HP 106B quartz frequency standard...the story so far
Hi All,
Well I'm doing stuff and have had an interesting journey so far and thought
it was time to share it with people who get it. My girlfriend just doesn't
understand...
First I want to give thanks to Stan Searing and Steve Roberts who emailed a
scanned manual and to tvb for the offer to send one in the post.
Here is the journey so far...
I asked a question of this group over two years ago what would be a good
thing to measure ABSOLUTE time at high precision. After all, with all the
gear I had there had to be something. No one came up with a good answer.
After wandering back to my old University and talking to the Professor of
mathematics and physics there, his answer has obvious...you need to study
pulsars.
So I'm doing a masters in astrophysics studying pulsars and I'm measuring
the arrival times of the pulses of the Vela pulsar to a resolution of 10
micro seconds using a hydrogen maser. Pretty pissweak accuracy in our realm
I have to admit, but there is a lot more to take into account. The arrival
time is dependent on frequency and also on the "stuff" (i.e. molecules)
between Earth and the pulsar. This makes 10 micro seconds pretty decent.
But I digress...
The Uni chucked out a 106B and I grabbed it. Stan and Steve (both) provided
me with a manual and I've been working on this beast.
First I had to remove the leaking NiCd batteries. A quick check of the power
supply rails then showed plenty of non DC rubbish so the filter capacitor
was suspect. I replace the 1000 uF cap and things started working way
better.
The main power rail instead of being slightly under 18v was a bit over 21. I
think this is the Q3 "regulator" issue (the regulator being a transistor and
probably destroyed by the bad batteries and capacitor) - but this is not
going to be an easy one to replace - a 2N1701.
Anyway, I let the voltage rail sit and the oven heat up for a number of days
and adjusted the frequency with the course control (all the nice fine
control mechanics had been removed) as close to 5 MHz as I could.
Things seemed great until I noticed the frequency steadily increasing over
time. Not matter what I set the frequency it kept increasing.
So I started playing with the inner oven control. I noticed this could be
adjusted to modify the change in frequency from positive to negative. As I
write I'm trying to find the spot where the frequency is level.
As an aside, this 106B has an extra board in it to generate a 1 PPS signal
and can be set with a sync signal. (Which seems to have a 200ms delay -
which is unfortunate) - I thought this was a local Uni thing but talking to
Stan he seems to have the same modification.
Updates to follow...
OK, I'm puzzled. Can someone with a good knowledge of OCXOs explain my
observation. This is my HP 106B double ovened quartz oscillator, but
I'm sure the theory applies generally.
It's easiest to show these observations as made-up but approximate
numbers an hour apart. Say the device is set to 5.000 000 000 MHz and
then measurements are done on a 5370B with a GPSDO as an external
reference. (Note that the 5370B has more jitter than shown below when
measuring frequency, but by following it for a minute or so you can
see what numbers it hovers around.)
Connect an HP rubidium (to prove it's not a measurement error):
Hour 0 1 2
3 4 5
5.000 000 000 5.000 000 000 5.000 000 000 5.000 000 000 5.000
000 000 5.000 000 000
My 106B as it currently is running:
5.000 000 000 5.000 000 001 5.000 000 002 5.000 000 003 5.000
000 004 5.000 000 005
If I turn the inner oven control in one direction (presumably the
hotter way) a fraction of a turn and then sit back and watch:
5.000 000 000 5.000 000 003 5.000 000 006 5.000 000 009 5.000
000 012 5.000 000 015
If I turn the inner oven control the other way (a bit further past the
original point):
5.000 000 000 4.999 999 995 4.999 999 990 4.999 999 985 4.999
999 980 4.999 999 975
What I don't understand is why changing the oven temperature cause the
frequency to continually increase or decrease. If you look at Quartz
temperature curves (and I'm presuming this is an AC cut since SC
wasn't invented until 1976) they show a frequency offset dependent on
temperature. But I'm not getting an offset, I'm getting a steady
increase - all the time.
Sure I can find my sweet spot where it moves minimally (as it is now)
but that is hard to improve and seems to be sitting on a knife edge.
