Original

article J . Optics (Paris), 1988, vol. 19, no 3, pp. 101-115

THE RING LASER GYRO

M. FAUCHEUX, D. FAYOUX, J. J. ROLAND

KEY WORDS : MOTSCLES

gyrometer,laser gyrometre,laser

Le gyrometre laser

SUMMARY : In this review paper, we recall the basic principles RESUME : Dans cet article de revuenous rappelons les principes
and advantages of the ring laser gyro (RLG) which is the most et les avantages du gyrometre laser en anneau (RLG) qui, parmi
widely developed of all optical gyroscopes based on the Sagnac tous les gyrometres optiques fondes sur I’effet Sagnac, fait I’objet
effect. de developpements et d’applications importants.
The main error sources which can affect RLG performances are Les principaux phtnomenes physiques qui peuvent limiter les
discussed, and among them gas flows in the active medium and performances du RLG sont exposes: notamment, les ecoulements
frequency synchronisation of counterpropagating waves or “lock- de fluides dans le milieuamplificateuret les couplagesentre
in”. ondes qui conduisent a la << zone aveugle n :
Particularattention is given to lockingphenomenon : the Uneattention
particulikre
est
donnee au
phenomene de
predominantinfluence of mirrorsbackscattering is emphasized couplageentreondes : I’influence prtdominantedesretrodiffu-
and means devised to overcome this problem are explained. sionsdesmiroirsformant la caviteestmise en evidenceet les
Some important design rules used in the RLG construction are differents moyens mis en Oeuvre pour surmonter ce probleme sont
defined.Tworepresentativeexamples of RLGfordifferent abordes.
applications are described including the triaxial RLG in a single En fonction des
applications
et
desperformancesvisees,
opticalblock. quelques regles de conception technologique sont definies. Deux
exemplesdeRLGsontdecritsdont le gyrometrelasertriaxial
((monobloc n.

-long
lifetime
a without
maintenance :
I. - INTRODUCTION - BASIC PRINCIPLES
thousands of hours,
- asmall size and mass,
The ideal sensor for inertial rotation would have - a low cost.. .
different characteristics like : For alongtimeallcommercialinertial rotation
- anoutput giving angleincrementswith high sensorshavebeen mechanicalgyroscopes. These
resolutionand accuracy(forinstance 1 arcsec. gyroscopes which use theinertiaproperties of a
resolution), spinningmass,suffer inherent limitationssuchas
- a well definedandstableinput axis (better accelerationsensitivityandanalogoutput,even if
rad), than theyhavereachedremarkableperformances.In
- insensitivity to acceleration effects and to en- some very sophisticated design, like the electrically
vironmental conditions such as temperature, with a suspendedgyroscopes,themesuredangularrate
stability betterthan 5 lo-’ rad/s or10-20/h(for sensitivity is 10- degreeperhour(aboutonearc
gyroscope performances rotation rate unit is usually second for one day).
degrees per hour instead of radians per second), The first optical
gyroscope was described by
- linearitywithangularrotationrate,over a Sagnac [l] in 1913, under the
form of ring
a
wide range up to hundreds of ‘ i s , interferometer.
102 M. FAUCHEUX,
D. FAYOUX,
J.J. ROLAND J . Optics (Paris), 1988,
vol. 19, I'1" 3

To be considered as an alternative to mechanical The optical path difference A L , between the two
sensors, optical gyroscopes had to wait fifty years for waves is :
the development of lasers or optical fibers.
Beforeadescription of ringlasergyroscopes,a AL=cAt=4An/c.
brief exposure of the operating principle - the so
called Sagnac effect - is given. It
appears
from
this
equationthat Sagnac
a
interferometer is not a very sensitive device :

I. - 1. Sagnac effect for A = 1 m 2 and 0 = 10"/h : A L 7.10 - l3 m .

A schematic diagram of the Sagnac interferometer In 1925,during a famous experiment, Michelson
is represented in figure 1. andGale [ 3 ] measuredtheearthrotationrate
Twooppositelydirectedbeams, (clockwise CW, (10"/h), with a rectangular path of 2.105 m2. For a
and counterclockwise CCW) arising from the .same visible light source the measured path difference was
source,propagate inside theinterferometeralong roughly a quarter of a fringe.
the same closed path. A modern design of aphasesensitiveoptical
gyroscope is represented by the optical fiber gyro-
scope :
To increase the surface area A without increasing
the size of theinterferometer,theopticalpath is
determined by multiturn
a loop of a long

1
(51 000 m ) single mode optical fiber.

CLOCKWISE
Even,for highly developedsystems [4],optical

Q
fibergyroscopes are presently limited to the 0,l"ih
performance range : this performance is convenient
WAVE for attitude reference but not for inertial navigation
systems.

I. - 2. Ring lasergyroscope (RLG) :

Another scheme devised to increase the sensitivity

of anoptical gyroscope is theintroduction of an
n=0 activelasermediuminaringcavity, as shown in
F'RINGE P A T E R N
figure 2 .
The basic interferometer is modified to become a
FIG. 1 - Classic Sagnac interferometer. The CW and CCW waves ring resonator with L as the perimeter.
interfere to produce the fringe pattern with shifts for R # 0.

