Академический Документы
Профессиональный Документы
Культура Документы
interferometry
We have developed an analytical model for the design and optimization of common-path interferometers
CPIs based on spatial filtering. We describe the mathematical analysis in detail and show how its
application to the optimization of a range of different CPIs results in the development of a graphical
framework to characterize quantitatively CPI performance. A detailed analytical treatment of the effect
of curvature in the synthetic reference wave is undertaken. We show that it is possible to improve the
linearity and fringe accuracy of certain standard interferometers by a modification of the Fourier filter,
and we propose and analyze a dual CPI system for the unambiguous mapping of phase to intensity over
the complete input phase range. 2001 Optical Society of America
OCIS codes: 070.2580, 070.6110, 100.2650, 100.5070, 120.3180, 120.5050.
choice of the Fourier filter parameters.
r21
x2y2r
expix, ydxdy Fig. 2. Schematic representation of the relationship between the
profile of the input aperture-diffracted Airy function and the spa-
tial profile of a generic Fourier filter see Fig. 1. The term is
expi , defined as the ratio of the radii of the filter R1 and the mainlobe of
the Airy function R2.
. (8)
The generally complex-valued and object-dependent
We can thus express in terms of r and fr such
term that corresponds to the amplitude of the fo-
that
cused light plays a significant role in the expression
for the interference pattern described by Eq. 7. Re- R1R2 0.611rfr, (10)
ferring to the discussion in Section 1, we are now able
to confirm that the frequent assumptionthat the where the factor of 0.61 arises from the radial dis-
amplitude of the focused light is approximately equal tance to the first zero crossing of the Airy function
to the first term of the Taylor expansion in approxi- corresponding to half of the Airy mainlobe factor of
mation 1 generally can result in misleading inter- 1.22.1 If we make this substitution in Eq. 9 and
pretations of the interferograms generated at the CPI then perform a series expansion in r, we obtain the
output. following expression for the SRW:
Of similar importance in the analysis of Eq. 7 is
the term gr that describes the spatial profile of the gr 1 J01.22 0.612J21.22
SRW, diffracted from the aperture formed by the on- rr2 0.61342J31.22
axis centered filtering region. It is the interference
between this SRW term, carrying the information 0.61J41.22rr4 . . . . (11)
about the filtering parameters, and the imaged phase In this expansion, the SRW is expressed in radial
object that generates the output interferogram. It is coordinates normalized to the radius of the imaged
important to obtain an accurate description for the input aperture. This can be scaled easily to allow for
SRW and thereby an accurate derivation for Eq. 7. a magnification within the imaging system, although
The zero-order Hankel transform1 followed by a se- for the remainder of our analysis a direct imaging
ries expansion in the spatial dimension r is used here operation is assumed. From Eq. 11 it is apparent
to describe the SRW. This is a relatively simple ap- that the SRW will change as a function of the radius
proach, which to the best of our knowledge21 has not of the central filtering region. In addition, it is clear
previously been applied to this problem. that the SRW profile is not necessarily flat over the
For a circular input aperture with radius r, we system output aperture. This is an important, yet
can describe the radius of the corresponding central often neglected, factor when one is determining the
phase-shifting region of the Fourier filter character- performance of a CPI.
ized by the parameters B and in terms of a radial In Fig. 3 we plot the input normalized SRW ampli-
spatial-frequency range fr. Thus we can obtain the tude as a function of the output radius coordinate r
following expression for the SRW by use of the zero- normalized to the system aperture radius r for a
order Hankel transform: range of different values. It can be seen that as
increases so does the strength of the SRW, which
fr
gr 2r J12rfrJ02rfrdfr. (9) from qualitative arguments can be expected see Fig.
