TV
Tom Van Baak
Sat, Nov 12, 2016 9:54 PM
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
- attached plot: TBolt-10day-fit0-e09.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each frequency plot is a free-running TBolt (no GPS, no disciplining). The X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What you see at this scale is that all the OCXO are quite stable. Also, some of them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days for a drift rate of 2e-10/day. It looks large in this plot but its well under the typical spec, such as 5e-10/day for a 10811A. We see a variety of drift rates, including some that appear to be zero: flat line. At this scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things to do. Zoom in and zoom out.
- attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale is still 10 days. Now we can see the drift much better. Also at this level we can see instability of each OCXO (or the lab environment). At this scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is 25x better than the 10811A spec. CH13, mentioned above, is not zero drift after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any of the 20 10-day plots. Your eye will tell you that the daily drift rate can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
- attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the Y-scale is now 1e-8 per division. Once a month I created a temporary thermal event in the lab (the little "speed bumps") which we will ignore for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also CH14 and CH16. In fact over 100 days most of them are logarithmic but the coefficients vary considerably so it's hard to see this at a common scale. Note also the logarithmic curve is vastly more apparent in the first few days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks linear if you're looking close enough. A straight highway may look linear but the equator is circular. So most OCXO drift (age) with a logarithmic curve and this is visible over long enough measurements. But for shorter time spans it will appear linear. Or, more likely, internal and external stability issues will dominate and this spoils any linear vs. log discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
1) attached plot: TBolt-10day-fit0-e09.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each frequency plot is a free-running TBolt (no GPS, no disciplining). The X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What you see at this scale is that all the OCXO are quite stable. Also, some of them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days for a drift rate of 2e-10/day. It looks large in this plot but its well under the typical spec, such as 5e-10/day for a 10811A. We see a variety of drift rates, including some that appear to be zero: flat line. At this scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things to do. Zoom in and zoom out.
2) attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale is still 10 days. Now we can see the drift much better. Also at this level we can see instability of each OCXO (or the lab environment). At this scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is 25x better than the 10811A spec. CH13, mentioned above, is not zero drift after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any of the 20 10-day plots. Your eye will tell you that the daily drift rate can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
3) attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the Y-scale is now 1e-8 per division. Once a month I created a temporary thermal event in the lab (the little "speed bumps") which we will ignore for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also CH14 and CH16. In fact over 100 days most of them are logarithmic but the coefficients vary considerably so it's hard to see this at a common scale. Note also the logarithmic curve is vastly more apparent in the first few days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks linear if you're looking close enough. A straight highway may look linear but the equator is circular. So most OCXO drift (age) with a logarithmic curve and this is visible over long enough measurements. But for shorter time spans it will appear linear. Or, more likely, internal and external stability issues will dominate and this spoils any linear vs. log discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
D
djl
Sat, Nov 12, 2016 10:16 PM
Interesting, Tom. I don't think I see any of those pesky grain boundary
shifts or readjustments in the lattice structure? If I remember, these
can cause instant shifts in frequency that do not heal?
Don
On 2016-11-12 14:54, Tom Van Baak wrote:
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it
provides an interesting set of data from which to make visual answers
to recent questions. Here are three plots.
- attached plot: TBolt-10day-fit0-e09.gif (
http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each
frequency plot is a free-running TBolt (no GPS, no disciplining). The
X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division.
What you see at this scale is that all the OCXO are quite stable.
Also, some of them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10
days for a drift rate of 2e-10/day. It looks large in this plot but
its well under the typical spec, such as 5e-10/day for a 10811A. We
see a variety of drift rates, including some that appear to be zero:
flat line. At this scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two
things to do. Zoom in and zoom out.
- attached plot: TBolt-10day-fit0-e10.gif (
http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The
X-scale is still 10 days. Now we can see the drift much better. Also
at this level we can see instability of each OCXO (or the lab
environment). At this scale, channels CH10 and CH14 are "off the
chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for
a drift rate of 2e-11/day. This is 25x better than the 10811A spec.
CH13, mentioned above, is not zero drift after all, but its drift rate
is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability)
are large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift
rate based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at
any of the 20 10-day plots. Your eye will tell you that the daily
drift rate can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules.
In a sense they each have their own personality. So one needs to be
very careful about algorithms that assume any sort of constant or
consistent behavior.
- attached plot: TBolt-100day-fit0-e08.gif (
http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the
Y-scale is now 1e-8 per division. Once a month I created a temporary
thermal event in the lab (the little "speed bumps") which we will
ignore for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift.
Also CH14 and CH16. In fact over 100 days most of them are logarithmic
but the coefficients vary considerably so it's hard to see this at a
common scale. Note also the logarithmic curve is vastly more apparent
in the first few days or weeks of operation, but I don't have that
data.
