Hi Folks,
During the build up of my 10 MHz DA project, my lovely OCXO failed! This provided an excuse to open it up and see what's inside. I was very curious as to if it was really a double oven and what the device's quality was like.
I got the device not so long ago on ebay from "beatjhunky" in Irvine CA. It feels heavy and is pretty large. When powered on, it may draw as much as 580mA on startup, which will soon get down to around 208mA steady-state. The device gets HOT, but that is expected of course. The output, unterminated, was just over 1Vpp. Without any trim, we were about 100 Hz out.
The failure mode is that the device now only draw about 10mA and the heaters do not kick on. The output still remains though.
Disassembly:
The outer metal can is soldered to the base plate with what seems to be standard solder. I used a heat gun to get the entire thing quite hot. You can see the metal's texture change when it starts to really get hot. I wasn't too worried about damaging it since the device is hot by nature and is designed to have thermal isolation inside. Using the heat gun, I got the entire thing extremely hot and then focused in on the edges. When I saw the solder start to flow, I used a razor blade to poke the solder down into the can (basically sliding it on the wall), and pretty soon I had major gaps on all sides. I then took a screw driver and pried it up and out.
Right away I noticed a lifted diode on the power line. However, this diode probably fell off due to my own carelessness and fixing it only got me to where it was before disassembly (10MHz out and no heaters).
There are two n-channel mosfets on either side of the first inner can, which are clamped to the can and have a thin piece of kapton tape as an insulator (the rear of the mosfets is a terminal). There's also a large power BJT that isn't heat-coupled to the can. This BJT I believe is used for the op-amp to drive the inner-most heater. The mosfet datasheet says it is designed to be driven with digital logic levels, so likely these outer heaters are bang-bang course heater control mainly for startup, driven by the control circuit without any proportional control (but I cannot be sure).
The board is multi-layered, with at least one inner layer. You can see traces that go off the side of the board, presumably to enable testing a large set of boards from fabrication before the boards are cut apart.
There are numerous parts on the outer rim of the board, including what appear to be some SMT transistors. You can also see a thermalcouple which is wired to the lid of the first inner can (small green wires in the photo). It's about 11k at room temp and goes to lower resistance when heated.
Well, I had to look inside the next can, because I didn't see anything else obviously broken. This can is soldered down to the PWB. The PWB is also cut in a bit of an island, likely to limit thermal and vibrations to the inner chamber. The solder was thin, and I was able to cut through it by starting at the PWB bridges and working slowly around it with a razor blade.
Inside we find another can, which is the crystal's sarcophagus, so to speak. It's connected using a 4-wire flexible cable to the main board (and all the traces are inner). There's also a large transistor bolted to the sarcophagus, which looks like it might be a BJT driven by the outside BJT (they share an emitter-to-base trace it seems), which is ultimately driven by a quad op-amp.
The quad op-amp sits directly beneath the sarcophagus, and the slightly-yellowed writing shows it's been tortured plenty. I checked for shorts between the pins and did not find any. The power traces buzzed out fine to the input power diode as well as ground. The op-amp is only $0.74 from mouser, but of course, shipping and tax and whatever, it'll end up costing as much as another OCXO by the time I order one.
I plan to power it on all opened up and see if I can determine what's going on with it. I suspect it is the op-amp since both heaters are failing.
Overall it seems like a very expensive and detailed design. It must be a lot of work to assemble something with all these layers and processes (tape, soldering, fasteners, over and over). I did not open the sarcophagus -- I think I might, for once, know better than to do so -- partially since it appears to still deliver a good 10 MHz signal. I was surprised to not see any insulating foam, as is common on these take-apart videos you find on youtube. However, I would guess that the inner most can does have foam around the crystal. And with all the layers and gaps, it's probably isolated enough. I also noticed a little glob of solder covering a hole on the outer-most can, and I wonder if that was to permit the can to be filled with some inert gas or something during assembly. The outer can does appear to possibly be hermetic, given that it is all soldered down and the pins appear to be ceramic-metal feedthroughs.
Anyone have troubleshooting suggestions? I'm going to run it and see if I can trace anything out. If I find a quad SOIC op-amp of similar spec anywhere I may try it out. I'll have to go over my boneyard of old circuit boards and see what I come up with.
--E
de W6EL
Home
Help
Search
Login
Register
