The invention concerns a method of correlating two or more optical signals as stated in the introductory portion of claim 1, as well as a device for correlating two or more optical signals and of the type stated in the introductory portion of claim 2.

Within the field of optical high rate systems it is often necessary to obtain knowledge of the shape of the signal pulses. If in an optical communications system it is no longer possible to detect a signal, the reason will fre¬ quently be interference with the signals en route in the system, e.g. dispersion fading. Usually, it is not pos¬ sible to use conventional measuring equipment, because typically the pulse widths of high rate systems will be less than 500 picoseconds and are expected to gradually approach pulse widths of the order of 5 femtoseconds.

It is well-known to obtain this information by correlating two signals - either two copies of the same signal (auto¬ correlation) - or two signals, where one is to be measured and the other is a well-known reference signal (cross- correlation). This is obtained by using a photoactive ele¬ ment in the form of a frequency doubling crystal which generates second harmonic oscillations of the incident light signal. This entails that the incident light, which usually has a wavelength of e.g. 1550 nm, is converted to light at a wavelength of 775 nm.

The intensity of the second harmonic oscillations depends upon the intensity of the incident light e.g. to the se¬ cond power. When the two signals are delayed temporally during the measurements, e.g. by changing the distance

travelled, the correlation between the two signals can be determined, and thus the pulse shape of the measured sig¬ nals can be determined by applying mathematical analysis to the measurement result.

However, the use of the frequency doubling crystals is associated with a plurality of drawbacks, because the com¬ ponents are quite expensive. The cystals are vulnerable to scratches and mechanical stress. Some of the crystals are moreover hygroscopic. For the crystals to be coupled to the optical system, it is necessary to use expensive bulk optical devices. To achieve the necessary phase matching, the components must be aligned very accurately, which means the use of an optical table in practice. This tech- nique is therefore only useful in connection with labora¬ tory tests and can be used only with great difficulty in connection with measurements in the field.

The article "Autocorrelation of short pulses using a single-mode fiber" by Ulf Osterberg and Walter Margulis in IEEE Journal of Quantum electronics, vol. 24, No. 10 Oct. 1988 discloses an autocorrelation principle, comprising measuring transmitted light from a single mode fibre, said light being provided by interaction between the incident light in the fibre and the electrons in the germanium atoms of the core. The germanium atoms are used for ob¬ taining a refractive index difference between the core and the cladding. The SHG principle has several drawbacks, the most essential one being that it operates only at three wavelengths, viz. 647.5 nm, 1064 nm and 1319 nm, and none of these is readily useful in connection with 1550 nm, which is the preferred wavelength in optical communica¬ tion. Another considerable drawback is that there is no linear correlation between the intensity of the input sig- nal and the fluorescence. The SHG autocorrelation is limited to a narrow wavelength region determined by the

phase matching requirements. This type of autocorrelation cannot be applied to pulses which are shorter than 1 ps.

The object of the invention is therefore to provide a method, which makes it possible to determine the correla¬ tion between two optical signals, and the method must be capable of being adapted to arbitrary wavelengths in the optical region, without using a crystal element which re¬ quires an expensive measurement setup.

This object is achieved as stated in the characterizing portion of claim 1. The method comprises using an active, optical wave conductor for generating fluorescence. The wave conductor is doped with active ions, which are known i.a. from optical amplifiers.

It has been found possible to replace the frequency doub¬ ling crystal by an active, optical wave guide, e.g. in the form of an active, optical fibre, the wave guide transmit- ting fluorescence which is intensity related to the arriv¬ ing signal. When this fluorescence is detected in spectral bands, e.g. in the blue region of the visible light, an image of the strength of the arriving signal can be formed. Current distance delay of the two correlated sig- nals makes it possible to generate an image of the corre¬ lation between the signals. The fluorescence is preferably produced by a multi-photon process in the active ions in the wave guide. The intensity of an N-multi-photon process ddeeppcends upon the N power intensity of the arriving sig- nal,

The invention moreover concerns a device as stated in the characterizing portion of claim 2, which makes it possible to perform the method stated in claim 1.

If the wave guide is an active, optical fibre, the detec¬ tor is preferably arranged as stated in the characterizing portion of claim 3. The doping material is preferably se¬ lected as stated in claim 4, providing a strong intensity- related fluorescence. The signals preferably have a pulse width as stated in claim 5, but other pulse widths may be used.

The distance delay of the optical signals is preferably provided by transmission fibres as stated in claim 6, since this enables easy coupling to the wave guide if it is an active, optical fibre. Then, a piezoelectric crystal may advantageously be arranged in one transmission fibre to control the delay, as stated in claim 7.

The device defined in claim 8 splits a signal into two identical signals, which makes it possible to autocorre- late the signal concerned. It is hereby possible to obtain information on the envelope of the signal as well as the frequency chirp of the enveloped carrier signal.

