Oscilloquartz 8663-XS OCXO

[brian_mk]

I have one of these OCXOs built into a 10MHz distribution amp.
Its powered from 12V DC. The control voltage is derived from a LH0070 precision 10V reference.

The problem is that the control voltage has to be set to the maximum 10V to get an exact 10MHz output.
With lower control voltages, the frequency is below 10MHz.
The OCXO was bought used from China so it may have been removed from another piece of equipment because it was out of spec.

I am wondering what my options are.
Used devices are still quite expensive to buy in the UK (eBay, Aliexpress).

Has anyone managed to open up the soldered case of one of these OCXOs without damage?
Is it possible to adjust or change an internal tuning capacitor to increase the frequency?

The only other option I can think of is to increase the tuning voltage to something just above 10V.
(10V is the maximum control voltage according to the 8663 spec sheet).
Increasing the tuning voltage would not be straight forward because the LH0070 is a fixed 10V reference.
It would mean cuts and straps to the PCB to add a couple of resistors.

[S57UUU]

There is very probably a trimmer cap inside. Some OCXOs have a hole through which you can adjust it.

I have opened and closed a few OCXOs, but not exactly your type. Whether you can successfully unsolder it depends on the internal construction, how the PCBs are fixed, etc.
My method was to fix the thing by the pins and heat the "cap" witt a hot air gun (paint removal, not SMD rework), and pull on the cap, gently, not to pull the pins. Once the solder flows, it usually goes off easily.
One problem can be SMD parts falling off,  if on bottom side, etc.

[Gertjan]

Hi brian_mk,

I have quite a few of these excellent Oscilloquartz 8663-XS OCXO's. The control voltage on my 8663's  is always between 4,7V and 5,5V.
So it seems yours has drifted (aged) out of tuning range....

just to check is not something is broken: What is the output of your reference voltage output? It should be around 7,5V.

Another check: does the oven still work? Startup current of the OCXO should be around 530...580mA. After warming up (about 10 minutes) the current should fall back to about 200...250mA.
If the oven stays too cold, the oscillator will not reach it's intended frequency.....

Best of luck, Gertjan.

[brian_mk]

There is a blob of solder on the side of the can. See attached photo.
Could that be covering a hole that gives access to an internal trimmer cap?

[Gertjan]

None All of my 8663's features such a solder blob....

[brian_mk]

I started to wonder if it could give access to a trim pot that adjusts the oven temperature but if you don't have a solder blob on your examples then maybe it's just where an earth braid was connected.
I also wonder if the inside could be filled with something like dry Nitrogen to avoid humidity issues?

[Gertjan]

I do not think the inside is filled with something like dry Nitrogen.

But why go into such details instead of first doing the basic checks, as I outlined above?

Also, keep in mind that these OCXO's are old. New these OCXO's were very expensive, the e-bay price is just a few percent of the original cost.
So, while it is very fortunate we can have these very good OCXO's for an incredible low price, the other side of the coin is that they are near the end of their life-cycle. I had one broken too. Of coarse it was the best of the bunch...

[brian_mk]

I had not checked Vref out from the OCXO because it is not used - the control voltage is derived from a separate three terminal reference.

Update: I just measured the OCXO Vref = 7.7v (spec indicates 7.8V).

I had already checked things like the supply current at start up and after the oven reaches temp.
The outside case of the oven gets pretty hot. I have not measured it but I would guess somewhere around 60->65 degC.

Update: Case temp measures about 60 degC.

Frequency measurements made using a counter with the EXT Ref Input fed from a GPSDO indicate a slight overshoot in frequency / temp during warm up. Then it settles back down and remains pretty stable.

Incidentally, why is it that the supply current drops suddenly by about a factor of two when the oven reaches temp?
If it uses some sort of proportional control system (PWM?) then you would expect the average current to drop gradually as it approaches the set temperature.

[ejeffrey]

For an idealized control system the optimal behavior is essentially to have 100% power until you have added enough heat and then drop to the maintenance power.  That's hard to get right and keep your control stable in the steady state and robust against external transients so realistic systems don't follow this directly, but it's normal to have a sudden drop in a well tuned system.

[Gertjan]

If you are curious what is inside the Oscilloquartz 8663-XS OCXO, see this topic: Inside the UCT 108663-1 10 MHz Double Oven OCXO.
This is the same OCXO. (it is a rebrand). I have two of these too.

And a correction: I checked my OCXO's, and they ALL have the solder blob on the side..... I think I just did not notice it before 

regards, Gertjan.

[brian_mk]

I modified the pcb to allow the OCXO reference voltage to be adjusted a couple of hundred mV above 10v.
I then carried out some tests on frequency stability.

