How to open an OCXO?

[rob77]

btw.. wouldn't chipquick or regular bismuth solder be the solution ? i think it would significantly lower the melting point of the original solder.

[rf-loop]

Thanks for that rf-loop and to everyone else - all good suggestions, but I just opened it. I used a more powerful heat gun and I had it attached with some screws to a piece of wire so I could wiggle it.

The bad news is that the wiggling resulted in 2 pins falling off + some internal components. By the look of it, some pads are broken as well - pretty weird.

Edit: the pads didn't came off, it's just bare copper. At least now I know where those components came from

This what happend to you  is just why I will use small size soldering gas torch IF better equipments are not available.
With torch and experience this what happend now to you, do not never happend if handle this kind of OCXO's.
This is because heating is enough fast and internal temperature do not rise to any damage level.
It need experience to adjust just enough heating to right place  and object need be fixed so that it can do enough fast so that do not need "burn" it. If outer case do not drop itselef it need know what is just right time to pull it. Also it need experienced eyes to look metal can surface to detect what is melting status. (melting point can see with eyes around of box surface.)
In some cases some area may also need cooling or isolate so that do not heat too much. (example bottom of can so that I/O pins do not damage depending how these are assembled (soldered to can).


Enough heating power to narrow solder area corner around case and fast heating (gas torch with very narrow fire)
Also it is best to heat just near soldered joint so that heat just this outer can and laterally so that bottom plate do not get high heating power.

If heat with big heat air gun using long time for heating is is like designed drive to damaage.. It need so much time and heat transfer go all around can and heat also all internals. All internals reach very easy damage temperature if use too long time.

Time a go I open expensive 5x5x5 cm OCXO  This model was hermetically sealed without any adjustment screw so that can not see inside at all for get any information about construction.

Legs up and bottom plate (now upside) fixed  so that I can cool center area of bottom plate (where is legs) using wet "ball" of boumwool. How I cool top of can (now downside) - of course using water, OCXO was legs up and top case down in water. My other hand was also there inside water so that I can pull down this can just after solder joint smelt. ~15 seconds small gas solder torch and it was open without any kind of damage. And this guite high level DOCXO get later his new life and better than before surgery.

[Fraser]

@RF Loop

Excellent comments on the right way to hat the can. As you say, practice is needed to perfect the process. Thanks for the very helpful information.

Aurora

[G0HZU]

I think you have to choose the best strategy for each type of package and also take into account the cost/availability of a replacement.

If I had to take apart the OPs (one off) OCXO I would have used a scalpel to cut away the solder from near a corner. Then I would have pried the case away a tiny bit (causing distortion of the case) and I would have worked my way around the whole OCXO like this.

The lid would end up looking a bit ripply and distorted but it would have used the least heat. i.e. I would have used a method based on the can opener approach that was suggested first. But instead of using a cutter I would just scrape the solder away and pry/peel the lid away breaking/peeling the solder seal as I went along.

If the device was really valuable I would have asked one of the techs at work to remove it as they do stuff like this much more often than me and have better cutting tools etc. But I can usually manage OK

Some of the bigger OCXOs can be taken apart fairly easily as they are fastened with screws. eg the huge OCXO below is from one of my old spectrum analysers

[jadew]

Sorry for the late reply, been busy with work.

@edpalmer42, I actually needed it, it was the OCXO from my frequency counter.

I managed to fix it, but not before playing puzzle with the pieces that fell off.

After I put it back together, it didn't work anymore, which is when I realized that I lost a few more parts (about 10 components fell off in total). This made me think that this was the original issue and that the parts didn't fell off because of something I did. Most components were barely hanging in there and there was very little (if any) solder on the pads.

Luckily, I had the picture I posted here and it helped me replace the missing components, after which I desoldered most of the other components (including the two chips), cleaned the pads and soldered them back together.

When I was playing puzzle the first time, I accidentally introduced a short, which I tracked to the heating element (a JE 800 darlington). I wanted to remove it, but it was glued, so I had to destroy it to see if that was causing the short - it wasn't.

Anyway, I cleaned up the spot where it was glued on, added some thermal paste and replaced it with a similarly spec'd darlington.

I put it back together just to realize that it no longer reaches 10 MHz, regardless of the voltage applied to the adjustment pin and no matter how hot it got (I experimented by adjusting the opamp divider network that handles the heating).