Attheoutput of the "ring" interferometer,the

l
GAIN MEDIUM
CW and CCW waves interfere to produce afringe
pattern which shifts if a rotation rate is applied along
an axis perpendicular to the planeof the beam path.
The twoCWandCCWexperience arelative
phasedifferenceproportionaltotherotationrate,
n.
This effect is based on thefact that, with respect to
inertialspace,the twocounter-propagatinglight
waves take different times to complete a trip around
a rotating closed path.
Ageneralapproach valid foranarbitraryinter- WAVE
ferometershape [ 2 ] leads to time
a difference
At proportional to 0 :
DETECTOR
At = 4 A 0 /c2
FIG. 2 - Schematic view of a ring laser gyroscope. When
where A is the area enclosed by the light path, and c R # 0 the frequency difference A f of the two CW and CCW laser
the velocity of light beams is detected.
J . Optics (Paris), 1988,
vol. 19, no 3 M. FAUCHEUX,D. FAYOUX,J. J. ROLAND 103

No externalsource is required : CWandCCW

laser waves are generated inside the resonator if two
conditions are realized :
* the number of wavelengths in the cavity must be
equaltoaninteger(resonanceconditionforthe
oscillation of one mode),
* the gain of the amplifying medium must exceed
cavity losses.
Theadjustment of the cavity lengthallows the
oscillating frequency f, of the laser wave to coincide
with the maximum gain frequency of the amplifying
medium.
If a rotation rate is applied to the ring cavity, the
optical path difference AL, arising from the Sagnac
effect, results in a frequency difference A f , between
two counter-propagating travelling waves.
Y
COMBINING PRISM
AL/L = Af/f (REFRACTIVEINDEX : n )

Leading to a frequency difference proportional to

the rotation rate applied along the sensitive axis of
the RLG :
Af = 4AL?/AL. FIG. 3 - Lasergyroreadoutoptics.

In this formula, the ratio K = 4 A/A L is called

the scale factor of the RLG, (in Hzlradls).
If A = 0,633 pm, assuming a square cavity side of
1and
mthe
numerical values
used before The fringe, spacing is x, = A / y .
( A = 1 m2, L? = 10”/h), the optical path difference Atrest,the fringe pattern is stable (A f = 0).
A L = 7.10-13 m, results in
a beat
frequency When the RLG is rotated, the fringe pattern moves
A f = 76.6 Hz which is easily measured. in a direction depending on the sense of rotation.
The RLG is a very sensitive and accurate device The sense of rotation is determined withatwo
for different reasons : arraysdetector : y is adjusted so thatthedistance
* the CW and CCW laserwaves have, in principle, between the sensitive arrays corresponds to a quarter
the same physical path and the same fluctuations : of a fringe (90” out of phase signals).
most frequency shifts due to external factors are of Fora givenintegrationtime t , thenumber of
reciprocal nature, leaving A f unchanged, intensitymaxima, or fringes,counted with the
* the very small spectral width of the laser allows detector are
measurement of beatfrequenciesdownto 10- Hz
eventhoughtheopticalfrequency is 5.1014 Hz N = j : A f . d t = 4 A /nA. L
dt=4A0/AL.
(A = 0,633 pm,for a HeNe gain medium).Long
term stabilitiesequivalent to 10- 30/hare actually
The output signal is proportional to the totalangle
achieved.
of rotation 0.
FurthermoretheRLG hasasimple and digital The RLG is a “self integrating”device.
readout as shown in figure 3. With the previous numerical values, the sensitivity
The two laser beams are combined, for instance, is aboutone fringe oronecountfor0,6 10- rad
withaprism. At the output of the prism, the two (z0,l arc second). For more realistic sizes of the
beamsarealmostcolinearandinterferetoforma RLG, the sensitivity is in the range of one count for
fringe pattern. a few arc seconds.
Assuming an equal intensity I , of the two beams,
the intensity of light on the detector is :
I. - 3. Advantages of the RLG

Even, from the description of the basic principles,

it can be seen that the RLG corresponds to some of
where (p,is aconstantphasedifference, y the the criteria of an ideal rotation sensor as discussed
angle between the output beams, and x a distance previously : digital output giving angleincrements,
measured along the detector. high sensitivity,stableinputaxis,easy detection
104 M. FAUCHEUX,FAYOUX,
D. J. J. ROLAND J . Optics (Paris), 1988,
vol. 19, no 3

over a wide range of measurement (from 10-3”/h to

thousands of “ i s ) , insensitivity toaccelerations(no
moving parts).
Most of these features are not obtained with other
opticalgyroscopes.
This explains the interest taken in the RLG soon
afterthe first experimentaldemonstration in1963
[ 5 ] , which immediately followed the construction of
the firstlasers. Ideal response curve mode pulling and pushing effect
Detaileddescription [6] and physical analysis [7] K---> K(1-a)
was published early. Some of these physical aspects
are presented in section 11.
An important problem had to be overcome before
Af4
theRLG becameausablerotationsensor : this
problem is knownas“lock-in” of frequencysyn-
chronization of the two travelling waves, associated
withback-scattering of elementsinsidethe cavity.
Thisproblemandvarioussolutionstosolve it are
discussed in section 111.
Manyyears of effortshave been necessary to non reciprocal effects - null shift mode locking effect
develop the RLG up to the stage of a commercial
product. FIG.4 - Sources of errors in the ring laser gyro.
Reasonsforthe delay are related to the lock-in
problem, and to stringent technical specifications of
some elements of theRLG.The technological
aspects are tackled in section IV.
During the development work, other advantages each resonance toward the center frequency of the
of the RLG appeared suchaslonglifetime(many laser gain. In fact this is the differential mode pulling
thousands of hours), low maintenancecosts,and which must be taken into account. This correction is
good adaptation to inertial navigationsystems. obtained by a first order polarisation calculation [9].
Another source of scale factor deviation is related
11. - SOURCES OF ERRORS IN THE RING toadifferentialmode pushing : thiseffectarises
LASER GYRO from the gain saturation for each beam due to the
holeburned by the other beam in the gain curve.
Ideally, the response curve of the RLG is linear as This is obtained by a third order polarisation calcu-
shown in figure 4(a). The frequencydifference is lation [9]. The pushing correction is opposite in sign
proportional to the rotation rate A f = K O , with a to thepulling correction andis an orderof magnitude
constant scale factor, K . smallerthanthe pullingcorrection.Inthecase of
Anyeffectwhichcauseadeviation from this He-Ne medium, themaximum scale factor correction
straight line must be considered as a source of error, is to the order of 10- ’.
which may limit the performance of the RLG as an Following Aronowitz & Lim [lo, 111the intensities
inertial sensor. In the conventional two mode active and phase self consistent equations of the RLG using
laser gyro there are three main typesof errors : scale the Lamb’s theory are, with some simplifications :
factor variations, nullshift and mode locking.
i =+[Q + (p + e)z] (1)
11. - 1. Scale factor variations Ai = c / L ( a - 2 pZ) Ai (2)