In general, any exponential or log or parabolic or circular curve
looks linear if you're looking close enough. A straight highway may
look linear but the equator is circular. So most OCXO drift (age) with
a logarithmic curve and this is visible over long enough measurements.
But for shorter time spans it will appear linear. Or, more likely,
internal and external stability issues will dominate and this spoils
any linear vs. log discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob
;-)
/tvb
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
--
Dr. Don Latham
PO Box 404, Frenchtown, MT, 59834
VOX: 406-626-4304
Interesting, Tom. I don't think I see any of those pesky grain boundary
shifts or readjustments in the lattice structure? If I remember, these
can cause instant shifts in frequency that do not heal?
Don
On 2016-11-12 14:54, Tom Van Baak wrote:
> There were postings recently about OCXO ageing, or drift rates.
>
> I've been testing a batch of TBolts for a couple of months and it
> provides an interesting set of data from which to make visual answers
> to recent questions. Here are three plots.
>
>
> 1) attached plot: TBolt-10day-fit0-e09.gif (
> http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
>
> A bunch of oscillators are measured with a 20-channel system. Each
> frequency plot is a free-running TBolt (no GPS, no disciplining). The
> X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division.
> What you see at this scale is that all the OCXO are quite stable.
> Also, some of them show drift.
>
> For example, the OCXO frequency in channel 14 changes by 2e-9 in 10
> days for a drift rate of 2e-10/day. It looks large in this plot but
> its well under the typical spec, such as 5e-10/day for a 10811A. We
> see a variety of drift rates, including some that appear to be zero:
> flat line. At this scale, CH13, for example, seems to have no drift.
>
> But the drift, when present, appears quite linear. So there are two
> things to do. Zoom in and zoom out.
>
>
> 2) attached plot: TBolt-10day-fit0-e10.gif (
> http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
>
> Here we zoom in by changing the Y-scale to 1e-10 per division. The
> X-scale is still 10 days. Now we can see the drift much better. Also
> at this level we can see instability of each OCXO (or the lab
> environment). At this scale, channels CH10 and CH14 are "off the
> chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for
> a drift rate of 2e-11/day. This is 25x better than the 10811A spec.
> CH13, mentioned above, is not zero drift after all, but its drift rate
> is even lower, close to 1e-11/day.
>
> For some oscillators the wiggles in the data (frequency instability)
> are large enough that the drift rate is not clearly measurable.
>
> The 10-day plots suggests you would not want to try to measure drift
> rate based on just one day of data.
>
> The plots also suggest that drift rate is not a hard constant. Look at
> any of the 20 10-day plots. Your eye will tell you that the daily
> drift rate can change significantly from day to day to day.
>
> The plots show that an OCXO doesn't necessarily follow strict rules.
> In a sense they each have their own personality. So one needs to be
> very careful about algorithms that assume any sort of constant or
> consistent behavior.
>
>
> 3) attached plot: TBolt-100day-fit0-e08.gif (
> http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
>
> Here we look at 100 days of data instead of just 10 days. To fit, the
> Y-scale is now 1e-8 per division. Once a month I created a temporary
> thermal event in the lab (the little "speed bumps") which we will
> ignore for now.
>
> At this long-term scale, OCXO in CH09 has textbook logarithmic drift.
> Also CH14 and CH16. In fact over 100 days most of them are logarithmic
> but the coefficients vary considerably so it's hard to see this at a
> common scale. Note also the logarithmic curve is vastly more apparent
> in the first few days or weeks of operation, but I don't have that
> data.
>
> In general, any exponential or log or parabolic or circular curve
> looks linear if you're looking close enough. A straight highway may
> look linear but the equator is circular. So most OCXO drift (age) with
> a logarithmic curve and this is visible over long enough measurements.
> But for shorter time spans it will appear linear. Or, more likely,
> internal and external stability issues will dominate and this spoils
> any linear vs. log discussion.
>
> So is it linear or log? The answer is it depends. Now I sound like Bob
> ;-)
>
> /tvb
>
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
--
Dr. Don Latham
PO Box 404, Frenchtown, MT, 59834
VOX: 406-626-4304
AK
Attila Kinali
Sat, Nov 12, 2016 10:25 PM
- attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale
is still 10 days. Now we can see the drift much better. Also at this level
we can see instability of each OCXO (or the lab environment).
These look like textbook examples of random walk frequency modulation.
As this is a random process it is not surprising that they look different
for each oscillator.
Do you know what caused the frequency jump of CH18?
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
Hehe. There are worse things than sounding like Bob :-)
Yes, depending on the data you show, it is not clear whether one
should do a linear or a logarithmic fit. But keep in mind that
a logarithmic curve can look like a linear curve depending on the
parameters. Also Depending on the exact properties of the noise
and the evaluation function, a linear fit might be better on a
logarithmic curve than a logarithmic fit. We are talking about
system identification under noise, after all. And things become
strange when noise involved. Even more so when it is not white
gaussian noise.
