CA1268541A - In-line fiber optic memory - Google Patents

Abstract

IN-LINE FIBER OPTIC MEMORY
Abstract of the Disclosure A fiber optic recirculating memory comprises a splice-free length of optical fiber which forms a loop that is optically closed by means of a fiber optic coupler. The coupler couples an optical signal input pulse to the loop for circulation therein, and outputs a portion of the signal pulse on each circulation to provide a series of output pulses. A pump source is included to pump the fiber loop with a pump signal having sufficient intensity to cause stimulated Raman scattering in the fiber loop, and thereby cause amplification of the circulating signal pulse. The fiber characteristics, coupler characteristics, and pump power are selected to yield a Raman gain which compensates for the total round-trip losses in the fiber loop, so as to provide an output pulse train of constant amplitude pulses. The invention may be implemented utilizing a standard coupler with a pump signal modulation technique. The pump signal is input to the loop as a series of pulses having a duration and a periodicity chosen to prevent overlap of recirculating pump pulses with input pump pulses until at least two circulations of the loop, which minimizes pump power fluctuations and will therefore enhance output signal stability.

Description

STANF-60B-Foleign ~6~

IN-LINE FIBER OPTIC MEM~Y

The ~ubject matter of this application is related to that in U.S. Patent 4,708,421 issued Nove~ber 24, 1987.
Back~round of the Invention The present invention relates generally to fiber optic optical memvry devices, and particularly to in-line fiber 10 optic memories which recirculate optical signals, and ,! which utilize non-linear effects for amplification of the recirculating signals.
Fiber optic recirculating memories typically comprise a loop of optical fiber and a fiber optic directional coupler for coupling light to and from the loop of optical fiber. An exemplary fiber optic recirculating memory is disclosed in U.S. Patent No. 4,473,27U, issued ~eptember 25, 1984, which is incorporated by reference herein. As disclo~ed in that patent, a 8ingle signal pulse is supplied a~ an input to the memory device. This pulse recirculates in the loop and a por~ion of the pulse is coupled out of the loop on each circulation to provide a serie3 o~ output signal pulses identical to the input ~ignal pulse, although at ~maller, gradually decreasing amplitudes. Such a device is particularly useful, for example, to provide a short term optical memory in a system where data is generated at a rate faster than ie can be accepted by a data processor, or to adjust time delay~ between sys~ems exchanging data (optical delay line) or to repe~t data (data regeneratior.). This device may also be u~i~i2ed as a p~Lse train generator, or as a filter. ~oreover, recirculating memory devices may be utilized as re-entrant fiber sensing loops for fiber optic gyro~copes.
Although ~he recirculating memory disclo~ed in the above-referenced patent represents a significant advance ~26~

in the art, and is quite advantageous for many optical memory applications, the optical memory capabilities of such devices are limited due to the fact that the circulating signal pulse is degraded by coupling a portion of this pulse on each circulation around the fiber loop to provide the train of output pulses. Further, fiber propagation losses attenuate the optical signal pulse as it propagates through the loop, causing additional losses. These losses cause the intensity of the signal pulse circulating in the loop to decay, which causes a concomitant decay in the series of output pulses, thereby limiting the number of useable output pulses and the lifetime of the optical signal memory.
In order to prevent the problems associated with decay of the circulating pulse, it has been proposed in the prior art to insert a fiber optic amplifier in the loop so as to amplify the pulse on each circulation to compensate ~or the loop losses. Such a system, for example, has been disclosed in U.S. Patent No. 4,136,929, to Suzaki, entitled "APPARATUS FOR GENERATING LIGHT PULSE TRAIN".
Fiber optic amplifiers, however, are typically discrete components, which must be spliced into the fiber loop.
Such splicing will cause additional loop losses which must be compensated for by the fiber optic amplifier.
~5 Accordingly, there is a need in the art ~or a low loss in-line recirculating optical memory which utilizes an amplification process capable of providing sufficient gain to compensate for the total round trip loop losses.
Summary of the Invention The present invention is an in-line fiber optic recirculating memory having a length of optical fiber, preferably splice-free, for receiving an optical signal pulse. The optical fiber forms a loop which is optically closed by means of an optical coupler, for example, a fiber optic directional coupler. The coupler ~ouples the optical signal input pulse to the loop for circulation ~2~ 5~

therein, and outputs a portion of the signal pulse on each circulation to provide a series of output pulses.
The present invention further includes a light source for generating an input pump signal comprising a series of pulses. Each of these pulses is serially input to the loop by the coupler~ The pulses circulate in the loop and generate Stokes light therein for amplifica~ion of the circulating optical signal pulse during each of plural circulations of the signal pulse. The optical wavelengths of the pump pulses and the optical signal pulse are selected such that the Stokes light has the same optical wavelength as the optical signal pulse. The coupling ratio of the coupler is greater than zero for the pump pulses so that each pump pulse makes plural circulations through the 1oop7 The length of the loop provides a loop transit time for the pump pulses such that each pump pulse in a series of pump pulses makes plural circulations in the loop before overlapping with any subsequent pump pulse in the series o~ pump pulses. This allows the amplitude of the pump pulses to decay prior to overlapping, thereby reducing phase noise effects in the loop.
In a preferred embodiment, the pump pulses are equally spaced and have the same time duration æo that the pump signal is periodic. The period of the pump signal is less than the amount of the loop transit time. Preferably, the time duration of each pump pulse is sub~tantially less than the loop transit time. In a preferred embodiment, the period of the pump signal is substantially equal to an integer multiple of the time duration of each pump pulse. Preferably, the sum of the period of the pump signal and the time duration of each pump signal pulse is substantially equal to the loop transit time.
The invention further includes a method for generating a series of output pulses from a single input pulse. A
loop, having a loop leng~h, is formed from a length of optical fiber that is closed by a coupl~er. An optical ~;Z68~

signal pulse having an optical signal wavelength is input into the optical fiber to circulate in the loop to provide a circulating optical pulse that makes plural circulations~
in the loop. The circulating optical pulse propaga~es S around the loop during a signal pulse loop propagation time which is determined by the optical signal wavelength and the loop length. A portion o~ the circulating signal pulse is tapped from the loop on each of the recirculations to provide the series of output pulses.
Additionally, a series of input pump pulses having an optical pump wave length are coupled into the fiber loop to cause stimulated scattering in the loop. The stimulated scattering generates Stokes light in the loop for amplification of the circulating signal pulse during each of the plural circulations by the signal pulsesO The pump pulses propagate around the loop in a pump pulse loop propagation time determined by the optical pump wavelength and the loop length. Each input pump pulse has a pump pulse time duration. The optical signal wavelength and the optical pump wavelength are selected such that the Stokes light has the same optical wavelength as the optical signal wavelen~th. Each pump pulse is spaced apart in time from the next succeeding pump pulse in a series of pump pulses so that each pump pulse makes plural circulations in the loop before overlapping with any - subsequent pump pulse. This allows the amplitude of the pump pulses to decay prior to any overlapping, thereby reducing phase noise effects in the loop.
In a preferred embodiment, the pump pulses are spaced apart in time at a selected time delay so that the pump pulses are periodic. The time delay is selected so that the pump pulse has a period less than the pump pulse loop propagation time. Preferably, the pump pulse time duration of each pump pulse is selected to be substantially less than the pump pulse loop propagation time. In a preferred embodiment, the period of the pump ~;:61~S~, pulses is selec~ed to be substantially equal to an integer multiple of the pump pulse time duration. Preferably, the pump pulse time duration and the pump pulse period are selected so that the sum of the pump pulse period and the S pump pulse time duration are substantially equal to the pump pulse loo2 propagation time.
The pump signal has sufficient intensity to cause gain, for example, Raman gain caused by stimulated Raman scattering (SRS), in the core of the fiber. Preferably, the optical fiber forming the loop is single mode fiber, which typically has a core diameter on the order of 5-10 microns. The small core diameter of single mode fiber is highly advantageous over the larger core diameter multimode fiber, since its small size provides increased optical power density in the fiber core for a given amount of pump light, and thus increases the optical gain.
In the preferred embodiment, the pump source has a coherence length which is short compared to the length of the fiber loop to ensure that any of the pump ligh~ that recirculates in the loop does not cause a resonant effect. Further, the pulse width of the signal pulse is preferably less than the loop length to ensure that the signal pulse does not interfere with itself while circulating in the loop.
Stimulated Raman scattering may be viewed as a three level laser emission process in which molecules of the active media (i.e., the fiber core) are excited from the ground level to an excited virtual level by absorbing input pump photons. Return of the excited molecules to an intermediate level results in the emission of photons, commonly referred to as "Stokes photons". These Stokes photons have a frequency downshifted with respect to the pump frequency by a characteristic amount, typically referred to as the ~aman shift. In the case of SRS in a fused silica fiber, the Stokes photons have a frequency which is about 13-15 terahertz (THz) below the frequency ~68~

of the pump light. In addition to the emission of Stokes photons, the transition from the intermediate level back to the ground level results in the emission of phonons.
~owever, such phonons are of little interest in the present invention, as they are quickly absorbed by the fiber and converted to heat.
In the present invention, the input signal pulse has a wavelength which is equal to the Stokes wavelength. When the signal pulse is injected into the fiber loop, such pulse will stimulate relaxation of the excited molecules and emission of photons at the Stokes wavelength. These stimulated photons have the same phase characteristics and frequency as the input signal pulse, and thus, provide coherent amplification of the input signal pulse.
The optical gain, and thus, the amount of amplification, depends upon the gain coefficient of the fiber (g); the optical power Ppc of the input pump signal;
the effective interaction area (Af) of the optical fiber (i.e., the area of mode or electromagnetic field overlap between the signal pulse and the pump signal); and the effective interaction length (Lf) of the fiber loop (i.e.~
the calculated length, less than the loop length, which accounts for pump power attenuation along the fiber loop). In addition, it has been found that for the above-described recirculating memory configuration, the gain isalso dependent upon the attenuation (a) of the fiber, the length (L) of the fiber loop, and the coupling ratio ~) of the coupler. These recirculating memory device parameters are coordinated in the present invention so that the gain provided by pumping the fiber compensates for the total round trip losses in the fiber loop. This loss compensation, therefore, provides an output pulse train of substantially constant amplitude pulses.
The invention may be implemented utilizing either an ordinary "standard" coupler or a specifically adapted "multiplexing coupler". The term "standard coupler" as ~Z~5~

