KVG OCXO failure

[Manul]

Hi, guys,

I have an OCXO with a weird failure. It goes like this:

1. Power up from cold state.
2. During temperature ramp up it crosses specific temperature point at which it goes mad: frequency suddenly drifts, huge distortion, phase random, amplitude random and then output goes to zero.
3. 1-2 seconds later it comes out of that failure zone and starts again nicely while continuing to warm up to it's temperature setpoint.

Completely stable and accurate otherwise. It is made by KVG in 2016, so not very old. The bad thing is that I can't seem to find this exact part number. Even more strange is that it seems none of KVG OCXO models matches the pinout (which I reverse engineered). Maybe it was custom order?

Anyway, so because the module is not hermetically sealed, I desoldered it's shield. Inside is a serious amount of electronics, double board construction. Seems quite complex. The heart is HC-49U looking XTAL.

Does it sound to you like a failing XTAL or more likely oscillator and other circuitry? I will mess around a bit more, but mostly at this point I would like to buy a new module, but there seems none. Will try to speak to KVG, but I don't have big hopes there.

[Manul]

The follow up. It is an interesting oscillator. Top PCB (heated, working temp around 75C) is actually a 20 Mhz sine oscillator. Bottom PCB is a divide by 2 sine output (final). No datasheet for this module (so far), so no real specs, but I expect this to be some ultra low phase noise thingy.

Below is oscillator section reverse enigeered, not 100% guaranteed, but seems logical, so might be correct afterall. A lot of passives not measured, of course might be done, but I don't want to start removing components for no reason at the moment.

What is clear after testing with no lid is that someone came and stole some Q.

It simply has marginal loop gain or so it seems. Without lid it works even slighly worse and sometimes has problem to start at all if cold. The complicating thing is that they use a lot of tuned circuits here. And I don't understand yet is it the crystal which deteriorated (I've seen such things happen) or is it the varactor or these tuned tanks. Amplitude control feedback is working as expected, it is just a DC voltage controlling the bias based on output stage amplitude.

Found some weird things like zero ohms resistor (jumper) across populated capacitor. Really had hopes that capacitor is cracked and shorted, but no. Also found small rf capacitor seemingly between grounds. Both of grounds goes into vias, one of them I really can't check without xray (4 layer PCB). Does not seem that they use ground separation or anything such fancy, so seemed weird, but oh well. All the components are glued with some kind of epoxy, even solder joints are covered, makes it hard to probe. Crystal is likely impossible to remove without risking serious destruction.

Tried injecting a signal, overall the passband of all stages together seems alright and centered, but that does not say much about the tuning of individual tanks. Thinking how should I do it. Add trimmer caps everywhere?

My theories:
1. XTAL Q deteriorated due to overdrive, temperature or vibrations
2. Varactor Q deteriorated
3. BJT deteriorated
4. LC tuned circuits marginal from factory due to component variations plus drifted due to aging
5. Some random failure which I haven't found

[hgl]

Here is a datasheet of a similar OCXO maybe it helps.
I have several 10MHz SC-cut KVG oscillators with the same pinout but without a label. They all work but with unknown performance.

[Stray Electron]

   Regarding the 2nd picture, is it my imagination or don't the solder joints on that CCA look pretty "crusty"?

[Manul]

Here is a datasheet of a similar OCXO maybe it helps.
I have several 10MHz SC-cut KVG oscillators with the same pinout but without a label. They all work but with unknown performance.

Yes, the pinout from that datasheet maches my module. Thanks a lot. Somehow I was not able to find anything that mached.

Regarding the 2nd picture, is it my imagination or don't the solder joints on that CCA look pretty "crusty"?

Well, yes, crusty. It is lead-free solder. Does not look very nice, but so far I haven't found dry joints. I work with good stereo microscope and test joints if suspicious. At the beginning I was hoping to find a bad joint and have quick success, but it seems no.

[Manul]

I'm thinking that I need to add extra series resistance to the crystal, lets say 50 Ohms to really kill the Q, then try to achieve oscillation. If I can manage that, remove series resistor and call it a day. Otherwise call it goneski. Because it is nothing worse then a marginal fix. What do you think?

Regarding the tuned LC networks. Do you think they all should theoretically be tuned for maximum gain and maximum output amplitude at the frequency of interest, or they might be deliberatly slighly out of tune for spurious filtering or with intention to achieve some performance specs? I don't think that I've seen a Colpitts design with tuned LC networks before.