Optical Society of America. V. S.

Sudarshanam, "Minimum detectable phase shift in spectrum-analysis techniques of

optical interferometric vibration detection," Appl. Opt. 31, 5997-6002 (1992). doi: 10.1364/AO.31.005997

Minimum detectable phase shift in

spectrum-analysis techniques of optical
interferometric vibration detection

V. S. Sudarshanam

The minimum detectable phase shift indicated in recent experimental reports of new linear spectrum-
analysis techniques of optical interferometric vibration detection is established as the direct consequence
of the 1/f noise voltage in the system components. The dynamic range and inaccuracy predicted by the
simple theoretical model presented is in good agreement with experimental measurements. The
conclusions of the analysis are compared with experimental reports of heterodyne shot-noise-limited
optical systems. With this effective tool the generic class of spectrum-analysis techniques can be
analyzed and relatively weighed to assess the effect of noise. This analysis is applicable to optical
interferometry in general, although the experiments specificallyinvolved fiber-optic modulators.

Introduction temperature and pressure randomly affect the path

Optical interferometric methods have been widely difference between the two beams of the interferome-
used for the detection of vibration and dynamic ter, the detected signal undergoes fading.8 This
displacement, both in noncontact and invasive config- process results in randomly fluctuating intensity
urations. 1-6 Noncontact methods are generally im- amplitudes of the vibration harmonics. As the phase
plemented with the use of a laser and conventional shift is detected through an instantaneous intensity
optical elements such as beam splitters and lenses. measurement, changes in the source intensity and
Optical fibers have been incorporated to serve as frequency and the polarization states of the interfer-
simple light-transmission links to provide a flexibility ing beams also cause unpredictable variations.
of arrangement. Alternatively, the fiber itself could Several demodulation schemes have been devised
be utilized to measure the vibration through the for dynamic phase detection in both conventional and
stress-optic effect. In such an invasive technique fiber-optic interferometry. The heterodyne3' 4 and
the optical fiber is bonded onto the structure whose active homodyne9 approaches offer attractive ways to
vibration is to be measured.7 For both the noncon- eliminate signal fading. Nevertheless, passive homo-
tact and invasive methods, the interferometer could dyne techniques are simpler in that the problems
be of the Mach-Zehnder, Michelson, or Fabry-Perot associated with the Bragg cell in the heterodyne
arrangements. This wide class of interferometric approach and the electronic reset glitch in the active-
vibration-measurement methods has a common limi- feedback approach do not exist.5 Spectrum-analysis
tation in that the output intensity is a cosinusoidal techniques form a generic class of passive phase-
function of the phase shift induced by the vibration. 5 6 detection schemes that is particularly suited to simple
Hence a direct linear measurement of the phase shift, harmonicvibration. 1-8 Conventional spectrum-anal-
and therefore the vibration, is generally not possible. ysis methods consist of the J 1 (max), J(null), J 1 1J 2 ,
This is particularly true if the amplitude of vibration J 1 /J 3 , and the Bessel recurrence relation methods.3
is considerably large. As fluctuations in ambient
These methods require the interferometer phase
difference to be biased at 900,which is the quadrature
The author is with the Fiber and Electro Optics Research point about which the intensity is linear with the
Center, Bradley Department of Electrical Engineering, Virginia phase.
Polytechnic Institute and State University, Blacksburg, Virginia A recent report 4 utilized the asymptotic relation-
24061-0111. ship, J 2(x) + J+ 1 2 (x) = 2x, between the Bessel
Received 5 December 1991. functions of the first kind to determine the amplitude
0003-6935/92/285997-06$05.00/0. x of the dynamic phase shift. This method involves
© 1992 Optical Society of America. measuring the photodetector voltage amplitude at