To establish the relation A f = K O we have con-

+ = W [ 1 +a(a” (p” T ’ ) Z ) ] (3)
sideredanempty cavity. IntheRLGthe cavity with
contains an active medium and we must consider its
dispersive properties.Thepresence of the gain Z = (II+ Z2)/2, Ai = (Z2 - Z1)/2,
mediumcauses the scale factor to deviate from its W = ~2 - w1 = 2 TKO,
ideal value : this scale factor variation is due to mode
pulling and pushing effects [g]. where
Because of the anomalous dispersion of the active ZI,Z2 : dimensionlessintensities of the twolaser
medium, the two counterpropagating waves, splitted beams,
infrequency by rotation,experienceadifferent o W : cavity angularfrequencyincludingSagnac
index of refraction. This leads to a mode pulling of effect,
J . Optics (Paris), 1988, vol. 19, no 3 M. FAUCHEUX,
D.
FAYOUX, J.
J. ROLAND 105

9 : instantaneous
phase
difference of the
waves,
(L/2 .TT : outputbeatfrequency A f .
TheLamb coefficients arewrittenassumsand
differences for example :

U = 1 / 2 ( ~+
1 ~ 2 )AU
, = 1 / 2 ( a 2- c l )

and A U = (w/2 k u ) U ' , the prime signifies differen-

tiation
with
respect
to Cl,?, where ll.*is the
detuning to the center gain curve normalized to the
Doppler width. V

a i: linear net gain p i : self saturation FIG. 5 - Typical arrangement to compensate for gas flow effects.
ei, : cross saturation u i : mode pulling
p i : mode pushing T ~ ,: crosspushing

a = 1/(2 ku) ku : Doppler width

11. - 2. a) Gas flow induced by D.C discharge
Fromequation (3) theexpression of the scale
factor is : Podgorsky & Aronowitz [l21 haveshown that
theirobservationsareconsistentwiththeso-called
K' = K(I - a) (4) Langmuir flow [13, 14, 151 in the active medium.
In a D.C. excited plasma, there is a movement of
with i3 > 0, where a is the scale factordeviation theneutralatomsfromthecathodetotheanode,
associated to mode pulling and pushing. because of the walls of the discharge tube, causing a
surpressure at the anode and balancedby a Poiseuille
flow, from the anode to the cathode at the center of
11. - 2. Null shift thetube.Overa crosssection thetotal flow rate
must be zero. The gas flow velocity depends on the
A null shift ie a frequency difference for zero input
discharge current and on the radius andlength of the
rate, arises fromanyanisotropy or non-reciprocal
tube.
effect in the cavity for the two counterpropagating
To avoid this problem, the usual way is to use two
waves (cf. fig. 4(c)). Hence the response of the gyro
dischargetubeswithoppositepolarities(cf. fig. 5).
becomes :
Thus, for each beam there is a compensation of the
Af = K'R +K,. (5) gas flow effects if the two capillaries are symetrical
andthe twodischarge currents equal, both witha
The main cause of anisotropy in the cavity i s the high accuracy. Any residualunbalance in the dis-
presence of a gas flow inside the amplifying medium. harge currents leading to a different velocity in the
Indeed when the light propagates through a moving two discharges will produce a null shift, as shown in
medium the index of refraction is different. the typical example of figure 6. In this example an
The maineffectcomesfromadispersioncurve unbalance of 1 p A leads to a null shift of 0,0032"/h
shiftfor each wave [12]. This is equivalenttoa fordischarge
a current of 0,3mA abovelaser
Doppler shift of the oscillating frequency of the two threshold.
waves which can be written as :

51.2 = 5 1 . 2 * kV/ku (6)

where V is the bulk velocity of the gas. The sign +
corresponds to an active atom flow in the direction
of positive rotation rate w 2 == w 1 (cf. fig. 5).
By using (6) morea generalform of ( 5 ) is
obtained :

Af = K'R -2KVa (7)

where a is themode pulling andpushingterm

defined above.
The gas flow may arise from discharge effects but FIG. 6 - Null shifr due to a currents unbalance, obtained with a
also from thermal effects. 33 cm perimeter laser gyro.
106 M. FAUCHEUX, J.J.
D. FAYOUX, ROLAND J . Optics
(Paris), 1988, vol. 19, no 3

Nevertheless this theory is incomplete and can not misalignement which leads to a different beam path
explain thevariation of sensitivity to the gas flow through the discharges is then a source of null shift
with the discharge currentasshown in figure 7. variations.These misalignements can bedueto
FollowingLeiby and Oskam [16, 171, the effect of thermal bending of the block, mirror tilt (in particu-
the
frictionnal
forces
between
the
neutral
and lar the mirror used for the path length control), non-
chargedparticleshas to be considered in the total planar beam path in the case of square gyros.
discharge tube. This effect produces a gas flow in the
Applied magnetic fields : Observations have
center of the tube from the cathode to the anode.In
shownthattheapplication of externalmagnetic
fact the gas flow velocity results from a combination
fields leads to a null shift. The detailed mechanism is
on the above two opposite flows, depending on the
not yet completely understood. This seems to result
discharge current [18]. Thus, a sensitivity near zero
from Faraday effect in the plasma which exhibits a
can be obtained with a specific discharge current as
non negligible Verdet constant (6 radicm. O e
shown in figure 7.
[21]). In order to produce a phase shift between the
counterpropagating waves through Faraday effect in
the active medium, a non linear polarization must be
sensitivity ( " / h / m A )
4T assumed.Someellipticpolarizationsshouldresult
from small mirror birefringence [22] due to residual
strengths in the mirrors occurring at the time of the
thin film deposition.

