How to remedy an intermittently starting crystal oscillator in oldish equipment

[DaJMasta]

I have a crystal in an Electrophysics 7290 camera that doesn't usually start on its own.  I ran into the issue before replacing a regulator and now the regulator works and the output is fine, but the crystal only occasionally starts.  My favorite part of the fault is that if I touch an oscilloscope probe to the one side of the crystal (connected to a scope, but the scope doesn't have to be on), I can get it to start.... with basically 100% reliability.  In further experimenting, I can get it to start by using a multimeter lead in the same way (also doesn't need to be on and can be floating), but what's better is that by touching the opposite side of the crystal I can get it to stop oscillating, though this doesn't work with the 10x scope probe (~13pF).  Helpful to me, the board has a whiny inductor or cap somewhere that will audibly tell me when the crystal starts running and the camera shows an image!

Normally, I'd replace the part and be done with it, but unfortunately this camera is full of oddball part numbers with no available documentation and the crystal in question is the extremely common 1.291500 MHz variety in a larger can form factor than even the larger crystal cans of today.  The chip that it connects to is marked but which turns up no search hits online, but given the outputs and crystal itself, it should be doing all the timekeeping required for analog video output.  There is a small load capacitor on one end of the crystal, but both connections go to the mystery IC and the case of the crystal has a grounding wire.  I've tried swapping the capacitor for larger and smaller values (20pF or so nominally, I tried 11pF and 30pF or so) and it doesn't seem to make a difference.  Also worth mentioning that it does not appear to be physical shock that is starting it - just tapping the board doesn't start it and prodding the pins with a nonconductive probe doesn't do anything, I've also reflowed the joints.

So I don't know if the crystal is going bad or the timing IC is going bad or what, but I know that replacing either will be a tremendous chore, so how do I add whatever parasitics it will need to start to the circuit to make it reliable?  I've also tested and see that a resistor to ground will get it to start on the capacitor side, so there is the option of a literal jump-start button for the crystal, but a no-manual-intervention option would be great.

[bdunham7]

Put an ice cube in a baggie and hold it against the crystal can.  If getting it cold starts it, replace the crystal, if you can find one.

[bob91343]

Getting down to basics, the problem is one of three:

1. bad crystal
2. insufficient oscillator gain
3. poor design

If you can open the crystal you can try cleaning it.  The frequency is pretty low but not excessively so, given that most watches have crystals at around one fourtieth of that.

[fzabkar]

A typical configuration that I see is where a 20pF (?) cap is connected between each pin and ground, and a 1Mohm resistor between the two pins.

[DaJMasta]

I had seen similar in a lot of cases, but for whatever reason this implementation seems to have a single 20pF cap on one side and nothing on the other, only connected to the IC pins otherwise.  I had tried a higher value resistor in parallel with the crystal, maybe 1M or so is worth a shot, and it's probably worth trying with a 1M to ground resistor too.

It occurs to me that I may actually be able to test the crystal with the very bottom range of a single port VNA, but removal is going to be a pain so it may take a few days to get to it yet.  I wonder if probing it is introducing a load or an interfering signal that is just enough of an impulse to give it that first oscillation.

[bob91343]

Substitute another crystal.

[DaJMasta]

Substitute another crystal.

Are you suggesting just any old crystal?  I guess I could drop in a 1MHz crystal to see if it oscillates, but I'm going to have to get at least close to the original frequency if I want to get in the right ballpark for the output image, and I don't really know how much headroom the unit has to run at a higher speed (like 2MHz) since this also seems to control the timing for the power supply switcher.

I was sort of hoping to find a workaround aside from replacing it to get it to start, since when it's going it does seem to be steady and stable.

[bdunham7]

I was sort of hoping to find a workaround aside from replacing it to get it to start, since when it's going it does seem to be steady and stable.

The last non-starting crystal I worked on I was able to jump-start it with a signal generator and an oscilloscope probe.  I could even get it to start without touching it at times.  You could put in a small add-on board with this oscillator and probably a 3V3 regulator if there is no 3.3V supply in the camera.  Just put the output of the oscillator close to your crystal, or maybe with an insulated wire touching one end or the other so there is ~1pF or so of coupling.  You might need to shut the supplemental oscillator off after the thing starts up.

https://www.mouser.com/ProductDetail/Silicon-Labs/510CCA1M28640AAG/?qs=qSfuJ%252Bfl%2Fd4ewuJDI0dfiw%3D%3D

I had a look and I've have to concur that your crystal is pretty much unobtanium.

