STREHL RATIO WITH ANISOPLANATISM
For a perfect correction the paths of the beacon signal and the
imaging or projected laser should be the same. In general, this
is not possible to achieve, and there is a degradation in
performance caused by time delays, displacement of the two paths
by translation and angle, and differences in wavelength of the
beacon and the measurement or projecting systems. The effects of
displacement, angular mispointing, time delay, and atmospheric
dispersion can each be treated as an anisoplanatic effect. In
fact, if all the effects are present simultaneously, they can be
combined to get a total offset of the measurement from the imaging
paths. In this section the effect of a general displacement on
the Strehl ratio is determined.
The Strehl ratio (SR) for a circular aperture [#!7!#] from the
Huygens–Fresnel approximation is
The integral is over a circular aperture of
unit radius,
(
α) is the structure
function, and
K(α) is a factor times the optical
transfer function given by
K(α) = ![$\displaystyle {{16} \over
\pi}$](img9.png) cos-1(α)-α 1-α2![$\displaystyle \left.\vphantom{ {1-\alpha ^2}
}\right)^{{1/ 2}}_{}$](img12.png) U(1 - α), |
|
|
(2) |
where
U
x
is the unit step function defined as
U(x) |
= |
1 for x≥0 , |
|
U(x) |
= |
0 for x < 0 . |
(3) |
To find the Strehl ratio, one must first determine the structure
function. It was found by Fried[#!4!#] for angular
anisoplanatism. If the source is collimated and a general
displacement is introduced, his expression for a wave propagating
from ground to space becomes
(αD) |
= |
2(2.91) k02 dz Cn2(z) (αD)5/3+d5/3(z) |
|
|
|
- 12 ![$\displaystyle \left\vert\vphantom{ {{\bbox \alpha} \kern 1ptD+{\bbox
d}(z)\,} }\right.$](img21.png) αD+ d(z) - 12 αD- d(z) ![$\displaystyle \left.\vphantom{
{\,{\bbox \alpha} \kern 1ptD-{\bbox d}(z)\,} }\right\vert^{{5 / 3}}_{}$](img25.png) , |
(4) |
where
Cn2(z) is the turbulence
strength as a function of altitude;
k0 = 2π/λ, where λ is the wavelength of operation; D is the
aperture diameter; and
d(z) is the vector displacement
of the two paths.
The sums of the terms in brackets almost cancel, thus causing
difficulties if one tries to evaluate this integral numerically.
The terms in the absolute-value sign are equal to
αD± d(z) = ![$\displaystyle \left[\vphantom{ {\left( {\alpha \kern 1ptD}
\right)^2\pm 2\alpha \kern 1ptD\,d(z)\cos \left( \varphi
\right)+d^2(z)} }\right.$](img29.png) αD ±2αD d (z)cos![$\displaystyle \left(\vphantom{ \varphi
}\right.$](img32.png) ![$\displaystyle \varphi$](img33.png) +d2(z) , |
|
|
(5) |
where is the angle
between
α and
d(z) . This
expression can be simplified and the numerical difficulties can be
eliminated by using Gegenbauer polynomials.[#!8!#] Their
generating function is
1-2ax+a2 = Cpλ(x) ap. |
|
|
(6) |
These functions are sometimes referred
to as ultraspherical functions because they are a generalization of
the Legendre polynomials Pn(t) , whose generating function is
The
Gegenbauer polynomials with the cosine of a variable as the
argument are given in Eq. (8.934 #2) of Ref. 8 and
can be rewritten as
where -
Γ
x
is
the gamma function. A particular Gegenbauer polynomial that is
required is
C2-5/6 cos( ) = 56 1- 13cos2![$\displaystyle \left(\vphantom{ \varphi
}\right.$](img32.png) ![$\displaystyle \varphi$](img33.png) ![$\displaystyle \left.\vphantom{ \varphi
}\right)$](img34.png) . |
|
|
(9) |
For
αD > d (z) , the terms in the
structure function can be expanded in Gegenbauer polynomials. The
zeroth- and all odd-order terms cancel. When the summation index
is changed by the substitution
p→2p the result is
(αD) = 2(2.91) k02 dz Cn2(z)![$\displaystyle \left\{\vphantom{ {d^{5/ 3}(z)- (\alpha \kern 1ptD)^{5/
3}\sum\li...
...)} \right]}\,\left[ {{{d(z)} \over {\alpha \kern
1ptD}}} \right]^{2p}} }\right.$](img51.png) ![$\displaystyle {d^{5/ 3}(z)- (\alpha \kern 1ptD)^{5/
3}\sum\limits_{p=1}^\infty ...