11. - 3. Mode locking

t
--l
3
A
By farthe most limitingfactor in theRLG
performance is the dependance of the scale factor K
FIG. 7 - Sensitivity of gas flow (induced by discharge effect) to the
ontherotationrate f2. It arisesfromlock-in or
discharge current (It = laser threshold current) obtained with a
33 cm perimeter laser gyro. frequency synchronisation of the counterpropagating
waves at low rotation rates [ l l ] .This phenomenon
common to alloscillators has been well studied in
thedomain of radio physics [23]. Whenthebeat
frequency A f is belowacriticalvalue,mutual
coupling of the two waves locks them together and
11. - 2. b) Gas flow induced by thermal effect causes themto oscillate atthesamefrequency,
making A f = 0. The RLG response curve takes the
A thermal gradient along the tube walls generates formshown in figure 4 ( d ) . Underthethreshold
a gas flow [19]. Indeed, during their collisions with rotation rate f l L , the RLG is insensitive, which may
the walls the particles are C thermalized >> and leave be described as a scale factor K ( f2 ) = 0, if f2 -= f2L .
the walls withmore or less momentum depending The coupling between the
counterpropagating
upontheir impingingdirection.There is anunba- waves arises mainly frombackscatteringproduced
lance of the net rate of momentum transfer inside a by mirrors or intracavity optical
elements. The
sheath of about one mean free path,which produces backscattering of each mirror is characterized by an
a gas flow inside the sheath from the hot point to the amplitudecoefficient ri andanadditionalphase
cold point, and by reaction a gas flow at the centerof angle (cf : section 111). A theoretical analysis
the tube from the coldpoint to the hot point. shows that, by using some simplifying assumptions,
The bulkvelocity at the center of the tube can thethree self-consistent equations may bewritten
reach 0,02 c d s for a gradient of 1"icm and a tube as :
diameter of 3 mm, leading to null shift of the order
of 0,02"lh. I = c / L [+
~ (p + e ) z ] + W l z COS 0 COS (8)

Ai = c/L(a - 2 pZ) Ai +0 1 Z sin c$ sin E (9)

11. - 2. c) Others effects
4 = wb - wl[sin c$ cos F + ( A i / Z ) cos c$ sin F ]

Misalignement : Gas flow velocity and lasergain (10)

have a non-uniform radial shape ; the laser beam has with
a radial gaussian intensity distribution ; then the gas
flow induced null shifts dependon theexact positions wl = 2(rc/L), E +
= ( E ~ ~ ~ ) / r 2=, +
(rl r 2 ) / 2 ,
of thebeams in thedischargetube [20]. Any + = + + . T T + &and Wb=W(l"a).
 J. Optics (Paris), 1988,
vol. 19, no
FAUCHEUX,
M.3 J. J. ROLAND
D. FAYOUX, 107

- For E = 0, (10) becomes : The lockingeffect is themajorproblem of the

RLG and it is clearthatuntil thisproblem is not
overcome,theRLGcannotbeusedasinertial
sensor.
the solutions of (11) are :
for W -= W [ ,4 = sin-l(wb/wl) and is constant,
making A f = 0 ; 111. - BACKSCATTERING AND SOLUTIONS TO
for W > W [ ,2 ?T A f = Wb[l - (W//Wb)2]1'2. LOCK-IN

- For E = ~ / 2 ,(10) becomes : 111. - 1. Backscattering

8, = Wb - w l ( A i / I ) COS 4 (12) The backscatteringmainsource is localized on
RLG cavity mirrors. The backscattered wave ampli-
after some algebraic [l11the solutions of (12) are : tude and phase parameters (noted hereafter r and E )
for W < w L ,4 = 1/2sin-l(wb/w,) and it is con- are dependent upon two factors :
stant, making again A f = 0 ; - the mirror scattering coefficient which affects
(with the amplitude r . The requirements are so stringent
(for high accuracy RLG,thetotalscatteringcoef-
ficient must be of the order of 10 ppm) that all RLG
manufacturershavemadeconsiderableefforts to
for produce in-house these very low scatter optical thin
films.
- the relative position of the mirrors in the cavity
which affects the vector arrangement of the waves
which exhibits a positive scale factor correction [ l l ] .
backscattered at each mirror as described below.
- In the general case 0<E < T /2, A f may be In a three mirror cavity (the principle is the same
written as as with 4 mirrors or more), thecw wave creates three
2 X Af = w b [ l - 1/2(W,
cos + ccw coherentbackscattered waves (one on each
mirror) which formtheoverall ccw backscattered
+ 1/2(w1 sin E ) ' / ( w , ' + W;)] (13) wave entering the RLG Eq. (10). The effect is the
same with the ccw wave.
in figure 8.
and the scale factor has the form depicted
Eachmirror issuedbackscatteredwave(MIBW)
has its ownparameters ri and which depend upon
the localization and nature of the scattering sources
with respect to the beam impact on the mirror. The
K'/K f globalbackscatteringfromeachmirrorcan be de-
scribed as a single virtual scattering source represent-
ing the coherent combination of the different scat-
tered waves on the mirror.
As shown on figure 9 and 10, the three MIBW will
createthe ccw backscattered wave by vectorial
combination R = R , + R , + R , with the three dis-
tances l , , l,, l , as parameters. As we considerthe
interference between waves created by three sources,
thesedistances aretakenbetweenthe successive
virtualscatteringsourcesandnotbetweenbeam
impact centers on the mirrors. For clarity, thevirtual
sources are localizedat thecenter of thebeam
FIG. 8 - Scale factor behaviour for 0 .=E C rr/2. impacts on figure 9. This will not change the results
aslongas we consider mirror translations.
If we considertheamplitudeandphase of the
three MIBW on mirror M1 at time t = 0, we find :
It is important to note that :
- the behaviour of the scale factor depends on R, = E r , e"',
the phase and amplitude of backscattering, R -
2 -
ei(E2+ 62)
with 42= - 4 T (I, + 1 2 ) / A ,
- the lock-in band depends also on the backscat-
tering phase E ; the nearer E to ~ / 2 the
, smaller the R, = Er, e'( E
'+ $3)
with 4, = -4 T~,/A,
lock-in band is. A = wavelength.
108 J.J.
M. FAUCHEUX,D. FAYOUX, ROLAND J . Optics (Paris), 1988,
vol. 19, no 3

