david@djwhome.demon.co.uk (David Woolley) writes:

In article <4096cbd5$1@news.cadence.com>,
Louis Scheffer <lou@cadence.com> wrote:

Louis wrote:

because the EIRP goes as N^2, so for constant EIRP, the power per module 

                           ^^^

Yes.  There is nothing magical.  The array with more modules is physically
bigger.  This is exactly the same as replacing a dish with another dish
with N times the area, and illuminating it with 1/N of the power.  The

                                                 ^^^
There's nothing magical about scaling gain with N, but you were 
claiming that the gain scales as N^2, if you use a discrete array
(see above quote - monospace font assumed).

If the elements added are the same power as the original, then
it does scale as N^2.  Take a single 1W omni-directional transmitter.  
It induces some voltage V in a far-away receiver.  Now add another 1W
transmitter in phase.  The received voltages add (linearity), so this
is the same as a 4W transmitter.  

 [ The total power decreases....]
But as 1/N, not as 1/N^2.

Agreed.  But the original argument is that Moore's law does not help.  It
does, though by 1/N, not 1/N^2.

Arrays are even more cost effective for transmitting than receiving.

Another factor is that it is difficult to justify a transmit only
antenna for this purpose.

Very true, though JPL is looking at transmit only facilities for a 
number of reasons.  First, the intermod between transmitted signals of
different frequencies causes problems for simultaneous reception. 
Second, some array antenna ideas (like the Allen Telescope Array) cannot
add a transmitter easily.

The need for cryogenic cooling is what drives the SKA to use dishes at

Interesting, because the SERENDIP receiver was swapped for one without
cryogenic cooling, but with a better noise figure, during S@H.

The are looking at this issue in gory detail.  Even though the uncooled
amps are good, they are even better at cryo temperatures, so you need
a lot less collecting area, which represents the bulk of the expense.

current technology, you can put about 100 small transmitters on a single

But the chip needs 100 times the power of a single tranmsitter.  

Yes, so the individual transmitters would need to be 10mw or so.  
This is roughly 1.5 volts into 75 ohms, so it's a good match for
CMOS, too.

Also the feed cables and all the joints on them probably represent the 
main source of failure modes (at the high end of frequency range
you might be able to use strip-line construction for the whole
10 by 10 sub-array, though - maybe a 1 metre square one for 21cm
is not too unrealistic).

I think you'd be better off at 3.5 cm like the Goldstone radar.  Then
each PC board is just 25 by 25 cm.

Also, whilst 1,000,000 elements would give the same feed point power
as Arecibo (although the optimal array, at 21cm) would be smaller than
Arecibo, although probably comparable to its effective area) you could
only produce 1 Arecibo equivalent beam for 1 watt per element; each
additional beam would mean adding another watt to the output of each
transmitter (more as input power).  Forming 100 beams would require
that your, 100 tranmitter, chip could handle 10kW (assume a fairly high
efficiency and one is still talking about a single chip dissipating as
much power as heat as a portable room heater.

No, you are much better off adding more modules and making the thing
physically bigger.  If you just make the array 10x bigger (keeping each
module the same) you get 100x the EIRP.  This can be split 100 ways
to form 100 beams.  (Of course each beam is only 1/10 the size, which
explains how 10x the input power can now create 100x the beams.) The
phase control gets more complex but not prohibitive.

chip on a single PC board.  There are only 10,000 of these, so any
reasonable MTBF will do, since one dead board won't cripple the system.

100,000 hrs is still multiple swapouts each day.  

Replacing 2.4 boards a day would require only a tiny fraction of the
current Arecibo staff.  Plus it can be at ground level, no climbing
required.

Incidentally the individual antennas couldn't be omnidirectional, as
such antennas are impossible to realise. They couldn't realistically be
ideal dipoles, for a ground based transmitter, as that could only be
achieved by putting the array over a 377 ohm per square ground plane,
which would waste half the power.  Putting dipoles above a real ground,
as well as producing a more directional per element beam, would produce
an array whose characteristics changed drastically after a rain shower.
I think, therefore, that you would need explicit reflectors/ground
plane, to make the behavour more repeatable and minimise losses.
(Anything except a metallic reflector, or a cryogenically cooled (<<12K)
377 ohm absorber would make the array unsuitable for reception.)

True, but there are a humongous number of papers this exact problem.
(Do an IEEE search for "printed antennas".  Most of them are some sort
of a patch or line over a ground plane.  They have about 6dbi gain 
at zenith, dropping to about 3db at 45 degrees.

    -Lou Scheffer