...hi \right)} \right]}\,\left[ {{{d(z)} \over {\alpha \kern
1ptD}}} \right]^{2p}}$](img52.png) . |
|
|
(10) |
It is this canceling
of the first two terms of the power series that would cause
numerical difficulties. Define a distance moment as
dm≡2.91 k02 dz Cn2(z) dm(z) |
|
|
(11) |
and a phase variance as
σ 2 = d5/3. |
|
|
(12) |
Unlike the calculation for Strehl ratio for
uncorrected turbulence and for corrected turbulence with tilt
jitter, an exact analytical solution cannot be found for
anisoplanatism. Fortunately, for adaptive-optics systems, the
Strehl ratio should be fairly high by design, which requires the
structure function to be small. This assumption allows one to
retain only the first term of the Gegenbauer expansion to give
(αD) = 2σ 2 - 2x, |
|
|
(13) |
where
x = d2 1- 13cos2![$\displaystyle \left(\vphantom{ \varphi
}\right.$](img32.png) ![$\displaystyle \varphi$](img33.png) ![$\displaystyle \left.\vphantom{ \varphi
}\right)$](img34.png) ![$\displaystyle \left.\vphantom{ {1- {\textstyle{
\slantfrac{1}{3}}}\cos ^2\left( \varphi \right)} }\right]$](img49.png) 56(αD)-1/3. |
|
|
(14) |
We justify this single-term approximation below by showing that it
produces a result close to the exact result.
...
The
Strehl ratio with the six term approximation is
If just
the first term in the last parenthetical expression is retained,
the result is equivalent to the extended Maréchal
approximation. It is shown below that the six-term approximation
is best for aperture sizes normally encountered. The angle
integral for the nth term, after use of the binomial theorem, is
proportional to
Φ(n) = ![$\displaystyle {1 \over {2\pi }}$](img1.png) d 1- 13cos2![$\displaystyle \left(\vphantom{ \varphi
}\right.$](img32.png) ![$\displaystyle \varphi$](img33.png) ![$\displaystyle \left.\vphantom{ \varphi
}\right)$](img34.png) = ![$\displaystyle {1 \over {2\pi }}$](img1.png) ![$\displaystyle \sum\limits_{{m=0}}^{n}$](img63.png) ![$\displaystyle \left(\vphantom{ \begin{array}{c} n \\
n-m\end{array} }\right.$](img64.png) ![$\displaystyle \begin{array}{c} n \\
n-m\end{array}$](img65.png) 3-m d cos2m![$\displaystyle \left(\vphantom{ \varphi
}\right.$](img32.png) ![$\displaystyle \varphi$](img33.png) , |
|
|
(16) |
where
Equation (4.641 # 4) in
Gradshteyn and Ryzhik[#!8!#] is
where
(2m - 1)!! |
= |
(2m - 1)(2m - 3)…(3)(1), |
(19) |
(2m)!! |
= |
(2m)(2m - 2)…(4)(2). |
(20) |
With these
relations, the angle integral is equal to
The values of interest to us are
Φ(0) = 1,
Φ(1) = 0.8333,
Φ(2) = 0.7083,
Φ(3) = 0.6134,
Φ(4) = 0.5404, and
Φ(5) = 0.4836. The
aperture integration for the nth term is proportional to
Y(n) = dα α1-n/3K(α). |
|
|
(22) |
This is a
generalization of the integral evaluated by Tatarski in Sec. 55,
Eq. (22) of Ref. 9. Its value is
The values of interest to us are
Y
0
= 1,
Y
1
= 1.402,
Y
2
= 2.087,
Y
3
= 3.396,
Y
4
= 6.419, and
Y
5
= 16.94. With these values for the integral, the Strehl
ratio approximation is
SR (1+0.9736 E+0.5133 E2+0.2009 E3+0.0697 E4+0.02744 E5)exp(-σ 2), |
|
|
(24) |
where
E = . |
|
|
(25) |
...
There is an error made in using this approximation for the central
part of the aperture that increases with each term in the
approximation. One has to determine whether this error is less
than or greater than the increased accuracy achieved over the
remainder of the aperture by using additional series terms. To
resolve these uncertainties, I compared the Strehl ratio, using
various numbers of terms, with exact calculations.
I calculated the Strehl ratio numerically for the case in which
the displacement does not vary with propagation distance. In
Fig.
are plotted the exact Strehl ratio versus
displacement for the Hufnagel–Valley 21 (HV-21) model of
turbulence[#!10!#,#!11!#,#!12!#] and the Strehl ratio from relation (24)
for
D/ro = 1, with only the unity term in parenthesis
(extended Marechal approximation) and with different numbers of
terms in the parenthesis.
...