0
2. However, the curvature also increases with in-
creasing , thus distorting the wave-front profile of
To simplify the analysis, we introduce the term , the SRW. From the point of view of optimizing a
which explicitly relates the radius of the central fil- CPI, we want the curvature of gr to be negligible
tering region R1 to the radius of the mainlobe of the over a sufficiently large spatial region of the system
Airy function R2 resulting from the Fourier trans- aperture centered around r 0. First, should be
form of the circular input aperture alone. This re- chosen so that gr 1, corresponding to the first
lationship defining is shown graphically in Fig. 2. zero crossing of the Bessel function J01.22, where
These constraints arise from the adoption of Ander- C 2cosC for C 2; 32,
sons maximum irradiance criterion.18 This implies
that we want to minimize unnecessary absorption of C 0 for C2; 32. (19)
light in the Fourier filter, which reduces both the
irradiance and the signal-to-noise ratio in the CPI The development of this complex filter space plot
output fringe measurements. makes it extremely simple to compare different filter
Any given filter can be defined explicitly by a given parameter selections that have been proposed and
value of the two parameters C and C. Therefore independently dealt with in the scientific literature.
we can use a single plot to display graphically the In Fig. 4 we superimposed the operating points for
location of a given filter within the complex filter some of the different CPI filter types that are com-
space. Such a plot is shown in Fig. 4 in which we monly used. The filter type, parameters, and corre-
plot the magnitude of the combined filter parameter sponding labels and references are summarized in
C versus its phase C. There are different families Table 1. The filter space plot provides not only a
of operating curves in this complex filter space, each convenient method of comparing different CPI archi-
of which can be traced out when a term such as BA1 tectures, but also forms a useful basis for exploring
is kept constant while is varied or vice versa these the effectiveness of new CPI configurations.
form the gridlike structure in Fig. 4. When the op- Referring to Fig. 4, it can be seen that many of the
erating curves for C are plotted in this way, it is filter types used in CPIs lie on or close to the lossless
relatively simple to identify particular operating re- operating curve defined by Eq. 19. The arrows on
gimes for different classes of filters. For example, the plot indicate the general regime in the complex
we are particularly interested in the operating curve filter space in which a given filter type can be oper-
for a lossless Fourier filter, a filter in which BA1 ated.
1, because this is a class of filter for which optical To the best of our knowledge, this identification
throughput is maximized. The lossless operating and labeling of known cases from the literature in a
curve is shown as the bold curve in Fig. 4. We can single general filter space representation has not
derive the expression for the shape of the lossless been demonstrated previously. We believe that it
operating curve by using the identity provides a useful basis for understanding the inher-
ent differences or similarities between existing filter-
expi 1 2sin2expi 2, (18) ing methods. We can use the complex filter space
plot as a framework on which to optimize the visibil- . For example, we can choose to analyze phase
ity and the irradiance of a CPI. perturbations modeled by a uniform random distri-
bution of phasors. In this case, the two terms and
A. Optimization of the Visibility and Peak Irradiance are related by the absolute value of a sinc function
Returning to the CPI optimization criteria of Ander- in the following way30:
son,18 we can denote the maximum and minimum
intensity levels at the interferometric output as Imax
and Imin for any given CPI. The visibility V is ex- uniform sinc 2. (24)
pressed in terms of the output intensity minima and
maxima such that Using this expression to substitute for in Eqs. 21
and 22 we can produce expressions for the visibility
Imax Imin
V . (20) and the peak irradiance, which depend only on the
Imax Imin input phase modulation depth and the combined fil-
ter parameter. After we derive expressions for the
The peak irradiance is defined as the maximum in
visibility and the peak irradiance, it is instructive to
the intensity at the output of the CPI and is therefore
compare their variation as a function of the phase
simply characterized by the term Imax.