First I waited until the frequency had stabilised with the top lid fitted to the case.
I measured 10,000,000.39 Hz using a counter with EXT Ref connected to a  GPSDO.
Then I removed the case lid and waited until the frequency stabilised.
I measured 9,999,999.97 Hz - a drop of 0.42Hz.
In both cases I checked the supply voltage and OCXO tuning voltage. They remaind unchanged.

To make sure the problem was not the GPSDO, I also took frequency measurements using the counter's internal standard.
This is uses an ultra high stability oven.
The frequency differences matched that using the GPSDO.

I then replaced the case lid. (I should add that the case lid has many perforation slots to allow air flow above the OCXO.)
The frequency gradually increased again to 10,000,000.46 Hz.

Update: Now reached 10,000,000.96 Hz

The conclusion has to be that the OCXO is not controlling it's internal temperature sufficiently well to compensate for external variations.

Is it just my particular device or am I expecting too much from an OCXO?

The 8663-XS spec indicates 4 x 10-9 over -20degC to + 70degC.

[eblc1388]

Incidentally, why is it that the supply current drops suddenly by about a factor of two when the oven reaches temp?
If it uses some sort of proportional control system (PWM?) then you would expect the average current to drop gradually as it approaches the set temperature.

It's probably the interaction of several temperature control loops. For your reference I attach a startup curve I measured back in 2014.

Don't be fool by the seemingly stable output 450 seconds after power ON, comparison with a locked RB standard shows that output frequency is still changing after 14 minutes. I would say to allow at least 20 minutes or more warm up time before taking any measurement. 

[eblc1388]

The conclusion has to be that the OCXO is not controlling it's internal temperature sufficiently well to compensate for external variations.

Try to do a plot of output frequency vs EFC voltage. The result should be linear and centered around 4.5~5.5V for a normal 8663 DOCXO 10.000MHz output. At least you'll know if the EFC circuit part is functioning correctly if the result is linear, albeit offset by a large amount.

[brian_mk]

Thanks for the charts.
It looks like they relate to a double oven so I wouldn't expect it to give quite the same results as my single oven

I assume the vertical axis on the frequency vs time from start up is in Hz?
If so, the resolution is not really sufficient to show anything less than about 1 Hz.
My Racal counter gives a resolution down to 0.01Hz.

Today I left the setup running for about 4 hours with the case lid fitted.
What I found was that the frequency continued to slowly increase above 10MHz.
I eventually adjusted the the control voltage to bring the frequency back down to exactly 10MHz.
The control voltage then measured 7.45V. This is a bit more like I would have expected.

I am surprised how long it takes for the frequency / temperature to stabilise. Is this normal?

The photo shows the test set up. The DUT is in the center, the GPSDO on the right.
You can see the ventilation slots in the distribution amp case lid just above the OCXO.

[Wolfgang]

I have one of these OCXOs built into a 10MHz distribution amp.
Its powered from 12V DC. The control voltage is derived from a LH0070 precision 10V reference.

The problem is that the control voltage has to be set to the maximum 10V to get an exact 10MHz output.
With lower control voltages, the frequency is below 10MHz.
The OCXO was bought used from China so it may have been removed from another piece of equipment because it was out of spec.

I am wondering what my options are.
Used devices are still quite expensive to buy in the UK (eBay, Aliexpress).

Has anyone managed to open up the soldered case of one of these OCXOs without damage?
Is it possible to adjust or change an internal tuning capacitor to increase the frequency?

The only other option I can think of is to increase the tuning voltage to something just above 10V.
(10V is the maximum control voltage according to the 8663 spec sheet).
Increasing the tuning voltage would not be straight forward because the LH0070 is a fixed 10V reference.
It would mean cuts and straps to the PCB to add a couple of resistors.

Some hints to this:
- Your DOCXO uses an SC-Cut crystal, and those age "upwards". See datasheet.
- If you cannot reach 10MHz you probably have a thermostat problem.
  SC OCXOs work on the rising edge of the frequency/temp curve, so if temp is too low frequency will be too.
- If the thermostat uses NTCs keep in mind that they age as well ...

[brian_mk]

Thanks for the input.

Having left the OCXO on for about 8 hours, the control voltage required to give 10MHz now measures 6.93V.
Hence my initial thoughts about it needing around 10V to give 10MHz were probably incorrect.
The small change I made to increase the 10V voltage reference by a couple of hundred mV was unnecessary.
As you suggest, I am beginning to think what I am seeing now is mainly a crystal aging process rather than being caused by temperature changes.
Right now, the frequency appears to be slowly falling rather than rising - although that may be because I removed the case cover for a few seconds while I measured the control voltage.
My intention is to leave the test running overnight and check in the morning.
I am also thinking that I should add a couple of sockets to the front or rear panel so I can monitor the control voltage without removing the cover.
A panel mounted 10 turn pot would also help.