Eventually, I ended up changing some resistors in a resistor network that, together with the voltage applied to the adjustment pin, was setting the frequency. As a side effect, the adjustment range is too wide so the auto-calibration of my frequency counter is a bit off (by 0.08 Hz), due to the resolution of the DAC used to apply the adjustment voltage.


Thank you all for the support and suggestions, next time I'll open up an OCXO I'll know how to do it properly

Oh, and I didn't sealed it this time. I only closed it and applied solder on two corners - this should make it easier to open if it starts acting up again.


Edit:
@G0HZU, after I heated it up and wiggled it, I used two screwdrivers to pry it open just like edpalmer42 suggested.

[G0HZU]

Edit:
@G0HZU, after I heated it up and wiggled it, I used two screwdrivers to pry it open just like edpalmer42 suggested.

Ah yes, I went back and saw his reply in post #11. That's how I would do it too. Usually, once you get a gap you can prise/crack the solder away at the expense of making the lid a bit ripply.

The reason I took apart the big Toyocom OCXO in my previous post was so I could alter the oven (equilibrium) temperature down a couple of degrees. That old OCXO only adjusts +/- 1Hz and it had aged off the end of the scale as it was nearly 30 yrs old.
Either the xtal had aged by 1Hz or the oven temp had 'aged'. It starts off +180Hz high when cold (this is normal) and the oven is supposed to reach equilibrium within the +/- 1Hz window range of the adjuster when the oven stabilises!

I found by dropping it just a few degrees I could set it back on frequency on the middle of the adjuster. So it looks like we had similar problems

[edpalmer42]

jadew and G0HZU:

You should never change the temperature of an OCXO in an attempt to change the frequency.

Every OCXO crystal has a unique temperature where it performs best.  Even two units of the same make, model, and design can have slightly different temperatures.  If you move away from that temperature, the frequency certainly changes, but the oscillator becomes more sensitive to ambient temperature changes.  Whether that's significant or not depends on how good your equipment is and how close you need your measurements to be.

All quartz oscillators change frequency with age.  Once they drift out of adjustment range, you have to do some reverse engineering to figure out how to modify the circuit to bring it back within range.  Usually that means increasing or decreasing the capacitance at a strategic point in the heart of the circuit.

G0HZU, in your case, since the frequency starts high and drops as the unit warms up, you want to adjust the temperature so that the frequency is as low as possible.  That's the correct temperature for your crystal.  Then you modify the circuit to bring the oscillator back on frequency.  The fact that the frequency starts high tells me that your oscillator probably has an AT type crystal in it.

FYI, there's another type of crystal called an SC crystal that's used in higher quality oscillators.  It starts slightly below the desired frequency and rises as it warms up.  For these crystals, you want to  adjust the oven for the highest frequency.

These two types of crystals are almost the only ones that you'd ever see in a modern oscillator.  If you're into vintage equipment, you may run across other types that are no longer used.

Ed

[jadew]

@Ed,

I only tried to change the temperature because I thought I messed it up when I replaced some of the components, but I guess I only messed up the frequency

The oscillator might actually be missing some parts, because the pads had clean copper on them, but I could figure out what would go there and the number of components that fell of initially was smaller than the number of free spots anyway. There were also some unpopulated footprints, but those were by design, so I figured that could be the case with the ones I couldn't match too.

Anyway, I don't mind the 0.08 Hz offset since I have better external references if I really need the accuracy.

[edpalmer42]

jadew,

Sounds good.  It's not unusual to see unpopulated pads due to things like unused options - different levels, dividers or multipliers to change the frequency, etc.  Another thing you often see in oscillators is sets of parallel pads.  Install a big component and then tune the value by adding components in parallel until the frequency, temperature, or level is correct.  It's hard to tell, but I think I can see a few of those in your picture.

Glad you got everything back together and running.  Well done!

Ed

[rf-loop]

G0HZU, in your case, since the frequency starts high and drops as the unit warms up, you want to adjust the temperature so that the frequency is as low as possible.  That's the correct temperature for your crystal.  Then you modify the circuit to bring the oscillator back on frequency.  The fact that the frequency starts high tells me that your oscillator probably has an AT type crystal in it.

FYI, there's another type of crystal called an SC crystal that's used in higher quality oscillators.  It starts slightly below the desired frequency and rises as it warms up.  For these crystals, you want to  adjust the oven for the highest frequency.