used herein, refers to a coupler in which the coupling ratio ~or the input signal pulse is close to that of the pump signal, such that the coupling ratios differ by no more than 0. 2. Conversely, a "multiplexing coupler" is a coupler which exhibits a sizeable difference in coupling ratio for the pump signal and signal pulse, such that the difference in coupling ratios is greater than 0.2. For a given input pump power, there is a set of coupling ratios for the pump signal and signal pulse which supplies the loop with the exact amount of pump power required to produce just enough gain to offset total round trip losses of the signal pulse such that the output pulses will have a constant amplitude.
If a standard coupler is used and it is tuned to have relatively high coupling ratios, the total round trip loop losses will be advantageously low thereby decreasing the amount of amplification or gain necessary to compensate for them. However, very little pump power will be launched into the loop, so that pump power utilization is quite inefficient. If, on the o~her hand, a standard coupler tuned to relatively low coupling ratios is used, a sizeable fraction of the pump power will be launched into the loop, so that the pump power i9 used efficiently to provide gain. However, a major fraction of the circulating pulse will be coupled out on each circulation, so that the total round trip loop losses will be quite high. Thus, if the coupler is set to yield high gain, the loop losses will be high, while, if the coupler is set for low losses, the gain will be low. Accordingly, there exists an optimum set of coupling ratios corresponding to a specific coupler tunin~ which minimizes the pump power required for compensation of total round trip losses. It has been observed however that with such mode of operation (i.e., using a standard coupler with a relatively high pump coupling ratio) the recirculation and interference of pump power in the loop causes enhanced pump power fluc~uations or noise which in turn cause gain fluctuations and ~hus sizeable signal intensity noise.
These fluctua~ions are due to interference between the recirculating pump electromagnetic fields of short coherence time and are commonly referred to as a phase noise effect.
As indicated above, a multiplexing coupler may be utilized, rather than a standard coupler. Preferably, the mutliplexing coupler is formed to have a large interaction length. There exists then a tuning position for the coupler for which the interaction length is substantially equal to an even number of coupling lengths for the pump signal and substantially an odd number of coupling lengths for the signal pulse, so that the coupling ratio for the pump signal is zero, or close to zero, while the coupling ratio Xor the signal pulse is close to unity, but not exactly unity. A coupling ratio of zero for the pump signal makes it possible to utilize Che entire amount of input pump power and thereby provides for maximum possible gain, while a coupling ratio close to unity (but not exactly unity) for the signal pulse provides for minimum round trip loop loss. Such combination of coupling ratios permits the invention to operate at very low input pump signal power, while still providing the proper amount of gain. In addition, the use of a zero or null pump coupling ratio suppresses the effect of pump power recirculacion, which, due to the phase noise, would cause gain fluctuations in the loop.
The invention may be implemented in either a forward configuration in which the pump signal and signal pulse are both input into the same end of the fiber loop such ! that the signal pulse propagates in the same direction around the loop as the pump signal, or it may be implemented in a backward configuration in which the pump signal and signal pulse are input into opposite ends of the loop, so that the signal pulse propagates in a ~26~

direction opposite that of the pump signal. In the preferred embodiment, a backward configuration is utilized, since it has been found that this advantageously averages out the effect of pump fluctuation frequencies (i.e., those larger than the reciprocal of the loop transit time) on the signal amplitude fluctuations.
The present invention suppresses the effects of phase noise, and thus, minimizes the signal intensity noise while using a coupler having a pump coupling ratio substantially greater than zero, such that the pump power is recirculated in the loop.
In the ~referred embodiment of the invention, the pump signal is input to the recirculating loop as a series or train of pump pulses having a selectable pulse width and a selectable delay between the beginning of each successive input pump pulse. The pulse width and delay are selec~ed so that each input pump pulse makes plural circulations in the fiber loop before overlapping with any subsequent pump pulse in the series oE pump pulses. This allows the amplitude of the recirculating pump pulses to decay before overlapping occurs and thus reduces the effect of their mutual interference. The pump power recirculating in the loop is ~hereby stabilized causing a sizeable reduction of gain noise, which in turn effects a drastic reduction of the amplified signal noise. Therefore, the invention improves the output signal stability when using a non zero pump coupling ratio, such as exists in the above discussed standard coupler.
Description of the Drawings These and o~her aspects of the present invention may be more fully understood ~hrough reference to the drawings in which:
Figure 1 is a schematic drawing of an embodiment of the optical memory of the present invention showing a signal pulse generator for introducing a signal pulse into the fiber loop, and a pump source for introducing a pump ~z~

signal into the fiber loop to cause Raman gain to amplify the signal pulse;
Figure 2 is a schematic drawing of an embodiment of the signal pulse generator of Figure l;
Figure 3 is a sectional view of one embodiment of a fiber optic directional coupler for use in the optical memory of Figure 1;
Figure 4 is a graph showing relative coupled power versus signal wavelength for a fiber optic coupler having a minimum fiber spacing of four microns, an offset of the oval facing surfaces of zero microns, and a fiber radius of curvature of 25 centimeters;
Figure 5 is a graph of relative coupled power versus signal wavelength similar to that of Figure 4, but for a fiber radius of 200 centimeters;
Figure 6 is a graph of relative coupled power versus signal wavelength for a fiber optic coupler having a minimum fiber spacing of four microns, a fiber radius curvature of 200 centimeters, and a selectable fiber offset;
Figure 7 is a schematic drawing showing the loop and fiber optic coupler of Figure 1 which illustrates the operation of the invention;
Figure ~ is a graph of the input pump power required to cause sufficient Kaman gain to make the train of output pulses substantially constant in intensity as a function of the coupling ratio of a standard coupler;
Figures 9a and 9b are grayhs of the signal coupling ratio as a function of the pump coupling ratio for a multiplexing coupler, with the input pump power required for constant amplitude output pulses as a parameter, for two different values of fiber attenuation.
Figure 10 is a schematic drawing of an alternative embodiment which provides the pump signal as a series of pulses to the recirculating loop to reduce phase noise.

6~5~

Figures lla, llb, 11c and lld are graphs of a pump input pulse train, a train of once-delayed pump pulses, a train of twice-delayed pump pulses, and a train of thrice-delayed pump pulses, respectfully, showing the timing S relationships between the pulse trains in the embodiment of Figure lO.
Description of the Preferred Embo_iments As shown in Figure 1, the optical memory of one embodiment of the present invention comprises a continuous, uninterrupted strand lO of optical fiber, preferably single-mode optical fiber, having an input end portion 12, a loop portion 14, and an ou~put end portion 16. At the ends of the loop portion 14, the single~mode fiber 10 is optically coupled together by means of a fiber optic, evanescent field, four port, directional coupler 20, having ports 1 and 2 on one side thereof, and ports 3 and 4 on the other side thereof. Tracing the fiber 12 from one end to the other, the fiber 10 first passes through ports 1 and 3, and then through port~ 2 and 4, so that the loop 14 extends from ports 3 and 2, while the input portion 12 extends from port 1, and the output portion 16 extends from port ~.
A signal pulse generator 22 is provided to selectively introduce an input pulse 21 into the input fiber portion 12. The coupler 20 couples a portion 24 o~ this pulse 21 to the loop 14 for recirculation therein. Each time the pulse 24 circulates in the loop, a portion is output through the output fiber portion 16 to provide a series or train of output pulses 26. These output pulses 26 are directed through a lens 27 to a beam splitter 29 which splits the output pulses 26 and directs a portion through a monochromator 28 to suppress Fresnel reflections and scattered pump light, so as to filter the output signal pulses 26. Af~er filtering by the monochromator 28, the output signal pulses 26 may be directed to a detector 30 which converts the optical signals to electrical signals ~Z~8~

for display on an oscilloscope 32. Those skilled in the art will recognize that a spike filter at the wavelength of the optical signal pulses 26 (or a dispensive element such as a prism or grating) may be substituted for the monochromator 28.
The embodiment of Figure 1 also includes a pump source 34 for producing a pump light signal 31. In the embodiment shown in Figure 1, the signal 31 is a pulse having a width which is at least as long as the desired length of the output pulse train ~6. In other words, the pump pulse 31 should have a duration which is at least as long as the desired number of output pulses times the loop transit time. In one particular embodiment of Figure 1, the duration of the pulse 31 is selected to be at least 50 times the loop transit time so as to yield at least S0 substantially identical output pulses 26. The pump source 34 of the embodiment of Figure 1 comprises a continuous wave Nd:YA~ laser having a CW power of 2 watts and a wavelength of 1.064 microns. A chopper 35 is included at the output of the pump source 34 to chop the pump signal into relatively long pulses (e.g., 0.5 msec). As is well known in the art, a chopper typically comprises a rotating disk having apertures to alternately block and pass the output light of a laser at a specific frequency. Those skilled in the art will recognize that an acousto-optic or electro-optic modulator ma~ be used instead of the chopper 35.
The pump signal 31 is directed from the pump source 34 to the beam splitter 29, where the light is split such that one portion (not shown) is directed to the monochromator 28, and the other portion 33 is directed to the lens 27 for introduction into ~he output end portion 16 of the fiber 10 for propagation through the loop 14.
The signal portion 33 in the end portion 16 will be referred to as the "input pump signal~'l The filter characteristics of the monochromator 28 are selected so ~;~6~5~

