Making TCXO or modding existing one for low -60C temp operation

[Leo Bodnar]

I have been trying to find commonly available TCXO that will work reasonably well below -40C.
What I have found through testing is that they are typically well corrected for their intended operational range to -40C and then literally fall off the cliff beyond that. 

0.5ppm TCXO usually stays well within that range until -40C and then from -45C to -55C it would be 5-10ppm off with rapidly increasing offset.

Has anybody tried creating a table or formula-based TCXO using an accurate temperature sensor bonded to XO?  I'd be happy with 1ppm vs temperature down to -60C.

Alternatively has anyone ever seen a way to modify stock (non-analogue) TCXO correction?  I assume they are either individually or batch-calibrated during production but I have never seen even a hint of how they are programmed during manufacturing.

Thanks
Leo

[Zero999]

Does it need to  be low power? If not, then how about adding a resistor near the crystal to heat it up, once the temperature falls below a certain threshold?

[Leo Bodnar]

Yes, ideally TCXO should be a low power (2-3mA.)
In the past I have used 1206 resistor soldered directly to TCXO metal lid (it's usually connected to GND) as a heater but it's quite difficult to figure out what exactly is happeneing with heat dustribution there since PCB sinks a lot of it via ground planes and tracks.  I can't say that this was a success.
Leo

Does it need to  be low power? If not, then how about adding a resistor near the crystal to heat it up, once the temperature falls below a certain threshold?

[Benta]

What's in the rest of the circuit? Other semiconductors?
Then please reveal where you get parts specified to -60C.

[richard.cs]

Is this for a product or a one-off? For work or hobby?

I would drop Golledge an email and ask them if they can do you a custom TCXO: https://www.golledge.com/ They're normally pretty responsive and have very reasonable prices, MOQs and lead times for custom products.

If you want to build one yourself as a one-off, and can't afford the power consumption of a heater I would start with a VCXO, characterise what it does over temperature with a fixed control voltage, and then load that as a lookup table into a cheap micro which drives the tune pin with a DAC or PWM. Thermistor based compensation also works but it's a bit of a pain to get right.

[Leo Bodnar]

This is for a small balloon tracker for 13k-14k float altitude.  Night time temperatures can drop to as low as -70C. 
Other parts of the circuit are reasonably fine but I'd like the TCXO to have less of a drift below -45C - it feeds the GPS and RF transmitter.
Leo

What's in the rest of the circuit? Other semiconductors?
Then please reveal where you get parts specified to -60C.

[TimFox]

If I remember correctly, military temperature range for ICs is -55 C to +125 C (as in the 5400 series equivalent to normal temperature 7400 series TTL).

[edpalmer42]

Can you run the TCXO off of a higher voltage?  e.g. run a 5V TCXO off a 12V source.  Use a resistor in series with the TCXO to drop the excess voltage and generate a bit of heat.  If necessary, use a shunt voltage regulator across the TCXO to maintain the 5V supply and maybe generate a bit more heat.  Adjust the series resistor to draw whatever current you want and generate whatever amount of heat you want.  Wrap everything in insulation to keep it warm.

Ignoring the shunt regulator, dropping 7V @ 3ma will only generate 21 mW of heat, but with good insulation, it might be enough to keep the TCXO above -40C.

[ZigmundRat]

I had a very good experience with Bliley a few months ago when I needed an OCXO for a vintage GPSDO. They have TCVCXO parts spec’ed to -55C, that’s getting close to your desired range. I’m not sure if they have plain TCXO with that spec or what your other requirements are. I would contact them, they were very helpful to me.

I think Rakon also has some low power TCXO that go to -55C, but again don’t know your exact requirements.

Of course you’ll have to deal with the likely parts shortages (oscillators being more difficult things to source these days, at least for us).

[Leo Bodnar]

Thanks for suggestions.

At the moment TCXO is used in on/off mode with a few percent duty ratio so it turns on for 20-30 seconds few times an hour.
Turning TCXO into OCXO needs more or less continuous power due to thermal inertia. I have considered aerogel-filled heater compartment for TCXO but I am not sure aerogel is still better than air at low air pressures (15kPa.)