Attila Kinali
--
Malek's Law:
Any simple idea will be worded in the most complicated way.
On Sat, 12 Nov 2016 13:54:14 -0800
"Tom Van Baak" <tvb@LeapSecond.com> wrote:
> 2) attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
>
> Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale
> is still 10 days. Now we can see the drift much better. Also at this level
> we can see instability of each OCXO (or the lab environment).
These look like textbook examples of random walk frequency modulation.
As this is a random process it is not surprising that they look different
for each oscillator.
> 3) attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Do you know what caused the frequency jump of CH18?
> So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
Hehe. There are worse things than sounding like Bob :-)
Yes, depending on the data you show, it is not clear whether one
should do a linear or a logarithmic fit. But keep in mind that
a logarithmic curve can look like a linear curve depending on the
parameters. Also Depending on the exact properties of the noise
and the evaluation function, a linear fit might be better on a
logarithmic curve than a logarithmic fit. We are talking about
system identification under noise, after all. And things become
strange when noise involved. Even more so when it is not white
gaussian noise.
Attila Kinali
--
Malek's Law:
Any simple idea will be worded in the most complicated way.
TM
Tom Miller
Sat, Nov 12, 2016 10:29 PM
Just out of curiosity, what is the age of each of these Tbolts? (i.e. date
codes?)
Thanks
----- Original Message -----
From: "Tom Van Baak" tvb@LeapSecond.com
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Saturday, November 12, 2016 4:54 PM
Subject: [time-nuts] quartz drift rates, linear or log
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides
an interesting set of data from which to make visual answers to recent
questions. Here are three plots.
Just out of curiosity, what is the age of each of these Tbolts? (i.e. date
codes?)
Thanks
----- Original Message -----
From: "Tom Van Baak" <tvb@LeapSecond.com>
To: "Discussion of precise time and frequency measurement"
<time-nuts@febo.com>
Sent: Saturday, November 12, 2016 4:54 PM
Subject: [time-nuts] quartz drift rates, linear or log
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides
an interesting set of data from which to make visual answers to recent
questions. Here are three plots.
BS
Bob Stewart
Sat, Nov 12, 2016 10:49 PM
So, are you measuring OCXO stability or EFC stability?
Bob
-----------------------------------------------------------------
AE6RV.com
GFS GPSDO list:
groups.yahoo.com/neo/groups/GFS-GPSDOs/info
From: Tom Van Baak <tvb@LeapSecond.com>
To: Discussion of precise time and frequency measurement time-nuts@febo.com
Sent: Saturday, November 12, 2016 3:54 PM
Subject: [time-nuts] quartz drift rates, linear or log
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
- attached plot: TBolt-10day-fit0-e09.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each frequency plot is a free-running TBolt (no GPS, no disciplining). The X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What you see at this scale is that all the OCXO are quite stable. Also, some of them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days for a drift rate of 2e-10/day. It looks large in this plot but its well under the typical spec, such as 5e-10/day for a 10811A. We see a variety of drift rates, including some that appear to be zero: flat line. At this scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things to do. Zoom in and zoom out.
- attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale is still 10 days. Now we can see the drift much better. Also at this level we can see instability of each OCXO (or the lab environment). At this scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is 25x better than the 10811A spec. CH13, mentioned above, is not zero drift after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any of the 20 10-day plots. Your eye will tell you that the daily drift rate can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
- attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the Y-scale is now 1e-8 per division. Once a month I created a temporary thermal event in the lab (the little "speed bumps") which we will ignore for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also CH14 and CH16. In fact over 100 days most of them are logarithmic but the coefficients vary considerably so it's hard to see this at a common scale. Note also the logarithmic curve is vastly more apparent in the first few days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks linear if you're looking close enough. A straight highway may look linear but the equator is circular. So most OCXO drift (age) with a logarithmic curve and this is visible over long enough measurements. But for shorter time spans it will appear linear. Or, more likely, internal and external stability issues will dominate and this spoils any linear vs. log discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
So, are you measuring OCXO stability or EFC stability?
Bob
-----------------------------------------------------------------
AE6RV.com
GFS GPSDO list:
groups.yahoo.com/neo/groups/GFS-GPSDOs/info
From: Tom Van Baak <tvb@LeapSecond.com>
To: Discussion of precise time and frequency measurement <time-nuts@febo.com>
Sent: Saturday, November 12, 2016 3:54 PM
Subject: [time-nuts] quartz drift rates, linear or log
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
1) attached plot: TBolt-10day-fit0-e09.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each frequency plot is a free-running TBolt (no GPS, no disciplining). The X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What you see at this scale is that all the OCXO are quite stable. Also, some of them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days for a drift rate of 2e-10/day. It looks large in this plot but its well under the typical spec, such as 5e-10/day for a 10811A. We see a variety of drift rates, including some that appear to be zero: flat line. At this scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things to do. Zoom in and zoom out.
2) attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale is still 10 days. Now we can see the drift much better. Also at this level we can see instability of each OCXO (or the lab environment). At this scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is 25x better than the 10811A spec. CH13, mentioned above, is not zero drift after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any of the 20 10-day plots. Your eye will tell you that the daily drift rate can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
3) attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the Y-scale is now 1e-8 per division. Once a month I created a temporary thermal event in the lab (the little "speed bumps") which we will ignore for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also CH14 and CH16. In fact over 100 days most of them are logarithmic but the coefficients vary considerably so it's hard to see this at a common scale. Note also the logarithmic curve is vastly more apparent in the first few days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks linear if you're looking close enough. A straight highway may look linear but the equator is circular. So most OCXO drift (age) with a logarithmic curve and this is visible over long enough measurements. But for shorter time spans it will appear linear. Or, more likely, internal and external stability issues will dominate and this spoils any linear vs. log discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
BC
Bob Camp
Sat, Nov 12, 2016 10:58 PM
Hi
One would guess that the OCXO’s all left the factory set to center at zero volts on the EFC. One thing
that is pretty easy to do is to look at the date code on the OCXO and the EFC voltage. That plus the
sensitivity (one could cheat and look at the frequency rather than EFC) will give you a guess at the total aging
of the OCXO since it was shipped.
Yes there are some holes in this. You don’t really know just how well somebody set the parts at shipment.
You don’t know how much they shifted when soldered into the board. You don’t know how much things
like shock transport shifted the part. Best guess is that all of these (taken individually) are in the “parts in
10^-8” region. Put another way … it all falls apart if the OCXO is still within < 1x10^-7 of it’s zero volt
frequency.
Bob
On Nov 12, 2016, at 4:54 PM, Tom Van Baak tvb@LeapSecond.com wrote:
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
- attached plot: TBolt-10day-fit0-e09.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each frequency plot is a free-running TBolt (no GPS, no disciplining). The X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What you see at this scale is that all the OCXO are quite stable. Also, some of them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days for a drift rate of 2e-10/day. It looks large in this plot but its well under the typical spec, such as 5e-10/day for a 10811A. We see a variety of drift rates, including some that appear to be zero: flat line. At this scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things to do. Zoom in and zoom out.
- attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale is still 10 days. Now we can see the drift much better. Also at this level we can see instability of each OCXO (or the lab environment). At this scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is 25x better than the 10811A spec. CH13, mentioned above, is not zero drift after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any of the 20 10-day plots. Your eye will tell you that the daily drift rate can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
- attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the Y-scale is now 1e-8 per division. Once a month I created a temporary thermal event in the lab (the little "speed bumps") which we will ignore for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also CH14 and CH16. In fact over 100 days most of them are logarithmic but the coefficients vary considerably so it's hard to see this at a common scale. Note also the logarithmic curve is vastly more apparent in the first few days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks linear if you're looking close enough. A straight highway may look linear but the equator is circular. So most OCXO drift (age) with a logarithmic curve and this is visible over long enough measurements. But for shorter time spans it will appear linear. Or, more likely, internal and external stability issues will dominate and this spoils any linear vs. log discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
<TBolt-10day-fit0-e09.gif><TBolt-10day-fit0-e10.gif><TBolt-100day-fit0-e08.gif>_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hi
One would *guess* that the OCXO’s all left the factory set to center at zero volts on the EFC. One thing
that is pretty easy to do is to look at the date code on the OCXO and the EFC voltage. That plus the
sensitivity (one could cheat and look at the frequency rather than EFC) will give you a guess at the total aging
of the OCXO since it was shipped.
Yes there are some holes in this. You don’t really know just how well somebody set the parts at shipment.
You don’t know how much they shifted when soldered into the board. You don’t know how much things
like shock transport shifted the part. Best guess is that all of these (taken individually) are in the “parts in
10^-8” region. Put another way … it all falls apart if the OCXO is still within < 1x10^-7 of it’s zero volt
frequency.
Bob
> On Nov 12, 2016, at 4:54 PM, Tom Van Baak <tvb@LeapSecond.com> wrote:
>
> There were postings recently about OCXO ageing, or drift rates.
>
> I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
>
>
> 1) attached plot: TBolt-10day-fit0-e09.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
>
> A bunch of oscillators are measured with a 20-channel system. Each frequency plot is a free-running TBolt (no GPS, no disciplining). The X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What you see at this scale is that all the OCXO are quite stable. Also, some of them show drift.