[wasyoungonce]

I've had a non starting crystal in a complex cct....replaced it and caps...no fix.....noticed my finger on the case it worked every time at pwr on.  Soldered wire earth to case ...fixed. 

[fzabkar]

I would be scared to price a one-off custom cut crystal, but 30 years ago I was able to obtain them for AU$25 from Hy-Q. Who knows, maybe they're cheaper today, especially if they're coming from China.

https://hy-q.com.au/crystals/crystals-order.htm

[bdunham7]

http://bomarcrystal.com/hc_51Ucrystal.html

[amyk]

Another option could be one of  those programmable oscillator modules.

[bob91343]

If it won't start, the Q is too low, or the circuit gain is too low.  There needs to be a perturbation for the amplifier to amplify in order to get started.  If you were using discrete parts it would be easy to adjust the gain.

Trouble is, if you artificially give it a perturbation and it runs, it could easily stop and need another.

[BreakingOhmsLaw]

Try turning it around (desolder and put it back in reverse).

[TerraHertz]

Is there a resistor in parallel with the crystal? And if so, is it a carbon composition type? (Which notoriously drift high with age.)

Perhaps the mystery IC uses a simple inverting amp, and needs the resistor to keep the amp input in the linear region.
If it's that kind of oscillator but no external resistor, maybe the chip has an internal resistor across the two pins for the crystal.

In either case, if the resistor drifts high, the oscillator can fail to start. Because the amp has reduced/no gain outside the linear region. But once running may get more or less workable average voltage to the input pin due to clipping.

Anyway, try adding a 1M resistor in parallel with the crystal? Or if the IC is something like TTL, it may need to be a much lower value resistor. 1K maybe?

[dietert1]

Normally a crystal should have two capacitors to Gnd and it doesn't work well if one of them is missing or to small.

Recently i had this in an ONKYO 8690 Amplifier, where the SPDIF inputs were not working reliably. For production they had replaced the crystal of the SPDIF receiver by a cheaper ceramic resonator, and forgot to increase the two capacitors. Replacing the two 20 pF capacitors by two 30 pF capacitors resolved the issue.

Regards, Dieter

[perieanuo]

hi,
start your oscillator, take a precise measure of his frequency then try to build an stand-alone oscillator with that freq, then feed that with some series resistor to your chip. it gotta work

[DaJMasta]

Well I did some testing and some playing around, and after a few hours of fine operation I think I've got a good replacement.

After considering the options to replace the crystal, I looked into buying a programmable oscillator and then placed an order - Digikey would sell me a programmed SGR-8002JC-SHC 5V oscillator for like $4, so it was certainly cheap enough to attempt.

While I waited for the part, I test the crystal on a VNA.... and it looked fine, so I guess this issue is the degraded feedback in the mystery chip, which bodes ill for the longevity of the system in case something is slowly failing open.

I tested with a signal generator what level and connection I needed, and while it looked like 5V would sometimes work, it needed a 5.3V amplitude or so on the sig gen, and it needed to be applied to both connection pins for the crystal to work reliably.  Playing with the phase on the second channel (the other pin) didn't seem to help anything, so I decided with the oscillator arriving I would build up a little mod board and just AC couple the output into both pins (since the pins are biased by the micro).

The oscillator arrived and I built up a little board that barely fit in the space where the relatively large crystal was.  Since there isn't actually a 5V rail in this camera, I included a TO-92 LM78L05, some bypassing, the oscillator, the AC coupling caps, and a filter.

The filter was made just through trial and error, and since I didn't know the impedance it was driving, to be sure the amplitude was in the right ballpark, I built in a load resistor and took a look on the scope when I was powering it up and making changes out of circuit.  I ended up with a series 22uH inductor (15uH was more efficient for power transfer, but then had a peak to peak voltage above the supply rail of the chip it was driving), about 2000pF of shunt capacitance, and a 7.5k Ohm load resistor.  Then I just used a couple of 0.1uF caps to AC couple into each side.  I played briefly with some copper tape for a shield, but grounding to the board itself stopped it from running, and grounding it to the board ground plane let it run for a bit, but then shut off.  Since the camera isn't really grounded (isolated wall wart) and I wasn't really seeing much noise on the output, I decided to just leave it off and call it done.


I can definitely recommend buying a programmable oscillator to build into a board to replace an oddball crystal value, they were surprisingly cheap, could be programmed precisely (they gave me 1.29155MHz on the nose!), and are small enough that with a more compact PCB could replace relatively small crystals (maybe not SMD or short can oscillators, but certainly the taller standard can size or larger), you just have to be able to build up your filter to make sure the drive voltage and waveform is good.