PHASE MODULATOR FOR FIBER-OPTIC GYROSCOPES CONTROLLED IN A ZERO-MEAN MANNER AND A FIBER-OPTIC GYROSCOPE

Abstract

The invention relates to a control system (100) for a fiber optic gryoscope, comprising a phase modulator (110) for modulating a phase of a light signal (115) and a control unit (120) for producing a control signal (125), by the value of which the phase is modulated and which is fed to the phase modulator (110). The control signal changes statistically and does not have an average value.

Claims

1-12. canceled

13. A control system for a fiber-optic gyroscope, comprising: a phase modulator for modulation a phase of a light signal; a control unit for generating a control signal, by the value of which the phase is modulated and that is fed to the phase modulator; characterized in that the control signal varies statistically and has an expected value of zero; and the control unit comprises a random number generator that generates random numbers r between 0 and q with a probability of 1/q; the control unit generates a primary signal having a value of x; the control signal has a value of (xx.sub.r); and it applies: x.sub.r=0 for x+r<q and x.sub.r=q for x+rq.

14. The control system according to claim 13, wherein the control unit generates the control signal having expected value of zero out of a deterministic primary signal.

15. The control system according to claim 13, wherein q=.

16. The control system according to claim 13, wherein q=.

17. The control system according to claim 14, wherein the primary signal is a superposition of a square wave signal and a ramping signal.

18. The control system according to claim 14, wherein the control signal has a word length that is by 1 bit wider than the word length of the primary signal.

19. The control system according to claim 13, wherein the phase modulator comprises a multifunctional integrated optical chip that receives the control signal.

20. A fiber-optic gyroscope having a control system according to claim 13 and comprising: a light source for emitting light having a predetermined wavelength; a beam splitter for splitting the light from the light source into two incoming beams that are fed into the phase modulator, and for superposing two outgoing beams coming out of the phase modulator to form a detection beam; a coil for guiding therein in opposite directions the incoming beams received by the phase modulator, before they are fed into the phase modulator as the two outgoing beams; a detector for measuring an intensity of the detection beam, which intensity is provided to the control unit; wherein the control unit determines from the measured intensity a rotation rate around a central axis of the coil and the statistical control signal having expected value zero.

21. The fiber-optic gyroscope according to claim 20, wherein the phase modulator is configured to modulate the phases of the two incoming beams and the phases of the two outgoing beams such that the intensity of the detection beam can be measured at predetermined working points.

22. The fiber-optic gyroscope according to claim 21, wherein the working points are subsequent points having maximal gradient of intensity of the detection beam plotted against the phase, and 4 or 8 working points are used.

23. The fiber-optic gyroscope according to claim 20, wherein the fiber-optic gyroscope is a fiber-optic Sagnac interferometer.

Description

[0034] These and further advantages of the invention are described in the following based on examples by using the accompanying figures. It shows:

[0035] FIG. 1A a schematic block diagram of a control system according an embodiment;

[0036] FIG. 1B a schematic block diagram of a control system according to a further embodiment;

[0037] FIG. 2 a schematic block diagram of data transfer in a control system according to an embodiment;

[0038] FIG. 3 a schematic block diagram of a fiber-optic gyroscope according to an embodiment;

[0039] FIG. 4 a schematic block diagram of a digital readout circuit according to the prior art;

[0040] FIG. 5 a schematic block diagram of data transfer in an control system according to an embodiment;

[0041] FIG. 6 a schematic block diagram of data transfer in a control system according to a further embodiment; and

[0042] FIGS. 7A bis 7C results of rotation rate measurements using a control system according to the prior art and control systems according to the embodiments of FIGS. 5 and 6.

[0043] FIG. 1A illustrates a control system 100 that is suitable for controlling a fiber-optic gyroscope. The control system 100 comprises a phase modulator 110 and a control unit 110.

[0044] The phase modulator 110 is configured to modify the phase of a signal 115 that passes through it. For example, the phase modulator 110 allows shifting the phase of a passing light beam by a predetermined amount. The phase modulator 110 may for example comprise a multifunctional integrated optical chip (MIOC). In the MIOC the phase of light is modulated by means of electrodes. Here, by using an electric field applied by means of the electrodes the effective index or the capability for guiding light can be influenced. This allows modulating or shifting the phase of passing light in a particularly easy manner.

