Deap Space Navigation

Dave Michelson wrote:
wrote:

Actually, this one is very close. If a 70 meter dish failed today,
I'd bet they'd do it. A new 70 meter class dish costs $100M, and an
array of smaller dishes has lots of advantages. It can be quite a
bit cheaper, depending on the details. It can be split to point at
many targets, it has fewer points of failure, it can be maintained a
little at a time, the size can be increased in smaller increments, it
can be spread out geographically to add weather and disaster
diversity, and so on....

The days of big dishes are numbered.

I agree with all of the above, but am curious about the cost.
It would take at least eight or ten 25-metre dishes to equal the
collecting area of a single 70-metre dish. (25-metres being a fairly
standard diameter reflector for use in radio astronomy arrays :-)


It appears the arrays are potentially quite a bit cheaper. This is a
fight between two factors - first, collecting area is cheaper for small
dishes, per m^2. (or expressed another way, the cost of big dishes
scales faster than diameter^2). On the other hand, a single big dish
needs only one feed/pre-amp, but an N element array needs N of them.

As a point of reference, recent DSN 35 meter antenna are about $45
million. The 70 meter dishes have not been built recently, but would
clearly be a lot more.

These tradeoffs can be made in many ways. For example, the ATA (Allen
Telescope Array) uses 6m dishes, the largest that can currently be
stamped in one step with satellite dish technology. You'd need at
least 136 of these to equal a 70m collecting area. However, you really
need more, since you can't afford 136 super-duper helium cooled masers,
and really high feed efficiency is hard in a small dish. So instead
you go with a liquid nitrogen cooled receiver, and a simpler feed. The
net result is double the noise temperature, so you need twice the
collecting area, or maybe 270 dishes. The Allen Telescope project
thinks that these will cost perhaps $50K each, in pseudo-mass
production, so that's only $13m for the signal collection. Then you
need to add the signals up, but that's not so expensive nowadays. But
the ATA dishes and feeds only go to 10-12 GHz or so, and JPL needs 32
GHz.

See
http://www.skatelescope.org/skaberke...tion%202.9.pdf
for a somewhat dated (2001) JPL analysis using small dishes like those
of the ATA.

Nowadays, JPL is tending towards a design with 12m dishes and better
receivers. This will win if you can find a way to stamp out the 12m
dishes at low cost. See
http://www.skatelescope.org/PDF/SKA2...aceNetwork.pdf

The pattern of a widely spaced array of large antennas would be a lot
more complicated than that of a single large antenna. In particular,
the grating lobes would be very noticeable. (Almost all conventional
arrays are closely spaced arrays of small antennas.) Having said, that,
this has been done before with the VLA and suitable use of feedback
control and RF-over-fibre links would solve lots of immediate problems.

This should not be much of a problem, since a spacecraft will look like
a point source, mostly against a dark sky. If you simply add the
beams with the right phase, you get the same central sensitivity as the
big dish, just crazier sidelobes. This may not be a problem for
spacecraft, but it is for astronomy, so to help with this, the ATA is
basically placing the antennas at random on their site (actually, they
try to get a good mix of short-long baselines, good U-V coverage, etc,
but it *looks* random) to reduce the grating lobes. The only case I
can see this being an issue with the DSN is when they try to track
spacecraft very close to a bright source such as the Sun.

The good news, of course, is that much technical experience and
expertise has developed over the past decade (in particular) from work
on multi-antenna radio astronomy projects ranging from the VLA to the
SKA (Square Kilometre Array). (Yes, there are significance differences
between aperture synthesis and conventional arrays, but by and large the
people involved could likely handle both or either as required.)

The communication job is much easier than the astronomy job, since you
are not forming images, just optimum response at one point. The people
skills are similar, as you say, but the astronomy work requires huge
correlators, and often high instantaneous bandwidths, where the
communications just needs tunable delay lines.

Lou Scheffer

wrote:

The pattern of a widely spaced array of large antennas would be a
lot more complicated than that of a single large antenna. In
particular, the grating lobes would be very noticeable. (Almost all
conventional arrays are closely spaced arrays of small antennas.)
Having said, that, this has been done before with the VLA and
suitable use of feedback control and RF-over-fibre links would
solve lots of immediate problems.

This should not be much of a problem, since a spacecraft will look
like a point source, mostly against a dark sky. If you simply add
the beams with the right phase,

My experience in a variety of contexts is that there's rarely anything
simple about phase that temperature, torsion, flexion, and a variety of
other maladies can't royally screw up :-\

you get the same central sensitivity as the big dish, just crazier
sidelobes....

Agreed, but so many of them and with very high SLLs relative to the main
lobe. I have visions of a classic interferometer pattern multiplied by
the pattern of a single dish. This all depends critically on the
separation between elements, of course.

With a larger aperture, albeit a sparsely sampled one, comes much
greater resolution. As a result, the pointing direction of the main lobe
will have to be much more carefully controlled. Fortunately the fine
tuning can be done exclusively by adjusting the phase delay to each
array element. As I mentioned previously, I suspect some sort of
feedback control would be required to achieve the desired pointing
accuracy. (Classic array signal processing wouldn't be appropriate,
though, because each element has insufficient sensitivity to see the
signal of interest.)

I see from the slides that you provided links to (thanks for that :-)
that the team of Caltech has conducted "uplink phasing tests to
determine the calibration burden." I assume that this means that
they're hoping to do it open loop but will resort to closed loop control
if necessary.

Neat stuff!

--
Dave Michelson