1 October 1992 / Vol. 31, No. 28 / APPLIED OPTICS 5997

 CH 1 FFT +00001
the heterodyne carrier frequency with and without -10.097dB
000.00 mHz
1.0200 kHz
5.1000kHz
vibration and that of the first-order spectra. In
comparison, an elegant technique named the J ... J4
method that provided a self-consistent direct linear
phase readout in a no-feedback, no-phase bias condi- m
tion was reported.5 The phase measurement was
unaffected by random fluctuations in the phase differ-
ence, the polarization states, or the source intensity.
The need for such a linear technique arises in multi-
mode-fiber interferometric sensor systems that are
prone to intensity fluctuations.7 The J 1 ... J 4 meth-
od was further modified in another report10 to extend
its usefulness without disturbing the basic Bessel Frequency, kHz
recurrence relationship used.
Fig. 1. Frequency spectrum of V(t) for an input voltage of 57
Although the two implementations involved dif-
mVms at 1.02 kHz to the piezofilm. Analyzer bandwidth is 20
ferent sets of optical and electronic elements, the Hz. CH1 FFT, fast Fourier transform of channel 1.
behavior of the phase measurement as a function of
the drive voltage to the fiber-optic phase-modulator
test element was essentially the same. Whereas the
mathematical expressions used to derive the phase where A = 2v, B = 2vb, v is the photovoltage from the
shift from the Bessel relations were exact and did not fiber output assumed equal for both the fibers, b is the
impose any lower limit as such, practical measure- fringe contrast or visibility, J0 (x) are Bessel functions
ments were limited to a minimum detectable phase of the first kind, and 4O(t)is the random phase drift
shift (MDPS) of 0.1 rad in both cases. Utilizing a caused by fluctuations in ambient temperature and
phase modulator to upshift the contribution of the Jo
component to non-dc frequencies, I found that the pressure.
Jo ... J 2 method8 led to an improvement by an order The signal analyzer converts V(t) into an amplitude-
of magnitude for the MDPS. This paper develops a frequency spectrum V(f). Figure 1 shows a plot of
simple theoretical model that establishes the ob- V(f) in decibels for a voltage of 57 mVrmsapplied to
served lower bound as the direct consequence of the the piezofilm at 1.02 kHz and an analyzer bandwidth
1/f noise voltage seen at the output. In excellent of 20 Hz. The peaks at the fundamental and odd
agreement with earlier experimental reports, the harmonics alone are visible because for this plot the
theoretical model also confirms the improvement in dc phase bias between the interferometer arms was
the MDPS for the Jo ... J 2 method and predicts the adjusted to 900, which is the quadrature point.
observed large error as a signal phase shift of 2.4 rad The voltage noise floor is determined to vary as 1/f,
is reached. With this simple tool the generic class of with peaks at regular intervals corresponding to the
spectrum-analysis techniques can be relatively com- line frequency of 50 Hz. A similar 1/f noise-voltage
pared to determine the dynamic range and accuracy spectrum, or equivalently 1/f 2 power spectrum, has
of phase detection in the presence of noise. It should been reported for similar systems by several research-
be emphasized that this paper addresses optical inter- ers. 9"1 ",2 The main sources for this noise are the
ferometry in general, although the experiments specif-
ically involved fiber phase sensors and modulators. semiconductor junctions in the photodetector and the
electronic circuitry following it. It is well known
Experimental Reports and 1 /f Noise Voltage that the generation-recombination noise in semicon-
A Mach-Zehnder homodyne interferometer formed ductors bears a 1/f 2 power spectrum.13 However, it
from two single-mode fibers is considered here as an has also been shown that different samples of lasers
example.5 The light from a He-Ne laser is split, exhibit varying levels of 1/f 2 noise power, with care
guided within the fibers, and recombined to form a taken to avoid any ambiguities in the measurement."
fringe pattern. An in-line fiber phase modulator is Thus the laser source also contributes to the 1/f noise
configured by bonding a bare section of the fiber onto voltage.'4 For values of x below 2.6 rad, where
a piezofilm. The piezofilm is driven by a sinusoidal Jl(x) = J 2 (x), the value of the Bessel function falls off
excitation of amplitude x at angular frequency w. with higher orders. Hence, for low values of x, V3
It is this amplitude x that is to be measured by a and V4 would be closer to the noise floor at their
photodetector accessing the fringe pattern. The de- respective frequencies as compared with V, and V2.
tector output is amplified and fed to a signal analyzer.
56 Here the Vi's stand for the voltage amplitude of the
The instantaneous voltage V(t) can be written as '
ith frequency component in Eq. (1). As the J, ... J4
V(t) = A + B[(JO(x) - 2J 2 (x)cos 2ot method utilizes the components V, to V4, the error
introduced by the noise floor gets translated into
+ 2J4 (x)cos4wt)cosX(t) - (2J,(x) inaccuracy in the measurement of x, resulting in a
x cos wt - 2J 3 cos 3wt)sin ( 0 (t)], (1) MDPS.