M2 M3

FIG. 11 - Two mirrors translation - modification of backscatte-

ring.
FIG. 9 - Backscattering sources.

intensities)andaservo loop controlling the mirror

positions.
The experiment is also in agreement with the
model for the case of the translation of three mirrors
with a constant cavity length which enables to rotate
independently the vectors R, and R,, and hence to
reach lower lock-in thresholds with the same set of
mirrors.
Using very high quality mirrors, lock-in threshold
E cw lower thantheearthrotationrate (15"ih) may be
FIG. 10 - Backscattered vectors composition, reached with these particular cavity configurations,
but this value is too high compared to the accuracy
requirements of 10- 30/h orless for inertial navigation
RLG. A way to circumvent the lock-inproblem is
still necessary (see next section). Nevertheless, these
Thissimplemodelpredicts thegeneralvariation cavity configurationcbntrol technics are useful1 to
of thebackscattering(andhence of the lock-in improvetheperformances of high sensitivitytwo-
thresholdvalue)
when
the cavity geometry is wave RLG even if equiped with low scatter mirrors.
changed. For example, the case of the translation of
two mirrors with a constant cavity length is treated
below. 111. - 2. Solutions to lock-in
In this case, the two mirrors are translated by the
same amount (x) along their normals, but in oppo- In order to operate the RLG out of the lock-in
site directions. Figure l1 shows the variation of the region,simple
a solutionwouldbetocreatea
distances I , , into I I; ;, , l;. Thephases frequency bias, between the two waves with a high
4 and 4, arechanged by theamount A 4 2 = value compared to the lock-in threshold. This bias
A 4 , = 2 n-x & / A . The backscattered vector R is could be realized either by rotating the RLG body
continuouslychangedwiththeperiod h / &. The (Sagnaceffect) or by introducinganon-reciprocal
experimentperformed witha 33 cm cavity length opticalpathlength(Faradayeffect,forinstance).
RLG (fig. 12) is ingood agreement withthisex- The problem with thiskind of solution is that the
pected variation. stability of the bias would be practically impossible
This shows that with the same mirrors, different to attain. Indeed, in the majority of applications, the
lock-in thresholds can be achieved and a minimum bias must be larger than 106"/h with an accuracy of
may be reached with an appropriate cavity configur- about 10- to 10-,"/h, that is, a relative instability of
ation. The RLG can be continuously operated under 10" to
this favorable configuration using an analysis of the The various solutions to the lock-inproblem are
non mixed output signals (contrarotating waves summarized in the diagram of figure 13.
J . Optics (Paris), 1988, vol. 19, n” 3 M. FAUCHEUX, J. J. ROLAND
D. FAYOUX, 109

I
Lock-in rate (O/h)

250
300

0 4 l
0 50 100 150 200 250
Mirror PZT Voltage (0,36 microns = 55 volts)

FIG. 12 - Two mirrors translation - constant cavity length.

Modes
TEMOO
/

Modes
TEMOO
3 or 4 mirrors
I
Biasing the RLG
2 + 2 waves RLG
+ TEM 10
3 or 4 mirrors
I

+ 4 waves RLG
Modes TEMOO
4 mirrors
Circular Dolarization
l I

h (P-polarization)
= yrd I 1- activity
l I
l
I

p+lpq ’

FIG. 13 - Solutions to the lock-in problem.

111. - 2. a) Two-wave RLG This bias can be created by :

The solution tocompensatefor lock-inconven- - mechanical oscillation of the RLG body (this
tional two-wave RLG is to create an alternating bias method - called “dithering” - is the most widely
[24]. The integration of the RLG output during one used solution),
bias cycle results in the cancellation of the bias effect - alternating a non reciprocal phase shift induced
andonlythenetrotationrate is measured.The by amagnetoopticaleffect(transverseKerreffect)
effects of analternating bias are discussedinsec- on a specific mirror [25],
tion I11 - 3. - or dithering two mirrorsalongtheirnormals
110 FAUCHEUX,
M. FAYOUX,
D. J. J. ROLAND J . Optics (Paris), 1988, vol. 19, no 3

and out of phase, to create an alternative Doppler This RLG concept presents many obvious advan-
shift between the twowaves, which reduces the lock- tages :
in threshold [26]. - the constant bias effect Afo is cancelled, avoid-
ing the need for high bias stability,
111. - 2. b) Four-wave RLG - the scale factor of the four-wave RLG is twice
that of a two-wave RLG with the same geometry,
Another concept of RLG uses the combination of - the RLG has no moving parts,
twoconstantbias, one reciprocaland one non- - the ZLAG configuration with out-of-plane
reciprocal, toenablethe oscillation of four waves cavity and Zeeman effect has no intracavity optical
well separated in frequency,avoidingthelock-in elements.
phenomenon [27].
The reciprocal bias is used to split the frequency of The disadvantages of the four-wave RLG are the
corotating modes of differentcircularpolarizations following :
(oneleftcircularpolarization - LCP - and one - themirrors must beisotropic to maintaina
right circularpolarization-RCP).Thiscanbedone circularpolarization,anappropriateout-of-plane
either by an opticallyactivematerialintroduced in configuration may reduce the anisotropy effect [28],
the beam path or by an out-of-plane cavity configur- - the output seems sensitive to cavity length and
ations.The cavity must have an evennumber of magnetic fluctuations [27].
mirrors to allow for simultaneousoscillation of Until
now,despite
many years of efforts to
differently polarized waves. developfour-wave RLG, theapplication of the
Thenon-reciprocal bias splits thefrequency of DILAG concept toactual systemsseems tobe
countertravelling modes of the same polarization. It problematic.
is achieved by Faraday effect in a cell located inside
the cavity (DILAG - Differential Laser Gyro) or 111. - 2. c) 2 + 2 - waveRLG
Zeeman effect on the gain medium itself (ZLAG).
The frequencydiagram of afour-wave RLG is A novel RLG concept has been introduced [29] to
presented in figure 14. The cavity contains a pair of eliminate the lock-in problem by using the oscillation
two-wave RLG(RLG 1 and RLG 2) with their of two weak TEMolmodes.The twoadditional
respective outputs obtained by mixing countertravel- modes create non-linear coupling with the main two
ling modes of equal polarization : TEMoo modes, thus keeping them fromlocking as in
the two-wave RLG (this principle is sometimes
Afl = Afo + K O , interpretedasan“opticaldithering”).Thetheory
Afz = Afo - K O . and experiments have not been so much investigated
in thiscase [30]. A self induced bias with a sign
Afo : frequencysplitinduced by thenon-recip- dependent of the direction of rotation was observed
rocal bias (Hz), butnot fairly understood.
R : input
rotation
rate
(radis). Nevertheless this RLG concept remains interesting
because it would lead to a simple plane cavity design
which does not require new optical elements, moving
parts, or magneticfields.
RLG 1 RLG 2

