Index Home About Blog
Date: 5 Feb 86 05:15:27 GMT
From: nike!caip!lll-crg!mordor!  (Jordan Kare)
Organization: S-1 Project, LLNL
Subject: Re: Phase Conjugate telescope

In article <> ST401385%BROWNVM.BITNET@WISCVM.ARPA writes:
>   >Why won't the phase conjugation technique work in reverse
>   >to build a large earth based telescope that removes the effects
>   >of atmospheric turbulence  ...  could make the Space Telescope
>   >obsolete.
>... There are two problems.  First... phase conjugation only works on
>monochromatic, coherent light (or at least light that is very nearly
>so).  More worrisome, though, is the fact that phase conjugation
>doesn't remove the distortion.  It antidistorts, so that repeating
>the passsage through the atmosphere cancels the distortion.

	Phase conjugation using non-linear optics (as discussed
in Sci. American recently) is (currently) limited to monochromatic
light and to some specific types of correction.  There is another
class of correction based on "adaptive optics": mirrors divided
into segments that can be moved (tilted) by electrical signals.
The "rubber mirror" project in the astrophysics group at Lawrence 
Berkeley Labs (where I got my degree) was an attempt to build
such a turbulence-correcting telescope.
	The size c of a "cell" of atmosphere over which starlight
is "coherent" (deflected the same way) is a few inches; the
"coherence time" over which such cells change is a few milliseconds
(and varies from place to place and night to night, just like 
telescope "seeing").  Thus, one needs (d/c)^2 mirror segments
to correct a telescope of size d -- a few tens to hundreds for
a good sized (say 4 meter) telescope -- and each segment must
be repositioned every few milliseconds.  The berkeley project
cheated by only worrying about 1 dimension, using 8 mirror segments
in a line to correct a modest (10 inch, I think) aperture in one
direction only.  The difference in path for different colors of light
is small as long as one is far from the horizon and not using too
broad a band, so the system works for white light.
	The problem is in figuring out where to move the mirrors.
It turns out that this is pretty easy if you are pointed at a bright
star; you just drive the mirrors one at a time to get the brightest
peak in the middle of the image.  The process converges to a "best" image
quite fast, and the electronics required are pretty modest.
	Unfortunately, one rapidly runs out of photons if the "reference"
star is dim (limit is about 8th magnitude, independent of
just about everything one can control, like aperture size), and the 
"field of view" for which the correction is good is very small -- and
there just aren't many things worth looking at that are that close
in the sky to 8th magnitude stars.  So the rubber mirror project got
dropped after proving (by resolving a close binary star) that the
principle worked.  So far, the problems appear to be fundamental.
	If you could supply the reference light, it would
indeed be possible to make diffraction-limited ground based telescopes
(possible, mind you, doesn't mean practical).  But remember that
anything in orbit (even geosync) would move 
rapidly relative to the fixed stars, so
you can't put your beacon on a satellite even if you could afford to.
Meanwhile, we'll just have to live with ten-meter light buckets
and 2000x2000 CCD detectors doing speckle imaging while we wait
(:-() for the Space Telescope.
				Jordin Kare

Date: 24 May 93 17:38:02 GMT
From: Dani Eder <>
Subject: Adaptive Optics (was Space Marketing)
Newsgroups: sci.astro,

In article <1993May21.170848.18186@ucsu.Colorado.EDU> fcrary@ucsu.Colorado.EDU (Frank Crary) writes:

>Exactly how small? Unless I'm very much mistaken, it can't be arbitrarily
>small and the size scales with the size of the primary. 5 - 10cm
>is typical of the current ~1m telescopes, but wouldn't you need
>something more like 50 cm for a 5-m primary?
>Does this mean you can have an arbitrarily small mirror, so long as
>you could also make the individual elements arbitrarily small? I'm
>not sure how you could: Can you actuate a 2mm mirror accurately
>enough. I assume the distance it would have to be moved would
>similarly be scaled down. Does that mean it has to be accurately
>moved distances on the order of a micron? 
>But how quickly can these hundreds of element arrays be adjusted? 
>The required time scale increases faster than linearly with 
>the observed frequency. All in all, I think my original remark
>was reasonably accurate: Large (~5m) adaptive telescopes working
>in the visible are still a long way away.

Last year I was involved with a NASA study on laser power beaming
to orbiting spacecraft (I was working the spacecraft end).  As
a result, I got to sit in on presentations on the current state
of the art in wavefront control.  The coherence scale of the
atmosphere is on the order of 6 cm on the ground.  The deformable
mirror can be any mirror in your optical train, but the
preferred location is the last surface before your imaging
electronics (CCD).  This way you can not only adjust for atmospheric
distortion, but also for any mechanical distortion caused by
the primary and secondary mirrors due to temperature and gravity.

The deformable mirror can be small, limited by diffraction onto
your imaging electronics, and by the ability to fabricate small
actuators.  The NASA program was looking at hexagonal segments
2 cm across for the deformable mirror.  Each segment has 3
peizo-electric acuators.  Quartz has the property of changing
shape under electric fields.  A sufficiently long rod of quartz
can provide the several tens of wavelengths of movement needed
to handle the distortions.  The positioning resolution acheived
is about 4 nanometers, which gives a tilt resolution of .2
microradian for a 2cm segment (or 0.04 arc second).

The 2 cm segment size was chosen for manufacturing reasons.
The concept is to have a control chip for each segment mounted
on the backplane, and driving the peizoelectric actuators.
The segment is mounted on the front side, and the actuators are
mounted to the back of the segment and to a structual backing.

The whole mirror has to be driven at about 200Hz, since that is
the characteristic time for the atmospheric distortions (it is
a function of wind velocity in the atmosphere, and the turbulence
it generates.

The current limiting factor on the adaptived optics is the computer
power to analyze the distorted guide star image and calculate how
to command the mirror segments.  For an 8m telescope with 0.1
arc second resolution, it would (in 1992) take 30 racks full
of signal processing chips to do solutions at 200 Hz (those are
19 inch wide by 6 foot high racks).  Where n is the number of
segments, the processing to solve the distortion goes as n*ln(n),
and n goes as diameter squared.  

Today, you can back down by restricting to a smaller mirror, or
by not solving as fast, thus losing some resolution.  But you
can in the lab today show real performance gain in the visible
for dozens to a few hundred elements, good for up to about a
1 meter telescope.  Hopefully in a decade or two the cost of the
processing chips will come down to where you can afford to drive
a big telescope.

Dani Eder

Dani Eder/Meridian Investment Company/(205)464-2697(w)/232-7467(h)/
Rt.1, Box 188-2, Athens AL 35611/Location: 34deg 37' N 86deg 43' W +100m alt.

Index Home About Blog