5998 APPLIED OPTICS / Vol. 31, No. 28 / 1 October 1992

 Theoretical Model the plot of Ax versus x intersects the straight line
Based on the Bessel recurrence relation, the J, . .. J4 Ax = x (see Fig. 2). Thus an MDPS of 0.13 rad is
method uses the following expression for measuring x: seen for K = 0.000587, which is in good agreement
with the experimentally measured values. It may be
x2= 24V 2V 3 /(V, + V3 )(V2 + V4 ), (2) worthwhile to find the relative magnitude of Ax with
respect to x. If Ax was considered to be small, the
where Vi = 2BgJi(x), g = cos 4 o (or sin +O)for i even derivative dx could replace Ax; the plots of x' = x + dx
(or odd). As a result of the 1/f noise voltage, the and dx as a function of x are also shown in Fig. 2.
measured value x' would deviate from the expected It can be seen that the approximation Ax = dx leads
value x according to to a large discrepancy between experimental and
expected values when x is below 0.2 rad. Hence Eq.
(XI)2 = 24(J2 (x) + K/2)(J3 (x) + K/3)/(Jl(x) (3) provides the correct model for the observed exper-
imental results.
+ J3 (x) + 1.33K)(J2 (x) + J4 (x) + 0.75K), (3)
Analysis
where K = AV,/2B and AV1 is the noise voltage at the
fundamental frequency. The value of K can be With Eq. (3) as its basis, Fig. 3 shows the error Ax as a
determined accurately from the spectrum in Fig. 1. function of x for three different values of K. The
The average phase-shifting coefficient5 was deter- standard deviation of the phase measured over 50
minedfromtheJ,(max),J .. . J 4, Jo. . . J2 , J(null), trials for each point in the experimental plot of Ref. 5
and ramp-input methods. The input voltage of 57 did not exceed 3 mrad. From Fig. 3 one sees a
mVrms corresponded to an x of 0.13 rad. With this dynamic range of approximately 1 to 4 rad within that
value, the ratio of signal and noise voltages at 1.02 accuracy, which is in good agreement with Ref. 5.
kHz in Fig. 1, and 4 = 900, K was determined as A larger dynamic range is possible with a lesser K.
0.000587. The K would be lesser for a higher &), e.g., in going
Figure 2 shows a plot of x' and the error Ax = from 1.02 kHz as in Fig. 1 to 30 kHz, K would
(x' - x) using Eq. (3) as a function of x. The second- decrease by 30 dB. Alternatively, a 10-dB increase
order least-squares fit for the experimental plot of in the difference between the signal and the noise
phase shift versus input voltage in Fig. 2 of Ref. 5 was voltage at 1.02 kHz decreases K to 1/5469 and
determined to be y = (2.66x - 0.078x2 - 0.317) for thereby increases the dynamic range by 2 rad for an
phase shifts above 0.5 rad. The intercept of -0.317 accuracy of 3 mrad. It has been shown15 that such
rad was specific to the piezofilm configuration used, a reduction in the noise floor is in fact possible with
whereas in general such an intercept may not be seen the use of two detectors shifted in the optical phase by
for all modulators. Therefore, the intercept was 1800.
corrected for and the best fit was utilized to replot the The effect of random phase drift on the measure-
experimental points on the changed axes of the ment of x is now studied. It is seen from Eqs. (1) and
measured phase versus the expected phase, as shown (2) that if V2 and V4 are zero at + = 900, V and V 3
in Fig. 2 here. The MDPS is defined as the value of x would be zero at + = 0. Therefore it is necessary to
for which Ax = x, or equivalently the point at which avoid measurements near P = nr/2, where n is an
integer. As 40omoves away from these points the