1.,“ -
A f l = Afo + KR

cw
L
Non reciprocal bias
*A;

Reciprocal bias
Af2=AfO-KR

ccw
I
111. - 3. Alternating bias (two-wave RLG)
The effect of the alternating bias is the introduc-
tion of a new term W , cos (w d t ) in theRLG
Eq. (lo), with W , = 2 TKO,.
LCP modes RCP modes R, and W d are the amplitude and pulsation (both
in radis),respectively, of the oscillating bias (or
Laser frequency dither).
The new equation can be analytically solved [30],
FIG.14 - Frequency diagram o f a four-wave RLG (not to scale).
resulting in a RLG beat note very close to the ideal
responseexceptfornarrowbandslocated around
multiples of w d and described by Bessel functions of
argument ( w m / w d ) . Figure15 shows the behaviour
The electronicdifference of the two readout of thealternatingbiased RLG with non-linearities
signals gives the output of the four-wave RLG : around zero rotation rate and harmonics of ad.
Around the zero rotation rate, the new (dynamic)
Afl-Afz=2KO. lock-inthreshold is proportionalto Jo( w , / w ~ ) .
J . Optics (Paris), 1988, vol. 19, no 3 D. FAYOUX,J. J. ROLAND
M. FAUCHEUX, 111

ment is created by asimple twisting bar driven to

resonance by piezoelectrictransducers. The mag-
neto-optic dither may provide this possibility if the
materials
present small magnetization switching
times.

111. - 4. RLG fundamental limit

The advances inlow scatter low losses mirror

technologyhavemade possible toreach very low
lock-in thresholds, so thatthe RLGrandom walk
performance may approach the ultimate limits set by
the quantum noise [31]. As spontaneous emission is
FIG. 15 - Alternating bias effect with no random noise body- always presentevenata low level comparedto
dithered RLG output vs input rotation rate. stimulated emission, the laser linewidth is inlarged.
The phase of the emitted radiation is randomized in
an uncorrelated way for the two or four waves of the
RLG, resulting in arandomization of theRLG
output [32]. This leads to a rotation angle uncertainty
Choosing the values of 0 , and W d so that W,/ w d is A 6 given by :
either large or close toazero of the Jo Bessel
function, makes Jo (wm/wd)< l and the dynamic AB,/ 4= V (K2 Q ) " ( ~ Y / P ) ~ ' ~
lock-inthreshold is much lower thanthat of the for a two-wave RLG or
undithered RLG.
This scale factornon-linearity may vanish is the A6, \/; = v ( K 4Q > - l ( h ~ / P ) l ' ~
dither
amplitude a, is randomly unstable. If for a four-wave RLG ,
w,/od haslarge
a value, smallfluctuations of
o mcause big relative changes of the Bessel functions, with
andthentheRLG average outputover several t = integration
time,
dither periods is K a , R = input rotation rate. The V = laser
frequency,
random noise may be applied directly to the dither K2 = two-wave RLG scale factor and
mechanism. TheRLGoutput is thenarandom K, = four-wave RLG scale factor
variable with an averagevalue close tothe ideal (= 2 K2),
Sagnacresponseandatime-dependentdispersion. Q = cavity quality factor, inversely pro-
In the case of the sinusoidally dithered RLG taken portional to the cavity losses,
here, the random noise introduces an uncertainty of P = power permode.
the RLG outputin the form of a standard deviation :
The reason for the \ , h factor between the express-
U (random walk) = 0 , ( 2 W, where t is the r r
ions of A 6 , t ' t and A 0 2 y l t is that the output of the
integration time. four-wave RLG is obtained by mixing fourfre-
This error is directlyproportionaltothenon- quencies(seesection I11 - 2. b), thenthespon-
dithered RLG lock-in threshold a,. It shows that a taneous emission induced noise in theoutput is
major way to increase the accuracy of adithered times more important than for a two-wave RLG.
RLG is the use of low scatter mirrors associated, if But, as K, = 2 K,, the uncertainty A 6 , is \ , h times
needed, with the special cavity configurations dis- less than A 6 2 if theotherparameters ( Q , P ) are
cussed in section I11 - 1. equal, which gives anotheradvantagetothefour-
With a typical value of W, = lo6 radis, the uncer- mode configuration.
tainty of the measured rotation angle is about : The fundamental limit of the RLG depends upon
the cavity size (scale factor), the mirror losses (Q)
A6/ 4; = & for a , = 100"/h or and the optical power ( P ) . Of course, we find here
A6/ 4= & for a,= 10"/h . the same influencing factors described by the Schaw-
low-Townes formula.
A more efficient dither would be a square wave Typical values of quantum noise limit are about
instead of a sine wave because the time spent in the 10- 4"/ \I% or less, which shows that a dithered two-
lock-in region would be much smaller, resulting in a wave RLG with a scale factor of lo5 Hziradis,
lower random walk coefficient. W , = lo6 radis and a lock-in threshold of about the
That kind of dither is difficult to implement on a earthrotationrate is nearlylimited by quantum
conventionalbody-dithered RLG where the move- noise.
112 M. FAUCHEUX, J.J.
D. FAYOUX, ROLAND J . Optics
(Paris), 1988, vol. 19, no 3

