit from a GPSDO that uses the higher frequency
output and thus eliminate the need for sawtooth correction data as the
correction data is already included in the fractional (jittering)
output frequency?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Hi
There are a lot of ways these modules do things. In most cases, the exact same offsets that are applied to the 1 pps get applied to a frequency output. The offset can only be applied once it is calculated. Often that happens once a second. It might be calculated more often, if so that’s rarely documented.
Since they are doing a “phase offset”, you get the same sort of glitch on a “frequency” output. Unless it’s calculated more often, there is no advantage.
Bob
On Feb 4, 2025, at 9:42 AM, Erik Kaashoek via time-nuts time-nuts@lists.febo.com wrote:
Many GPS modules can output either a PPS with a sawtooth problem (and possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this is where I'm not sure) the opportunity of providing more information to the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1 PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10 lower than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g. locked phase) where the 10 kHz output seems to go down half a decade per decade
Is this correctly interpreted?
Could an application that requires only frequency stability (such as a 3 cm repeater) benefit from a GPSDO that uses the higher frequency output and thus eliminate the need for sawtooth correction data as the correction data is already included in the fractional (jittering) output frequency?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
This is my take. As I recall it from looking at the u-blox GNSS devices, internally they use a 48MHz TCXO as a timebase. This is divided down to give the output frequency using a direct digital synthesiser - i.e. a register with a maximum value of 48e6 which is increment with a value N every clock cycle. If N=1 the register overflows at 1Hz, if N=2 at 2 Hz, etc. However the 48MHz is not usually phase locked to the GPS signal, so the 1Hz time is corrected occasionally to align with a true second at least to the precision possible with a 48MHz clock - i.e. ~21ns. It thus exhibits a jitter and the jitter frequency depends on the exact frequency of the TCXO. The sawtooth signal if present provides an estimate of the error at each pulse which can be used to correct for it.
If you select an N that is an integer factor of 48e6 then the output frequency only has the same 21ns jitter as the 1Hz. However if N is not an integer factor then additional jitter appears. 48e6 has a lot of factors so there is a wide choice of minimum jitter frequencies, but certain obvious frequencies such as 10MHz isn't one of them. There are GNSS devices - I believe that some of the u-blox ones for example - where the (VC)TCXO is phase locked to the received signal so the "basic jitter" is eliminated - in effect these are a GPSDO in a package.
I'm not sure how this relates to your ADEV observations but it could be that for the GPS device used 10kHz wasn't integer related to the basic clock frequency. I don't think there is anything in the GPS specs that specifies the clock frequency.
-----Original Message-----
From: Erik Kaashoek via time-nuts time-nuts@lists.febo.com
Sent: 04 February 2025 14:43
To: time nuts time-nuts@lists.febo.com
Cc: Erik Kaashoek erik@kaashoek.com
Subject: [time-nuts] Does a higher GPS module output frequency have any advantages for usage in a GPSDO?
Many GPS modules can output either a PPS with a sawtooth problem (and possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this is where I'm not sure) the opportunity of providing more information to the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1 PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10 lower than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g.
locked phase) where the 10 kHz output seems to go down half a decade per decade Is this correctly interpreted?
Could an application that requires only frequency stability (such as a 3 cm repeater) benefit from a GPSDO that uses the higher frequency output and thus eliminate the need for sawtooth correction data as the correction data is already included in the fractional (jittering) output frequency?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe send an email to time-nuts-leave@lists.febo.com
I did some measurements on ZED-F9T GPS module outputting 100 Hz, 10 kHz
and 1 MHz
It seems to jump between two output frequencies at least 5 times per
second when outputting 100 Hz and about 50 times per second at least
when outputting 10 kHz.
This was measured with a frequency counter with a 20 ms gate time and
the difference between the two frequencies is about 8e-7 of the output
frequency and thus independent of the output frequency
It seems like a fractional divider is used to generate the output
frequency derived from the internal clock (48 MHz?)
When outputting 1 Mhz , fractional spurs are visible on the spectrum
analyzer. First spur offset is 22 Hz.
My SA is not good enough to see fractional spurs at lower output
frequencies.
With the counter set to 1 s gate time the jumping is no longer visible
and you measure the fractional frequency being the average output
frequency of all the jumps.
Would you happen to know if the LEA-M8F which has a 30.72 MHz oscillator also has the 48M oscillator?
From: john.haine--- via time-nuts time-nuts@lists.febo.com
Sent: Friday, February 7, 2025 4:21:17 AM
To: 'Discussion of precise time and frequency measurement' time-nuts@lists.febo.com
Cc: 'Erik Kaashoek' erik@kaashoek.com; john.haine@haine-online.net john.haine@haine-online.net
Subject: [time-nuts] Re: Does a higher GPS module output frequency have any advantages for usage in a GPSDO?