3
X10-

ht
DI

0.8 4

0.7 9
3

0 0.6 e.. o 2 + o
+ 0

'~0.5 X 1

0.4- 0I t + + +++.
0.03 -1 0 o +

o +
0.2 o. -2 o +

0.1 -3

0* -4 o
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Phase shift x (rad) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Fig. 2. Measured phase shift x' and the error Ax = (x' - x) as a Phase x (rad)
function of the expected phase shift x. Dashed curve, Ax; solid Fig. 3. Error Ax in the measured phase shift as a function of the
curve, dx; dotted curve; Ax = x; asterisks, experimental values; phase shift x for different values of K. Asterisks, K = 1/1701;
open circles, x' = x + Ax; crosses, x' = x + dx. open circles, K = 1/3400; crosses, K = 1/5469.

1 October 1992 / Vol. 31, No. 28 / APPLIED OPTICS 5999

accuracy of the measurement becomes higher. To The numerals dividing K in Eqs. 6(a)-6(e) are based
see this phenomenon one needs to write the measured on the signal modulation frequency of 400 Hz and
4
phase x' as a function of o for a given x: reference modulation frequency of 2 kHz, with K
measured at 1.02 kHz as described earlier.
x')2 = 24(QJ2 + K/2)(PJ3 + K/3)/[(Jl + J3 )P In the followingdiscussionit is assumed that P =
Q = cos(,r/4). This assumption is based on the insen-
+ 1.33K][(J2 + J4 )Q + 0.75K], (4) sitivity of the phase measurement proved earlier,
within a certain inaccuracy, with respect to variations
where P = sin 4o and Q = cos 4O. Figure 4 shows the in 4O. It has been verified that the MDPS, as defined
measured phase x' as a function of kofor x = 2.405 earlier, is lower for higher K for any value of L.
rad for three values of K. It is seen that (a) for K = Figure 5 shows a plot of the signal phase-shift error
0, the measurement of x by the J, ... J 4 method is
AM as a function of the reference phase shift L for
independent of +o as indicated by the ideal Eq. (2), different values of the signal phase shift M with K =
and (b) the range of 410over which the error is below a 1/1701. It is seen that for the value of L = 1 rad
certain limit, say ±3 mrad, is larger for a lesser K and chosen for the experiments in Ref. 8, a range of M
is dependent on the value of x.
from approximately 0.01 to 2 is measurable within an
The Jo . . . J 2 method8 is now analyzed with the error of 2 mrad. The experimentally observed
model tested above. Basically this method consists inaccuracy was ±1 mrad for most of this range. As
in superimposing the signal phase shift of M rad, stated in Ref 8, Fig. 5 shows that M can be measured
which is to be measured, on a sinusoidal reference to within an accuracy of 1 mrad independent of the
phase modulation of L rad at a higher frequency. value of L, which can be chosen over a wide range.
The following experimental observations are verified The experimental fact that the error AM becomes
specifically: (a) the MDPS of 0.01 rad and (b) the large as the value of M reaches 2.4 rad is seen from
larger error near M = 2.4 rad. The measured phase Fig. 6. This phenomenon happens because Jo(M)
shift M' for the Jo ... J2 scheme can be written as (see would approach zero closer as M crosses 2 rad. In
Ref. 8 for ideal M) good agreement with the experimental value of ±1
2/(G' + T'),
M' = 2(G'H'I') 2J'"-1 (5) mrad for the inaccuracy, Fig. 6 shows an error of
similar magnitude for L = 1.5 rad over the range ofM
where of interest. However, the experimental value of 10
mrad for the MDPS was far higher than the 1 mrad
G'= 2BPJ,(L)Jo(M) + K/2, predicted by Fig. 6. This higher value was to be
expected because the function generator driving the
H' = 2BQJ,(L)J,(M) + K/2.4, piezofilm had a 2% instability for each range. Thus,
for the lowest range of voltage of 0.2 Vrmsused in Ref.
1' = 2BPJ2 (L)J,(M) + K/4.4, 8, the instability in the phase shift turns out to be 9
+K/4, mrad. It should be noted that the Jo ... J 2 method
2(L)JO(M)
J' = 2BQJ was relatively insensitive to the error caused by
T'= 2BPJ,(L)J2 (M) + K/2.8, random phase drifts away from the points of disconti-
nuity.
.