>
> For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days for a drift rate of 2e-10/day. It looks large in this plot but its well under the typical spec, such as 5e-10/day for a 10811A. We see a variety of drift rates, including some that appear to be zero: flat line. At this scale, CH13, for example, seems to have no drift.
>
> But the drift, when present, appears quite linear. So there are two things to do. Zoom in and zoom out.
>
>
> 2) attached plot: TBolt-10day-fit0-e10.gif ( http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
>
> Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale is still 10 days. Now we can see the drift much better. Also at this level we can see instability of each OCXO (or the lab environment). At this scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is 25x better than the 10811A spec. CH13, mentioned above, is not zero drift after all, but its drift rate is even lower, close to 1e-11/day.
>
> For some oscillators the wiggles in the data (frequency instability) are large enough that the drift rate is not clearly measurable.
>
> The 10-day plots suggests you would not want to try to measure drift rate based on just one day of data.
>
> The plots also suggest that drift rate is not a hard constant. Look at any of the 20 10-day plots. Your eye will tell you that the daily drift rate can change significantly from day to day to day.
>
> The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
>
>
> 3) attached plot: TBolt-100day-fit0-e08.gif ( http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
>
> Here we look at 100 days of data instead of just 10 days. To fit, the Y-scale is now 1e-8 per division. Once a month I created a temporary thermal event in the lab (the little "speed bumps") which we will ignore for now.
>
> At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also CH14 and CH16. In fact over 100 days most of them are logarithmic but the coefficients vary considerably so it's hard to see this at a common scale. Note also the logarithmic curve is vastly more apparent in the first few days or weeks of operation, but I don't have that data.
>
> In general, any exponential or log or parabolic or circular curve looks linear if you're looking close enough. A straight highway may look linear but the equator is circular. So most OCXO drift (age) with a logarithmic curve and this is visible over long enough measurements. But for shorter time spans it will appear linear. Or, more likely, internal and external stability issues will dominate and this spoils any linear vs. log discussion.
>
> So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
>
> /tvb
> <TBolt-10day-fit0-e09.gif><TBolt-10day-fit0-e10.gif><TBolt-100day-fit0-e08.gif>_______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
CS
Charles Steinmetz
Sat, Nov 12, 2016 11:51 PM
Yes, depending on the data you show, it is not clear
whether one should do a linear or a logarithmic fit.
When the purpose is correcting a GPSDO local oscillator during holdover,
it depends on how long one expects to trust the corrected frequency.
Practical realities make it pointless to trust corrections longer than a
day or so, if that long. At that scale, these all look pretty linear.
NB: That may not be the case for a TCXO, or a crappy OCXO. But then,
it would probably be pointless to correct these for longer than an hour
or so, at which scale they also may look linear.
Best regards,
Charles
Attila wrote:
> Yes, depending on the data you show, it is not clear
> whether one should do a linear or a logarithmic fit.
When the purpose is correcting a GPSDO local oscillator during holdover,
it depends on how long one expects to trust the corrected frequency.
Practical realities make it pointless to trust corrections longer than a
day or so, if that long. At that scale, these all look pretty linear.
NB: That may *not* be the case for a TCXO, or a crappy OCXO. But then,
it would probably be pointless to correct these for longer than an hour
or so, at which scale they also may look linear.
Best regards,
Charles
J
jimlux
Sun, Nov 13, 2016 12:12 AM
On 11/12/16 1:54 PM, Tom Van Baak wrote:
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
<snip>
The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
<snip>
I think this is a general rule: when they make precision oscillators
(USOs) for spacecraft, they make a fairly large batch (a couple dozen)
to some intermediate level of assembly, and then they run them for a
while and watch them, and from that, they pick the "good" ones. (where
good is somewhat mission dependent).
The good ones go through the rest of the assembly process into the "box
level" and, then, from those, the "really good" ones are selected.
On 11/12/16 1:54 PM, Tom Van Baak wrote:
> There were postings recently about OCXO ageing, or drift rates.
>
> I've been testing a batch of TBolts for a couple of months and it provides an interesting set of data from which to make visual answers to recent questions. Here are three plots.
> <snip>
> The plots show that an OCXO doesn't necessarily follow strict rules. In a sense they each have their own personality. So one needs to be very careful about algorithms that assume any sort of constant or consistent behavior.
>
>
> <snip>
I think this is a general rule: when they make precision oscillators
(USOs) for spacecraft, they make a fairly large batch (a couple dozen)
to some intermediate level of assembly, and then they run them for a
while and watch them, and from that, they pick the "good" ones. (where
good is somewhat mission dependent).
The good ones go through the rest of the assembly process into the "box
level" and, then, from those, the "really good" ones are selected.