that the por~ion of the pump signal directed towards the monochromator 28 will be blocked to prevent it from reaching the detector 3~.
The pump source 34 and signal pulse generator 22 are connected by respective lines 36, 38 to a synchronizer 40. The synchronizer 40 synchronizes the chopped pump signal pulses to the signal pulses from the pulse generator 22, such that the pump light enters the loop 14 just prior to the signal pulses so that the pump light is present in the loop when the signal pulses arrive.
Preferably, the coherence length of the pump source 34 is relatively short compared to the length of the fi'ber loop 14 to ensure that any of the pump light that recirculates in the loop 14 does not interfere with itself. Further, the width of the signal pulse 21 produced by the generator 22 should preferably be less than the length of the ~iber loop to ensure that the circulating pulse 24 does not interfere with itself as it circuLates through the loop 14.
The wavelength and amplitude of the input pump signal 33 are selected to cause stimulated scattering in the fiber loop 14 at the pulse signal wavelength. As discussed in more detail hereinafter, Ruch stimulated scattering produces photons at a wavelength referred herein as the "Stokes wavelength". In the case of stimulated Kaman scattering in silica glass ibers, the Stokes wavelength i8 1.12 microns for a pump wavelength of 1.064 microns. The photons produced by this stimulated scattering amplify the circulating pulse 24 as it propagates around the fiber loop 14~ However, for such amplification to occur, it is important that the signal pulse ~1 (and the circulating pulse 24) be at a wavelength equal to the Stokes wavelength.
To insure that the signal wavelength i5 the same as the Stokes wavelength, the signal pulse generator 22 comprises a laser 42 which produces light at a waveleng~h ~61~5g~L

identical to that of the pump source 34 (e.g., 1.064 microns), and a signal generating loop 44, formed of the same type of fiber as the loop 14, as shown in Figure 2.
By way of example, the laser 42 may be a ~-switched Nd:YAG
laser having an output wavelength of 1.064 microns, a peak ! power of 120 watts and a switching rate of 50 ~z, so as to generate relatively thin (500 nsec) high power pump pulses which are widely spaced (e.g., a spacing greater than the pump signal duration). ~y way of example, the loop 44 may have a length of about l km. The light from the source 42 is input to one end of the loop 44 through a lens 45 to generate stimulated scattering in the loop 44 such that a wave having the Stokes wavelength of e.g. 1.12 microns is output, together with the pump light, at the other end of the fiber. The Stokes wave and ~he pump light are then directed through a lens 46 to a diffraction grating 48 which separates the two wavelengths into two diverging rays. A diaphragm 49 with a slit therein is positioned to block the 1,064 micron pump light, but pass the 1.12 micron Stokes light to provide the signal pulse 21 (Figure l) which is input to the input end portion 12 of the ~iber lO (Figure 1) through a lens 41. Thus, the signal pulse generator 22 provides signals at exactly the same optical wavelength as the Stokes light generated in the fiber loop 14 (Figure 1).
Those skilled in the art will recognize that a simpler set up may be achieved by replacing the grating 48 with a spike filter at the Stokes wavelength.
A preferred fiber optic directional coupler for use as the coupler 20 (Figure l) in the optical memory of the present invention is shown in Figure 3. As illustrated therein, this coupler includes two exemplary strands 50A
and 50B of a single mode fiber optic material mounted in longitudinal arcuate grooves 52A and 52B, respectively, formed in optically flat, confronting surfaces of rectangular bases or blocks 53A and 53B, respectively.

~6~

The block 53A with the strand 50A mounted in the groove 52A will be referred to as the coupler half 51A, and the block 53~ with the strand 50B mounted in the groove 52B
will be referred to as the coupler half 51B.
S The arcuate grooves 52A and 52B have a radius of curvature which is large compared to the diameter of the fibers 50, and have a width slightly larger than the fiber diameter to permit the fibers 50, when mounted therein, to conform to a path defined by the bottom walls of the grooves 52. The depth of the grooves 52A and 52B varies from a minimum at the center of the blocks 53A and 53~, respectively, to a maximum at the edges of the blocks 53A
and 53B, respectively. This advantageously permits the fiber optic strands~ 50A and 50B, when mounted in the grooves 52A and 52B, respectively, to gradually converge toward the center and diverge toward the edges of the blocks 53A, 53B, thereby eliminating any sharp bends or abrupt changes in direction of the fibers 50 which may cause power loss through mode perturbation. In the embodiment shown, the grooves 5~ are rectangular in cross-section, however, it will be understood that other suitable cross-sectional contours which will accommodate the fibers 50 may be used alternatively, such as U-shaped cross-section or a V-shaped cross-section.
2S At the centers of the blocks 53, in the embodiment shown, the depth of the grooves 52 which mount the strands 50 is less than the diameter of the strands 50, while at the edges of the blocks 53, the depth of the grooves 52 is preferably at least as great as the diameter of the strands 50. Fiber optic material was removed from each of the st~ands 50A and 50B, e.g., by lapping, to form respective oval-shaped planar surfaces, which are coplanar with ~he confronting surfaces of the bloc~s 53A, 53B.
These oval surfaces, where the fiber optic material removed, will be referred to herein as the fiber "facing surfaces". Thus, the amount of fiber optic material that ~Z6i3~

has been removed increases gradually from zero towards the edges of the blocks 53 to a maximum towards the center of the blocks 53. l'his ~apered removal of the fiber optic material enables the fibers to converge and diverge gradually, which is advantageous for avoi~ing backward reflection and excess loss of light energy.
In the embodiment shown, the coupler halves 51A and 51B are identical, and are assembled by placing the confronting surfaces of the blocks 53A and 53B together, so that the facing surfaces of the strands 50A and 50~ are juxtaposed in facing relationship~
An index matching substance (not shown), such as index matching oil, is provided between the confronting surfaces of the blocks 53. This substance has a refractive index approximately equal to the refractive index of the fiber cladding, and also functions to prevent the optically flat surfaces from becoming permanently locked together. The oil is introduced between the blocks 53 by capillary action.
An interaction region 54 is formed at the junction of the strands 50, in which light is transferred between the strands by evanescent field coupling~ It has been found that, to ensure proper evanescent field coupling, the amount of material removed from the fibers 50 must 'be carefully controlled so that the spacing between the core portions of the strands 50 is within a predetermined "critical zone". The evanescent fields extend into the cladding and decrease rapidly with distance outside their respective cores. Thus, sufficient material should be removed to permit each core to be positioned substantially within the evanescent field of the other. If too little material is removed, the cores will not be sufficiently close to permit the evanescent fields to cause the desired interaction of the guided modes, and thus, insufficient coupling will result. Conversely, if too much material is removed, the pro~agation characteristics of the fibers i4gl will be altered, resulting in loss of light energy due to mode perturbation. However, when the spacing between the cores of the strands 50 is within the critical zone, each strand receives a significant portion of the evanescent field energy from the other strand, and good coupling is achieved without significant energy loss. The critical zone includes that area in which the evanescent fields of the fibers 50A and 50B overlap with sufficient strength to provide coupling, i.e., each core is within the ev~nescent field of the other. ~owever, as previously indicated, mode perturbation occurs when the cores are brought too close together. For example, it is believed that, for weakly guided modes, such as the HE~1 mode in single mode fibers, such mode perturbation begins to occur when sufficient material is removed from the fibers 50 to expose their cores. Thus, the critical zone is defined as that area in which the evanescent fields overlap with sufficient strength to cause coupling without substantial mode perturbation induced power loss.
The extent of the critical zone for a particular coupler is dependent upon a number of interrelated factors such as the ~arameters of the fiber itself and the geometry of the coupler. ~'urther, for a single mode fiber having a step-index profile, the critical zone can be quite narrow. In a single mode fiber coupler of the type shown, the required center-to-center spacing between the strands 50 at the center of the coupler is typically less than a few (e.g., 2-3) core diameters, Preferably, the strands 50A and 50B (1) are identical to each other; (2) have the same radius of curvature at the interac~ion region 54; and (3) have an equal amount of fiber optic material removed therefrom to form their respective facing surfaces. Thus, the fibers 50 are symmetrical, through the interaction region 54, in the plane of their facing surfaces, so that their facing surfaces are coextensive if superimposed~ This ensures ~2~

.

that the ~wo fibers 50A and 5~ will have the same propagation characteristics at the interaction region 54, and thereby avoids coupling attenuation associated with dissimilar propagation characteristics.
S The blocks or bases 53 may be fabricated of any suitable rigid material. In one presently preferred embodiment, the bases 53 comprise generally rectangular blocks of fused quartz glass approxîmately 1 inch long, 1 inch wide, and r).4 inch thick. In this embodiment, the fiber optic strands 50 are secured in the slots 52 by suitable cement, such as epoxy glue. ~ne advantage of the fused quartz blocks 53 is that they have a coefficient of thermal expansion similar to that of glass fibers, and this advantage is particularly important if the blocks 53 and fibers 50 are subjected to any heat treatment during the manufacturing process. Another suitable material for the block 53 is silicon, which also has excellent thermal properties for this application.
The coupler of Figure 3 includes four ports, labeled A, B, C, and D, which correspond to the ports 1, 2, 3, and 4, respectively, of the coupler 20 in Figure 1 When viewed from the perspective of Figure 3, ports A and B
which correspond to strands S~A and 50B, respectively, are on the left-hand side of the coupler, while the ports C
2S and D, which correspond to the strands 5~A and 50B, respectively, are on the right-hand side oi the coupler.
For the purposes of discussion, it will be assumed that input light is applied to port A. This light passes through the coupler and is output at port C and/or port D, depending upon the amount of power ~hat is coupled between the strands 5~. In this regard, the term "coupling ratio"
is defined as the ratio of the coupled power to the total output power. In the above example, the coupling ratio would be equal to the ratio of the power at port D to the sum of the power output at ports C and D. This ratio is also referred to as the "coupling efficiency", and when so ~Z6~