[0045] From the control unit 120 the phase modulator 110 obtains a control signal 125 by means of which the phase modulator is controlled. Besides providing the control signal 125, the control unit 120 may have also further functions. For example, the control unit 120 may be a computer processor such as a CPU that controls all or a part of the processes in a system or a device such as a fiber-optic gyroscope in which the control system 100 is used.

[0046] The amount of phase modulation is determined by the size of the control signal 125. For example, in the MIOC the control signal may be the voltage to be applied at the electrodes and the phase shift may be directly proportional to the voltage. The constant of proportionality is the electro-optic amplification factor of the used modulation structure. This factor can be determined for the used modulation structures. If one scales the control signal 125 with the inverse of this electro-optic amplification factor, the size of the resulting control signal 125 will therefore be equal to the phase shift, i.e. the value of the control signal 125 will directly indicate the phase shift. The compensation of the electro-optic amplification factor may for example be achieved by a digital analog (DA) conversion within the control unit 120.

[0047] The control signal 125 that is used for controlling the phase modulation is a statistical signal. This means that in contrast to deterministic signals its value is not uniquely determined, but follows a specific probability distribution. In each calculation cycles of the control unit 120 the value of the control signal 125 is therefore determined out of a set of values, to each of which values a probability for its occurrence is assigned. The control signal has therefore a statistic, i.e. not previously definable, fluctuation.

[0048] The statistic of the control signal 125 is here characterized in that the mean value of the control signal 125 is zero. The control signal 125 is hence zero-mean. A result of these characteristics of the control 125 is that in the frequency spectrum of the control signal 125 frequencies near zero are strongly suppressed. Due to this, it is possible to operate phase modulators such as MIOCs with the control signal 125, whose response behavior for small frequencies differs from their response behavior for large frequencies. For example, small electro-optic amplification in the MIOC at small frequencies can be made unrecognizable by using the statistical, zero-mean control signal 125 that does not have frequency components in the frequency range that is less amplified in the MIOC or comprise such components only strongly suppressed. This allows stable and reliable operation of a phase modulator having different response behaviors at small and large frequencies.

[0049] The control signal 125 may here be generated from a primary signal that is in itself deterministic. The primary signal has therefore a value that is assigned to it by the control unit 120, based on which thereafter the control signal 125 is generated. This determines in addition to being zero-mean a further feature of the control signal 125, i.e. that the control signal is based on the primary signal. This allows controlling the phase modulator 110 effectively by using as original signal that is used to set the phase modulation to a specific value the primary signal which is then modified such that a statistic, zero-mean control signal 125 is obtained.

[0050] The primary signal may for example be a superposition of a square wave signal having at least four temporarily fixed signal values and a ramping signal. Here, the temporarily fixed signal values of the square wave signal are used to generate by jumping between different signal values a sudden, discontinuous phase shift in the phase modulator 110. In contrast, the ramping signal is used to generate continuously growing phase shifts or to compensate phase shifts acting from outside onto the signal 115.

[0051] Exemplarily, FIG. 1B shows a schematic block diagram of the control system 100 in which the control unit 120 comprises a random number generator 130. The random number generator 130 generates random numbers r in the range of 0 to q, wherein q may be an arbitrary real number. The number of digits of the numbers r and q is only limited by the computational capacity of the control unit 120 or by the word length of the numbers r and q. Each random number r may be generated in the random generator with the probability 1/q (or q for q<1).

[0052] The primary signal generated by the control unit 120 may have the value x and may vary with time. Depending on the temporarily variable size of the statistic random number r and the deterministic primary signal x and the temporarily constant number q the statistic parameter x.sub.r is formed by the control unit 120. x.sub.r has a value of 0 for a value of (x+r), that is smaller than q, and a value of q, if a value of (x+r) is larger or equal to q. x.sub.r is the statistic rounding of the primary signal x.