IV. - APPLICATIONS AND TECHNOLOGICAL Our description here will be limited to two rep-
DESCRIPTION OF THE RLG resentativeexamplesdesignedfordifferentappli-
cations :
IV. - 1. Applications 1. the "standard" body dithered RLG with trian-
gular cavity
The RLG has become a competitive candidate for 2 . aparticulartriaxial RLG in a single optical
most applications in inertial navigation and guidance block.
systems. Sensors adapted to awide range of perform-
anceshave been developped. IV. - 2. a) Body dithered RLG with triangular
For a given quality of components and construc- cavity :
tion, the performance of the RLG is related to its
size : (seeschematic view of figure 16 andthephoto-
Sensitivity and accuracy (long term stability of the graph of figure 17 showingatypicalexample of a
beal signal, in "/h) are primarily determined by the 33 cm perimeterRLGdesignedforinertial navi-
size of the optical cavity. gation systems).
RLG with stabilities down to 10-40/h and perime- The basic structure is acarefullymachined solid
ters up to 60 cm can be used for high performance block.A glass ceramicmaterialwitha very low
navigationsystems(forinstance in ships). coefficient of thermal expansion is used. The three
Small missiles guidance systems use RLG sensors mirrors forming a triangular cavity are mounted to
having stabilities in the range of 10"ih with perime- the block by optical contact. This provides stability
ters as smallasa few cm. and ruggedness.
Navigation for
commercial
aircrafts, the most The multidielectric coated mirrors have very high
important application of RLG, requires stabilities in reflectivity. Absorption is less than (100 ppm)
the
neighbourhood of 10- 2"/hto achieve the and total scattering close to 10 ppm.
traditionalonenautical mile perhour system per- This corresponds to a substrate roughness of 1 or
formance. 2 A rms, and to specific high quality coatings needed
The perimeter of the cavity (square or triangular to reduce backscattering and lock-in.
form) is close to 30 cm. An example is shown in the One of themirrors is sphericalforstabilityand
photograph of figure 17. optimization of the cavity. The cavity perimeter is
adjusted by means of apiezoelectrictransducer
IV. - 2. Description of the RLG attached to a mirror ; its position is controlled with a
servo loop designed so that the laser oscillates at its
Manyvariations in theconstruction of RLG are maximum power for a given discharge current : this
observed depending on the method to prevent lock- allows foranearconstantopticalfrequencyand
in (body dithered RLG, magnetic mirror, DILAG... scale factor of the RLG.
as discussed in section 111) and on particular design The amplifyingmedium is a 10:1 mixture of He
choices of eachconstruction, influenced by appli- and Ne excited by a dc discharge. The transition is
cations(size,shape,materials, electronicscircuits, generally the 0.633 pm line of Ne. Equal mixture of
packaging.. .). two isotopes 20Ne and 22Neis necessary to avoid gain
However, in view of the previousdiscussion of induced mode competition of the oppositely directed
possible error sources: some important design rules laser beams.
are common to all realizations : The solid block provides a vacuum sealed reservoir
- define a very stable cavity construction, for the gain medium.Capillaries are drilled in the
- minimize the number of intracavity elements, blockalong thebeampath.Thedc discharges
- use very high quality optical components (mir- through these capillaries are symetricbetween two
rors), anodes and one cathode. Currents are regulated to
- reduce the perturbations arising from the gain avoid gas flow inducedeffects.
medium (optical, thermal, electrical., .). The design of the RLG, for a lifetime more than
This explains the systematic use of a He-Ne medium, 10 000 hourstakesaccount of plasmaaction on
mirrors, purity of gas andcurrent density onthe
- avoid non-reciprocal effects, providing symetry
cathode.
and isotropy of all elements with respect to the two RLGreadout andmechanicaldithering ofthe
contrarotating laserbeams(in all environmental whole optical block have been described before.
conditions),
- suppress reflexions of output beams back into IV. - 2. b) Triaxial RLG in a single optical block.
the cavity and oscillation of transverse modes,
- reduceexternalperturbationsandtheir vari- Forsome
applications,like
attitude
reference
ations such as : thermal gradients, magnetic sources, systems for tactical missiles, small size and mass of
mechanical constraints on the optical block.. . the system are required.
J . Optics (Paris), 1988, vol. 19, no 3 M. FAUCHEUX,D. FAYOUX,J. J. ROLAND 113

FIG. 16 - Schematic top view of a ring laser gyroscope with its various components.

FIG. 17 -A 33 cm perimeter body dithered Ring Laser GFro (SFENA Company product f;)r inertial navigation system).
114 M. FAUCHEUX,D. FAYOUX,J . J . ROLAND J . Optics (Paris), 1988, vol. 19, no 3

Since the accuracy of the RLG is size dependent, A particular arrangement using only one cathode
the design of suchsystems with three singleaxis andthreeanodes, allows forthecompensation of
sensors is rather difficult. Langmuir flow effects in the gas discharges as shown
The construction o f atriaxial RLG in a single in figure 186.
optical block leads to a better compromise between The manufacturing of the glass ceramic block
accuracyandsmallsize. includes mainly :
An interestingconfiguration of suchatriaxial - drilling 12 capillaries for theoptical beam
optical block [33j, is presented in figure 28a : three path,
square form optical cavities with orthogonal axis can - machining 6 flat surfaces for positionning the
be defined with only six mirrors (instead of nine for mirrors.
threetriangular singleaxis RLG). Eachmirror is
Some advantages and drawbacks of this design are
used with perpendicular planes of incidence, for two
straightforward :
opticalcavities.
The three square optical beam path determine a - size,volumeand mass of theoptical compo-
regular octahedral geometrical form. nents are reduced. Compared to three independent
optical cavities of the same size, the volume of the
triaxial optical blocs is reduced by a factor of two.
- The angularstabilitybetween theorthogonal
sensitive axis of the sensors is improved, due to the
high mechanical rigidity of the triaxial block.
- Onlyone mechanicalsystem is neededfor
dithering the wholeopticalassembly.
- The reduction of number of electrodes leads to
simplifiedelectric currentcontrol circuitand less
power consumption.
- Thecost, associated tothe minimization of
components number is reduced.
Since each mirror is used for two optical cavities,
the simultaneous adjustment of cavitiesresonances
is more difficult andrequires specialtoolingand
appropriatemethod.
- Scattering of mirrors could
induceenergy
interaction of beamsfromdifferentcavitiesand
affectperformances.