. , g |
2.408 nns . . , .

*.o..o. 0+
*~~~
0.015
2.407 L
f~ 0.01-
2.406k
*2. 40 ~ 1:: :. :0
00 0 : : 0
oO 0.005 -
I I 4 0 0 ; I
0 I 'f x
2.405- .' .. °° o oW@' + + + 0
-a 0 v
~~ ~~~~~~ 0 + + +

C 0.
i 2.4044 .
-0.01-
2.403 F
-0.015-

2.402 L- 0.5 - 1.5 2 25 § -0.02 )

0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 3.5

Phase drift (a ( x r rad) Reference phase L (rad)

Fig. 4. Measured phase shiftx' as a function of the random phase Fig. 5. Error AM in the signal phase shift measured by the Jo ...
shift kofor x = 2.405 rad. The horizontal line at x' = 2.405 rad J2 method as a function of the reference phase shift L, for different
corresponds to K = 0. Asterisks, K = 1/1701; open circles, K = values of signal phase shiftM; K = 1/1701. Exes, M = 0.01 rad;
1/5469. Gaps in the plots indicate discontinuities whose values asterisks, M = 0.3 rad; open circles, M = 1.5 rad; crosses, M = 2
have not been plotted here. rad.

6000 APPLIED OPTICS / Vol. 31, No. 28 / 1 October 1992

 U.UID, , , , , , C ,
The phase shift x can then be derived by (a) consider-
ing 3 either of the first sideband components at fl + 0
0.01 or f - frequencies, both of which would be
proportional to x, or (b) considering16 the power ratio
2 0.005 of the first sideband, P1, and the carrier frequency, P0 ,
using the relation
+ 000.t
++
0
Pl/Po = [Jl(x)/Jo(x)] 2 = x/2]2 . (9)
R
I + +

-0.005 + 0 0 For a heterodyne Mach-Zehnder shot-noise-lim-

-8
a ited interferometer, the signal photocurrent i and
ZZ ~~~~~~~~~-. . I
the shot-noise current i, can be written as3
-0.01