SS
Scott Stobbe
Sun, Nov 13, 2016 1:56 AM
Those are wonderful plots :)
I vaguely recall that a 1ppm frequency shift is approximately equivalent to
the mass transfer of one molecular layer of a crystal. So at some point
your counting atoms if there was no noise, thermal disturbance, mechanical
disturbance...
On Sat, Nov 12, 2016 at 5:00 PM Tom Van Baak tvb@leapsecond.com wrote:
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides
an interesting set of data from which to make visual answers to recent
questions. Here are three plots.
- attached plot: TBolt-10day-fit0-e09.gif (
http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each
frequency plot is a free-running TBolt (no GPS, no disciplining). The
X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What
you see at this scale is that all the OCXO are quite stable. Also, some of
them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days
for a drift rate of 2e-10/day. It looks large in this plot but its well
under the typical spec, such as 5e-10/day for a 10811A. We see a variety of
drift rates, including some that appear to be zero: flat line. At this
scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things
to do. Zoom in and zoom out.
- attached plot: TBolt-10day-fit0-e10.gif (
http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale
is still 10 days. Now we can see the drift much better. Also at this level
we can see instability of each OCXO (or the lab environment). At this
scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in
CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is
25x better than the 10811A spec. CH13, mentioned above, is not zero drift
after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are
large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate
based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any
of the 20 10-day plots. Your eye will tell you that the daily drift rate
can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a
sense they each have their own personality. So one needs to be very careful
about algorithms that assume any sort of constant or consistent behavior.
- attached plot: TBolt-100day-fit0-e08.gif (
http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the
Y-scale is now 1e-8 per division. Once a month I created a temporary
thermal event in the lab (the little "speed bumps") which we will ignore
for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also
CH14 and CH16. In fact over 100 days most of them are logarithmic but the
coefficients vary considerably so it's hard to see this at a common scale.
Note also the logarithmic curve is vastly more apparent in the first few
days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks
linear if you're looking close enough. A straight highway may look linear
but the equator is circular. So most OCXO drift (age) with a logarithmic
curve and this is visible over long enough measurements. But for shorter
time spans it will appear linear. Or, more likely, internal and external
stability issues will dominate and this spoils any linear vs. log
discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Those are wonderful plots :)
I vaguely recall that a 1ppm frequency shift is approximately equivalent to
the mass transfer of one molecular layer of a crystal. So at some point
your counting atoms if there was no noise, thermal disturbance, mechanical
disturbance...
On Sat, Nov 12, 2016 at 5:00 PM Tom Van Baak <tvb@leapsecond.com> wrote:
> There were postings recently about OCXO ageing, or drift rates.
>
> I've been testing a batch of TBolts for a couple of months and it provides
> an interesting set of data from which to make visual answers to recent
> questions. Here are three plots.
>
>
> 1) attached plot: TBolt-10day-fit0-e09.gif (
> http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
>
> A bunch of oscillators are measured with a 20-channel system. Each
> frequency plot is a free-running TBolt (no GPS, no disciplining). The
> X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What
> you see at this scale is that all the OCXO are quite stable. Also, some of
> them show drift.
>
> For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days
> for a drift rate of 2e-10/day. It looks large in this plot but its well
> under the typical spec, such as 5e-10/day for a 10811A. We see a variety of
> drift rates, including some that appear to be zero: flat line. At this
> scale, CH13, for example, seems to have no drift.
>
> But the drift, when present, appears quite linear. So there are two things
> to do. Zoom in and zoom out.
>
>
> 2) attached plot: TBolt-10day-fit0-e10.gif (
> http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
>
> Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale
> is still 10 days. Now we can see the drift much better. Also at this level
> we can see instability of each OCXO (or the lab environment). At this
> scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in
> CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is
> 25x better than the 10811A spec. CH13, mentioned above, is not zero drift
> after all, but its drift rate is even lower, close to 1e-11/day.
>
> For some oscillators the wiggles in the data (frequency instability) are
> large enough that the drift rate is not clearly measurable.
>
> The 10-day plots suggests you would not want to try to measure drift rate
> based on just one day of data.
>
> The plots also suggest that drift rate is not a hard constant. Look at any
> of the 20 10-day plots. Your eye will tell you that the daily drift rate
> can change significantly from day to day to day.
>
> The plots show that an OCXO doesn't necessarily follow strict rules. In a
> sense they each have their own personality. So one needs to be very careful
> about algorithms that assume any sort of constant or consistent behavior.
>
>
> 3) attached plot: TBolt-100day-fit0-e08.gif (
> http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
>
> Here we look at 100 days of data instead of just 10 days. To fit, the
> Y-scale is now 1e-8 per division. Once a month I created a temporary
> thermal event in the lab (the little "speed bumps") which we will ignore
> for now.
