10MHz.
(c) University of Southampton 1994
All-fibre acousto-optic frequency shifters have in the past been based on fibres supporting two
modes. Resonant coupling between the modes occurs when their beat-length matches the
acoustic wavelength, the coupled light being
frequency-shifted [1-3]. These devices can also
function as tunable filters, switches and modulators. When they are used as single-sideband
frequency shifters, mode convertors and filters are necessary to separate the residual carrier
from the shifted signal [1], and give a single-mode output. Furthermore, the maximum frequency
shifts are limited acoustically (by the relatively large fibre diameter) to
10MHz.
We have previously reported acousto-optic interactions in the narrow uniform waist of a tapered single-mode fibre [4]. The taper waist is in fact a multi-mode waveguide, supporting many cladding modes which fill the fibre. Light entering the taper waist in the fundamental mode can be coupled to the second mode (and vice versa), with a frequency-shift, by a flexural acoustic wave. There is complete overlap between the acoustic and optical waves, and hence a low acoustic power requirement, in contrast to other schemes [1-3]. The design is versatile because the nature of the interaction region at the taper waist is fixed only when the taper is made, and it can be operated at acoustic frequencies of up to hundreds of MHz by making the waist diameter as small as a micron.
This single taper device suffers from the disadvantage that the frequency-shifted light coupled into the second mode is not guided by the single-mode fibre, and is therefore lost. It can be recovered if a tapered two-mode fibre is used; but it is then necessary to incorporate mode convertors and filters to yield a single-mode output, as with a frequency shifter made using ordinary untapered two-mode fibre [1].
We describe here a new design of acousto-optic frequency shifter, in which the single fibre taper is replaced with a ``null'' taper coupler. This is a fused taper directional coupler which is so phase-mismatched that the maximum coupling is zero. The mechanism of the acousto- optic interaction is unchanged, because the waist of the taper coupler is a similar waveguide to the waist of a single taper. However, the frequency-shifted light is not now lost, but emerges from one of the coupler's output fibres.
A fused taper coupler, or beam-splitter, is made by heating and stretching two parallel fibres together in a small flame. In general, some of the light entering the coupler in one fibre is cross-coupled to the other fibre (the coupled wave), while the rest remains in the first fibre (the throughput wave). If the coupler is made from a pair of identical single-mode fibres, any coupling ratio from 0% to the maximum of 100% is possible as the coupler is elongated. With dissimilar fibres (or if one has been pre-tapered), the maximum coupling can be less [5] than 100%. In an ideal null coupler, the fibres are so mis-matched that the maximum coupling ratio is effectively zero; that is, the passive null coupler does not function as a beam- splitter at all. Light launched in one fibre evolves adiabatically into just the fundamental mode of the coupler waist, emerging from the same fibre at the exit (Figure 1a). Similarly, light launched in the other fibre evolves into the second mode of the waist. This behaviour has been described as ``mode splitting'' in planar waveguides [6], and arises critically from the optical properties of the coupler's taper transitions.
Despite the lack of interaction between the light waves in the two fibres, they do interpenetrate and overlap at the waist of the null coupler. Hence a travelling flexural acoustic wave can cause resonant forward-coupling between the fundamental and second modes of the coupler waist, with a frequency shift, if the beat-length of the two modes matches the acoustic wavelength. The two modes emerge via different fibres, so if light enters in one fibre then a pure frequency-shifted wave leaves in the other fibre (Figure 1b - in this case, there is a down-shift). No mode convertors or filters are necessary to separate shifted and unshifted waves, because any residual unshifted light emerges from the first fibre.
We made a null coupler for 
= 633nm operation using a pair of dissimilar fibres, with
diameters of 60
m and 80m and cut-off wavelengths of 500nm and 650nm respectively.
The second fibre was not single-mode at 633nm, but in our experiments we always launched
light into the first fibre, and checked that the output from the second fibre was in the
fundamental mode. The fibres were held in parallel and then heated and stretched together in
a travelling flame. The final waist was 25mm long, uniform, and about 6m in diameter,
with short taper transitions each 25mm long. Control of these dimensions was achieved by
varying the flame's travel distance during coupler elongation [7]. The waist cross-section was
circular; this gives a good overlap between the two optical modes, is reproducible, and
resembles the waist of a single tapered fibre (which is simple to analyse
theoretically [4]). The
excess loss was about 0.1dB and the maximum coupling ratio was 1:400 (maximum coupling
ratios as small as 1:6000 were seen in other null couplers).
A flexural acoustic wave was excited in the coupler waist using a PZT disc with a concentrator horn, driven by an RF signal. The horn was fixed to the pair of untapered fibres at one end of the coupler (Figure 2), in such a way that the plane of the acoustic wave coincided with the plane of the coupler. The acoustic wave travels along the fibres, through the coupler's taper transition, and into the coupler waist. This arrangement has the advantages that the acoustic wave is focused by the transition, and is unidirectional in the interaction region. Light at 633nm from a polarised HeNe laser was launched into one input fibre of the coupler via a polarisation controller. The optical powers emerging from the two output fibres were monitored, while a resonance was sought by changing the RF frequency.
This polarisation dependence is undesirable, but can be masked by making the coupler waist slightly non-uniform. This broadens the resonance for each polarisation, because the resonance condition now varies along the waist. If this broadening exceeds the polarisation splitting, there will be an overlap where both polarisations are efficiently frequency-shifted. Such a frequency shifter would behave in an effectively polarisation-insensitive manner for restricted combinations of optical wavelength and acoustic frequency.
The frequency shift was measured by inserting the device (taking the output from the second fibre) into one arm of a Mach-Zehnder interferometer. A Bragg cell up-shifted the wave in the other arm by 80MHz. The detected beat signal was monitored on an RF spectrum analyser, Figure 4. The main beat component is visible near 82MHz, corresponding to a frequency down-shift equal to the acoustic frequency. Also visible above the noise floor are beat components with the carrier frequency (80MHz) and the image sideband (near 78MHz). These are both about 30dB below the principal component. The purity of this output was little changed when the drive voltage was reduced from the value for maximum conversion, though of course the total amount of light dropped. The output from the first fibre was unshifted, as expected. In contrast with the situation depicted in Figure 1b, here the input light excites the second mode of the coupler waist, and interacts with a counter-propagating acoustic wave. However, this reverses the sign of the frequency shift twice, so the net result is again a down-shift.
m in the diameter of the coupler waist is sufficient to account for this.
Indeed, the spectral bandwidth can be increased if necessary by deliberately making the
coupler waist non-uniform, as discussed above in the context of polarisation dependence.
This work was supported by the U.S. Department of the Air Force through the European Office of Aerospace Research and Development, London.