There is a lot more to this than I ever imagined.
As a retired engineer I am finding it a fascinating subject.
I found an interesting PDF that delves into it in detail:-

https://www.scribd.com/document/289237197/quartz-crystal-resonators-and-oscillators-for-frequency-control-and-timing-applications-pdf

[Wolfgang]

Thanks for the input.

Having left the OCXO on for about 8 hours, the control voltage required to give 10MHz now measures 6.93V.
Hence my initial thoughts about it needing around 10V to give 10MHz were probably incorrect.
The small change I made to increase the 10V voltage reference by a couple of hundred mV was unnecessary.
As you suggest, I am beginning to think what I am seeing now is mainly a crystal aging process rather than being caused by temperature changes.
Right now, the frequency appears to be slowly falling rather than rising - although that may be because I removed the case cover for a few seconds while I measured the control voltage.
My intention is to leave the test running overnight and check in the morning.
I am also thinking that I should add a couple of sockets to the front or rear panel so I can monitor the control voltage without removing the cover.
A panel mounted 10 turn pot would also help.

There is a lot more to this than I ever imagined.
As a retired engineer I am finding it a fascinating subject.
I found an interesting PDF that delves into it in detail:-

https://www.scribd.com/document/289237197/quartz-crystal-resonators-and-oscillators-for-frequency-control-and-timing-applications-pdf

If the OXCO can be brought into range at 6-7V that means the crystal is fine.
Running OXCOs open is not a good idea because this may overload the heaters that are not built for massive heat losses due to an open case.
Some days to burn in and settle is OK, a good power supply is also essential.

[brian_mk]

I decided to carry out further investigation and plotted a graph of frequency error and OCXO case temperature against time.
See attached image.
The measurements were obtained with the instrument lid removed.
The OCXO case temperature was measured using a digital thermometer with heat sink compound between the probe and OCXO case.
I would guess it is only accurate to around +/- 1 degC.
After about 1 hour, both frequency and temperature appeared to stablise with the OCXO case temperature reading a steady 60 degC.

I then adjusted the frequency to exactly 10MHz and fitted the instrument lid.
When I checked again a few hours later I found the OXCO case temperature had risen to 65degC and the frequency had increased by 0.85Hz.
My conclusion is that there appears to be a problem with the OCXO internal temperature regulation for this particular device.
The specification indicates it should be much better than this.
I am not sure what (if anything) I can do about it.
If I replace the OCXO with another used device from China (cost of around 40GBP), it may be no better.

[Wolfgang]

If its an SC cut than 60°C is way too low. Below 80 and 90 is normal. So maybe you do have a heater problem.

[brian_mk]

If its an SC cut than 60°C is way too low. Below 80 and 90 is normal. So maybe you do have a heater problem.

From what I have read, a typical OCXO maintains an internal temperature of 75degC to wiithin +/- 1degC.
That makes sense because the 8663-XS operating range is specified from -20degC to +70degC.
The crystal temperature needs to be maintained above the maxiumum external temperature to allow it to be controlled.

I have read conflicting information about adding external thermal insulation around the OXCO can.
The link below suggests adding a foam cover to reduce air flow:-
https://www.microchip.com/content/dam/mchp/documents/VOP/ApplicationNotes/ApplicationNotes/Optimizing%20the%20OX-601%20and%20OX-502%20Performance.pdf

To try this, I added a 6mm thick foam jacket around the can. Initially I left the top of the jacket open.
After waiting a few hours for the frequency to stabilise (with the instrument lid in place), I measured the OCXO temp as 62.5degC.
The tuning voltage required to give exactly 10MHz had dropped to 5V.
That doesn't make much sense to me.
If the OXCO is regulating it's internal temperature correctly, then the tuning voltage required for 10MHz should remain about the same.
I would also expect it to take less time to reach a stable temperature and frequency.

Update:
I tried adding another piece of foam to act as a top cover to the thermal jacket.
After leaving for a couple of hours, the OXCO can temp was 71degC.
At this temperature, I was unable to adjust the frequency down to anything close to 10MHz even with the tuning control voltage reduced to 0V.
I am baffled.

[PCB.Wiz]

You could try logging the current going into the OCXO as well, so you can see the control action.
That should show you where the internal thermal feedback regulator has control, and where it is outside that band.