This is my take. As I recall it from looking at the u-blox GNSS devices, internally they use a 48MHz TCXO as a timebase. This is divided down to give the output frequency using a direct digital synthesiser - i.e. a register with a maximum value of 48e6 which is increment with a value N every clock cycle. If N=1 the register overflows at 1Hz, if N=2 at 2 Hz, etc. However the 48MHz is not usually phase locked to the GPS signal, so the 1Hz time is corrected occasionally to align with a true second at least to the precision possible with a 48MHz clock - i.e. ~21ns. It thus exhibits a jitter and the jitter frequency depends on the exact frequency of the TCXO. The sawtooth signal if present provides an estimate of the error at each pulse which can be used to correct for it.
If you select an N that is an integer factor of 48e6 then the output frequency only has the same 21ns jitter as the 1Hz. However if N is not an integer factor then additional jitter appears. 48e6 has a lot of factors so there is a wide choice of minimum jitter frequencies, but certain obvious frequencies such as 10MHz isn't one of them. There are GNSS devices - I believe that some of the u-blox ones for example - where the (VC)TCXO is phase locked to the received signal so the "basic jitter" is eliminated - in effect these are a GPSDO in a package.
I'm not sure how this relates to your ADEV observations but it could be that for the GPS device used 10kHz wasn't integer related to the basic clock frequency. I don't think there is anything in the GPS specs that specifies the clock frequency.
-----Original Message-----
From: Erik Kaashoek via time-nuts time-nuts@lists.febo.com
Sent: 04 February 2025 14:43
To: time nuts time-nuts@lists.febo.com
Cc: Erik Kaashoek erik@kaashoek.com
Subject: [time-nuts] Does a higher GPS module output frequency have any advantages for usage in a GPSDO?
Many GPS modules can output either a PPS with a sawtooth problem (and possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this is where I'm not sure) the opportunity of providing more information to the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1 PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10 lower than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g.
locked phase) where the 10 kHz output seems to go down half a decade per decade Is this correctly interpreted?
Could an application that requires only frequency stability (such as a 3 cm repeater) benefit from a GPSDO that uses the higher frequency output and thus eliminate the need for sawtooth correction data as the correction data is already included in the fractional (jittering) output frequency?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe send an email to time-nuts-leave@lists.febo.com
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Erik,
Could you please label the axes so we could know what you're showing?
Thanks,
Dana
On Fri, Feb 7, 2025 at 7:16 AM Erik Kaashoek via time-nuts <
time-nuts@lists.febo.com> wrote:
I did some measurements on ZED-F9T GPS module outputting 100 Hz, 10 kHz
and 1 MHz
It seems to jump between two output frequencies at least 5 times per
second when outputting 100 Hz and about 50 times per second at least
when outputting 10 kHz.
This was measured with a frequency counter with a 20 ms gate time and
the difference between the two frequencies is about 8e-7 of the output
frequency and thus independent of the output frequency
It seems like a fractional divider is used to generate the output
frequency derived from the internal clock (48 MHz?)
When outputting 1 Mhz , fractional spurs are visible on the spectrum
analyzer. First spur offset is 22 Hz.
My SA is not good enough to see fractional spurs at lower output
frequencies.
With the counter set to 1 s gate time the jumping is no longer visible
and you measure the fractional frequency being the average output
frequency of all the jumps._______________________________________________
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Hi
All of these basic GPS modules have one TCXO / XO in them for doing “all the GPS stuff”. On rare occasions they might have a 32.768 KHz “real time clock” oscillator. That is pretty rare and contributes nothing to the GPS processing or the RF / PPS outputs.
Bob
On Feb 7, 2025, at 8:18 AM, Ed Marciniak via time-nuts time-nuts@lists.febo.com wrote:
Would you happen to know if the LEA-M8F which has a 30.72 MHz oscillator also has the 48M oscillator?
From: john.haine--- via time-nuts time-nuts@lists.febo.com
Sent: Friday, February 7, 2025 4:21:17 AM
To: 'Discussion of precise time and frequency measurement' time-nuts@lists.febo.com
Cc: 'Erik Kaashoek' erik@kaashoek.com; john.haine@haine-online.net john.haine@haine-online.net
Subject: [time-nuts] Re: Does a higher GPS module output frequency have any advantages for usage in a GPSDO?