i = xbIB)' 12
, (10)
-0.015L
0 0.5 1 1.5 2 2.5 3 3.5 4
in,= (4eIB) , (11)
Signal phase M (rad)
Fig. 6. Error AM in the signal phase shift measured by the Jo ... where I is the photocurrent from each of the interfer-
J 2 method as a function of the signal phase shift M, for different ing beams, e is the electronic charge, and B is the
values of reference phase shift L; K = 1/1701. Asterisks, L = 0.3 detection bandwidth. The MDPS is obtained by
rad; open circles, L = 1.5 rad; crosses, L = 3.7 rad.
equating Eqs. (10) and (11) for a signal-to-noise ratio
(SNR) of 1, and assuming b = 1, as
The analysis presented above explains all the exper-
imentally observed characteristics of the J ... J 4 and MDPS = (8eB/I) 12 ,
the Jo ... J 2 methods. It should be pointed out that = 4(hvBlqP o)"2 ,
several expressions similar to Eqs. (2) and (5) could be (12)
utilized to determine the dynamic phase change. where v is the unshifted optical frequency, h is the
The noise model presented here predicts the MDPS Planck's constant, 'q is the detector quantum effi-
and the accuracy for each such technique and permits ciency, and Po is the laser power. In comparison
a judicious choice of the expression to be used for a with a homodyne Michelson interferometer, the fac-
given application. tor 4 in Eq. (12) arises because (a) the Mach-Zehnder
Comparison with Heterodyne Shot-Noise-Limited configuration has half the sensitivity of the Michelson
Systems
one, and (b) the heterodyne method utilizes only one
of the sidebands.3 Substituting the typical numeri-
It would be worthwhile to compare the MDPS achiev- cal values16 for the photomultiplier detection of P =
able from the homodyne J ... J 4 method with that of 17 nW, B = 1 kHz, -t = 5%, and hv = 3.1 x 10-19 Ws,
a heterodyne shot-noise-limited system. In a hetero- we see that the MDPS becomes 2.41 millirad. It is to
dyne interferometer, where the optical frequency in be emphasized that this MDPS holds only when the
one of the arms is shifted by Q, the instantaneous J, component alone is considered. The reason the
voltage V(t) can be written in a like manner to Eq. (1): MDPS (0.1 rad) for the homodyne J ... J 4 method is
much higher than this value is that whereas the noise
V(t) = A + B Jn(x)cos[(f - n)t + 0(t)]. has a 1/f behavior, the Bessel function magnitude at
2 (7) successive frequencies follows Eq. (8) so that the SNR
n=-x
for successive frequencies decreases rapidly for x <<
It is seen from Eq. (7) that the random phase term 1. However, phase detection by considering only the
10 (t), which now appears as the phase constant of J, component is susceptible to source-intensity
each V(f ), would not affect the measurement of x if changes and visibility fluctuations. It is in overcom-
one utilizes the digital Fourier transform of V(t) and ing this serious problem through linearization by the
reads only the magnitude of the voltage V f ) at each J, ... J4 method that the MDPSis sacrificed.
frequency. It should be pointed out that the approx- From the experimental results of Ref. 16 we have,
imation J + J 0 +,21 2 = 2/(Qrx)utilized in the hetero- for a displacement of 24 nm corresponding to a phase
dyne system of Ref. 4 holds good only for large values shift of 0.336 rad, a shot-noise level of 0.23 nm and a
of x. Moreover, the application of the noise analysis SNR of 52.5 dB for the Jo component. The corre-
presented here in the earlier sections to this expres- sponding value of K determined from these values is
sion has shown that the derivation of x from such a 0.0023. Utilizing this value and modifying the noise
relation results in an oscillatory error and is therefore analysis presented here under the assumption that
not considered here. For comparison of the MDPS shot noise is the same for all frequencies, one can see
here, assume that the signal phase shift x is very a MDPS of 0.0045 rad for the J ... Jo method of Ref.
small, x <<1, so that the following approximation for 16. This is in excellent agreement with the value of
the Bessel function can be used: 0.0032 rad calculated directly from the observed noise
floor level and = 14 jim-l in Ref. 16. In compari-
jn(X) Xn/(2n/n!). (8) son, the MDPS for the J ... J 4 method for a