[0053] The control signal 125 is generated from the primary signal x and the value x.sub.r and has the value (xx.sub.r). Due to the probability distribution of the random number and the rounding parameter q it can be easily verified that the control signal 125 is zero-mean. The size of the control signal 125 depends therefore on the size of the primary signal x and the random number r equal to the primary signal x or reduced by the amount q with respect to the primary signal x. The larger the (deterministic) primary signal x the more likely it is that the control signal 125 is reduced by q.

[0054] Due to this, low frequencies comprised in the primary signal x are strongly suppressed, as the statistic jumping between the values of x and (xq) leads to an increase of the frequency of the primary signal x.

[0055] In a fiber-optic gyroscope the phase modulation is used to carry out the measurements necessary for the determination of rotation rates and specific working points of an interferometer characteristic that indicate the interference of two light beams superposed in the fiber-optic gyroscope.

[0056] Regarding the selection of the sequence of working points there exist some degrees of freedoms that has to be used appropriately. Selecting these working points by the phase modulator has the following goals: [0057] 1. Without modulation peaks of the cosine shaped interferometer characteristic would be selected that have zero gradient. Hence, the sensitivity of the fiber-optic gyroscope would be zero and no directional information would be present. In order to avoid these disadvantages, points having a maximum gradient are selected. [0058] 2. If only points having the same sign were selected, an applied rotation rate would lead to a DC voltage signal that would be suppressed during consecutive amplification. Hence, working points having alternating signs are selected. In this manner a readout signal is created that is within the pass band of the following amplification units. [0059] 3. If the working points were selected such that positive and negative gradients alternate periodically, a correlation between the control signal 125 and further signals that are necessary for operating the fiber-optic gyroscope would emerge, which would lead to insensitive bands for small rotation rates. Hence, the sequence of the signs of gradients of the working points (modulation) has to be selected such that this correlation becomes zero. [0060] 4. The modulation must be carried out such that the electro-optic amplification factor is compensated for arbitrary input rotation rates of the sensor as was described above.

[0061] If one chooses q= the phase changes statistically about the value n, which leads to a change to a working point having opposite gradient and changes hence the demodulator reference. For q=2 the phase changes statistically by the value 2, which leaves the sign of the gradient unchanged.

[0062] FIG. 2 illustrates a circuit-wise realization of generating a zero-mean control signal 125 (expected value E(xx.sub.r)=0). The primary signal may be time-dependent and chosen from the range x(t) [0,q). The control signal may e.g. have a word length of 12 bit, wherein the weighting of the most significant bit (MSB) is q/2. The uniformly distributed random r, whose word length may also be 12 bit and which is generated in a random number generator 130, is added to this signal x(t) in the adder 140. The condition x+rq is indicated by the occurrence of a carryover. It is not necessary to use the sum itself. If one complements the 12 bit of the signal x by this carryover as a new MSB and interprets the 13 bit number obtained in this manner as two's complement number, the control signal 125 obtained in this manner will be zero-mean, since the new MSB (carryover) has the weight q.

[0063] The primary signal may also have a different word length. Then, the primary signal is also complemented by the carryover as MSB to obtain the control signal 125, for which reason the control signal has one bit more than the primary signal.

[0064] The control signal 125 having the value (x-x.sub.r) and satisfying E(xx,)=0 may then be supplemented to a DA converter 150 that controls in turn the phase modulator 110. Here, different values of q are possible, e.g.: [0065] 1. q= and resulting therefrom a dynamic range of the phase modulator 110 of 2 (called 2 modulation in what follows) [0066] 2. q=2 and resulting thereform a dynamic range of the phase modulator 110 of 4 (called 4 modulation in what follows).

[0067] According to a further embodiment the control system 100 described above is used in a fiber-optic gyroscope for determining rotation rates, e.g. in a Sagnac interferometer. FIG. 3 illustrates such a fiber-optic gyroscope 200.

[0068] The design of fiber-optic gyroscope 200 of FIG. 3 corresponds to the typically used design. A light source 201 emits light of the wavelength and frequency =2c/, wherein c is the speed of light.

[0069] The light waves travel through a coupler 202 and are then split in a beam splitter 203 into two partial beams. Both partial beams run through a phase modulator 210 that provides an additional phase modulation. Due to this between the two beams a phase shift of (t)=c.sub.1.Math.u.sub.(t) is generated. Here, u is a control voltage of the phase modulator 210 and c.sub.1 its electro-optic amplification factor. The negative sign for the resulting phase difference is selected arbitrarily.

