Reliability Crystal vs PLL oscillators

I am wondering if there are any studies for oscillator circuit selection for the space (Satellite) applications. Frequency drift to failure rate. I am assuming that PLL has more failures then the crystal based circuits due to number of active components used in the PLL circuits.

I will be appreciated to any personal experience or University studies.

Thanks
Levent

There are some standard methods to calculate the failure rate. It's a little complex. I found them in a radar engineering textbook.
Good luck.

The thing is most PLL's have crystals as a reference or other use anyway, so the question is probably a little moot

Dave

Peregrine Semiconductor makes rad hard PLLs for space applications. The attachment has some reliability data for their PLLs.

MIl-Std-Hdbk-833 has a base rate for each part number based on history and stress factors based on many design environmental limits, such as operating temp range etc. So the qty pf parts is not as important. Of course actual vendor data is better than generic calcuations, where a failure if analyzed and fixed in a design change then re-verified can be censored giving long run device -hrs between failures. Crystals come in many sizes and shapes, so military or aerospace requirements ought to factor if the part has a ruggedized mount on the quartz and the power level is well under limits if in series mode. But vibration can shift even the most rugged crystals. I had to design one for Doppler tracking in a small rocket (Black Brandt IV with Nike booster) in '76.

I chose a Vectron SC cut ruggedized mount OCXO with a spec. of 1e-12 short term and 1e-11 long term.. That still wasn't good enough and I needed 1e-10 tested at 100g/10mS sock, 50g acceleration and 15g vibration. . I had only 1/4" gap allowed for my enclosure budget and it took me a month to test and pass the right material , like low Q Solethane and CMOS packing foam to absorb enough to protect the OCXO from drift.

Lead foil shields may help for your longer term journeys..
Trust but verify.

Folks,

Thanks a lot for the replays.

I was looking this problem from a little bit different angel. However your postings are showing more dimensions of the problem and I think I need to think more. Last time when I calculate modular failure probability it was 30 years ago as far as I remember it was higher numbers when you add more components.

If we look the major components of a classical UHF oscillator 300 ? 500 MHz we will have a crystal, 2 ? 3 transistors (BJT,FET) and a MMIC for output buffering. Basically we are looking 4 active components and two critical passive components (Crystal and Helical filter). On the other hand PLL will have a crystal oscillator module (1 or 2 transistors and a crystal) , PLL, maybe a pre-scaller, one MCU to control PLL, VCO and buffer amplifier. My best estimate is 2 IC and 3 active components with a crystal. Number of components are nearly 50% more and silicon area is nearly doubled in the PLL solution. Just looking from number of active components and the silicon area which is open to space radiation will increase failure probability gives classical oscillator better marks. If we add SunnySkyguy's vibration stress and count the number of soldering points to the probability equation classical Oscillator is at least two fold better choice then the PLL approach.

Am I right on this assumptions or am I missing any other factor which can favor PLL based solutions.

Thanks again,

- - - Updated - - -

pstuckey : Thanks a lot this is a very good document.

Not all components have the same FIT factors . Last time I did it was 35 yrs ago.. there must be better data now. Consult 883 handbook.

- - - Updated - - -

COrrection: I meant to say Mil-Std-883 and Mil-HDBK-217

Thanks a lot this is a very good source.

hi Levent,
as it is mentioned before peregrine`s pll s are nearly the only choice for space applications.
many space apps today use a radhard and stable crystal and multiply the output to reach higher frequencies you may use this approach for UHF band. as many space design do so nowadays.
by using a radhard transistor you can design your own oscillator and use a PFD from hittite or macom, those two vendors and some other also supply hirel parts.

regards
kagan

I believe for Satellite they buy very expensive, and high performing Crystals

ref journals;
RADC-TR-85-229 ADA164747 Reliability Prediction for Spacecraft
MIL-HDBK-978B 1 Sep 89 NASA Parts Application Handbook

But I highly recommend you seek the supplier of choice I used in 1976.
Vectron: The R-Type crystal oscillators are quartz-stabilized, TTL or CMOS selectable,square-wave generators. They are packaged in low-profile, hermetic ceramic, 4-pin DIPs. Long-term reliability exceeds 20 million hours mean time between failure (MTBF).