>
> At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also
> CH14 and CH16. In fact over 100 days most of them are logarithmic but the
> coefficients vary considerably so it's hard to see this at a common scale.
> Note also the logarithmic curve is vastly more apparent in the first few
> days or weeks of operation, but I don't have that data.
>
> In general, any exponential or log or parabolic or circular curve looks
> linear if you're looking close enough. A straight highway may look linear
> but the equator is circular. So most OCXO drift (age) with a logarithmic
> curve and this is visible over long enough measurements. But for shorter
> time spans it will appear linear. Or, more likely, internal and external
> stability issues will dominate and this spoils any linear vs. log
> discussion.
>
> So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
>
> /tvb
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
BC
Bob Camp
Sun, Nov 13, 2016 3:53 AM
Hi
Exact info on mass transfer is a bit complicated. A 5 MHz 5th overtone is
a bit thicker and more massive than a 100 MHz 5th. Both are thicker (and
more massive) than a 100 MHz fundamental. On top of that the blank is not
equally sensitive to mass at all points on it’s surface. Finally, gold has a bit more
mass than hydrogen. A layer of one is not quite the same as a layer of the other.
All that said, The standard “gee wiz” number is that 1 ppb is an atomic layer
on a 5 MHz thrid. Given all of the hand waving, it’s a back calculated number
based on calibrating the crystal with a thin film of gold (under these conditions ….
on that design … calculated after XXX beers ...).
Bob
On Nov 12, 2016, at 8:56 PM, Scott Stobbe scott.j.stobbe@gmail.com wrote:
Those are wonderful plots :)
I vaguely recall that a 1ppm frequency shift is approximately equivalent to
the mass transfer of one molecular layer of a crystal. So at some point
your counting atoms if there was no noise, thermal disturbance, mechanical
disturbance...
On Sat, Nov 12, 2016 at 5:00 PM Tom Van Baak tvb@leapsecond.com wrote:
There were postings recently about OCXO ageing, or drift rates.
I've been testing a batch of TBolts for a couple of months and it provides
an interesting set of data from which to make visual answers to recent
questions. Here are three plots.
- attached plot: TBolt-10day-fit0-e09.gif (
http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
A bunch of oscillators are measured with a 20-channel system. Each
frequency plot is a free-running TBolt (no GPS, no disciplining). The
X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What
you see at this scale is that all the OCXO are quite stable. Also, some of
them show drift.
For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days
for a drift rate of 2e-10/day. It looks large in this plot but its well
under the typical spec, such as 5e-10/day for a 10811A. We see a variety of
drift rates, including some that appear to be zero: flat line. At this
scale, CH13, for example, seems to have no drift.
But the drift, when present, appears quite linear. So there are two things
to do. Zoom in and zoom out.
- attached plot: TBolt-10day-fit0-e10.gif (
http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale
is still 10 days. Now we can see the drift much better. Also at this level
we can see instability of each OCXO (or the lab environment). At this
scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in
CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is
25x better than the 10811A spec. CH13, mentioned above, is not zero drift
after all, but its drift rate is even lower, close to 1e-11/day.
For some oscillators the wiggles in the data (frequency instability) are
large enough that the drift rate is not clearly measurable.
The 10-day plots suggests you would not want to try to measure drift rate
based on just one day of data.
The plots also suggest that drift rate is not a hard constant. Look at any
of the 20 10-day plots. Your eye will tell you that the daily drift rate
can change significantly from day to day to day.
The plots show that an OCXO doesn't necessarily follow strict rules. In a
sense they each have their own personality. So one needs to be very careful
about algorithms that assume any sort of constant or consistent behavior.
- attached plot: TBolt-100day-fit0-e08.gif (
http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
Here we look at 100 days of data instead of just 10 days. To fit, the
Y-scale is now 1e-8 per division. Once a month I created a temporary
thermal event in the lab (the little "speed bumps") which we will ignore
for now.
At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also
CH14 and CH16. In fact over 100 days most of them are logarithmic but the
coefficients vary considerably so it's hard to see this at a common scale.
Note also the logarithmic curve is vastly more apparent in the first few
days or weeks of operation, but I don't have that data.
In general, any exponential or log or parabolic or circular curve looks
linear if you're looking close enough. A straight highway may look linear
but the equator is circular. So most OCXO drift (age) with a logarithmic
curve and this is visible over long enough measurements. But for shorter
time spans it will appear linear. Or, more likely, internal and external
stability issues will dominate and this spoils any linear vs. log
discussion.
So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
/tvb
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hi
Exact info on mass transfer is a bit complicated. A 5 MHz 5th overtone is
a bit thicker and more massive than a 100 MHz 5th. Both are thicker (and
more massive) than a 100 MHz fundamental. On top of that the blank is not
equally sensitive to mass at all points on it’s surface. Finally, gold has a bit more
mass than hydrogen. A layer of one is not quite the same as a layer of the other.