[0070] Then, both beams travel in opposite directions to each other through a fiber having total length L.sub.0 that is wound to a coil 204 having radius A and that rotates by angular velocity with respect to inertial space. Due to the Sagnac effect an additional phase shift of .sub.s=.Math.S with S=4RL.sub.0/(c) between the two beams occurs. The runtime of light through the fiber coil be T.sub.0. After both beams have passed through the coil 204, there is a phase shift between them of (t).Math.S(tT.sub.0). Both beams travel then again through the phase modulator 210, but this time with exchanged function such that as further component the phase (t) is added with positive sign. The two beams going out of the phase modulator 210 are then brought to interference in the beam splitter 203 with a total phase shift of .Math.S+(t)(tT.sub.0).

[0071] After its unification the light wave travels as a detection beam again to the coupler 202 where a part of the detection beam is guided to the detector 205. There, a readout voltage u.sub.det=c.sub.0 cos(.Math.S+(t)(tT.sub.0)) is generated that depends on the phase shift of the interfering light beams. The constant c.sub.0 depends on the mean light power at the receiver, on its sensitivity, and on the amplification in following units.

[0072] The remaining part of the circuitry in FIG. 3 serves as control unit 220 and has the purpose to bring the Sagnac interferometer 200 into a state that allows an evaluation of the detector signal u.sub.det for the purpose of determining the rotation rate by supplying appropriate signals to the phase modulator 210.

[0073] The signal u.sub.det generated by the detector 205 is supplied to a first amplification unit 221 having an adjustable amplification a.sub.0. Due to this, the signal is brought to a defined level a.sub.0u.sub.det and afterwards digitalized by an AD converter 222. The obtained signal x.sub.AD is provided to a digital evaluation circuit 223 that generates a signal y.sub.DA. This output signal, which corresponds to the primary signal, is converted in a DA converter 224 into an analog voltage and after multiplication by an adjustable amplification factor a.sub.1 supplied to the phase modulator 210 at a second amplification unit 226. For adjusting the amplification advantageously a multiplying DA converter is provided, whose reference voltage is used for influencing the amplification.

[0074] Typically, the electro-optical amplification factor c.sub.1 is compensated by the DA converter 224 and the second amplification unit 226.

[0075] The digital evaluation circuit 223 and the DA and AD converters 222, 224 operate with clock cycle T.sub.0, which is the runtime of the light through the coil 204. Therefore, there is a closed signal path. The digital evaluation circuit 223 provides at specific, selectable times output values y.sub.for the rotation rate, y.sub.a0 for the amplification factor a.sub.0 of the input path, and y.sub.a1 for the amplification factor a.sub.1 of the output path. All these values are averaged values that are provided to a processor 227 for further processing. In addition, the digital evaluation circuit 223 is provided with a clear command after each readout of the averaged output values by the processor 227 or a timing circuit, which serves for resetting the internal averaging unit.

[0076] The processor 227 calculates from the pre-averaged values y.sub., y.sub.a0, and y.sub.a1 after optional further filtering the measurement value and the digital signals necessary for adjusting the amplification factors a.sub.0 and a.sub.1, which influence via a first supporting DA converter 228 and a second supporting DA converter 229 the corresponding first and second amplification units 222, 224.

[0077] A digital evaluation circuit 223 according to the prior art is sketched in FIG. 4.

[0078] Since via a.sub.1 and c.sub.1 a relation of the digital data word y.sub.DA and the optical phase is established, it is possible by appropriately choosing a.sub.1 such that a.sub.1c.sub.1=1 holds to achieve that the single bits in the data word y.sub.DA corresponding to the primary signal corresponds to phase shifts at the modulator having the size .Math.2.sup.k. In order to simplify the following explanations these values .sub.k=.Math.2.sup.k are directly associated to the place values of the bits of the digital data word. Except for y.sub.DA, this definition shall also apply to all digital phase words of the digital evaluation circuit 223, i.e. also for the data words s.sub.i, i=1, . . . , 8, s.sub.3, s.sub.5, y.sub.a0, y.sub.a1 and y.sub.106 in FIG. 4. This means that in deviation from the convention the numerical value of a data word s having the bits .sub.k, k=1, . . . , m is calculated according to