, g used, is typically expressed as a percent. Thus, when the term "coupling ratio" is used herein, it should be understood that the corresponding coupling efficiency is equal to the coupling ratio times 100. For example, a coupling ratio of 0.5 is equivalent to a coupling efficiency of 50%1The coupler may be "tuned" to adjust the coupling ratio to any desired value between zero and 1.0, by offsetting the facing surfaces of the blocks 53. Such tuning may be accomplished by sliding the blocks 53 1~ laterally relative to each other, so as to increase the distance between the fiber cores.
The coupler is highly directional, with substantially all of the power applied at one side of the coupler being delivered to the other side of the coupler. That is, substantially all of the light applied to input port A is delivered to the ports C and D, without contra-directional coupling to port B. Likewise, substantially all of the light applied to port B is delivered to the ports C and D. Eur~her, this directivity is symmetrical, so that substantially all of the light applied to either port C or input port D is delivered to the ports A and B. Moreover, the coupler is essentially non-discriminatory with respect to polarizations, and thus, preserves the polarization of the light. Thus, for example, if a light beam having a vertical polarization is input to port A, the light cross-coupled from port ~ to port D, as well as the light passing straight through from port A to port C, will remain vertically polarized.
The coupler is also a low loss device, having insertion or throughput losses typically on the order of

2-3 percent. The term "insertion loss", as used herein, refers to the real scattering losses of light passing through the coupler, from one side to the other. For example, if light is applied to port A, and 97% of that light reaches ports C and D (combined), the insertion loss would be 0~03 (3~). The ter~ "coupler transmission" is ~6 defined as one minus the insertion loss. Thus, if the insertion loss is 0.03 (37O)~ the coupler transmission is 0.97 (97%) Further details regarding this coupler are disclosed 5in U.S. Patent No. 4,536,058 issued on August 20, 1985, and U.S. ~atent No. 4,4~3,52~, issued on January 15, 1985. In addition, the coupler is described in the March 29, 1980 issue of Electronics Letters, Vol. 16, No. 7, pa~es ~60-261.
In general, the coupler depicted in Figure 3 ~xhibits wavelength dependencies such that its coupling ratio may be different for two signals having a large ~avelength separation. ~owever, for the small wavelength separation between the pump signal 33 (Figure 1) and the signal pulse 21, the coupling ratios for the coupler of Figure 3 are ordinarily clo~e to each other (i.e., within 0.2).
The coupler of Figure 3 may, nevertheless, be specially adapted to operate a~ a "multiplexing coupler", ~uch that the coupler exhibits significantly different coupling ratios for diferent wavelengths, even if the wavelength separation is quite small. ~y properly selecting the radius of curvature of the fibers and the core spacing therebetween the coupler can be made to ~S provide virtually any desired coupling ratio for substantially ang pair of ~avelengths.
To further ~xplain ~hi3 aspect of the invention, it will be recalled that the coupler 20 operates on eva~es~ent ~ie~d coupling principles in which guided modes of the str~d~ 5~ interact through their evanescent fields to c~u~e l~ght ~o be transferred between the strands 50 at the interaction region 54. The amount of light transferred i~ dependent upon the proximity and orieDtation of the cores 8S well as the effective length 35 o~ the in~ceract ion region 54 . The length of ~:he ~n~eraction region 54 is dependen~ upon ehe radius of 6 ~ 5 curvatu~e of the fibers 50, and, to a limited extent, the core spacing, although it has been found that the effective length of the interaction region is substantially independent of core spacing~ However, the "coupling length" (i.e., the length within the interaction region 54 which is required Eor a single, complete transfer of a light signal from one fiber to the other? is a function of core spacing, as well as wavelength.
If the length of the interaction region 5~ is increased, and/or core spacing decreased, so that the coupling length is shorter than the e~fective interaction length, a phenomenon referred to herein as "overcoupling"
will occur. Under these circumstances, the light will transfer back to the strand from which it originated. As the interaction length is further increased, and/or the core spacing further decreased, the effective interaction length becomes a greater multiple of the coupling length, and the ligh~ transfers back to the other strand. Thus, the light may make multiple transfers back and forth between the two strands as it travels through the region 54, the number of such transfers being dependent on the leng~h of the interaction region 54, the light wavelength, and the core spacing.
This phenomena permits selection of virtually any two coupling ratios for any two signals of different wavelengths. ~or example, by properly choosing the geometrical parameters for the coupler 20, one signal wavelength may be substantially totally coupled, while a second signal wavelength remains substantially uncoupled.
To illustrate this wavelength dependence, Figure 4 provides a plot of coupled power versus signal wavelength in the visible and near infrared spectrum for a particular coupler geometry. For this coupler configuration, the effective interaction length of the coupler is an odd multiple cf the coupling length for the wavelength 720 nm, but an even multiple of the coupling length for the waveleng~h 550 nm. Thus, the wavelength 720 nm will be 100% coupled, while the wavelength 550 nm will be effectively uncoupled, yielding a wavelength resolution of 170 nm. Other wavelengths exhihit different coupling efficiencies~ For example, 590 nm has a coupling efficiency of about 5-10% and 650 nm has a coupling efficiency of about 80-85%.
As the number of coupling lengths within the effective interac~ion length increases, the resolution of the multiplexing coupler is enhanced. Thus, by increasing the radius of curvature to increase the effective interaction length so that it becomes a higher multiple of the coupling length, resolution is improved. This result is illustrated in Figure 5, which is comparable to the graph of Figure 4, except that the radius of curvature has been increased from 25 cm to 200 cm. As expected, Lhis increase in radius improves the coupler resolution near 600 nm from approximately 170 nm in the 2S cm radius case of Figure 4 to approximately 60 nm in the 200 cm case.
After the resolution of the coupler has been set in accordance with the particular wavelengths of interest, the coupler may be tuned to precisely adjust the coupling length for the wavelengths to yield the desired coupling efficiencies. This is accomylished by offsetting the fibers by sliding the blocks 53A, 53B (Figure 3) relative to each other in a direction normal to the axis of the fibers 50A, 50B. ~uch an offset has the effect of increasing the core spacing. If the required offset is small, it will not upset the resolution.
To illustrate the tunability of multiplexing couplers, Figure 6 provides a plot of relative coupled power versus wavelength for three increasing values of fiber offset (~
microns, 055 microns, and 1.0 microns). The curve is seen to shift toward increasing wavelengths as the offset increases, while the period of oscillation (or resolution) ~6~

remains virtually unchanged. In this partlcular example ~n which the radius of curvature was 20~ cm and the minimum core-to-core spacing was 4 microns, a one micron offset shifted the curve by approximately 45 nm~
Additional details of the above-described multiplexing coupler may be found in U~S. Patent No. 4,556,279 lssued on December 3, 1~85. The wavelength dependencies of the above-described coupler are further discussed in an article by Digonnet, et al., entitled "ANALYSIS OF A
TUNABLE SINGLE MODE OPTICAL FIBER COU~LER", IEEE Journal of ~uantu~ Mech~nics, Vol. ~-18, No. 4 (April, 1982).

The embodiment of Figure 1 of the present invention can utilize the above-described multiplexing coupler and can also utilize a standard nonmultiplexing coupler. If a mul~Lple~ing coupler is utilized, the coupler is preferably adapted ~o exhibit overcoupling for either the pump wavelength (e.g., 1.064 microns) or ~he sLgnal pulse wavelength (e.g., 1.12 microns), so as to provide a very low coupling ratio for the pump signal 33 ~nd a very high c~upling ratio for the signal pulce 21. Except for the type o coupler used, bo~h embodiments of Figure 1 are the ~ame.
2S The above embodiments o~ Figure 1 will be discussed further in reference to Figure 7, which shows the fiber 10 and coupler 20 of Figure 1 without the associated component~, for clarity of illustration. The device of Figure 7 will first be described in terms of operation ~ithout ~he p~,p &ou~ce 34 ~Fi~re 1).
~ ~he ~bsence of the pump 34, the inpue signal pulse 21 p~opagaees along the input fiber portion 12 to port 1 of the coupler, where this pu~se is split by the coupler into the circulating pulse 24, which exits the coupler through port 3, ~nd a pulse (noe shown) which exits the coupler 20 thr~ugh por~ 4. Depending on the coupling ~1,~

~Z6BS41 ratio of the coupler 20, the portion of the pulse 24 exiting port 4 can be substantial. Accordingly, it would be preferable to utilize a switchable coupler as the coupler 20, if available, so that the signal pulse 5 coupling ratio can be switched to a lower value when the signal pulse 21 is initially coupled to the loop 14, and switched back to a high value before the circulating signal 24 completes its first circulation. It should be recognized, however, that the device will function in the intended manner regardless of whether the coupler 20 is switchable.
After the pulse 24 exits port 3 of the coupler, it circulates through the loop 14 and arrives at port 2 of the coupler, where a portion of the pulse 24 is coupled to port 3, to again circulate in the loop 14, while another portion is output from port 4 as the pulse 26a. For purposes of explanation, it will be assumed that the coupling ratio of the coupler 20 for the circulating signal 24 is 0.9, and for simplicity, it will be assumed that the loop 14 and coupler 20 are lossless. If it is further assumed that the pulse 24 has a normalized value of 1.0 on its first circulation abou~ the loop 14, it follows that 10% of the pulse 24 will exit port 4 of the coupler, yielding a normalized value of 0.1 for the first output pulse 26a. The remaining 90~ of the light will be coupled to port 3 for propagation about the loop a second time, thus yielding a normalized value of 0~9 or the pulse 24 on its second circulation. When the pulse arrives back at port 2 of the coupler, 10% of the light will again exit port 4 of the coupler to form the second pulse 24b, which is delayed from the pulse 26a by an amount equal to the round-trip transit time of the fiber loop 14. The amplitude of the pulse 26b will be 10% of the 0.9 normalized value of the second recirculating pulse, or 0.09. On the third circulation, the pulse 24 will have a normalized value of 0.81 (90X x 0.9). After ~6~

completing the third circulation, 10% of the 0.81 normali~ed amplitude of the pulse 24 will be coupled to for~ a third output pulse 26c, which is separated from the second output pulse 26b by the round-trip transit time of the loop~ and has a normalized value of 0.081. By continuing these calculations the normalized amplitude values of subsequent pulses may be found. For example, the fourth pul~e 26d will have a normalized amplitude of .073, while the fifth pulse 26e will have a normalized amplitude of ~066. Accordingly, it can be ~een that each ti~e ~he pulse 24 circulates in the loop 14, an output pulse i~ generated such that a train of output pulses 26 is provided at the f iber output end portlon 16. Of particular significance i8 the fact t~at the output pulses 26 gradually decreaQe in amplitude, thereby limiting the lifetime of the optical memory (i.e., the number of useable output pulses 26). Additional details a~ to the lifetime of the decaying pul~e train 26 may be found in U.S. Patent No. 4,473,270, issued September 25, 1984 to Shaw.
The present inven~ion advantageously amplifies the pulse 24 on each ~irculation through the loop 14 to compen~ate for round-trip losses ~uch that the pulse 24 ha~ a normalized value of 1.0 each time it arrives back at port 2 of the coupler, thereby providing an output pulse train 26~1 ~o 26-5 having a constant amplitude. This is acc~mpliQhe~ without splicing any discrete amplifying components i~o the lo~p ~4, but ra~her by utilizing the core of the fi~er 14 a~ ~he active ampl~fying medium.
A~ was di~cussed in reference ~o Figure l, the pump source 34 (Figure 1) inputs the pump signal 33 into the output end portion 16 of the fiber 10 for introduction into the loop 14 through the coupler 20. The roupler 20 ~plit~ the pump ~ignal in an a~ount depending on the coupling rstio such ~hat a portion ou~puts the coupler through port 1 and another portion outputs the coupler ~3 ~26~5~

