Re: MOSFETS not as trad-hard as promised threaten NASa space mission to Europa

How do mosfets fail from radiation?

Enough rads to kill a mosfet would be bad news for people.

Radhard electronics is as much a problem of design as of component and technology selection. In MOSFETs, the nett effect is positive charge getting trapped in gate oxide, resulting in threshold voltage shifting increasingly negative. In bipolar transistors, radiation creates recombination centres in the base region, resulting in a reduction of transistor beta at low bias currents.

The message is obvious: In MOS circuitry, anticipate threshold voltage drift, so apply wide gate voltage swings for switching MOSFETs and bias point feedback for linear circuitry.

In bipolar circuits, anticipate a drop of beta and use generous bias currents. Avoid ICs with lateral PNPs, which are only barely good enough to start with. It's a good rule to choose old designs, because those were designed in an era when parameters were less well controlled, and consequently admitted more variability in parameters before failing.

This is a gross simplification, of course, but it gets you quite far on the way towards rad-hard design,

Jeroen Belleman

Are you able to choose old bipolar designs? With MOSFETs, the vendors regularly do die shrinks etc. that you don't have good visibility to unless you have a record of all the Part Change Notifications (PCNs) that have occurred throughout the part's history.

That's a problem, indeed. The manufacturer may change the chip so that it still answers to the original specs without telling us, and that may indeed completely change its radiation hardness.

I was in the fortunate -if uncomfortable- situation of never making large series of devices, and several times was able to find enough old stock.

I remember being shocked by the fact that the newer TTL chips, F, AS, ALS, what not, tended to die after 40Gy or so, while the old standard and LS TTL would still work reliably with more than a kGy.

Rad hard design has its challenges.

Jeroen Belleman

I've been thinking about designing a radiation resistant counter. Basically feed in a 32.768 kHz clock and be able to pick off bits that will tell me when a certain amount of time has elapsed since the last power on of the spacecraft. Just AND the higher order bits together if I'm looking for intermediate times. So if I want to get up into the length of days, I'm going to need > 32 bits. Any ideas on how to do this?

A handful of flip flops and a SN74LV8154 32 bit counter gets me up to 35 bits terrestrially. What about space?

HCC4060B Rad-hard 14-stage counter/divider w/ oscillator + HCC4020B Rad-hard 14-stage binary/ripple counter + HCC4024B Rad-hard 7-stage binary/ripple counter gets me 35 bits, but is that nuts?

Are there radiation resistant PLDs or tiny FPGAs that get used for something like this? Are the old discrete logic families radiation resistant enough? I have some experience with radiation resistant memories and large SoCs, but not lower level logic.