[00001]$s=\underset{k=1}{\overset{m}{.Math.}}.Math.{}_k.Math.{}_k,{}_k=.Math.2^k.$

[0079] Here, .sub.1 is the least significant bit (LSB) and .sub.m is the MSB of the data word. For the data word y.sub.DA having the bits .sub.k, k=1 . . . m holds

[00002]$y_{\mathrm{DA}}=\underset{k=1^{}}{\overset{m^{}}{.Math.}}.Math.{}_k^{}.Math..Math.2^k.$

[0080] Since =a.sub.1c.sub.1y.sub.DA for a.sub.1c.sub.1=1 the phase shift at the phase modulator 210 equals =y.sub.DA. Then, it holds in this case that

[00003]$=\underset{k=1^{}}{\overset{m^{}}{.Math.}}.Math.{}_k^{}.Math..Math.2^k.$

[0081] As will be explained, m=0.

[0082] The input signal x.sub.AD provided by the AD converter 222 is fed as internal signal S.sub.1 to an input of the adder ADD.sub.1. Here, depending on a demodulation signal d.sub.2 that can take the value 0 or 1, a weighting with 1-2d.sub.2 is carried out, i.e. with +1 or 1. The demodulation signal d.sub.2(i) is the modulation signal d.sub.2(i) that is retarded by n cycles by retarder V.sub.2 and generated by a random number generator M, i.e. d.sub.2(i)=d.sub.2(in). The parameter n is adjustable within predetermined limits and serves for a runtime adaption to the external signal path. The signals d.sub.2 or d.sub.2 can each assume two states (0 or 1). For d.sub.2=0 an addition occurs at the unit ADD.sub.1, while for d.sub.2=1 a subtraction of the value S.sub.1 occurs. The other input of the adder is connected to a register pair RP.sub.1 into which two predetermined values +d and d are stored. The test parameter d is fed as an additional signal into the main control loop as will be indicated later with the aim to measure its loop amplification and to control this amplification by means of a supporting control loop that influences a controllable amplifier to a predefined set value. The test signal d superposed to the used signal has to be sufficiently small in order to avoid overdriving of the external gyroscope path. As will be explained, for correctly adjusted amplification an exact compensation of this test signal occurs such that the measurement position of the sensor is not influenced. For selecting the respectively desired value a select input s is present that is controlled by a signal d.sub.1. The selected value active at the adder input is (2d.sub.1-1).Math.d. Then,

s.sub.2(i)=2d.sub.1(i)1).Math.d(2d.sub.2(i)1).Math.s.sub.1(i).

[0083] The signal d.sub.1 is generated analogously to d.sub.2 by n-times retardation by means of V.sub.1 of the signal d.sub.1. The signal d.sub.1 is generated by a random number generator D that is independent of M. The sum s.sub.2 generated in ADD.sub.1 is supplied to the inputs of two averaging units further described below with ADD.sub.5 and ADD.sub.6 as well as to the input of an adder ADD.sub.2. The output of the adder ADD.sub.2 is supplied to a register chain REG.sub.1 and fed back as a signal d.sub.1 retarded by n cycles to the other input of the adder:

s.sub.3(i)=s.sub.3(in)+s.sub.2(i).

[0084] In addition, s.sub.3 also is input to the averaging unit having ADD.sub.7 that is explained further below as well as to the adder ADD.sub.3. At the other input of ADD.sub.3 the aforementioned signal d.sub.2 that is generated by the random number generator M is fed in with significance . To the place values having smaller significance (/2, /4, . . . ) of the same input the selectable output of a register pair RP.sub.2 with the pre-stored values /2+d and /2d is connected. The selection is carried out by the aforementioned signal d.sub.1 that is generated in the random number generator D. Then,

s.sub.4(i)=s.sub.3(i)+/2+d.sub.2+(2d.sub.11).Math.d.

[0085] From the summation signal s.sub.4 of adder ADD.sub.3 at the digit tr all bits having a significance of 2 and higher are separated. This process corresponds to a modulo 2 operation.

