Re: [Gumstix-users] Radioactive tolerance of different boards

[Bill G. <bg...@bi...>]

cro...@fi... wrote:
> Hi there all,
>
> I'm wondering if anyone has any data regarding the radioactive tolerance of the different gumstix consoles and expansion boards. I mean, exposing it to radiation levels found in space to be specific. 
>   

That actually isn't very specific.  :)  What kind of radiation?

I've done testing with accelerated protons (not with Gumstix boards), 
which are commonly found in low-earth-orbit (LEO) altitudes.  Through 
that process I learned that there are other kinds of radiation you have 
to know about, too.  Some are easy to find in terrestrial labs, others 
not.  Some are a concern for certain kinds of missions, others not.  And 
then in LEO you have to worry about things like atomic oxygen, which 
eats a board alive due to corrosion, and pretty serious magnetic fields 
that might confuse the coils in your switching power supply.

The boards themselves won't generally be a problem, except for the 
radioactivity they pick up during testing.  It's the _components_ on the 
boards that you have to pay attention to.  Some types of digital logic 
circuits are immune, others will "latch up" at the least disturbance.  
And you find both kinds in an XScale chip, or any other 
relatively-sophisticated VLSI component.

Generally speaking, unless you absolutely torch your chips with 
completely unnatural levels of ionizing radiation (thereby turning the 
chip into something other than silicon), the main failure mode is 
latchup and subsequent overheating of the target transistor(s).  The 
effect is usually an open circuit thereafter.  For storage devices like 
SRAM and particularly SDRAM, lower levels of radiation from things like 
electrons or protons cause single-event-upsets in the storage elements, 
i.e. a bit "flips" to a state other than the one you originally wrote.  
Nondestructive to the circuit, but they obviously affect your data.

If you have sensitive power consumption monitoring circuits, you can 
detect the latchup due to increased current demand and forcibly 
power-cycle the circuit to release the latch.  You can't just interrupt 
the power, you usually want to "crowbar" the Vcc line to GND to suck all 
the energy out of the circuit as quickly as possible.  In a project I 
worked on, we had to constantly regulate the setting to a comparator 
circuit that would trip the power supply protection if we exceeded a 
preset current consumption level.  We were constantly adjusting the 
comparator setpoint based on what the program was doing at the time, so 
as to keep the threshold just slightly above our expected demand.

I haven't looked at Gumstix's boards in detail for this kind of 
question, but I would be very, very surprised to find anything that 
could do this in their products.  It's kind of an unusual requirement, 
and something that a terrestrial application probably would _not_ 
consider to be a feature.

So then there's the question as to the cost of failure, and whether it 
makes sense to spin a board that can prevent it to a greater degree than 
an unprotected circuit.  You could probably fly an unmodified Gumstix 
for a short mission, but it would end up being a short mission whether 
you planned it that way or not.  :)

How short?  Well, that goes back to your first question.  Could be mere 
seconds, if Nature's aim was particularly good that day.  :)



b.g.

-- 
Bill Gatliff
bg...@bi...