These two types of crystals are almost the only ones that you'd ever see in a modern oscillator.  If you're into vintage equipment, you may run across other types that are no longer used.

Yes, this can understand easy with reading this very simplified description from Isotemp.
http://www.isotemp.com/wp-content/uploads/2011/06/Understanding_Ovenized_Oscillators.pdf

Specially with some SC cut XTAL's finding turn point may be difficult. In this pdf there is only one simplified curves temp vs freq. There may be even this kind of SC cut (depending more deep data about cutting angles etc)  where can not at all easy find turn point. Also temperature may be much higher up to over 120 celsius. (ref Hewlett-Packard)

Idea is to find point where small temperature change do not change nearly any amount of frequency.
Long time aged OCXO may have drift out from this right temp set point.  If want very good stability OCXO this temp set point need find carefully. Also every individual XTAL is different. Mass produced cheap OCXO's they only use average good set point for manufacture lot. Then one OCXO may have good and one may have low ambient temp stability. In most cases this can improve if do this long time needed work for finding right temperature set point (crap OCXO - who want use week work hours for crap OCXO fine adjust specially if temperature control is crap) So, in crap OCXO case it can just do inside some hour and changing slowly temp to around this right setpoint. Finding exact point need long time watching with good equipments and statistics. No one want do it with low grade OCXO.

Here can read one real example:
http://www.realhamradio.com/z3801a-turning-point.htm

[G0HZU]

jadew and G0HZU:

You should never change the temperature of an OCXO in an attempt to change the frequency.

Every OCXO crystal has a unique temperature where it performs best.  Even two units of the same make, model, and design can have slightly different temperatures.  If you move away from that temperature, the frequency certainly changes, but the oscillator becomes more sensitive to ambient temperature changes.  Whether that's significant or not depends on how good your equipment is and how close you need your measurements to be.

Hi Ed
I agree with what you are saying but you have to consider my (real world) position here... 

In my case I was looking at a big old (mega expensive and obsolete/unobtanium) Toyo OCXO that was nearly 30 years old. As I said in a previous post I don't know if the change in frequency is due to the oven(s) temperature ageing across 28years or the crystal (or both).

I've been using the analyser it was fitted to since 1990 (analyser was new in 1986) and over this time the OCXO freq has gradually changed until it went off the range of the adjuster pot a couple of years ago. It only has to age/shift about 1Hz to drift off the range of the adjuster. Despite this, I carried on using it but the search was on to try and find a healthy replacement in 'slow time'. I finally found one last year in another TR4172 analyser that was sold as faulty/spares.

So the replacement OCXO is now in my main analyser. However, I managed to repair the faulty TR4172 analyser so I had a go at tweaking my old (aged) OCXO. I either had the choice of altering the 'fine' resistor to set the oven point slightly lower or I would have had to take the OCXO apart even further because the main xtal is buried inside the inner heated metal core.

I decided that this was too high a risk and just altered the temperature of the oven slightly.  For all I know it was the oven temperature that had aged about 3 degrees over the 28 years.

The analyser it is fitted to was one of the most expensive analysers on the planet in the 1980s (much more expensive than a HP8568B)

The tweaked one showed really good stability over several hours of use in the repaired TR4172 (only drifted a few mHz) and this would be for quite a change in ambient temperature inside the spectrum analyser chassis as it sees quite a temperature gradient inside the analyser over the course of several hours.

[G0HZU]

Hope Jadew doesn't mind me posting this stuff here but last year I logged the warmup profile of the aged OCXO last Autumn before I meddled with it.
It's a 5MHz OCXO but the analyser doubles it to 10MHz. So the counter reads it as a 10MHz OCXO. I wound the adjuster pot way back past the mid point as it wasn't as stable with the adjuster point at the endstop. I guessed where the mid point was but it looks like I over wound it a bit

So in this test it is about 5Hz off frequency. At 10MHz it probably adjusts about +/-3Hz on the trimmer. Looking at the numbers in the text file below it drifts about 2.2Hz past the equilibrium point when warming up then rises back up 2.2Hz before reaching equilibrium. So the lowest point is actually at a hotter oven temperature.