[0086] The remaining bits are fed to the input of the phase integrator consisting of ADD.sub.4 and REG.sub.2. The summation output s.sub.5 of ADD.sub.4 contains also only bits having a significance of smaller than 2. The output is retarded by REG.sub.2 by one cycle and fed back to the other input of the adder. The carryover bit C that is generated by the addition is provided as signal d.sub.3 to the retardation chain V.sub.3. Then,

[00004]$s_5\left(i\right)=\mathrm{mod}.Math..Math.2.Math.\left[s_5\left(i-1\right)+\mathrm{mod}.Math..Math.2.Math.\left[s_4\left(i\right)\right]\right]$$d_3\left(i\right)=\frac{s_5\left(i-1\right)+\mathrm{mod}.Math..Math.2.Math.\left[s_4\left(i\right)\right]-s_5\left(i\right)}{2.Math.}$

[0087] Simultaneously, at the output of REG.sub.2 the output signal y.sub.DA serving as primary signal is output to the DA converter.

[0088] The signals s.sub.2 or s.sub.3 are as described above provided to three averaging units. These are accumulators that can be reset from outside that sum the signal that is to be averaged over a predetermined time period of m cycles.

[0089] The averaged rotation rate value y.sub. is generated by accumulating s.sub.3 with ADD.sub.7 and REG.sub.5:

[00005]$y_{}=\underset{i=1}{\overset{m}{.Math.}}.Math.s_3\left(i\right)$

[0090] The adjustment parameter y.sub.a0 is generated by an accumulation of s.sub.2 carried out by ADD.sub.5 and REG.sub.5, wherein an additional weighting of s.sub.2 by +1 or 1 is carried out that depends on d.sub.1:

[00006]$y_{a.Math..Math.0}=\underset{i=1}{\overset{m}{.Math.}}.Math.s_2\left(i\right).Math.\left(2.Math.d_1^{}\left(i\right)-1\right).$

[0091] Analogously y.sub.a1 is generated by a weighted accumulation of s.sub.2 by ADD.sub.6 and REG.sub.4 that depends on d.sub.3. d.sub.3 is the signal d.sub.3 that is retarded by n cycles by V.sub.3 and formed out of the carryover bit C having significance 2 in the adder ADD.sub.4 of the phase integrator:

[00007]$y_{a.Math..Math.1}=\underset{i=1}{\overset{m}{.Math.}}.Math.s_2\left(i\right).Math.\left(2.Math.d_3^{}\left(i\right)-1\right).$

[0092] Here, it is at first assumed that the factors a.sub.0 and a.sub.1 are adjusted such that a.sub.0c.sub.0=1 and a.sub.1c.sub.1=1 applies. Moreover, due to the properties of the converters also n1 dead times shall be taken into account. Then,

x.sub.AD(i+n)=cos(.Math.S+y.sub.DA(i+1)y.sub.DA(i)).

As shown in FIG. 4 y.sub.DA(i)=s.sub.5(i) and y.sub.DA(i+1)=s.sub.5(i) apply. In addition,

s.sub.4(i)=s.sub.5(i)s.sub.5(i)+k.Math.2

applies.

[0093] The deviation by k.Math.2 occurs due to the modulo 2 operation at tr. The term k.Math.2 can be omitted in the argument of the cosine function due to its periodicity. Then,

s.sub.1(i+1)x.sub.AD(i+n)=cos(.Math.S+s.sub.4(i)).

At first, it is assumed that in both register pairs RP.sub.1 and RP.sub.2 d=0. Then:

s.sub.4(i)=s.sub.3(i)+/2+d.sub.2)

and due to cos(x+/2)=sin(x) as well as sin(x)=sin(x+) and s.sub.1=x.sub.AD:

s.sub.1(i+n)=sin(.Math.S+s.sub.3(i)).Math.(2d.sub.2(i)1).

On the other hand:

[00008]$\begin{array}{c}{s}_2\left(i+n\right)=.Math.-s_1\left(i+n\right).Math.\left(2.Math.d_2^{}\left(i+n\right)-1\right)\\ =.Math.s_1\left(i+n\right).Math.\left(2.Math.d_2\left(i\right)-1\right)\end{array}$

Then, it follows:

s.sub.2(i+n)=sin(.Math.S+s.sub.3(i)).