The only other thing to note is that my power supply rail is a few
volts too high (see previous post). Could this be the cause?
If what I'm seeing is normal I'd love to have it explained to me!
Regards,
Jim
Jim
I may be talking out of the wrong end of my digestive tract, but what
timescale were you using for your measurements?
If you are seeing a continual drift with observations over an hour or two,
I'd look to see what you have over 24 hours or more. Is it possible that
the drift you are seeing is caused by regular diurnal variation in the
temperature of the inner oven which the system cannot quite control enough.
OTOH I could be barking up completely the wrong tree.
Dave
Hi Jim,
Maybe 5 hours is insufficient time for the oven to stabilize at each new
temperature setting? If I remember correctly, plotting all the data (and
waiting for oscillator/oven to stabilize at each incremental temperature
setting took me >30 days to complete the whole plot? ...This was 25 years
ago, so my recollection of exact time required for the whole project might
be off a bit.
Regards,
Greg
Greg Burnett wrote:
Hi Jim,
Maybe 5 hours is insufficient time for the oven to stabilize at each new
temperature setting? If I remember correctly, plotting all the data (and
waiting for oscillator/oven to stabilize at each incremental temperature
setting took me >30 days to complete the whole plot? ...This was 25 years
ago, so my recollection of exact time required for the whole project might
be off a bit.
here.
In contrast, good quality OCXOs still have their long term specs
conditioned by a 30 day heat-up period. It is not the oven as such which
heat-up but rather the crystal "aging" mechanisms that got started again.
It's a good exercise to make a long log of the heat-up from power on to
a couple of weeks.
Cheers,
Magnus
Are you just seeing the transient as it reaches a new equilibrium of
temperature after which it would reduce to a background drift rate or is
this the observation after it has stabilized?
Joe
-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Jim Palfreyman
Sent: Thursday, August 06, 2009 12:38 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] HP 106B quartz frequency standard...the story so
far
OK, I'm puzzled. Can someone with a good knowledge of OCXOs explain my
observation. This is my HP 106B double ovened quartz oscillator, but I'm
sure the theory applies generally.
It's easiest to show these observations as made-up but approximate numbers
an hour apart. Say the device is set to 5.000 000 000 MHz and then
measurements are done on a 5370B with a GPSDO as an external reference.
(Note that the 5370B has more jitter than shown below when measuring
frequency, but by following it for a minute or so you can see what numbers
it hovers around.)
Connect an HP rubidium (to prove it's not a measurement error):
Hour 0 1 2
3 4 5
5.000 000 000 5.000 000 000 5.000 000 000 5.000 000 000 5.000 000
000 5.000 000 000
My 106B as it currently is running:
5.000 000 000 5.000 000 001 5.000 000 002 5.000 000 003 5.000 000
004 5.000 000 005
If I turn the inner oven control in one direction (presumably the hotter
way) a fraction of a turn and then sit back and watch:
5.000 000 000 5.000 000 003 5.000 000 006 5.000 000 009 5.000 000
012 5.000 000 015
If I turn the inner oven control the other way (a bit further past the
original point):
5.000 000 000 4.999 999 995 4.999 999 990 4.999 999 985 4.999
999 980 4.999 999 975
What I don't understand is why changing the oven temperature cause the
frequency to continually increase or decrease. If you look at Quartz
temperature curves (and I'm presuming this is an AC cut since SC wasn't
invented until 1976) they show a frequency offset dependent on temperature.
But I'm not getting an offset, I'm getting a steady increase - all the time.
Sure I can find my sweet spot where it moves minimally (as it is now) but
that is hard to improve and seems to be sitting on a knife edge.
The only other thing to note is that my power supply rail is a few volts too
high (see previous post). Could this be the cause?
If what I'm seeing is normal I'd love to have it explained to me!
Regards,
Jim
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.
J. L. Trantham wrote:
Are you just seeing the transient as it reaches a new equilibrium of
temperature after which it would reduce to a background drift rate or is
this the observation after it has stabilized?
