[time-nuts] Assistance needed to understand some V_OCXO stability concepts.

Bob kb8tq kb8tq at n1k.org
Sat Oct 17 16:48:14 UTC 2020 

Hi

Oven supply sensitivity and oven temperature sensitivity are two different (and 
independent) aspects of the oscillator. One magnitude of one parameter does not
predict the magnitude (or direction) of the other parameter.  Indeed changes in 
either voltage or temperature will have an impact, that’s about all you can be 
sure of.

When you design an OCXO, you have a number of components that are sensitive
to voltage and / or temperature. There is no way to put them all at a single point. 
The heat from the heater and heat from the outside will hit different parts in different 
ways. Since they are different distances from the temperature sensor(s) the control
point will have a different impact on each of them. 

The net result is that you balance the oven gain and set point for optimum temperature
performance. In a modern part temperature runs are done to optimize both parameters. 
The oven is then set to those “correct” values. Any change in set point or gain is
likely to degrade the performance. Gain is impacted by changing the insulation on
the oven.

If you wish to do an enclosure, you need some “room” around the part for air circulation. 
The more circulation, the closer you can get to the rated upper limit on the part. In a 
closed box with little moving air, a 20C to 30C “offset” may be needed ( = a 50 C OCXO 
might only work to 50 - 30 = 20C. ). With a fan (even a small one) you could get much
closer to the rated upper limit. Fans add their own problems so that may be a win / loose
sort of proposition. 

=====


Best approach to care and feeding of an OCXO:

Put it on an independent linear regulator. If oscillator and oven have separate inputs,
put each on its own regulator. 

Generate the EFC from a voltage reference. Do not use the “other” supply voltages 
to derive the EFC. EFC is much more sensitive than the other pins. 

Leave it on power all the time.

Keep it loaded with the correct load. 

Buffer the output before it goes anywhere.

Keep drafts away.

Keep the temperature as stable as you can.

Older parts will (likely) be sensitive to pressure and humidity. There isn’t a lot you
can practically do about this. Going to a newer part that is better sealed / designed 
is generally the best alternative. (Yes, there are exceptions to this…..)

=====

The longer slope in your plot is (likely) aging. It is not uncommon for older parts
to take a *long* time to stabilize. Stabilization times dimensioned in months are
not uncommon. I have seen units that have taken > 9 months to settle down. 