1 October 1992 / Vol. 31, No. 28 / APPLIED OPTICS 6001

 5. V. S. Sudarshanam and K. Srinivasan, "Linear readout of
heterodyne shot-noise-limited system can be found by
dynamic phase change in a fiber optic homodyne interferome-
extrapolating the plot of Fig. 1 to higher frequencies ter," Opt. Lett. 14, 140-142 (1989).
where the assumption of shot noise is valid. As the 6. V. S. Sudarshanam and K. Srinivasan, "Universal dynamic
value of K dcreases by 30 dB in going over to phase calibration technique for fiber optic interferometric
frequencies higher than 30 kHz, the MDPS is deter- sensors and phase modulators," Opt. Lett. 14, 1287-1289
mined to be 0.065 rad. Thus the MDPS increases by (1989).
more than an order of magnitude when one measures 7. V. S. Sudarshanam, "Multimode fiber axial strain sensor
x from four components instead ofjust one. utilising end reflection interference," J. Mod. Opt. 39, 615-
624 (1992).
Conclusion analysis technique for
8. V. S. Sudarshanam, "New spectrum
A simple yet powerful theoretical model based on the interferometric vibration measurement," Opt. Commun. 88,
1/f noise voltage in system components has been 291-294 (1992).
presented to predict the minimum detectable phase 9. K. P. Koo and G. H. Sigel, Jr., "An electric field sensor utilising
shift, dynamic range, and accuracy of spectrum- a piezoelectric polyvinylidene fluoride (PVF2) film in a single
analysis techniques of interferometric phase detection. mode fiber interferometer," IEEE J. Quantum Electron.
The excellent agreement with experimental reports QE-18, 670-675 (1982).
indicates the versatility of the analysis for predicting 10. W. Jin, L. M. Zhang, D. Uttamchandani, and B. Culshaw,
the noise behavior of, and choice of optimum values "ModifiedJ,.. .J 4 method for linear readout of dynamic phase
for reference phase modulation in, such spectrum- changes in a fiber optic interferometer," in Fiber Optic Sensors
analysis techniques. IV, R. T. Kersten, Proc. Soc. Photo-Opt. Instrum. Eng. 1267,
153-157 (1990).
The author thanks R. 0. Claus for constant encour- 11. J. H. Cole, "Low frequency laser noise of several commercial
agement. This research was partially supported by lasers," Appl. Opt. 19, 1023-1025 (1980).
the Center for Innovative Technology, Virginia Poly- 12. A. Dandridge, A. B. Tveten, R. 0. Miles, and T. G. Giallorenzi,
technic Institute and State University. "Laser noise in fiber optic interferometer systems," Appl.
Phys. Lett. 37, 526-528 (1980).
References 13. F. N. Hooge, T. G. M. Kleinpenning, and L. K. J. Vandamme,
1. H. A. Deffarari, R. A. Darby, and F. A. Andrews, "Vibrational "Experimental studies of 1f noise," Rep. Prog. Phys. 44,
displacement and mode shape measurement by a laser interfer- 480-532 (1981).
ometer," J. Acoust. Soc. Am. 42, 982-990 (1967). 14. 0. Ishida, "Delayed self heterodyne measurement of laser
2. B. J. Pernick, "Self-consistent and direct reading laser homo- frequency fluctuations," IEEE J. Lightwave Technol. 9, 1528-
dyne measurement technique," Appl. Opt. 12,607-610 (1973).
1533 (1991).
3. Y. Ohtsuka, "Dynamic measurements of small displacements 15. A. Dandridge, "Noise reduction in fiber optic interferometer
by laser interferometry," Trans. Inst. Meas. Control 4, 115-
systems," Appl. Opt. 20, 2337-2339 (1981).
124 (1982).
16. R. Daendliker and J.-F. Willemin, "Measuring microvibrations
4. C. S. Vikram and T. E. McDevitt, "Simple spectrum analysis in
laser Doppler studies of sinusoidal vibrations," Opt. Eng. 28, by heterodyne speckle interferometry," Opt. Lett. 6, 165-167
922-925 (1989). (1981).

6002 APPLIED OPTICS / Vol. 31, No. 28 / 1 October 1992