I think doing the type of long-time log that I proposed may help in
getting to know the behaviour better. One of my counters has a nifty
real-time update distplay and I set it to 5 sec time-base and see how it
slowly updates. If it was even niftier it would scroll the display
continously as an option, but you can't get it all.
Oh, when I described means to estimate drift parameters out if the
frequency derivate, then again the need for zero dead time is there,
between the frequency or time samples (whichever is used). Picket fence
techniques should be used. The TADD-2 may come in handy. Oh, got to play
some more with mine.
Cheers,
Magnus
Jim
If your data is accurate and complete,
Looks like your OXCO's aging rate changes from
+2e-10 per Hr to +6e-10 per Hr to -1e-9 / Hr
when you change it's oven temperature a little.
Hard to believe your limited data and that the effect is not caused by something else, EXCEPT:
My New Tbold OXCO has been doing almost the exact same thing as you described over the last week.
Its aging rate will switch from
-5*e-10 per Hr, (-1e-8 per day) to
+2.5e-10 per Hr (+0.5e-9 per day).
and All I do is change its case temperature a few Deg.
And Do note these aging rates are 10 to 100 + times what they should be, and have no sign of getting better.
so even if they have not had time to fully stabilize as some suggest, THEY ARE still BAD.
In my case this is consistent and repeatable over minutes, hrs, and days, I have not tried weeks yet,
and there is a magic temp in the center that is under a deg wide that the aging rate seems to be about zero.
In my case the value of the two Aging rates does not seem to be a function of temperature or time.
The temperature change causes more like a switch between the two values with a very narrow zero aging rate range in the middle.
For the non believers a have attached one of many of the Lady Heater plots that I have recorded over the last week
The OXCO effect is independent of the Tbolt's mode, It can be in discipline mode, hold over mode or disable mode but of course the way the Tbold responses to the OXCO aging is dependant on which mode it is in.
The bright side is, If I can not get it replaced, at least it becomes a great Kalman testing tool.
Because of all the data I have taken and varied test, I now know a lot more about how the Kalman filter responses.
My Tbolt's Kalman filter has a first order correction factor for aging, Plus its temperature correction factor.
If I'm careful with what data I feed it, so as not to confuse it with the difference between temperature effects and time effects,
It can learn a new Aging rate in 4 hours, and still know it has a zero TC temperature.
If I am not careful, It confuses temperature effects and times effects and its responds is way off in holdover.
warren
----- Original Message -----
From: "Jim Palfreyman" jim77742@gmail.com
To: "Discussion of precise time and frequency measurement" time-nuts@febo.com
Sent: Wednesday, August 05, 2009 10:38 PM
Subject: Re: [time-nuts] HP 106B quartz frequency standard...the story so far
OK, I'm puzzled. Can someone with a good knowledge of OCXOs explain my
observation. This is my HP 106B double ovened quartz oscillator, but
I'm sure the theory applies generally.
It's easiest to show these observations as made-up but approximate
numbers an hour apart. Say the device is set to 5.000 000 000 MHz and
then measurements are done on a 5370B with a GPSDO as an external
reference. (Note that the 5370B has more jitter than shown below when
measuring frequency, but by following it for a minute or so you can
see what numbers it hovers around.)
Connect an HP rubidium (to prove it's not a measurement error):
Hour 0, 1, 2, 3, 4, 5
5.000 000 000
5.000 000 000
5.000 000 000
5.000 000 000
5.000 000 000
5.000 000 000
My 106B as it currently is running:
5.000 000 000
5.000 000 001
5.000 000 002
5.000 000 003
5.000 000 004
5.000 000 005
If I turn the inner oven control in one direction (presumably the
hotter way) a fraction of a turn and then sit back and watch:
5.000 000 000
5.000 000 003
5.000 000 006
5.000 000 009
5.000 000 012
5.000 000 015
If I turn the inner oven control the other way (a bit further past the
original point):
5.000 000 000
4.999 999 995
4.999 999 990
4.999 999 985
4.999 999 980
4.999 999 975
What I don't understand is why changing the oven temperature cause the
frequency to continually increase or decrease. If you look at Quartz
temperature curves (and I'm presuming this is an AC cut since SC
wasn't invented until 1976) they show a frequency offset dependent on
temperature. But I'm not getting an offset, I'm getting a steady
increase - all the time.