I haven't logged it since the mod to the oven but I also logged the replacement (healthy) OCXO a few days later.
This one warms up and doesn't overshoot by 2.2Hz like the faulty one did. EDIT: Yes it does also overshoot by about 3Hz. I misread the data the first time I looked at it.

This is the other data file below.

Note: Ignore the second column of frequency data as it is an averaged number and isn't relevant to an OCXO in warmup. Just look at the first column of frequency data. The initial column of data is time in seconds.

I suspect that the first OCXO has aged both in the oven temperature (the other way!) and also the crystal has aged a bit?

[G0HZU]

It's the first time I've really looked at this data since I took it last year and you can see that the original OCXO isn't quite as stable as the replacement as it jumps around by the odd LSD.

I suppose I ought to fire up the other TR4172 with this original (tweaked) OCXO and log it with the revised oven temperature to see if the overshoot dip has now changed. The oven won't get as hot so may not see the true dip anymore? But the weather is far too hot here in the UK to do anything!

[edpalmer42]

So in this test it is about 5Hz off frequency. At 10MHz it probably adjusts about +/-3Hz on the trimmer. Looking at the numbers in the text file below it drifts about 2.2Hz past the equilibrium point when warming up then rises back up 2.2Hz before reaching equilibrium. So the lowest point is actually at a hotter oven temperature.

I haven't logged it since the mod to the oven but I also logged the replacement (healthy) OCXO a few days later.
This one warms up and doesn't overshoot by 2.2Hz like the faulty one did. EDIT: Yes it does also overshoot by about 3Hz. I misread the data the first time I looked at it.

This is the other data file below.

Note: Ignore the second column of frequency data as it is an averaged number and isn't relevant to an OCXO in warmup. Just look at the first column of frequency data. The initial column of data is time in seconds.

I suspect that the first OCXO has aged both in the oven temperature (the other way!) and also the crystal has aged a bit?

The overshoot is partially due to the design of the oven controller and partially due to the crystal's response to the temperature change.  I don't think that you can infer that the temperature is low.  All AT crystals respond like that - even if the temperature is adjusted correctly.

Your final point is a good one.  I've never found any info to say if a crystal's turning point (sometimes called inflection point - the 'correct' temperature) changes with age.  The components that measure and adjust the oven's temperature can certainly drift with age.  Likewise, crystals are guaranteed to drift in frequency with age.  Sometimes the frequency drifts up, sometimes down.  Sometimes the drift changes direction!  The only constant is change! 

Ed

[G0HZU]

The overshoot is partially due to the design of the oven controller and partially due to the crystal's response to the temperature change.  I don't think that you can infer that the temperature is low.  All AT crystals respond like that - even if the temperature is adjusted correctly.

Yes I think you are right. I have several logs of my various OCXO warmup characteristics here. I had a rummage through them and quite a few show the overshoot effect. I've never really studied the data closely before. I mainly keep the warmup profiles because I'm mostly interested in the relative warmup time of each one.

I have some little DIL14 OCXOs that warmup (to be usably close) in 30 seconds and others that are huge and take 45minutes to get within about 0.1Hz.

Here's an old image of the repaired TR4172 analyser with my fudged OCXO swapped over into it. I managed to set the frequency agreement to within about 1mHz (according to the counter that is) and I recall that it didn't drift very much wrt the counter. The counter has a very similar Toyo OCXO package inside and it has the same spec for stability and ageing.

So this fudged OCXO is hopefully going to be fine for this spare analyser. I really only bought the second TR4172 for spares (mainly the OCXO!)

The spares/repair TR4172 analyser cost me about £250 which is probably a fraction of the cost of a new OCXO. The bonus is that I managed to repair the analyser at zero cost so I have a fully working spare analyser for my much cherished TR4172

[edpalmer42]

Your TR4172 looks like a nice piece of equipment, although not exactly portable!  I love the way they phrase it in the specs - "less than 50 kg".  I don't see an external reference input connector.  Too bad.  You could get either a Rubidium standard or GPSDO to use as a house standard and have various counters, generators, and analyzers all syncronized together and guaranteed accurate to a fraction of a Hertz.

Some months ago I picked up an Advantest R3465 Spectrum Analyzer that was sold as 'not working - for parts only'.  It's still a work in progress, but it's now alive and somewhat usable.  They seem to build nice equipment, although I wish it was easier to find detailed service manuals.  I think I've been spoiled by the old HP manuals.

Ed