[0094] The digital evaluation circuit 223 is a closed control loop that tries to keep the controlled deviation (.Math.S+s.sub.3(i)) as small as possible. If this value that occurs in the argument of the sine function is small, the sine can be approximated by its argument and:

s.sub.2(i+n)=.Math.Ss.sub.3(i)

or, in its z transform form:

s.sub.2(z)=z.sup.n(.Math.S+s.sub.3(z)).

[0095] The unit following thereafter consist of ADD.sub.2 and REG.sub.1, has the transfer function

[00009]$\frac{S_3\left(z\right)}{S_2\left(z\right)}=\frac{1}{1-z^{-n}},$

and closes the control loop. From the last two equations one obtains by eliminating the term S.sub.2(z) the relation

S.sub.3(z)=z.sup.n.Math.S.

[0096] The signal s.sub.3 is therefore proportional to the rotation rate . The averaging unit consisting of ADD.sub.7 and REG.sub.5 generates therefrom the signal y.sub..

[0097] The previous explanation assumes that the condition a.sub.1c.sub.1=1 is satisfied. A specific supporting control loop is used to control a.sub.1 until this requirement is satisfied. Here, one uses the fact that the modulo 2 operation carried out digitally will generate an additional error signal, if in the interferometer the phase does not jump exactly by the value 2 corresponding to the modulo operation. The phase acting at the phase detector is

.sub.d(i+1)=.Math.S+a.sub.1c.sub.1(s.sub.5(i)s.sub.5(i1)).

[0098] If the product a.sub.1c.sub.1 deviates from its ideal value 1, to the ideal detector phase the phase error will be added:

.sub.e(i+1)=(a.sub.1c.sub.11)(s.sub.5(i)s.sub.5(i1))

[0099] After demodulation this phase error occurs as additional rotation rate signal. This error signal is therefore the scale factor deviation that is modulated by s.sub.5(i)s.sub.5(i1). But it also holds:

s.sub.5(i)s.sub.5(i1)=mod2[s.sub.4(i)]2l.sub.3(i).

[0100] The right side of this equation can be interpreted as two's complement number having the sign bit d.sub.3. Then, d.sub.3 is the sign of the signal s.sub.5(i)s.sub.5(i1) that modulates the scale factor deviation (a.sub.1c.sub.11). The error modulated in such a manner occurs after n cycles at connection point s.sub.2 and may be modulated by the sign d.sub.3(i) that is also retarded by n cycles in order to deduce a control parameter for a.sub.1. This is carried out by the averaging unit consisting of ADD.sub.6 and REG.sub.4. The additional demodulation is carried out via the -control input of the adder. The average signal at the output y.sub.a1 is therefore a measure for the deviation of the factor a.sub.1 from its set value (given by a.sub.1c.sub.1=1) and is used to adjust the factor to its set value.

[0101] For the stability of the main control loop it is necessary that the loop amplification has the correct value that is determined by a.sub.0c.sub.0=1. In order to always satisfy this condition a supporting control loop for adjusting a.sub.0 is provided. For =0.sub.2 the signal that is retarded by n cycles is s.sub.3. For =0 and a.sub.0c.sub.01 it holds

s.sub.2(i+n)=a.sub.0c.sub.0s.sub.3(i).

[0102] In order to automatically find a measure for the deviation of the factor a.sub.0 from its ideal value in the register pair RP.sub.2 a small test value +d and d is stored in addition to the value /2. Due to this, in addition to s.sub.3 a test signal 2d.sub.1(i)1.Math.d is input into the adder ADD.sub.3 that is controlled in its sign by the random number generator D. If one is interested only in the influence of the test signal, then it applies that

s.sub.2(i+n)=a.sub.0c.sub.0.Math.(2d.sub.1(i)1).Math.d.

[0103] If in the register pair RP.sub.1 the same test values +d and d are stored, the test signal (2d.sub.1(i+n)1).Math.d will be added to s.sub.2(i+n) and it applies that

s.sub.2(i+n)=(1a.sub.0c.sub.0)(2d.sub.1(i)1(.Math.d.