This is my take. As I recall it from looking at the u-blox GNSS devices, internally they use a 48MHz TCXO as a timebase. This is divided down to give the output frequency using a direct digital synthesiser - i.e. a register with a maximum value of 48e6 which is increment with a value N every clock cycle. If N=1 the register overflows at 1Hz, if N=2 at 2 Hz, etc. However the 48MHz is not usually phase locked to the GPS signal, so the 1Hz time is corrected occasionally to align with a true second at least to the precision possible with a 48MHz clock - i.e. ~21ns. It thus exhibits a jitter and the jitter frequency depends on the exact frequency of the TCXO. The sawtooth signal if present provides an estimate of the error at each pulse which can be used to correct for it.
If you select an N that is an integer factor of 48e6 then the output frequency only has the same 21ns jitter as the 1Hz. However if N is not an integer factor then additional jitter appears. 48e6 has a lot of factors so there is a wide choice of minimum jitter frequencies, but certain obvious frequencies such as 10MHz isn't one of them. There are GNSS devices - I believe that some of the u-blox ones for example - where the (VC)TCXO is phase locked to the received signal so the "basic jitter" is eliminated - in effect these are a GPSDO in a package.
I'm not sure how this relates to your ADEV observations but it could be that for the GPS device used 10kHz wasn't integer related to the basic clock frequency. I don't think there is anything in the GPS specs that specifies the clock frequency.
-----Original Message-----
From: Erik Kaashoek via time-nuts time-nuts@lists.febo.com
Sent: 04 February 2025 14:43
To: time nuts time-nuts@lists.febo.com
Cc: Erik Kaashoek erik@kaashoek.com
Subject: [time-nuts] Does a higher GPS module output frequency have any advantages for usage in a GPSDO?
Many GPS modules can output either a PPS with a sawtooth problem (and possible sawtooth correction information) or a other frequency.
The PPS has the advantage that it clearly fixes the top of the second and, when used in a PLL mode, prevent the GPSDO from drifting its phase.
But a higher output frequency (which will jitter) could have (and this is where I'm not sure) the opportunity of providing more information to the GPSDO.
I checked this by measuring the ADEV of a GPS module output, both at 1 PPS and at 10 kHz. The 10 kHz output starts at tau=1 s a factor 10 lower than the 1 PPS.
But the slope of the 1 PPS ADEV is one decade down per decade (e.g.
locked phase) where the 10 kHz output seems to go down half a decade per decade Is this correctly interpreted?
Could an application that requires only frequency stability (such as a 3 cm repeater) benefit from a GPSDO that uses the higher frequency output and thus eliminate the need for sawtooth correction data as the correction data is already included in the fractional (jittering) output frequency?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe send an email to time-nuts-leave@lists.febo.com
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Dana,
Hope the picture makes it this way
This is cycle time, measured with 40 ps resolution, of about 3500
individual cycles of the output clock set to 10 kHz.
It seems the cycle time is set to the integer divide of the system clock
closest to the requested output frequency.
Every 460 (or sometimes 470) cycles a individual output cycle is
stretched (or shortened) with about 8 ns to move the average output
frequency to the requested output frequency.
The number of cycle insertions/deletions depend on the fractional
difference between integer divided system clock and the requested clock.
If these are very close you could get a hanging bridge I guess as there
is no need to insert a cycle for a long time.
On 7-2-2025 15:47, Dana Whitlow via time-nuts wrote:
Erik,
Could you please label the axes so we could know what you're showing?
Thanks,
Dana
On Fri, Feb 7, 2025 at 7:16 AM Erik Kaashoek via time-nuts <
time-nuts@lists.febo.com> wrote:
I did some measurements on ZED-F9T GPS module outputting 100 Hz, 10 kHz
and 1 MHz
It seems to jump between two output frequencies at least 5 times per
second when outputting 100 Hz and about 50 times per second at least
when outputting 10 kHz.
This was measured with a frequency counter with a 20 ms gate time and
the difference between the two frequencies is about 8e-7 of the output
frequency and thus independent of the output frequency
It seems like a fractional divider is used to generate the output
frequency derived from the internal clock (48 MHz?)
When outputting 1 Mhz , fractional spurs are visible on the spectrum
analyzer. First spur offset is 22 Hz.
My SA is not good enough to see fractional spurs at lower output
frequencies.
With the counter set to 1 s gate time the jumping is no longer visible
and you measure the fractional frequency being the average output
frequency of all the jumps._______________________________________________
time-nuts mailing list --time-nuts@lists.febo.com
To unsubscribe send an email totime-nuts-leave@lists.febo.com
time-nuts mailing list --time-nuts@lists.febo.com
To unsubscribe send an email totime-nuts-leave@lists.febo.com