[Beowulf] Server Sky - Internet and computation in orbit

Lux, Jim (337C) james.p.lux at jpl.nasa.gov
Sat Jun 22 19:43:55 PDT 2013

On 6/22/13 4:58 PM, "Lawrence Stewart" <stewart at serissa.com> wrote:

>I've been imagining that these thinsats are a sandwich:  Solar cells on
>the sun side, computation in the middle, and a radiator on the back.
>It might be best to have two sheets, with a vacuum gap between the solar
>cells and the computation, to prevent conduction.  Presumably the solar
>cells would have thin film filters to reflect (not absorb!) the radiation
>except at wavelenghts with good PV efficiency.
>Insolation at Earth orbit is about 1 KW/m**2, and PV efficiency about
>20%.  Suppose that
>with the filters, you can reject 60% while absorbing 20%.  So for 1 m**2
>you get about 200 watts electric and 200w heat, total 400w to dispose of
>via radiation.
>According to (that excellent source!) Wikipedia, at 307K with good
>emissivity,  you can
>radiate about 1000 w from the 2 m**2 area.

You're facing earth sometimes (e.g. If you're over local noon) so you
don¹t get that good.

A good working number is 300-400W/square meter to cold space and 100W to

You aren't always directly facing the sun, so you have to take cos of the
angle into account.  And the chipsats/thin sats are using solar pressure
for movement, so there's going to be tradeoff of how you orient from a
thrust standpoint to how you orient for power generation.

All this stuff is pretty easy to calculate, though, a couple dozen lines
of matlab kind of thing.

Don't forget that if you radiate RF power, that helps get rid of some
energy. (e.g. A typical power amplifier is 20% efficient, so you put in 5
Watts of DC power, 4 of which gets rejected as heat, 1 of which gets
radiated away)

Pick the right materials and you can run substantially hotter which helps
the radiation cooling (GaAs or GaN or SiC are all attractive in one way or

You could also run your silicon fairly hot, if the life doesn't have to be
decades (with cheap launches, presumably you just keep feeding the cloud).

SiGe is also very attractive (no latchup for one thing, and quite rad
hard, although the total dose issue probably isn't a big deal.. You're
taking lots of radiation at the proposed height, so you either can take
all those Rad(Si) and keep working, or you can't.

A lot of the "radiation tolerance" has to do with whether the part meets
the original databook specs after taking the dose.  If you don't care
about parameter drift but it still works, then relatively soft parts might
work.  For instance, if you need a transistor gain of 10, and the part
you've got has a beta of 200, then you can take a 95% hit in performance,
and the circuit still works.

Also, it's really more about "knowledge of what happens" that makes space
parts expensive. It costs a lot to take a batch of parts and run them over
time/temperature/radiation so that you can predict what will happen and
what the failure modes are.  If you don't care about failures, or you're
bold, or you have some way to test them cheaply, then you can do better.
For most NASA missions, they prefer to have the "risk" on the science, not
on infrastructure and components, so they tend to stick with parts with
known properties, rather than spend money on testing new parts.
Commercial customers, particularly in the small sat world (e.g. Surrey's
customers) have a different trade space.  And commercial customers doing
things like comsats which might have 50 or 100 radios have large enough
volumes to justify testing (or design of custom parts), but those parts,
and their data, is a competitive advantage and not published.

>Consequently, it appears to me that you can have a thinsat that is solar
>powered and radiatively
>cooled, with an equilibrium temperature below room temperature.

Yes.. A sphere painted white or black, for instance, gets very cold.  A
shiny metal sphere will get hot (just like your seatbelt latch in your car
or a handrail). It's all about the alpha/epsilon ratio.

If you look at early satellites (e.g. Explorer 1, or OSCAR 1), they're
stripes of shiny metal and paint, with the ratio adjusted to hold the
right temperature.

>Of course my figures could be way wrong, but it seems plausible.  To do
>better you just need a bigger radiator on the back.
>I suppose the radio is a phased array of antenna elements or even
>integrated lasers with mems mirror steering.
>Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>To change your subscription (digest mode or unsubscribe) visit

More information about the Beowulf mailing list