-~6-through port 2 for propagation through the loop 14. For example, assuming the coupling ratio for the pump signal 33 is the same as that previously assumed for the signal pulse 21, namely 0.~, it follows ~hat 90% of ~he light will exit port 1 and be lost, while the remaining 10% will be available to propagate through the fiber loop and pump the fiber loop 14. The portion of the pump signal which propagates in the fiber loop 14 is depicted in Figure 7 as the arrow labelled "loop pump signal". In the embodiment shown, the loop pump signal propagates in a direction opposite that of the circulating pulse 24. The pump signal duration for this embodiment is at least equal to the number (N) of desired constant amplitude output pulses (e.g., 26-1 through 26-N) times the round-trip transit time of the loop. In other word~, the input pump signal pulse 33 has a sufficiently long duration such that it can be considered as a continuous wave for the desired number of circulations of the circulating pulse 240 The loop 14 is thus continuously activated by pump light throughout the duration of the constant amplitude pulse train output 26-1, 26-2 . . . 26-N. The optical power of the pump 4ignal 33 is selected to provide sufficient optical power for the portion of this signal 33 propagating in the fiber loop 14 to cause stimulated scattering. As discussed above, such stimulated scattering will generate photons at the S~okes wavelength, 90 as to amplify the circulating slgnal pulse 24 as it propagates through the loop 14.
Although various types of stimulated scattering processes are available for use in the present invention, the preferred embodiments utilize stimulated Raman scattering. As is well known in the art, stimulated Raman scattering is a phenomena in which coherent radiation is generated by optically pumping the molecules of a material, such as an optical fiber, into an excited vibrational state. The process may be viewed as a three-level laser emission process in ~hich molecules of the active media are excited from the ground level to an excited virtual level by absorbing input pump photons.
Return of the excited molecules to an intermediate level results in ~he emission of photons, commonly referred to as "Stokes photons" which have a characteristic wavelength, commonly referred to as the "Stokes wavelength". In the case of S~S, the Stokes photons have a particular optical frequency relationship to ~he pump light which depends upon the molecular structure of the core of the optical fiber. For a preferred embodiment, which utilizes silica glass fibers, this ~requency relationship causes the Stokes photons to be shifted in frequency relative to the pump signal by an amount, referred herein as the "Stokes shift", which, for SRS, is typically about 13 to 15 THz. The Stokes shift in SRS is due to the difference in energy between the incident pump photon and the vibrational level of the molecule, It should be noted that the transition of the excited molecules from the intermediate level back to the ground level results in emission of phonons, although such phonons are of little interest in the present invention, as they are quickly absorbed by the fiber and converted to heat.
As discussed above, the input signal light 21, 24 has a wavelength which is equal to the Stokes wavelength.
When the circulating signal pulse 24 is injected into the loop 14, such pulse 24 will stimulate relaxation of the excited molecules and emission of photons at the Stokes wavelength. The stimulated photons have the same phase characteristics and frequency as the pulse 24, and thus, amplify the circulating pulse 24.
Those skilled in the art will understand that stimulated scattering processes are quite complex, and that the above description of amplification using stimulated Raman sca~tering is somewhat simplified. For example, those skilled in the art will recognize that the 5~

28~

lifetime of the excited vibrational states of the molecules is extremely short. Accordingly, a number of the excited molecules will return ~o the ground state spontaneously, resulting in the spontaneous emission of Stokes photons. These spontaneously emitted Stokes photons can themselves be amplified by the same process as described above for the input signal pulse, such that a secondary wave is created in the loop, referred to herein as a "Stokes background wave". Since the fiber 10 provides an optical guiding structure, two such Stokes background waves will be generated, and these two waves will counterpropagate in the optical fiber loop 14. As shown in Figure 7, one of the Stokes background waves travels around the loop in the same direction as the pump signal (referred herein as the "forward" Stvkes background wave), while the other travels in the opposite direction around the loop relative to the pump signal (referred herein as the "backward" S~okes background wave). Since the backward Stokes background wave propagates in the same direction through the loop 14 as the circulating signal 24, the output 26 of the optical memory will be comprised of a superposition of light from the amplified signal pulse 24 and light from the backward Stokes background wave. Thus, the backward Stokes background wave may be ~5 considered as background noise in the optical memory - output 26. It should be noted that the forward Stokes background wave has no significant effect on the output 26, since it propagates in a direction opposite to the circulating signal 24.
If the optical gain provided by the stimulated scattering in the fiber loop 14 is sufficiently high, these Stokes background waves may grow rapidly in amplitude, due to the fact that they will be amplified in the same manner as the circulating pulse 24, thereby increasing the noise level and adversely affecting the signal to noise ratio. Fortunately, it has been found that, by properly controlling the optical gain provided by the stimulated scattering process in the loop 14, the growth of the background Stokes wave can be main~ained at a level, such that the background Stokes power is several orders of magnitude below that of the circulating signal 24 for a large number of circulations (e.g. thousands).
The optical power of the input pump signal 33 should be selected to provide sufficient pump power in ~he fiber loop 14 such that the stimulated scattering process yields just enough amplification (e.g., Raman gain) to compensate for the total round-trip loop losses of the circulating pulse 24. A portion of these round-trip loop losses is due to fiber attenuation which causes propagation losses in the opti~al fiber loop 14. These propagation losses cause some degradation of the pulse 24 as it propagates around the loop 14. The attenuation (a) of an optical fiber is typically expressed in dB per kilometer. Thus, the total propagation losses per circulation (in dB) due to fiber attenuation are equal to aL, where L is length of the fiber loop 14 in kilometers. Additional losses are caused by the fact that a portion of the pulse 24 is lost from the loop 14 on each circulation du~ to tapping by the coupler 20 to form the outyut signals 26. For the configuration disclosed, such coupling losses will be a function of l - ns. where tls is the coupling ratio of the coupler 2~ for the signal pulse 24. The total round-trip loop losses (~) of the optical signal pulse 24 may be expressed as follows:

~ = (1 - tls) [l-exp(-asL) ] (1 ~2~6~

Where as is the attenuation coefficient at the Stokes wavelength, expressed in km~l through the following conversion: as(km 1) = -as(dB/km)/(lO log~Oe)O
Equation 1 thus indicates the fraction of the optical 5 power of the pulse 24 which is lost on each circulation through the loop 14. The fraction of optical power remaining in the loop 14 will be referred to as the loop transmission (Tloop), and is given by.

Tloop = ns~s (2) where Ts is the fiber transmission for the signal pulse 24 (i.e., exp(-asL)).
Those skilled in the art will understand that there is an additional source of loss which was not discussed above, namely, the coupler insertion losses. However, for the coupler 20, these losses are qui~e small (e.g., 0.15 2Q db), and thus may be ignored. However, if a coupler having high insertion losses i8 used as the coupler 20, these coupler insertion losses should be taken into account, for example, by adding the insertion losses to the fiber propagation losses.
For a given input pump power Pp, the Raman optical gain (G) in the fiber loop 14 may be expressed as:

G = exp [ ~ (I - `lp~p) ]

where ~p is the coupling ratio of the pump signal, g is the ~aman gain coefficient, A~ is the effective interaction area accounting for the pump signal and signal pulse 24 mode overlap in the fiber 14, and Lf is the effective fiber length of interaction area (explicitly Lf = (1-Tp)/~p), where Tp is the fiber transmission at the pump wavelength, and ap is the attenuation coefficient`at the pump wavelength. In order to maintain the recirculating pulse 24 at constant intensity, the gain (G) should equal the total loop loss (~), such that:

GTloop = 1 (4) It may thus be found that the amount of optical power (Ppc) of the input pump signal 33 which is necessary to maintain constant intensity of the recirculating pulse 24 is given by:

A~ e pL
Ppc ~ (asL - log ns) 1 - ~p If the signal pulse coupling ratio (ns) and the pump signal coupling ratio (~p) are substantially the same, Equation 5 may be rewritten as:

P = ~ (a L - log n) -1 ~ ne p (6) where ~ is the coupling ratio of both the pump signal and the signal pulse (i.e- ns = np = n) Equation 6 is a transcendental equation, and thus, the coupling coefficient cannot be expressed directly as a function of the pump power (Ppc)~ ~owever, for any given set of fiber parameters (a, g, Lf, Af), the pump power Ppc defined by Equation (6) may be plotted as a function of the coupling ratio, as shown in the curves of Figure 8~
In developing the curves of Figure 8, it was assumed that s~

the fiber attenuation for the signal light (~s) and the fiber attenuation ~or the pump light (ap) were equal (i.e., aS=ap=a), which is a very good approximationO
The curves of Figure 8 illustrate the coupling ratio as a ~unction of the input pump power, Ppc (in watts) necessary to achieve an optical gain which will compensate for to~al loop losses, for various values of fiber attenuation (~) assuming, in accordance with Equation ~, that the coupler ~0 exhibits substantially the same coupling ratios for the pump light and the signal light.
It was also assumed that the loop length (L) was 810 meters. As shown in Figure 8, the optical input pump power (Ppc) required for loop loss compensation varies dramatically with the coupling ratio (n). Indeed, the required pump power (Ppc) increases toward infinite values for very low coupling ratios and for very high coupling ratios. For intermediate coupling ratios, the required pump power Ppc decreases dramatically. Further, there is an optimum coupling ratio for e~ch value of fiber attenuation . In this exemplary set of curves, the curve ïO corresponds to a fiber attenuation of 1.8 dB per kilometer, while the curve 72 and 74 correspond to attenuations of 1.0 dB per kilometer and 0.5 dB per kilometer, respectively. It may be seen that, in general, the required pump power Ppc decreases as the fiber attenuation decreases. Further, there is an optimum coupling ratlo for each of the curves 70, 72, 74, corresponding to the point of zero slope, at which the required pump power Ppc is a minimum. Significantly, as the fiber attenuation decreases, these optimum coupling ratios increase, whereas the required pump power Ppc decreases. Thus, by properly selecting the coupling ratio for the particular fiber utilized, the required pump power Ppc can be reduced and minimized.
In the yresent invention, it is preferable that the pump power utilized be nc more ~han about 2 watts. It has s~

been found, for example, that pump powers in excess of 2 watts tend to cause "burn out" of the coupler 20~ in that the index matching oil between the blocks 53 tends to burn, thereby damaging the coupler. Although such coupler "burn out" might be alleviated by fusing the fiber together without oil therebetween, high optical powers are disadvantageous for other reasons. For exa~ple, they may cause undesirable non-linear effects (e.g., multi~Stokes effect) in the fiber 10~ Further, high power laser instability may cause enhanced amplitude variations in the output pulse train 26, as opposed to lower power lasers.
Moreover, such amplitude instability of high power lasers tend to cause the circulating Stokes background waves to build up very rapidly in the fiber loop. This buildup of t5 the Stokes background can significantly reduce the signal to noise ratio of the output pulses and/or cause coupler burn out.
Accordingly, the coupling ratio of the coupler 20 should be selected such that the input pump signal 33 optical power Ppc required for loop loss compensation is no greater than about 2 watts. This relationship may be defined by rewriting Equation 6 as:

A (asL - log n) T-~ ~ (7) It will be recognized that Equations 6 and 7 assume that the coupling ratios for the pump light and the signal light are the same. Thus, these equations are accurate to a good approximation only in the event that a "standard"
coupler is utili~ed.
If the signal pulse coupling ratio and the pump signal coupling ratio are not close to e~ch other, e.g., as in the case of a multiylexing coupler, Equations 6 and 7 may ~26~

not be sufficiently accurate. In such event, the pump power Ppc required for constant amplitude output pulses 26 should be determined in accordance with Equation 5, above, which may be rewritten to express the signal coupling ratio (~s) as a function of the pump coupling ratio (~p), with the pump power Ppc as a parameter, as follows:

~s = exp [ asL - P

Equa~ion 8 defines a family of curves, as shown in Figures 9a and gb. Figure 9a illustrates the signal coupling ratio (~s) at the signal wavelength of 1.12 microns as a function of the pump coupling ratio (~p) at the pump wavelength of 1.064 microns for various values of Ppc~ with a fiber attenuation of 1.8 dB per kilometer.
Figure 9b shows the same relationships as Figure 9a, but for a fiber attenuation of 0.5 dB per kilometer. As in Figure 8, it was assumed in Figure 9a and 9b that aS=ap=a, which is a very good approximation. It may be seen from the curves of Figures 9a and 9b that the pump power Ppc is, in general, at a minimum when the signal coupling ratio is close to one and the pump coupling ratio is equal to zero. Indeed, for any given signal coupling ratio, the pump power Ppc for that coupling ratio is at a minimum when the pump coupling ratio is equal to zero.
Conversely, for any given pump coupling ratio, the pump power Ppc decreases as the signal coupling ratio increases.
It will be understood that a high signal coupling ratio advantageously decreases the round-trip loop losses, since less optical power is coupled from the loop to form the constant amplitude output pulses 26. Although it is advantageous to select the signal coupling ratio so that it i8 close to unity, it should preferably not be exactly unity, otherwise there will be no light coupled from the loop to form t~e outpu~ pulse train ~6. On the other 5~

hand, the pump signal ratio should preferably be zero, or close to zero, so as to couple maximum pump power to the loop, and thereby efficiently use such pump power to cause Raman gain in the loop.
S From the foregoing, it will be understood that, although the invention ~ay be implemented utilizing either a standard coupler or a multiplexing coupler, a multiplexing coupler is advantageous in that it provides the gain necessary to achieve constant amplitude output pulses 26 at a lower optical power of the input pump signal 33. Ln the case of a multiplexing coupler, the signal coupling ratio and pump coupling ratio may be selected independently of each other so as to achieve low loop loss while still coupling substantially all of the pump light signal to the loop for maximum pumping efficiency. In the case of a standard coupler, the pump and signal coupling ratios are close to each other, and thus, a compromise must be drawn between the loop losses and the amount of pump power coupled to the loop. As discussed in reference to Figure 8, there is a particular coupling ratio compromise which makes it possible to minimixe the pump power Ppc~
Regardless of whether a standard coupler or a multiplexing coupler is used, the coupling ratios should preferably be selected so that the pump signal 33 optical power Ppc is no greater than 2 watts, for the reasons discussed above. From Equation 8, it may be found that this result is achieved when the signal coupling ratio is related to the pump coupling ratio as follows:

~s ~ exp [ asL ~ A (1 -~pL)]

Although the pump power required for cons~ant amplitude output pulses 26 may be readily calculated from the foregoing equations, it may be advantageous to adjust the pump power Ppc downwards very slightly to suppress any build-up of the Stokes background waves. As discussed previously, these Stokes background waves are amplified with the signal pulse, and their growth can significantly reduce the signal to noise ratio. It is believed that growth of the Stokes background waves is related to the stability of the pump laser, such that the greater the pump signal fluctuations, the higher the rate of growth of the Stokes background waves. For pump lasers having amplitude fluctuations of no more than about 10% from a nominal value, the amount of downward adjustment of the pump power necessary to suppress build-up of the Stokes background waves is typically very small, such that the output pulses 26 are still substantially constant in amplitude over a large number of circulations. It is expected that use of highly stable pump sources at appropriate wavelengths will substantially reduce any problems relating to build-up of the Stokes background waves, and will permit the memory of the present invention to achieve millions of circulations without significant degradation of the recirculating signal pulses and build-up of background noise. E'urther, due ~o the low pumppower requirements of the present invention, it is expected that the invention may be implemented using high power laser diodes. Such laser diodes are advantageous in that, through proper control of the driving current, the amplitude of the laser output can be made highly stable.
In addition, laser diodes have the advantage of permitting the invention to be compactly packaged in a relatively small unit.
Although preferred embodiments of the invention utilize stimulated Raman scattering, it will be recognized that other types of scattering processes may be used ~2~

alternatively. For example, those skilled in the art will understand that the invention may be implemented utilizing stimulated Brillouin scattering, or four photon mixing.
Stimulated Raman scattering is preferred becawse (1) it does not have any phase matching recluirements, (2) it can be implemented in both forward and backward configurations, and (3) the ~aman gain curve is relatively large so that frequency matching betweert the signal pulse and the pump generated Stokes light does not have to be as precise, thereby making it possible to use a signal pulse having a larger bandwidth~ Additional advantages of Raman scattering will be apparent to those skilled in the art~
An exemplary embodiment corresponding to the arrangement shown in Figure 1 was tested. In the experimental set-up, the signal generating loop 44 (Figure 2) was formed from 1 km length of non-polarization holding fiber. The pump pulses from the laser 42 (Figure 2) had a peak power of 120 watts and a repetition rate of 50 Hz, thus generating 500 n~ signal pulses 21. Each signal pulse 21 had an intensity of about 50 mW.
The loop 14 (Figures 1 and 7) was formed of 810 meters of 6 micron core diameter single mode non-polarization holding fiber. The coupler 20 was a standard coupler presenting coupling ratios of 0.48 and 0.66 at the pump and Stokes wavelengths, respectively.
The pump laser 34 was a 2 watt cw Nd:YAG laser. The output of this laser was chopped in 0.5 ms pulses synchronized to the output of the Q-switched YAG laser used for pulsed signal generation. The intensity of the input pump signal 33 was about 708 mW.
The fiber attenuation was 2.26 dB per km, the interaction area was 38.3 ~m2, and the interaction length was 676 meters. The Kaman gain coefficient was experimentally determined to be 8.5 x l0-14 meters per watt.