FIG. 18 - Triaxial RLG in a single optical block.

a) schematic view of the three square optical cavities in orthogonal
planes using six mirrors(cavities CDEF, ACBE, ADBF).
b) Hc-Nc plasma discharge configuration using 1 cathode and 3
anodes : the electrical discharge induced gas flow is compemated FIG. 19 - Triaxial Ring Laser Gyro in a single opticalblock,
for the three square cavities. designed by SFENA.
J . Optics(Paris), 1988, vol. 19, no 3 M. FAUCHEUX,
D. FAYOUX,
J. J. ROLAND 115

This kind of coupling is not observed, at least for compared to single axis RLG, is a good example of
mediumrangeperformancesystems.. the state of the art RLG technology required for the
Thisparticulartriaxial RLG design,withsize, adaptation of opticalrotationsensorstoinertial
mass,volume and costsreduced by afactor of 2 systems.

REFERENCES

[ l] SAGNAC (G,).- 1913, C.R.Acad.Sci., Vol 157, 708. [l91 COCCOLI (J. D.) & HELFANT (S.). - May 1978, Proceedings
[ 21 POST(E. J.). - 1967, Rev. Mod. Phys., Vol 39, 475. of the IEEE, National Aerospace and Electronics Confe-
[ 31 MICHELSON (A. H.)& GALE (H. G.). - 1925, Astrophys. rence.
J . , Vol 61, 140. [20] JIANG YANAN. - 1981, Jiguang, Vol8, N" 4, (1-8).
[ 41 BERGH (R. A.) etal. - 1981, Opt.Lett., Vol 6, 198. [21] TOBIAS (I.)& WALLACE (R. A.).- 1964, Phys. Rev., Vol
[ 51 MACEK(W. M.) & DAVIS (D. T.M.). - 1963, Appl. Phys: 134, A549.
Lett., Vol2, 67. [22] MARTIN (G. J.). - 1984, SPIE, Vol 478, Fiberoptic and
[ 61 KILLPATRICK (J.). - 1967, ZEEE Spectr., Vol4, 44. laser sensors 11, 110.
[ 71 ARONOWITZ (F.). - 1971, in "Laser Applications", edited [23] ADLER(R.). - 1973, Proceedings of the IEEE, Vol 61,
by M. Ross,(Academic, New York), pp. 113-200. N" 10.
[ 81 SARGENT (M,) et al. - In "Laser physics", Addison Wesley, [24] KILLPATRICK (J.). - 1968, US Patent 3 373650.
1974. [25] MACEK(W.M.). - 1974, US Patent 3 851 973.
[ 91 ARONOWITZ (F.). - 1965, Phys.Rev., Vol 139, A 635. [26] PATTERSON (R. A.) et al., 1984, SPIE Vol487, 78.
[lo] ARONOWITZ (F.) & LIM(W. L.). - 1979, Optical Enginee- [27] CHOW(W. W.) etal. - 1980, ZEEE J . Q.E, Vol QE-16,
ring, Vol 18 No 4, 376. No 9, 918.
[l11 ARONOWITZ (F.) & LIM(W. L.).- 1977, ZEEE J. o f Q . E . , [28] SANDERS (V. E.). - 1981, US patent 4 247 832.
V01 QE-13, N" 5. [29] SCULLY (M. 0 . ) et al. - Aug. 1978, Opt.Let., Vol3, 43.
[l21 PODGORSKI (T. J.) & ARONOWITZ (F.). - 1968, IEEE J. of [30] CHOW(W. W.) er al. -Jan. 1985, Rev. Mod. Phys., Vol57,
Q.E., Vol QE-4, n' 1. N" 1, 61.
[l31 LANGMUIR (I). - 1923, J. FranklinInst., Vol 196, 751. [31] SCHAWLOW(L.)& TOWNES (C. H.). - 1958, Phys.Rev.,
[l41 DRCYVESTEYN (M. J.). - 1922, Physica, Vol 2, 255. Vol 112, 1940.
[l51 RUTTENAUER (A.) ; - 1922, Z . Phys., Vol 10, 269. [32] DORSCHNER (T. A.) et al. - Dec. 1980, IEEE J . Q. E., Vol
[l61 LEIBY (C. C.) & OSKAM(H. J.). - 1967, Phys. of Fluids, QE-16, N" 12, 1376.
Vol 10, N' 9, 1992. [33] BERNELIN(B.).- Patent N" 8006298, SFENACompany,
[l71 FISCHER (J.).- 1977, Phys. of Fluids, Vol20, n' 2, 216. March 1980.
[l81 LE TOUZE(G.) & FAUCHEUX (M.). - To be published. (Manuscrit received in april6, 1988.)

L'ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE

MET AU CONCOURS UN POSTE DE :

PROFESSEUR D'OPTIQUE APPLIQUEE

Dtlai d'inscription : 31 mars 1989

Entrke en fonction : a convenir

Les personnes interesskes voudront bien demander le dossier relatif a ce poste au :

Secrttariat gknkral de 1'Ecole Polytechnique

Fkdtrale de Lausanne - CE - Ecublens - 1015 LausannelSuisse

m OmMasson, Paris, 1988