Bob



> On Oct 17, 2020, at 4:54 AM, Joe & Gisela Noci <jgnoci at gmail.com> wrote:
> 
> Hi to all,
> Still persevering with my GPSDO - this stuff is very addictive...
> I may be able to have a loan Rubidium reference for a few weeks which will
> assist me with comparative measurements - I hope!
> 
> In the meantime, the effects of ambient temp changes are very evident in
> the change of oven control voltage, and therefore the EFC voltage as well -
> Exactly what the effect is on frequency remains to be seen once I have the
> above reference.
> However, I would in any case like to try improve on this situation. The OCXO
> ( HP_00105-6013) is covered in a 25mm thick polystyrene 'box' inside the
> enclosure of the entire unit. The unit external shell is 2mm aluminium, and
> so does radiate internal heat, etc.
> I would like to try a sort of external oven arrangement -  remove the OCXO
> from the system unit enclosure, and fit it into a dedicated enclosure that
> forms
> a hopefully better temp shield around the OCXO, and make this housing
> internally temp controlled.
> The temp control would not have to be very accurate, maybe a few (2?)
> degrees C, all in an effort to improve on the ambient temp effects I am
> seeing.
> My lab ambient changes from 15 degC to 26degC over a winter 25hour period
> typically, and does reach 28degC in summer..
> 
> Has anyone reading this had any experience in designing something like
> this? It is not so simple as there not a great temp differential to play
> with.
> I guess the second 'oven' would need not only a heating capability, but
> cooling as well - to try stabilize at maybe 30 degC internally?
> 
> I have dug in the archives, and a lot of googling took place, but this sort
> of info is scarce!
> 
> An annotated screen shot shows the daily temp variation effects on the oven
> control voltage, and on the DAC control voltage.
> Scale is in 10's of seconds horizontally. Oven volts are actually the volts
> across the control transistor of the HP oven , so high volts = less heating
> DAC volts are in millivolts
> The rising trend in oven volts ( reduced heating) over the 48something
> hour period shows the gentle increase in daily ambient temps as we are
> warming up here.
> The DAC control voltage is synced to the oven control voltage - as the oven
> voltage increases so does the DAC voltage.  I have not implemented a means
> of actually measuring the oven temp, but that may be needed - to be sure
> that the temp is in fact being controlled properly - as I have no idea of
> the actual temp change for that oven voltage change. I also hope that there
> are no supply power return path issues -
> that the DAC voltage is perhaps affected by the change in oven current and
> so the system tries to compensate...
> 
> Also, a disturbing trend in the DAC voltage - the voltage is going more
> negative all the time, even though the oven volts are in a positive trend -
>   ie, the oven is 'cooling' more with increasing ambient, which should
> make the DAC volts trend increase at the same time, but it is trending more
> negative - I wonder if this is not an issue of crystal aging...
> 
> The attached screenshot is around 48 hours long, but a 2 week long capture
> at 10second rate has the DAC voltage decreasing from 5250mV to the present
> 4900mV
> Need to measure oven temp!
> 
> MANY things to dig into!
> 
> Regards
> Joe
> 
> 
> 
> On Fri, Oct 2, 2020 at 5:35 PM Bob kb8tq <kb8tq at n1k.org> wrote:
> 
>> Hi
>> 
>> On an oscillator that is 50 years old, there likely has been some aging of
>> the crystal
>> and other components. As you tune the crystal back onto frequency, you can
>> easily
>> impact the EFC sensitivity. On *any* older device, “measure” is very much
>> the right
>> way to do it.
>> 
>> Bob
>> 
>>> On Oct 2, 2020, at 7:44 AM, Joe & Gisela Noci <jgnoci at gmail.com> wrote:
>>> 
>>> Hello Tom.
>>> Thanks for your info.
>>> 
>>> My oscillator comes from an HP-5061a, which I obtained from a company as
>>> the tube was dead.
>>> The log book indicates a start date of 1971, so yes, not a youngster!
>>> I have the full set of original manuals for this 5061A , but the
>> oscillator
>>> specs are not detailed to the KV level, nor is the actual voltage level
>>> range of EFC indicated in the manual.
>>> The full circuits are there, etc.
>>> In the 5061A , the EFC was a +ve voltage from an accurate reference, via
>> a
>>> front panel multiturn pot, onto the +EFC input, and then the loop control
>>> voltage from a op-amp/hybrid, which is fed from a split 30volt supply.
>>> So it was not possible to discover what range the control voltage out of
>>> the op amp covered ( not in the circuit descriptions) . I finally
>>> discovered in the calibration descriptions that the front panel pot had
>> to
>>> be at midrange ( -7.5v) and the coarse
>>> control trimcap in the OCXO adjusted to set the control voltage (+EFC
>>> voltage) to +5V....
>>> 