~2~5~

A total of about 13~ pulses of substantially constant amplitude were observed during the duration of the pump signal (0.5 ms). This corresponds to an optical path of nearly 105 km or a time delay of 500 microseconds between the first output pulse and the last output pulse.
Accordingly, the embodiment of Figure 1 of the present invention provides an optical memory having a significantly longer lifetime than prior art optical memories. Further, it is believed that this is the first optical memory utilizing in-line amplification through the non-linear properties of optical fibers.
As discussed above, it is important that the pump power in the recirculating loop be maintained at or below the cri~ical pump power in order to minimize the build-up of ,Stokes noise. It has been established experimentally that pump power fluctuations about the critical level cause gain fluctuations in the recirculating loop that result in the dramatic build-up of the Stokes intensities. Although instability of the pump source is a principal cause of such gain fluctuations, it has been found that gain fluctuations in the recirculating loop can also be attributed to phase noise in the pump power in the loop caused by the interaction of the pump signals circulating in the loop with pump signals entering the loop from the pump source. This phase noise is associated with a low coherence optical signal, such as is preferred for the pump signal in the present invention, and results in a s~ructured intensity noise in the pump power in the loop. This intensity noise i9 minimal at low pump signal coupling ratios, and thus, the noise can be avoided through use of a multiplexing coupler having a pump coupling ratio of zero. ~lowever, for pump coupling ratios substantially greater than zero, particularly intermediate pump signal couplin~ ratios, such as those used for some embodiments of the present invention, pump power fluctuations due to phase noise can result in broadband ~L26~

-3g-modulation of the gain in the loop which causes significant signal noise and build-up of Stokes noise.
Figure 10 illuscrates a preferred embodiment of the above described invention which reduces the phase noise and thus reduces the build-up of Stokes noise in the loop. The phase noise reduction is accomplished by utilizing a pump signal modulation scheme in which the pump power is input in the form of pulses which have a duration and are spaced in time in relation to the loop transit time to reduce the in~eraction between the recirculating pump light and the pump light input to the loop.
The embodiment of Figure 10 comprises a signal generator 100, shown in dashed lines, which corresponds generally to the signal generator 22 described above in connection with Figure 2. The signal generator 100 advantageously comprises a Q-switched Nd:YAG laser 102 which advantageously operates at a wavelength of 1~064 microns, for example, to provide pump light; an optical fiber signal generating loop 106 that receives the pump light from the laser 102; a lens 110 through which the light generated by the laser 100 i8 input to the signal generating loop 106; a lens 112 through which the light from the signal generating loop 106 is output; and an interference filter 120. The signal generator 100 operates in a manner similar to that described above in connection with Figure 2 to produce a pulse signal. The 1.064 micron pump light entering the signal generating loop 106 stimulates 1.12 micron Stokes waves which exit the loop 106 through the lens 112. In an exemplary embodiment, the Q-switched laser 102 is operated at a low repetition rate, for example 31 Hz, and with a relatively low duty cycle ~o generate relatively narrow pulses, for example 280 nanosecond pulses.
The interference filter 120 has a 30 nanometer bandwidth and suppresses the 1.064 micron pump signal 5~L

pulses and transmits only the 1.12 micron Stokes signal generated in the signal generating loop 106. The resulting 1.12 micron signal pulses from the interference filter 120 are input through a lens 132 into a first end of a fiber 134 which ~orms ~ recirculating loop 130. The recirculating loop 130 generally corresponds to the recirculating loop lO of Figure 1. The fiber 134 preferably comprises a continuous, uninterrupted length of a single-mode optical fiber 134. A coupler 136 is interconnected in the manner described above in connection with Figure l so that the fiber 134 forms the loop 130.
Preferably, the fiber 134 is a nonpolarization-preserving single-mode fiber having a 6 micron core and a 720 nanometer cutoff wavelength. The coupler 136, ~hich generally corresponds to the coupler 20 in Figure 1 and Figure 3, has four ports 1, 2, 3, and 4, and has a coupling ratio greater than zero. The signal propagates in the :~iber 134 from the interference filter 120 to the coupler 136 and enters the coupler 136 via the port l. A
coupled portion of each 1.12 micron signal pulse entering the coupler 136 via the port 1 from the interference filter 120 exits the coupler via the port 4 as a first 1.12 micron signal output pulse~ An uncoupled portion of the 1.12 micron signal pulse entering the coupler via the port I exits the coupler 136 via the port 3 and travels - around the loop portion of the fiber 134 and reenters the coupler 136 via the port 2. An uncoupled portion of the 1.12 micron signal pulse entering the coupler 136 from the loop portion via the port 2 exits the coupler 136 via the port 4 as a second 1.12 micron signal outpuc pulse, and a coupled portion of the 1.12 micron signal pulse entering the coupler 136 via the port 2 exits the coupler 136 via the port 3 and recirculates through the loop portion of the fiber l30. Thus, each 1.12 micron input signal entering the loop 130 via the port 1 of the coupler 136 will recirculate in the loop 130 as described above so ~6~5~

that a portion of the signal is output after the completion of each transit of the loop 130 of the fiber 134. The output portions of the circulatin~ signal comprise a series or train of 1.12 micron signal outpu~
S pulses that are spaced apart in time by an amount equal to the loop transit time.
A loop pump source 140 is provided that generates a pump signal at a wavelength which causes Raman gain, which, for a fused silica fiber and a signal wavelength of 1.12 microns, is 1.064 microns. The pump signal from the loop pump source 140 is input to the loop 130 via the port 4 of the coupler 136 so that the pump signal recirculates around the loop 13~ in the direction opposite the direction of recirculation of the 1.12 micron signal pulses described above (i.e., the pump signal is counterpropagating with respect to the 1.12 micron signal pulses). The pump signal causes stimulated 'Kaman scattering in the fiber 134 of the loop 130 which amplifies the recirculating signal pulses as described above. Those skilled in the art will understand that Raman gain can be produced in fibers at a number of frequencies in which the pump frequency is separated from the signal frequency by the amount o~' the Raman shift (e.g., 13-14 Terahertz). The choice of frequency for the described em'bodiment was due in part to the ready avai'lability of a laser pump source at 1.064 microns.
In the embodiment ~hown in Figure 10, the loop pump source 140 comprises a polarized Nd:YAG continuous wave (CW) laser 150 which operates at 1.064 microns. The output o~ the CW laser 150 is passed through an acousto-optic modulator 152 which modulates the amplitude of the pump light. The acousto-optic modulator 152 is driven by an enable pulse generator 154 which is controlled by a gate generator 156. The gate generator 156 is controlled by an amplifier 158 which is connected to the output of a photodetector 160. The photodetector 160 receives light ~6 ~

from a beam splitter 162 which is positioned in the optical path of the light from the ~-switched laser 10~ in the generator 100 so that the photodetector 160 provides an active output signal when the Q-switched laser 102 S generates the pump signal for the signal generating loop 106 . The active output signal ~rom the amplifier 158 causes the gate generator 156 to generate an active gating pulse which is thus synchronized with the signal pulses which are produced in the signal generating loop 106. The gate generator 156 generates a gating pulse having an adjustable pulse width. Thus, the gate generator 156 operates in a manner similar to a one-shot multivibrator to produce a single gating pulse which begins in synchronism with the pump signal generated by the Q-switched laser 1~ and ends in accordance with the selected pulse width. The gating pulse from the gate generator 156 controls the enable pulse generator 154.
When the gating pulse is active, the enable pulse generator 154 generates a series of periodic enabling pulses having a selectable width (i.e., time duration) and a selectable repetition rate or periodicity (i.e., time duration between the beginning of a pulse and the beginning of the next successive pulse). The periodic enabling pulses from the pulse generator 154 control the acousto-optic modulator 152. When the periodic enabling pulses are active, the acousto-optic modulator 152 gates the pump signal from the CW laser 150 to the output of the pump source 140 to create a series of pump pulses which enter the loop 130 via the port 4 of the coupler 13S.
When the periodic enabling pulses are inactive, the acousto-optic modulator 152 blocks the pump signal from the CW laser 150 so that no pump signals are output from the pump source 14U. 'rhe series of enabling pulses and thus the series of pump pulses will continue to occur so long as the gating pulse generated by the gate generator 156 is ac~ive. These pump pulses are illustrated in -~3-Figure 11a. In the preferred embodiment of Figure 10, the pulse widths and repetition rates of the periodic enabling pulses from the pulse generator 154 and thus the widths and repetition rates of the pump pulses gated through the acousto optic modulator 152 are preferably selected so that the pump pulses entering the loop 130 do not overlap with the first few recirculations of pump pulses which have previously entered the loop 130 as will be discussed more fully below in connection with Figures 1la-1ld.
In the preferred embodiment of the inven~ion illustrated in Figure 10, the pump source 140 is optically isolated from power backscattered by the loop 130 with a Glan polarizer 170 and a quarter-wave plate 172. A fiber polarization controller 174 is advantageously placed on the pump input end of the loop 130 and is used to adjust the backscattered light polarization and thus maximize the a~ount of backscattered power which is rejected by the Glan polarizer 170. Since the optical fiber 134 is preferably nonpolarization~preserving, and since the coupler coupling ratio is substantially polarization-independent, the polarization controller 174 does not sub.stantially affect pump power input to the loop and thus does not substantially affect the signal recirculations in the loop 130. A lens 176 is also preferably included to direct the output signal from the polarization controller 174 to the quarter-wave plate l72.
The Glan polarizer 170 also acts as a beam splitter that directs a portion of the output signal from the loop 130 to a monochromator 180. The output of the monochromator 180 is directed onto a photodetector 182 which provides an electrical representation of the output signal from the loop 130. In the experimental set up, the electrical output from the photodetector 182 is provided to an oscilloscope 184 so that the signal can be analyzed. When incorporated into a system, the electrical output from the photodetector 182 can be provided to other ~ ~ 6 -4~-portions of the system to provide a representation of the pulses output from the loop. Alternatively, the optical output pulses from the monochromator 180 can be provided directly to optical fiber components of a system.
As mentioned above, phase noise reduction is accomplished in the embodiment of Figure 1~ by modulating the light produced by the pump source 140 to provide a series of pump pulses having a duration and spacing selected in relation to the transit ~ime of the fiber loop 130 to reduce the interaction between the pump light recirculating in the loop 130 and the pump light input to port 4 of the coupler 136. A specific example o a pulsed wave form utilized to reduce phase noise, and thereby suppress build up of the Stokes waves, is illustrated in Figures 11a and 11b. Figure 11a depicts an exemplary train of pump pulses referred to herein as "pump input pulses" which are produced by the source 140 and input to port 4 of the coupler 136. These pump input pulses, which are shown in Figure 11a with respect to time, include a first pump input pulse 202, a second pump input pulse 204, a third pump input pulse ~06, and fourth, fifth, sixth and seventh pump input pulses 208, 210, 212 and 214, respectively. Each pump input pulse has a width (i.e., time duration) of approximately one-fourth the loop transit time. l'he loop transit time is shown as tdelay~
Thus, each pump pulse has a width of approximately 1/4 tdelay~ The pump input pulses 202-214 have a periodicity (i.e., time between the beginnings of successive pump input pulses) selected to be substantially equal to 3/4 tdelaY' When an input pump pulse reaches port 4 of the coupler, a portion is coupled ~o the loop 130 to form a circulating pump pulse which is delayed in time on each circulation by an amount tdelay~ A circulating pump pulse which has circulated once through the loop 130 will be referred to as a "once-delayed" pump pulse. Similarly, a pulse which has circulated twice through the loop 130 will be referred to as a "twice-delayed" pump pulse, and so forth, Figure 11b illustrates a train of once-delayed pump pulses with respect to time and with respect to the pump input pulse train of Figure lla. The first pump pulse entering the loop (e.g., the first pump input pulse 202) circulates through the loop and arrives back at the coupler 136 after a delay of tdelay as a once-delayed first pump pulse 202'~ Since the delay time through the loop 130 is greater than the time between pump input pulses, the once-delayed first pump pulse 202' does not overlap with the second pump input pulse 204 and instead arrives at the coupler 136 with a delay o~ tdelay/4 with respect to the second pump input pulse 204. Similarly, the second pump input pulse 204 will be delayed by the time tdelay and will arrive at the coupler 136 as a once-delayed second pump pulse 204' with a delay o~ tdelay/4 with respect to the third pump input pulse 206.
2~ Additional once-delayed pump pulses 20~', 206', 208', 210', etc., in the train of once-delayed pump pulses will arrive at the coupler 136 with corresponding delays with respect to the pump input pulses in the train of pump input pulses. None of the pulses in the train of once-delayed pump pulses overlap with any of the pulses in the train o~ pump input pulses.
Figure llc illustrates a train of twice-delayed pump pulses with respect to time and with respect to the pump pulse trains of Figures lla and llb. ~fter an additional delay o~ tdelay with respect to the once-delayed first pump pulse 202', a twice-delayed first pump pulse 202"
arrives at the coupler 136. The twice-delayed first pump pulse 202" does not o-~erlap with any of the pump input pulses (Figure 1la) or with any of the once-delayed pump pulses (Figure 11b). Similarly, a twice-delayed second pump pulse 204", a twice-delayed third pump pulse 206", ~2~

