A mode-locked fiber laser gyroscope (MLFLG) consists of a gain medium, and a fiber cavity with a Sagnac loop interferometer. The loop interferometer serves as a rotation sensing element and as a second mirror in the cavity. The reflectivity of the loop mirror was modulated by a phase-difference modulator. When the frequency of the phase-difference modulation is equal to the longitudinal mode spacing of the laser, mode-locked pulses can be obtained. Two reciprocal pulses per cavity round-trip time are generated at the moments of maximum reflectivity because the loop reflectivity is modulated twice for one period of phase-difference modulation. Thus a Sagnac phase shift according to a rotation rate is converted to a pulse timing shift in time domain.
The MLFLG built with an Er$^{+3}$ doped fiber as the gain medium exhibited strong gain competition between two consecutive pulses, which resulted in serious errors in the measurement of pulse intervals (i.e. rotation rate). The strong gain competition is mainly due to a slow gain recovery time after the gain is depleted by one of two pulses.
To overcome the gain competition, we used as the gain medium an antireflection coated diode laser (LD) whose gain recovery time is short enough compared to pulse intervals. For stable reciprocal operation of the Sagnac loop mirror with reduced polarization drift, a highly birefringent polarization-maintaining fiber was used for the entire fiber cavity. The injection current was set to be below the threshold current of the solitary LD to be operated as an optical amplifier. The amount of optical feedback from the fiber circuit into the LD was about -10dB.
We could obtain a stable mode-locked optical pulse train without gain competition even when the Sagnac loop was rotating. The two reciprocal pulses were nearly the same in shape and amplitude. The stability of the pulse train could be estimated from the measurement of rf spectral contents. The ratio of the 1st harmonic component to the 2nd harmonic component was less than -60dB, indicating that the amplitude difference between two consecutive pulses would be less than 0.1%. The change of the time intervals between two consecutive pulses was measured as a function of rotation rates with much improved accuracy by a time interval analyzer. The scale factor of the gyroscope obtained by pulse interval changes agreed with the theory.
The equivalent rotation rate noise was 4deg/hr/$\sqrt{Hz}$ which corresponded to 10$\mu$rad/$\sqrt{Hz}$ of phase noise. Long-term drift of the gyroscope signal was measured. Without any polarizer in the laser cavity, The drift was rather large with about 600deg/hr. Drift could be reduced to 100deg/hr when a polarizer was inserted between the LD and the fiber cavity to suppress feedback of the light containing error signals caused by polarization cross-coupling.
Polarization analysis for the laser predicts that reciprocal condition can be automatically satisfied. The large drift seems to come from the fact that the amount of a spontaneous emission was rather large (about -10dB of the total emission) because the laser operated near threshold. The spontaneous emission can introduce errors without a good polarizer.