>>> Thanks again!
>>> 
>>> I will persevere!
>>> 
>>> Kind Regards
>>> Joe
>>> 
>>> 
>>> 
>>> On Fri, Oct 2, 2020 at 12:41 PM Tom Van Baak <tvb at leapsecond.com> wrote:
>>> 
>>>> Joe,
>>>> 
>>>> The 00105 oscillator was used in early hp 105 quartz frequency
>>>> standards. So the specifications, theory of operation, photos and
>>>> schematics for the 00105 "brick" oscillator are all in the hp 105 op/svc
>>>> manual (April 1968).
>>>> 
>>>> There are two copies on Didier's site. Go to
>>>> http://www.ko4bb.com/manuals/ and type 105ab into the search box.
>>>> 
>>>> The EFC+ and EFC- inputs are a little different from a modern OCXO. The
>>>> hp105 provides a rear EFC input that works from -5 V to +5 V and changes
>>>> df/f by "greater than 4e-8" over that range. [1] So that would be 4e9 /
>>>> V which at 5 MHz is 0.02 Hz / V. Check the schematics for how they
>>>> implement it; the manual provides all the details. It may be an example
>>>> for your GPSDO. Your 0.033 number sounds ok to me.
>>>> 
>>>> In general, as you have done, it's always best to measure rather than
>>>> read the manual, especially since the manual uses the word "greater
>>>> than", and also since some of these 00105 oscillators are approaching 50
>>>> years old. If you have a frequency counter it should be simple to
>>>> measure the frequency during a voltage sweep, make a plot, and determine
>>>> the range, the slope, and also the linearity. The nice thing about EFC
>>>> experiments is that you don't even have to wait for the oscillator to
>>>> fully stabilize.
>>>> 
>>>> /tvb
>>>> 
>>>> [1] EFC paragraph from page 2-4 of hp 105 manual (02479-1)
>>>> 
>>>> 
>>>> On 10/1/2020 11:53 PM, Joe & Gisela Noci wrote:
>>>>> Jim,
>>>>> some further info -
>>>>> The OCXO I use -  HP_00105-6013 - does not have a lot of info available
>>>> - I
>>>>> have searched all over! I have found no definitive info on its KV, etc,
>>>> so
>>>>> had to determine it myself.
>>>>> 
>>>>> This I did as follows ( the complexity of my process was to try
>> 'really'
>>>>> know if it is correct, and if the OCXO behaves according to the correct
>>>>> Physics!)
>>>>> 
>>>>> The two jpegs,
>>>>> *KV determination @5V and 6V.jpg* and *KV determination @9V and
>> 10V.jpg*
>>>>> show the phase detector wrap around during HOLD mode, with the DAC
>> output
>>>>> set to specific voltages.
>>>>> 
>>>>> For these tests the OCXO was left running for 24hours to warm up and
>>>>> stabilize as best it could.
>>>>> 
>>>>> The 9V_10V and 5V_6V tests differ in the following:
>>>>> 
>>>>> *     For the 9V_10V test:*
>>>>> 
>>>>> The -EFC Voltage was set to -10V
>>>>> The +EFC DAC voltage was first set to 8V and the phase detector
>> monitored
>>>>> on a 'scope. The coarse adjust trim cap in the OCXO was then adjusted
>> so
>>>>> that the phase detector voltage was close to 1volt and 'not' drifting.
>>>> This
>>>>> was monitored and adjusted over a 3 hour period to verify reasonable
>>>>> stability.
>>>>> 
>>>>> Then the detector voltage was recorded with DAC set to 9V.
>>>>> Then the DAC voltage was set to +10V and the detector voltage recorded
>>>>> again.
>>>>> The phase detector wrap around time were 620sec @ 9V and 300sec @ 10V
>>>>> 
>>>>> I believe this equates to -
>>>>> @9V t=620sec so 10/620 = 0.0162Hz
>>>>> @10V t=300sec so 10/300 = 0.0333Hz
>>>>>   *So KV = 0.0333-0.01612 = 0.017 Hz/V*
>>>>> 
>>>>> *        For the 5V_6V Test:*
>>>>> 
>>>>> The -EFC voltage was set to -7V.
>>>>> With the DAC voltage @+4V first, the OCXO coarse adjust was set so the
>>>>> phase detector voltage was stable.
>>>>> Then the DAC was set to +5V, which gave a wrap-around time of t= 550sec
>>>>> Then with the DAC V= +6V, t=200sec.
>>>>> *This gives a KV of 0.032 Hz/V*
>>>>> 
>>>>> This all makes sense since the varicap is operating on a different part
>>>>> capacitance curve slope - at the lower voltages the capacitance is
>>>> greater,
>>>>> so the delta_V effect will be greater.
>>>>> 
>>>>> This verifies that the KV values are valid and sensible.
>>>>> 
>>>>> I made use of the 0.033Hz/V settings, as that is what my loop gains are
>>>>> calculated for.
>>>>> 
>>>> 
>>>> _______________________________________________
>>>> time-nuts mailing list -- time-nuts at lists.febo.com
>>>> To unsubscribe, go to
>>>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>>>> and follow the instructions there.
>>>> 
>>> _______________________________________________
>>> time-nuts mailing list -- time-nuts at lists.febo.com
>>> To unsubscribe, go to
>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>>> and follow the instructions there.
>> 
>> 
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at lists.febo.com
>> To unsubscribe, go to
>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>> and follow the instructions there.
>> 
> <OCXO Oven trend.jpg>_______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.

More information about the Time-nuts_lists.febo.com mailing list