Crystal oscillator in a vacuum?

[hex4def6]

Hi all,

I have a problem with a 32.768 kHz crystal oscillator circuit that's being conformally coated in a vacuum.

It appears that on some devices, the crystal stops working after the coating. *However*, it seems that the "bad" devices "recover", which is extremely wacky to me.

Any thoughts on what could be happening?

Thoughts I've had so far:

* Coating is dielectric; maybe it is affecting the load capacitance
* Crystal is affected by vacuum, and maybe stops working (But aren't they hermetically sealed?). Maybe the conformal coating "locks in" the vacuum / no-op state, and it slowly relaxes once back in atmospheric pressure
* Coating has some sort of charge attached to it
* Vacuum is causing crystal package to be stressed in some way


Any ideas? This is just a basic 32k crystal, glass top plastic casing... The conformal coating is some sort of proprietary stuff that I don't know much about, but isn't conductive.


-Stewart

[Benta]

but isn't conductive

Are you sure? Bear in mind, that a 32 kHz crystal only requires a drive level of around 1 uW, which means that there's normally a 0.5...2 Mohm resistor in series with it.
Non-conductive is not quite the same as isolating, even a smallish leakage current could make trouble.

[jmelson]

Well known problem!  Epoxies either shrink or expand on curing, damaging sensitive components like inductors and crystals.  The mechanical deflection relaxes as the entire epoxy potting completes the curing process.  So, it is common to dip the assembly, or sometimes just coat the sensitive few parts with a soft material like RTV first, before the potting.  This allows the epoxy to deflect without cracking the fragile parts.

Jon

[Benta]

It's not potting, it's conformal coating. I would not expect mechanical stress, it's just a very thin layer.

[rhb]

I'd suspect parasitic capacitance produced by the partially cured coating.  Try measuring the intertrace capacitance of a PCB before you dip it and then over time as it cures.  If a crystal housing were that sensitive to mechanical stress, changes in barometric pressure would affect the operation of hermetically sealed crystals.

The best test would be put tests leads in place of the crystal on an otherwise bare board and measure that.

[Wimberleytech]

This is just a basic 32k crystal, glass top plastic casing...

Can you offer more specifics on the package?  Maybe provide a part number. 

[BrianHG]

This is just a basic 32k crystal, glass top plastic casing...

Can you offer more specifics on the package?  Maybe provide a part number.

I think also the conformal coating material is important as well.
Also, what is the oscillating circuit.  For example, Microchips ULP 32Khz crystal oscillator's drive circuit built into their PIC24 and PIC32 series is notoriously weak and has trouble starting up on occasion due to even some flux residue or track length on your PCB.  With such a PCB problem, even temperature or humidity can tilt the difference at powerup.  Especially on non-cleaned prototype PCBs.

[ConKbot]

In a vacuum? Is this an acrylic/urethane/etc 2 part resin coating? Or something that is "done" instantly like parylene?

Either way, run a vacuum cycle without coating, see if the problem is there. If it is, it's a vacuum related problem. If it isn't, it's a coating related problem.

Epoxies can definitely be slightly conductive before curing. I've seen potting conduct enough current to light a LED when in a peculiar situation there was enough surface area of the 2 appropriate circuit nodes exposed.

[hex4def6]

In a vacuum? Is this an acrylic/urethane/etc 2 part resin coating? Or something that is "done" instantly like parylene?

Either way, run a vacuum cycle without coating, see if the problem is there. If it is, it's a vacuum related problem. If it isn't, it's a coating related problem.

Epoxies can definitely be slightly conductive before curing. I've seen potting conduct enough current to light a LED when in a peculiar situation there was enough surface area of the 2 appropriate circuit nodes exposed.

Bingo -- it's parylene. I was being needlessly obtuse

I've tried the suggestion you had, and it appears that the crystal still oscillates in a vacuum... Was really hoping that was the problem.

I'm wondering if it's possible that there is a "pocket" formed under the crystal that gets subjected to the vacuum, then sealed by the parylene. When atmospheric pressure is restored, you now have a vacuum sucking down on the crystal?

The thing that gets me is that the circuit "recovers"... Some of them will fail to oscillate straight out of the vacuum chamber, but if you give it a day (or 3), it seems to just work. I was thinking somehow pressure was being equalized...

[rhb]

I'd have a very close look at the oscillator circuit.  I think that @BrianHG is pointing in the right direction.  You've got a marginal oscillator circuit and the coating is enough to keep it from starting.  That's likely to cause long term problems in the field even for units that work fine right after coating.

[ejeffrey]

A 32 kHz tuning fork crystal will work fine in vacuum.  If you have too much loop (no gain control) gain you might break it by over-driving without air to provide damping, but people take them out of the can and use them as AFM probes in vacuum.  Its most likely either the electrical properties of the coating (parasitic capacitance or leakage conductance) making the circuit not oscillate.

If the package isn't sealed, there is a possibility that applying the coating, then breaking the vacuum is causing the pressure to push the coating material into the package and getting somewhere that is mechanically damping the crystal.

But the most likely I would say is the electrical properties of the coating.  Because of the high impedance it is easy to mess up the oscillator circuit.

[David Hess]

In my experience integrated gate oscillators can be very fickle and are often poorly characterized.  I have seen problems where the ICs needed to be sample tested to verify proper operation of the oscillator over temperature.  For this reason, I either now do this myself to qualify the IC used for the gate oscillator, and sometimes this needs to be done for every lot, or even better, use an external discrete transistor based oscillator which will be much more consistent.

It would not surprise me if the original oscillator had so little operating margin that the vacuum encapsulation process caused it to fail.

[hex4def6]

Hi all,

Thanks for the suggestions.

It's looking like what's actually killing us is the "adhesion promoter" they use. I'm not yet sure what exactly it is, but a googling shows that A174 Silane is a common one.

We had them try a couple of experiments with / without the adhesion promoter, and it seems like just parylene is fine, but the combo causes some units to fail.

We don't use an Rl or Rs in the oscillator circuit, and both the crystal vendor and IC vendors have OK'd the C_l value we use. So, I'm not sure what the next thing to try is.

Would having an R_l help in this situation? Is it possible there is a "parasitic" R_l that is disrupted by the adhesion promoter deposition?

[SL4P]

I may have asked this elsewhere, but are we talking about a canned oscillator, or a crystal and supporting caps etc?

In dubious environments go with the former.

[hex4def6]

It's a crystal and its support caps.

The support caps in question seem to be lower than I would have used (~5pF), but that's been blessed as I said by both IC and crystal manf.

I have found that I'm able to measure the crystal directly on the support cap on the (output?) side of the inverter. On the input side, it seems to collapse the oscillation. I'm not too suprised by that, it's 8pF of passive probe after all.

Would having an R_L across the xtal help in any way?

I agree going with the canned oscillator would be the easy way to go.