The device defined in claim 9 enables using a reference signal well-known in advance, by means of which the cross- correlation between the reference signal and the measure- ment signal is determined. If the reference signal is a delta signal, the cross-correlation between the two sig¬ nals will be an image of the signal to be determined. The device defined in claim 10 uses the cross-correlation between two signals, one of which is a reference signal, for generating a clock signal for the other signal. This technique may be used in a receiver unit for optical high rate signals. On the basis of a received data signal it is hereby possible to generate a clock signal for it.

The invention will be described more fully below in con¬ nection with preferred embodiments and with reference to

the drawing, in which

fig. 1 shows an energy level diagram of Er 3+ ions;

fig. 2 is an outline diagram of an embodiment of an auto- correlator according to the invention;

figs. 3 and 4 show the autocorrelation of a signal col¬ lected with the principle shown in fig. 2 and a theoreti- cally computed image of the autocorrelation of the same signal, respectively;

fig. 5 is an outline diagram of an embodiment of a cross- correlator according to the invention; and

fig. 6 is an outline diagram of a preferred embodiment of a clock generator according to the invention.

Fig. 1 shows the level diagram of Er 3+. The preferred embodiment of the invention uses light having a wavelength in the region around 1550 nm (1500-1600 nm), which is used for optical communication. When Er is used as the active material and light at 1550 nm as excitation energy, exci- tation of Er 3+ i obtained to a first excited level desig-

4 nated I.,,,. Owing to the long life (about 11 ms) of this level a considerable population is obtained in this level even for small average powers (typically 0.1 mW) of light at 1550 nm. Er 3+ can now be excited from this first ex¬ cited level by multi-photon processes. This type of pro- cesses is described in Fahrad Faisal: "Theory of Multi- photon Processes," Plenum Press. As an example, the dia¬ gram in fig. 1 shows a three-photon absorption process

4 from this first excited level I. .~ . Such a process re¬ quires simultaneous, coherent interaction of three photons with the Er ion; but it is not necessary that the Er ion includes energy levels corresponding to the absorption of

one or two photons. The Er ion will now be excited to a higher excited state, e.g. the level designated 4G _11/2 in the figure. From this level or after a radiation-less transition to a lower level, the Er ion can now transmit blue light having a wavelength around 400 nm.

The probability of the process described here and thus the intensity of the transmitted blue light depend upon the peak intensity of the incident light to the third power. However, for a multi-photon process to proceed, the arriv¬ ing photos must be in phase, since, otherwise, destructive interference will occur.

It is essential to note that the intensity of the blue fluorescence from the Er ions in the multi-photon process described here depends on the peak intensity of the inci¬ dent light to the third power. The process is not restric¬ ted to Er ions or light having a wavelength around 1550 nm. Most rare earths have several closely spaced levels which make them useful for generating light by multi- photon processes. Nor is it essential that a three-photon process is used. For example a two-photon process in erbium will lead to fluorescence at about 550 nm. The pro¬ bability of such a process depends on the intensity of the incident light to the second power. Generally, n-photon processes (n an integer) can be used in rare earths, where the intensity of the fluorescence light depends on the peak intensity of the incident light to the n power.

For example in Er doped fibres there are processes compet¬ ing with the mentioned two- and three-photon processes.

Cooperative upconversion, where two closely spaced Er ions

4 both excited to the level I., ,» can exchange energy, so that one ion decays to the basic state, I.,..-, while the other ion is excited to a higher level, where the electron either decays by transmission of IR light or decays with-

out radiation. Such a process greatly depends on the Er ion concentration. To avoid such processes an optical fibre or wave conductor must not be too highly doped with Er, since the individual Er ions can then exchange energy. Typically, the concentration of Er ions in an optical fibre is to be around 1.3 x 10 ¹⁸ cm-3. If an Er doped op¬ tical fibre is used for detection of light by a multi- photon process, typical parameters of the fibre will be: the fibre core will have a diameter of about 3 μm, and the refractive index difference between the core and the clad¬ ding will be about 0.01 to 0.03.

Fig. 2 shows an embodiment of an autocorrela or for detec¬ tion of pulse width and phase chirp. A periodic flow of pulses is transmitted into an optical fibre 21 called the input fibre. The flow of pulses is divided in a fibre op¬ tical coupler 24, such that equal amounts of the intensity of the light are transmitted into two other optical fibres 22 and 23, which constitute two arms of a fibre optical interferometer. A piezoelectric crystal 26 capable of stretching the fibre is inserted in one arm 22. This is merely an example of a method by means of which the op¬ tical distance in one arm may be varied with respect to the other. Other distance regulating elements may be used. The two pulse flows are united again with a second fibre optical coupler 25, but now the pulses in the two flows will be delayed with respect to each other owing to the different distances they have travelled. Finally, the light pulses are transmitted into an Er doped optical fibre 27. For example the blue light from a three-photon process is detected from here using a sensitive detector, e.g. a photomultiplier tube (PMT) 28. An optical filter 29 may be arranged in front of this so that only the blue light is detected.

The intensity of the blue fluorescence light detected in this manner now depends on the overlap which will exist between the pulses from the two flows in the interfero¬ meter. If the overlap is great, i.e. pulses from the two flows arrive simultaneously at the coupler 24 (the delay time is 0), there will be a high intensity of light and thus a strong blue fluorescence, but if the overlap is small (i.e. the delay time is great) no or only little light can be detected.