Total available energy going into the night is around 200J and cannot be extracted quickly - due to high ESR.  For a 14 hour night this is equivalent to continuous 4mA for everything.  Of course, the night length is unpredictable and can be anything - up to a few weeks in case of polar regions.  Therefore, TCXO needs to be able to start up on a short notice and at unpredictable intervals.

The other option is to qualify TCXO drift below -40C and then correct GPS to approximate offset.  I am just not sure if GPS I use (Ublox) accepts 100ppm clock offsets, I know from experience that up to 10ppm is fine. After GPS locks on it reports actual TCXO offset so RF stage can be corrected in software.

Abracon has -55C 0.5ppm TCXOs in AST3TQ series but they are relatively high power (10mA at 3.3V)

Leo

[TimFox]

Is it possible to wrap the temperature-sensitive sections thoroughly in lightweight thermal insulation, so that the section starts out at reasonable temperature (combination of temperature at launch and heat load from intermittent power-on time) and stays above -50 C or so for the duration of the measurement, relying on the temperature compensation to cover the evolution of temperature?

[Leo Bodnar]

It's definitely possible TimFox - in theory.
In practice you are facing a heat loss from the surface of the thermal package through radiation and convection.
Radiation (both ways) plays a big role in high altitude/space heat loss and you can easily have supercooled object with temperature lower then ambient air (or what is left of it.)

One of the best insulators (and readily available) is aerogel, it has better thermal insulation property than air at 100kPa but I am not sure how it's stacking up against air at lower pressure levels.
I think I played with numbers in the past and found that you can indeed make, say, ball shaped thermal package for TCXO and its heater.
I need to look the calculations up again but for 10cm ball and 5mm source package you can get 20C differential when using aerogel.
10cm sphere is very impractical (the whole tracker is only 50x10mm and weighs about 5-9 grams) but it's not impossible.
I am going to look into this again.  I have some fine grain aerogel granules to experiment with.

Leo

[NiHaoMike]

Maybe have a small bag of water in contact with the crystal? Once it reaches 0C, it will release heat as it freezes.

[ejeffrey]

I think aerogel should work very well at low ambient pressure, I expect it's advantage over air actually improves at low pressure.  The reason is that the thermal conductivity of a gas is proportional to density * mean free path * molecular velocity.  In a normal gas the mean free path is proportional to 1/density so the conductivity is roughly constant at constant temperature.   The reason aerogel has lower conductivity than air is because it is mostly just air but the mean free path is reduced.  That reduction becomes more significant at lower density when the mean free path at ambient is larger.  The aerogel also blocks convection which is usually more significant than pure conduction.

A thin coating of aerogel probably won't do much for radiative transfer since I think it is mostly transparent at the IR wavelengths of interest at low temperature. If that turns out to be a significant factor it can be dramatically reduced with a single layer of aluminized mylar.

[David Hess]

I have been trying to find commonly available TCXO that will work reasonably well below -40C.
What I have found through testing is that they are typically well corrected for their intended operational range to -40C and then literally fall off the cliff beyond that.

The reason for that is that the crystal is cut to have a minimum of variation over the intended operating temperature range, usually centered at room temperature like with the AT cut.  It is not simply a matter of matching the temperature compensation circuit to the crystal.  So a crystal oscillator or TCXO intended to operate at a nominal temperature below 40C would use a different cut, and be a specialty item because of low volumes.

I think getting any reasonable result unless you want to use an oven will require temperature compensating your own oscillator.  If you can get a crystal cut for the temperature range you are interested in, then it could be relatively easy.

[Leo Bodnar]

There must be a reason why NASA uses payload heaters rather than cryogenic electronics.

My experiments have showed that I can possibly achieve 20C temperature rise of -40C TCXO over ambient if I use combination of increased self-heating, extremely low conduction (0.02mm diameter wires) and Aerogel insulation.
Time will tell.

Leo

[David Hess]

Cryogenic electronics by themselves would not help NASA because then they would need coolers some of the time, and heater are simpler.  They need circuits which operate over a wider temperature range rather than just a cooler temperature range.