All that said, The standard “gee wiz” number is that 1 ppb is an atomic layer
on a 5 MHz thrid. Given all of the hand waving, it’s a back calculated number
based on calibrating the crystal with a thin film of gold (under these conditions ….
on that design … calculated after XXX beers ...).
Bob
> On Nov 12, 2016, at 8:56 PM, Scott Stobbe <scott.j.stobbe@gmail.com> wrote:
>
> Those are wonderful plots :)
>
> I vaguely recall that a 1ppm frequency shift is approximately equivalent to
> the mass transfer of one molecular layer of a crystal. So at some point
> your counting atoms if there was no noise, thermal disturbance, mechanical
> disturbance...
>
> On Sat, Nov 12, 2016 at 5:00 PM Tom Van Baak <tvb@leapsecond.com> wrote:
>
>> There were postings recently about OCXO ageing, or drift rates.
>>
>> I've been testing a batch of TBolts for a couple of months and it provides
>> an interesting set of data from which to make visual answers to recent
>> questions. Here are three plots.
>>
>>
>> 1) attached plot: TBolt-10day-fit0-e09.gif (
>> http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e09.gif )
>>
>> A bunch of oscillators are measured with a 20-channel system. Each
>> frequency plot is a free-running TBolt (no GPS, no disciplining). The
>> X-scale is 10 days and the Y-scale is 1 ppb, or 1e-9 per Y-division. What
>> you see at this scale is that all the OCXO are quite stable. Also, some of
>> them show drift.
>>
>> For example, the OCXO frequency in channel 14 changes by 2e-9 in 10 days
>> for a drift rate of 2e-10/day. It looks large in this plot but its well
>> under the typical spec, such as 5e-10/day for a 10811A. We see a variety of
>> drift rates, including some that appear to be zero: flat line. At this
>> scale, CH13, for example, seems to have no drift.
>>
>> But the drift, when present, appears quite linear. So there are two things
>> to do. Zoom in and zoom out.
>>
>>
>> 2) attached plot: TBolt-10day-fit0-e10.gif (
>> http://leapsecond.com/pages/tbolt/TBolt-10day-fit0-e10.gif )
>>
>> Here we zoom in by changing the Y-scale to 1e-10 per division. The X-scale
>> is still 10 days. Now we can see the drift much better. Also at this level
>> we can see instability of each OCXO (or the lab environment). At this
>> scale, channels CH10 and CH14 are "off the chart". An OCXO like the one in
>> CH01 climbs by 2e-10 over 10 days for a drift rate of 2e-11/day. This is
>> 25x better than the 10811A spec. CH13, mentioned above, is not zero drift
>> after all, but its drift rate is even lower, close to 1e-11/day.
>>
>> For some oscillators the wiggles in the data (frequency instability) are
>> large enough that the drift rate is not clearly measurable.
>>
>> The 10-day plots suggests you would not want to try to measure drift rate
>> based on just one day of data.
>>
>> The plots also suggest that drift rate is not a hard constant. Look at any
>> of the 20 10-day plots. Your eye will tell you that the daily drift rate
>> can change significantly from day to day to day.
>>
>> The plots show that an OCXO doesn't necessarily follow strict rules. In a
>> sense they each have their own personality. So one needs to be very careful
>> about algorithms that assume any sort of constant or consistent behavior.
>>
>>
>> 3) attached plot: TBolt-100day-fit0-e08.gif (
>> http://leapsecond.com/pages/tbolt/TBolt-100day-fit0-e08.gif )
>>
>> Here we look at 100 days of data instead of just 10 days. To fit, the
>> Y-scale is now 1e-8 per division. Once a month I created a temporary
>> thermal event in the lab (the little "speed bumps") which we will ignore
>> for now.
>>
>> At this long-term scale, OCXO in CH09 has textbook logarithmic drift. Also
>> CH14 and CH16. In fact over 100 days most of them are logarithmic but the
>> coefficients vary considerably so it's hard to see this at a common scale.
>> Note also the logarithmic curve is vastly more apparent in the first few
>> days or weeks of operation, but I don't have that data.
>>
>> In general, any exponential or log or parabolic or circular curve looks
>> linear if you're looking close enough. A straight highway may look linear
>> but the equator is circular. So most OCXO drift (age) with a logarithmic
>> curve and this is visible over long enough measurements. But for shorter
>> time spans it will appear linear. Or, more likely, internal and external
>> stability issues will dominate and this spoils any linear vs. log
>> discussion.
>>
>> So is it linear or log? The answer is it depends. Now I sound like Bob ;-)
>>
>> /tvb
>> _______________________________________________
>> time-nuts mailing list -- time-nuts@febo.com
>> To unsubscribe, go to
>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