The autocorrelation function of the pulses can be detected by varying the optical distance in the one arm 22 in the fibre optical interferometer. During the detection the optical distance is varied in the one arm 22 in the fibre optical interferometer in a plurality of steps. The fluo¬ rescence intensity is detected at each step. To ensure correct measurement, the measurement in each step must not be of a shorter duration than the life of the fluores¬ cence, i.e. detection is to be performed at each point for a longer duration than 10 μs in the embodiment discussed. Typically, it is necessary to measure on a pulse train having at least one million pulses for each step, which, of course, depends on the repetition frequency.

An example of such a measurement is shown in fig. 3. This figure shows an experimental measurement, in which the fibre interferometer is replaced by bulk optical instru¬ ments. This setup is not illustrated, but corresponds in principle to the setup shown in fig. 2, but owing to the understanding the invention is explained in connection with a fibre optical interferometer. The actual detection is performed using an Er doped optical fibre. Fig. 4 shows a theoretical computation of the corresponding experiment. There will be a background of light for long delay times (0.8 ps), and for a delay time 0 there is a strong peak in the envelope of the measured light intensity. Owing to

different phase of the light through the light pulse used, something resembling a raised background will be seen for medium delay times, about 0.4 ps. The envelope function of the measured fluorescence provides a measure of the origi- nal pulse width. In this case it will be seen that the width of the autocorrelation function of the pulses is about 0.6 ps, and at the same time it is possible to give a measure of the phase chirp which is present in the ori¬ ginal pulse at the raised background for medium delay times.

Typical pulse widths that can be detected with the method described here is 100 ps to 0.05 ps.

When very short pulses, e.g. 0.1 ps, are detected, disper¬ sion in the fibre is to be taken into consideration, be¬ cause when passing through the fibre the pulse broadens. The fibres used can be designed such that the dispersion is minimum and does not destroy the pulse prior to the detection.

The theory of interferometrical autocorrelators prepared for bulk optical instruments and for frequency doubling crystals can be applied directly to the autocorrelator "type described here. This theory is described e.g. in K.L. Sala et al. IEEE, Journal of Quantum Electronics, Vol. QE- 16, No. 9, pp 990-996, 1980.

It should be mentioned e.g. that the contrast ratio, i.e. the radio of the peak at 0 in the delay time to the back¬ ground, depends on the type of multi-photon process which is used. The contrast radio is 2 ^{l ~} , i.e. for a three- photon process the contrast ratio is 32.

Fig. 5 shows the principle in a cross-correlator. Here the pulse is not correlated with itself, but with a known

other pulse. It is an advantage that the known pulse is short with respect to the pulse it is desired to measure, since the signal then obtained as fluorescence directly reflects the shape of the studied pulse. The test pulse may e.g. be a 0.5 ps pulse to measure the shape of a 50 ps pulse. In principle, a test pulse shaped as a delta func¬ tion is preferred, since the studied pulse then occurs directly by the cross-correlation.

The cross-correlator shown in fig. 5 corresponds in prin¬ ciple to the autocorrelator, there being merely used dif¬ ferent pulse flows in the two arms 41 and 42 in the inter¬ ferometer. Short pulses are used in the one arm 41 as test pulses to measure on the pulses which are transmitted into the other arm 42 of the interferometer. The pulse flows in the two arms 41, 42 are combined by a fibre optical coup¬ ler 45 and are directed into an active fibre 46. Of course, it must be demanded that the repetition frequen¬ cies of the two pulse trains are the same or at any rate commensurable. Detection of the fluorescence from the multi-photon process in the active fibre 46 takes place in the same manner as in the autocorrelator by means of a filter 47 and a sensitive detector 48.

Fig. 6 shows a special use of a cross-correlator. In opti¬ cal communications systems information is transmitted in the form of light pulses. These are transmitted with a firm repetition frequency. Bits in the form of 1 and 0 are transmitted as light pulses, so that e.g. l's are repre- sented by a light pulse, while an 0 is represented by no light pulse. When such pulses are to be detected, the arrival time of the pulses must be known. It is therefore necessesary to have a clock locally where the detection takes place, synchronized with the pulses to be received. This can be obtained by transmitting, by means of a fibre optical coupler 52, part of the bit flow from a fibre 51

into a first arm 54 in the cross-correlator. The second arm 55 of the cross-correlator is coupled to a pulse flow having a repetition frequency which is equal to or commen¬ surable with the bit flow. The local clock can now be syn- chronized with the delay time at which maximum fluores¬ cence is obtained from the cross-correlator, which is ge¬ nerally designated 511. This problem of making a clock re¬ generator is particularly important in very fast communi¬ cations systems where the repetition frequency is equal to or above 10 GHz. Since the life of the fluorescence is of the order of a few μs, it will be seen that the delay time is measured averaged over many bits, and the alternating 0's and 1's in the pulse flow will therefore not give rise to variations in the fluorescence output from the cross- correlator.