Sure I can find my sweet spot where it moves minimally (as it is now)
but that is hard to improve and seems to be sitting on a knife edge.
The only other thing to note is that my power supply rail is a few
volts too high (see previous post). Could this be the cause?
If what I'm seeing is normal I'd love to have it explained to me!
Regards,
Jim
Hi All,
Thanks very much for your interesting replies.
I'm beginning to think I'm not letting it settle between adjustments.
And besides, The thing has been powered off for at least a decade I
would suspect and maybe I'm just watching the crystal adjust to micro
changes in temperature as it warms up. It's been on for about 10 days
with only about 30 min off time in between. I shall leave it on the
bench for a month or two before changing anything again.
Now my next question is to those with double oven experience. The
service manual says the oven is "not field repairable". And there is
minimal description on how to get the oven settings right.
The container has 61.3 C on the side - obviously the point of
inflection for the temperature curve of the crystal. There are two
test points which connect to either side of a thermistor buried deep
inside the inner oven. The label next to it say 432 ohms. The
resistance it should be at the correct temperature. I'm currently
running at 422 ohms (hotter) - that's the temperature that causes
minimal drift in frequency (at the moment - maybe this will change
after further burning in).
The outer oven has no such thermistor. How do I know that is set
correctly? If it is too high, I'd assume it would heat the inner oven
too much and not allow you to drop the temperature in the inner oven.
If it is too low I'd guess the inner heater would have to use more
power and do more work to keep it ok. What is the ideal outer oven
temperature and how is this worked out?
In all this, keep in mind my power rails are running a few volts high
and I'm not sure the effect this is having in the oven.
Regards,
Jim Palfreyman
Hi
Regardless of weather it's a single or double oven, you want the
crystal to be at it's "turn temperature". If you have a BT cut
crystal, that's going to be the highest frequency you find as you move
the oven around. With either a double or single oven, you should be
able to "map out" a parabolic plot of frequency versus temperature.
That's only close to "right" since the electronics contribute
something to the temperature performance. To really get it right you
need to vary the external temperature and see what happens ....
On a double oven, the outer oven is set for some "nominal" amount of
inner oven current. The idea is to keep the inner oven controller from
either cutting off or saturating as the external temperature is
varied. Often the inner oven is set to the "low side" of it's power
range.
Bob
On Aug 6, 2009, at 6:24 PM, Jim Palfreyman wrote:
Hi All,
Thanks very much for your interesting replies.
I'm beginning to think I'm not letting it settle between adjustments.
And besides, The thing has been powered off for at least a decade I
would suspect and maybe I'm just watching the crystal adjust to micro
changes in temperature as it warms up. It's been on for about 10 days
with only about 30 min off time in between. I shall leave it on the
bench for a month or two before changing anything again.
Now my next question is to those with double oven experience. The
service manual says the oven is "not field repairable". And there is
minimal description on how to get the oven settings right.
The container has 61.3 C on the side - obviously the point of
inflection for the temperature curve of the crystal. There are two
test points which connect to either side of a thermistor buried deep
inside the inner oven. The label next to it say 432 ohms. The
resistance it should be at the correct temperature. I'm currently
running at 422 ohms (hotter) - that's the temperature that causes
minimal drift in frequency (at the moment - maybe this will change
after further burning in).
The outer oven has no such thermistor. How do I know that is set
correctly? If it is too high, I'd assume it would heat the inner oven
too much and not allow you to drop the temperature in the inner oven.
If it is too low I'd guess the inner heater would have to use more
power and do more work to keep it ok. What is the ideal outer oven
temperature and how is this worked out?
In all this, keep in mind my power rails are running a few volts high
and I'm not sure the effect this is having in the oven.
Regards,
Jim Palfreyman
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.