modulation depth. We are then able to compare sys-
For any given input phase perturbation with a dy-
tematically the performance of a generic CPI system
namic range of , we can derive an expression for
for different ranges of phase perturbations at the
the visibility and the peak irradiance by combining
input. Using the complex filter space plot as a
Eqs. 15 and 20:
framework, we are able to display the variation in the
V max cos min cosK C1 K C visibility and the peak irradiance at the CPI output
as a function of the complex components C and C of
max cos min cos1, (21) the combined filter parameter. By using a pair of
complex filter space plots, we can thus show the op-
Imax A21 K C2 2K Cmax cos, (22) timum filter type for a given phase perturbation to
where maximize the visibility or the peak irradiance. In
Fig. 5 we show a set of contour plot pairs for the
min cos mincos C, visibility and the peak irradiance with a range of
input phase distribution depths from a 8 to d
max cos maxcos C, 2; 2. 32. These are shown in the form of normalized
(23) gray-scale contour plots in which the maximum value
is represented by the darkest region. All four sets of
These expressions for the visibility and the peak ir- complex filter space plots cover the same range for
radiance can be simplified in some special cases, the parameters C and C, from 0 2 rad and 0 4,
which we discuss below and in Section 5. However, respectively. The operating curve for the lossless
first let us consider an analysis of the general case as filter given by Eq. 19 is shown on all the plots. It is
given in the full versions of Eqs. 2123. It is ap- evident in Fig. 5a that a small does not provide
parent that a general analysis over the entire domain much overlap between regions of high visibility and
covered by the complex filter space plot of Fig. 4 is strong peak irradiance. For the larger values of ,
complicated by the appearance of the variables such as those in Figs. 5b and 5c, an overlap can be
and in the expressions for the visibility and the found between the regions of high visibility and peak
peak irradiance. The variable K is held fixed K irradiance. However, there is no overlap between
12 in this analysis for high fringe accuracy, as was regions in which these values are maximized, illus-
discussed in Section 3. Instead of arbitrarily choos- trating that there will always be a trade-off between
ing combinations of and before an optimization the absolute optimization of these two terms. It is of
analysis for the filter parameters, we model the input interest to note that, as the value of increases
phase perturbations by a random distribution of pha- further to 32 see Fig. 5d and beyond, then the
sors over the phase range generating a spatial overlap between high values of the visibility and the
average or wave-front Strehl ratio29 corresponding to peak irradiance diminishes.
The upward shift toward higher values of C and the Imax A21 K C2 2K C. (28)
broadening of the maximum visibility region in the
filter space domain can be understood when we derive Another interesting observation related to the contour
the expression for C in the case of optimal visibility: plots in Fig. 5 is that C 2 2 provides a good
compromise filter in most cases, both for obtaining high
CV1 K 1 expi i , 2; 2. visibility and for strong peak irradiance. Moreover, this
(25) corresponds to a lossless phase-only filter with and
A B 1 as is evident from Eq. 14 or Eq. 18. Using
In this case, we use the fact that the minimum in- numerical simulations, Anderson18 also found the loss-
tensity will have to be zero within the interval to less -phase filter to be a good overall compromise for a
obtain optimal visibility. When we combine Eq. 21 wide range of input phase disturbances.
with Eq. 25, the reason for the shift and the broad-
ening of the maximum visibility region in Fig. 5 with B. Optimal Unambiguous Phase-to-Intensity Mapping
increasing becomes clear. In the limiting range For phase perturbations in the range we have
where 3 2, we can disregard the phasor part of the possibility of optimizing both the visibility and
Eq. 25 and we then find that the irradiance while simultaneously obtaining a
monotonically increasing mapping of the input phase
CV1 CV1 K 1, 2. (26) to the output intensity. Unfortunately this excludes
We find the corresponding expressions for the visibil- relatively simple filters such as the lossless -phase
ity and the peak irradiance in this limiting case to be filter that was convenient in the above analysis in
which the unambiguous mapping of phase to inten-
V 2K C1 K C1, (27) sity was not the primary concern.
To obtain the required monotonic increase in the This corresponds to a filter consisting of the following
phase-to-intensity mapping we must map the respec- amplitude and phase values obtained by use of Eq.
tive minima and maxima of the input phase distur- 16:
bance to the corresponding limits in the output
intensity distribution. We therefore obtain the fol- BA1 1 2K 1 cos 2 K 212,
lowing expression for the visibility:
sin1BA1K 1 sin 2. (32)
V I 2 I 2I 2 I 21
Another interesting constraint enforced on the filter
2 sin 2sinCK C1 K C to ensure an unambiguous phase-to-intensity map-
2 cos 2cosC1, (29) ping is that the available phase space for C is reduced
as increases, approaching its maximum value of
where the corresponding peak irradiance is given by . In fact, it can be shown that the range of C that we
can use for a given is
Imax I 2 A2expi 2 iC K C2. (30)
C 2; 2, (33)
As in the analysis of Subsection 4.A, we can again
derive an analytic expression for the combined single where the single value C 2 is obtained for the
filter parameter C that generates optimal visibility largest possible unambiguous range .
for the unambiguous phase-to-intensity mapping The constraints on the values that C can take
over the range : given by expression 33 can be identified easily on
the complex filter space plots for the visibility and the
CV1 K 1 expi i 2. (31) irradiance. In Fig. 6 we show sets of such plots in
K 12 f C 2 expi2. (39)
This corresponds to the following choice of the basic
filter parameters:
BA1 5 f A 15, B 1,
sin 25.