and so forth, do not overlap with any of the pump input pulses or wi~h any of the once-delayed pump pulses.
Figure lld illustrates a train of thrice-delayed pump pulses with respect to time and with respect to the pulse trains of Figures 11a, llb, and 11c. After an additional delay of tdelay with respect to the twice-delayed first pump pulse 202", a thrice-delayed first pump pulse 202"' arrives at the coupler 136. As illustrated, the thrice-delayed first pump pulse 202"' overlaps the fifth pump input pulse 210 in the pump input pulse train (Figure l l a) . Subsequent thrice-delayed pump pulses 204"', 2~6"', etc., overlap with the sixth, seventh, and so forth, pump input pulses. Although the thrice-delayed pump pulse train overlaps the pump input pulse train, the effec~ of the thrice-delayed pump pulse train is minimized because the pulses in the thrice-delayed pump pulse train of Figure 11d are substantially attenuated by the recirculations in the loop 130. For example, in one embodiment, the pump coupling ratio ~p and the loop attenuation for the pump signal may be such that each pulse is reduced by one-half during each recirculation (i.e., the intensity of the once-delayed first pump pulse 202' is one-half the intensity of the first pump input pulse, the intensity of the twice-delayed first pump pulse 202" is approximately one-half the intensity of the once-- delayed first pump pulse 202', etc.). Thus, for this example, the thrice-delayed first pump pulse 202"' has an intensity of one-eighth the intensity of the fifth pump input pulse 210. Thus, the interaction between the pulses in the pump input pulse train and the circulating pulses in the thrice-delayed pump pulse train is substantially reduced, and the pump intensity noise caused by the interference of the recirculating pump power with the input pump power, both of which have a high phase noise, is also substantially reduced. The pump intensity noise and power fluctuations can be fur~her reduced by -~7-increasing the number of pulses before overlapping occurs, which can be done by reducing the pulse widths of the pump pulses. For example, in an alternative embodiment (not shown), each pulse can have a pulse width of ~delay/5 and a repetition rate o~ 4/5 tdelay~ In this alternative embodiment, there will be no overlap of a circulating pulse with a pump input pulse until the sixth pump input pulse arrives at the coupler 136. Using the above example, at the time that the first circulating pulse overlaps with the sixth pump input pulse the first circulating pulse would have traversed the loop four times, thereby reducing i~s intensity to approximately 1/16 of the intensity of the input pulses.
The effective pump power in the loop Peff is a summation of the effects of each of the recirculating pump pulses in the loopO It can be shown that for the example described above, the steady-state pump power corresponding to Ppc in ~he continuous pump embodiment i9 ~1/3)peff/(1-np)~ In the embodiment of Figure 10, with a pump coupling ratio of 0.42, the steady-state pump power is 0.57 times Peff. In an experimental set up used to test the present invention, the gate generator l 56 was adjusted to provide a pulse width of between 1 millisecond and 3 milliseconds. The pump pulses generated by the pulse generator l 54 had widths of 930 nanoseconds and were spaced apart by 2.78 microseconds corresponding to 1/4 tdelay and 3/4 tdelay for a delay time in the loop 130 of approximately 3. 72 microseconds. Using this modulation, the measured effective pump power PeiE was 1.08 watts, corresponding to coupling ratios np = O. 42 and ~s =
0.6S. This corresponds to a theoretical value of approximately 530 milliwatts for the steady state power Ppc for the same coupler ratios. This is in general agreement with the calculated required peak power (i.e., 530 milliwatts X 1/0. 57 = 922 milliwatts. A 3 millisecond delay, corresponding to more than 800 recirculations in 5~

-4~-the loop 130 and thus corresponding to an optical fiber path longer than 600 kilometers, was obtained with a pump source having a 3 percent r.m.s. power stability. It is believed that additional recirculations can be obtained by S using a power supply with greater stability.

Claims (11)

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:

1. A fiber optic recirculating memory, comprising:
a length of optical fiber for receiving an optical signal pulse, said fiber forming a loop;
a coupler, having a coupling ratio, for coupling said optical signal pulse to said loop to provide a circulating optical signal pulse which makes plural circulations in said loop, said coupler additionally outputting a portion of said circulating signal pulse from said loop on each of said plural circulations to provide a series of output pulses; and a light source for generating an input pump signal comprising a series of pump pulses, said coupler serially inputting each of said pump pulses to said fiber loop to circulate therein and optically pump said optical fiber to generate photons in said loop for amplification of said circulating optical signal pulse during each of said plural circulations of said signal pulses, the optical wavelength of said pump pulses and said optical signal pulse selected such that said photons have the same optical wavelength as said optical signal pulse, the coupling ratio of said coupler greater than zero for said pump pulses such that each of said pump pulses makes plural circulations in said loop, the length of said loop providing a loop transit time for said pump pulses such that each of said pump pulses of said series of pump pulses makes plural circulations in said loop before overlapping with any subsequent pump pulse in said series of pump pulses to allow the amplitude of said pump pulses to decay prior to said overlapping and thereby reduce phase noise effects in said loop.

2. The fiber optic recirculating memory as defined in claim 1, wherein said pump pulses are equally spaced and have the same time duration so that said pump signal is periodic, the period of said pump signal being less than the amount of said loop transit time.

3. The fiber optic recirculating memory as defined in claim 2, wherein said time duration of each of said pump pulses is substantially less than said loop transit time.

4. The fiber optic recirculating memory as defined in claim 3, wherein said period of said pump signal is substantially equal to an integer multiple of the time duration of each said pump pulse.

5. The fiber optic recirculating memory as defined in claim 4, wherein the sum of the period and the time duration is substantially equal to said loop transit time.

6. A method of producing a series of output pulses, comprising the steps of:
providing a loop comprised of a length of optical fiber, said loop having a loop length;
coupling an optical signal pulse having an optical signal wavelength into said loop to circulate plural times in said loop, said circulating optical pulse propagating around said loop in a signal pulse loop propagation time determined by said optical signal wavelength and said loop length;
tapping a portion of said circulating signal pulse from said loop on each of the circulations to provide said series of output pulses;
coupling a series of input pump pulses having an optical pump wavelength into said loop of fiber to pump said fiber to generate photons in the loop at said optical signal wavelength for amplification of said circulating signal pulse during each of said plural circulations, said pump pulses propagating around said loop in a pump wavelength and said loop length, said input pump pulse having a pump pulse time duration; and spacing each pump pulse apart in time from the next succeeding pump pulse in said series of pump pulses so that each pump pulse makes plural circulations in said loop before overlapping with any subsequent pump pulse to allow the amplitude of each of said pump pulses to decay prior to said overlapping, thereby reducing phase noise effects in said loop.

7. The method defined in claim 6, wherein said pump pulses are spaced apart in time by a selected time delay so that said pump pulses are periodic, said method additionally comprising the step of selecting said time delay so that said pump pulse has a period smaller than said pump pulse loop propagation time.

8. The method defined in claim 7, additionally comprising the step of selecting said pump pulse time duration of each pump pulse to be substantially smaller than said pump pulse loop propagation time.

9. The method defined in claim 8, additionally comprising the step of selecting said period of said pump pulses to be substantially equal to an integer multiple of said pump pulse time duration.

10. The method defined in claim 9, additionally comprising the step of selecting said pump pulse time duration and said pump pulse period so that said sum of said pump pulse period and said pump pulse time duration is substantially equal to said pump pulse loop propagation time.

11. In a fiber optic recirculating memory device having a loop comprised of optical fiber, a method of pumping said memory device to cause amplification of a signal propagating in said loop, said method comprising:
inputting a series of pump pulses into said loop for recirculation therein;
utilizing said pump pulses to optically pump said optical fiber to generate photons in said loop to cause said amplification of said signal; and spacing each pump pulse apart in time from the next succeeding pump pulse so that each pump pulse makes plural circulations in said loop before overlapping with any subsequent pump pulse.