[0104] Then, at the connection point a component of the test signal weighted with (1a.sub.0c.sub.0) is present. This component is filtered by the averaging unit ADD.sub.5 and REG.sub.3, whose input signal s.sub.2 is additionally weighted with the sign of the test signal. Then, the average signal y.sub.a0 is a measure for the deviation of the product a.sub.0c.sub.0 from 1 and can be used for adjusting a.sub.0 to its set value.

[0105] Using the above-described digital evaluation circuit 223 that is integrated in the control unit 220 it is possible to determine the rotation rate as well as the parameters necessary for its determination. In addition, the evaluation circuit 223 outputs that the signal y.sub.AD as primary signal that directly influences the phase modulation. This primary signal is neither statistical nor zero-mean.

[0106] A possible modification of the digital evaluation circuit 223 according to the example of the circuit-wise realization of FIG. 2 is illustrated in FIG. 5. FIG. 5 illustrates the elements D, ADD.sub.3, ADD.sub.4 and REG.sub.2 of FIG. 4 that are necessary for output of the signal y.sub.DA.

[0107] Instead of supplying the signal y.sub.DA directly to the DA converter 224, the signal y.sub.DA is modified by means of a random number generator 230 and an adder 240 as described above with respect to FIG. 2 The signal y.sub.DA corresponds here therefore to the primary signal x.

[0108] The circuitry illustrated in FIG. 5 is used for a dynamic range of the phase modulator 210 of 2, i.e. for q=.

[0109] FIG. 6 illustrates a possible realization of the generation of a statistical zero-mean control signal based on the digital evaluation circuit 223 known from the prior art. Also in this case the signal y.sub.DA (primary signal x) is modified by the random number generator 230 and the adder according to the method described above with respect to FIG. 2 to generate a statistical, zero-mean control signal.

[0110] In difference to the circuitry of FIG. 5 the dynamic range of the phase modulator in the circuitry of FIG. 6 is 4, i.e. q=2 applies. Due to this, one obtains in contrast to the case of q= further degrees of freedom by selecting the working points and the control loop parameters such that the circuit design can be simplified substantially. For example, it is not necessary for q=2 to feed back the signal generated by the random generator 230 and the adder 240 via a gate of several XOR elements 560 to the digital evaluation unit 223 as illustrated for the case q= in FIG. 5.

[0111] In FIGS. 7A to 7C it is based on diagrams illustrated schematically how the measurement precision of a fiber-optic gyroscope at rotation rates equal to zero improves, if one uses a control system that uses a statistical, zero-mean control signal for phase modulation.

[0112] FIG. 7A illustrates in its upper diagram the rotation rate in /h against time for fiber-optic gyroscope having a control system according to the prior art at rotation rate 0/h. In this case for a system functioning without error it can be expected that only statistically distributed noise occurs, i.e. that the values of the rotation rate have a Gaussian distribution. As can be clearly seen the measured curve has accumulations at the value 0/h, i.e. a lock-in effect occurs that biases the measurement systematically. The measurement results do not follow only a Gaussian distribution. This can be also clearly seen in the lower diagram of FIG. 7A that is a histogram of the different measurement values. It can be clearly seen that no Gaussian distributed noise is shown, as a significant accumulation at 0/h is present.

[0113] FIGS. 7B and 7C show the same diagrams for the circuitries of FIG. 5 and FIG. 6, respectively, i.e. for a statistical, zero-mean control signal generated as described above having q= or q=2, respectively. It can be clearly seen that the measurement is only influenced by statistic noise, i.e. the histogram of the measurement value is a Gaussian curve having a center at 0/h and the measurement values do not have an accumulation at 0/h. Similar improvements occur for comparisons of measurements that are carried out for small rotation rates such as for example for gyro compassing in which components of Earth's rotation rate of 15/h have to be measured as exactly as possible.

[0114] It is therefore possible to reduce the lock-in effect at 0/h by means of a control system for fiber-optic gyroscope that generates a control signal or phase modulation, which signal is statistical and zero-mean, and to enhance in this manner the reliability of the measurement results at rotation rates close to 0/h.