1
(40)
and the resulting intensity distribution Fig. 7. Graphical phasor chart analysis for the generalized Hen-
ning method. The zero point of the quadratic intensity scale I is
Ix, yHenning 251 sin x, y. (41) fixed at the point where 32 on the unit phase circle. The
K12 intersection of I with the unit circle directly yields the output
intensity value, which corresponds to the input phase of the CPI.
This result can be improved further if we also take This mapping is approximately linear for the range 3 about the
into consideration the correct input-dependent aver- points 0 and indicated by the dashed curves.
age value of 1 so that we obtain
C K 1 expi2. (42) brief summary of the phasor chart and its use are
given in Appendix A.
The filter parameters now have the values
In Fig. 7 it can be seen that, for the phasor chart
BA1 1 K 212 f A 1 K 212, representation of the generalized Henning method,
we fix the zero point for the quadratic intensity scale
B 1, sin11 K 212. (43) I at the point 32 on the unit circle corre-
sponding to the filter parameters given by Eqs. 42
which generates the resulting intensity distribution and 43. It is then possible to directly read out
corresponding pairs of values for the phase-to-
I x, ygeneralized 21 K 211 sin x, y. intensity mapping in the generalized Henning case
Henning
(44) as is illustrated in Fig. 7. Using the phasor chart,
we can verify graphically that the range for linear
We call this technique with improved linearity the
mapping is approximately 3 around 0 indi-
generalized Henning method. In practice, it would
cated by the dashed curve on the chart. An addi-
require an adaptive Fourier filter controlled by a
tional linear mapping region of 3 around is
sensing unit that continuously measures the
also indicated. We can confirm the linearity ob-
strength of the focused light intensity, 2, and pro-
tained in the phase-to-intensity mapping in the pha-
vides this as feedback to determine the selection of
sor chart by considering the mapping as the graphical
the parameters in the adaptive filter. With such a
intersection points of the quadratic intensity scale
filter in place, we would be guaranteed optimal lin-
with the dashed curve. By simple trigonometry we
earity for any given phase perturbation covering a
find that the squared length L2, measured along the
range up to 3, which is the approximately lin-
intensity scale intersecting the unit circle at any
ear region of the term 1 sin .
given point, is given by
From Eq. 42 it can be seen that the inclusion of
the C-parameter solutions for the generalized Hen-
ning method would correspond graphically to a trac- L2 cos2 sin 12 1 sin . (45)
ing of a vertical line parallel to the C axis and fixed
at C 2 in the filter space plot of Fig. 4. It This expression has exactly the same functional form
should be noted that the filter operating points form- as obtained in Eq. 44.
ing the grid in Fig. 4 have an irregular distribution,
which explains why an apparently simple C solution B. Linear Phase-to-Intensity Mapping over the Phase
generates the need for such an adaptive filter in Unity Circle
which one component is fixed and the other two are Now that we have demonstrated that it is theoreti-
functions of , as shown in Eq. 43. cally possible to improve the linear range of the
It is possible to illustrate the operating regime for phase-to-intensity mapping for the CPI through a
the generalized Henning phase-contrast method by generalization of the Henning method as described in
use of a two-dimensional phasor chart. This chart Subsection 5.A, it is interesting to contemplate how
was developed originally as a graphical method for linearity might be extended to generate a mapping
the analysis of CPI systems.15 It has been applied for a larger range of input phase values than 3.
subsequently to the optimization of systems based on One possible solution for the extension of a linear
the generalized phase-contrast technique, including phase-to-intensity mapping regime would be the ad-
among others the analysis of array generators.16 A dition of a second CPI working in conjunction with