JPH0626937A - Measuring method of optical pulse waveform - Google Patents

Abstract

PURPOSE:To shorten operation time and operation quantity. CONSTITUTION:An input optical pulse is incident to a Michelson interferometer, and the interfered light is subjected to high speed Fourier transformation. A proper phase phie(omega) is given to a basic wave optical spectrum E0(omega) to perform reverse high speed Fourier transformation (S3), the transformation result E(t)phie(t) is subjected to square operation (S4) to calculate the spectrum of double higher harmonics (S5), and the calculated U(omega) is compared with the measured double higher harmonic spectrum U0(omega). When the error is a determined value or more, phase correction is conducted to return to step S2, and the above is repeated. A Boltzmann machine type neural network of multilayer structure is used for this phase correction, phases phie(omega0)--phie(omega255) corresponding to, for example, 256 spectra are inputted to correct all weighs of the neural network on the basis of the error in step S6, and phases phie'(omega0)--phix'(omega255) are obtained as the output of the neural network.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical pulse waveform measuring method for measuring the waveform of an extremely narrow optical pulse.

[0002]

2. Description of the Related Art Conventionally, when measuring the waveform of a light pulse having a very narrow width, even if the light pulse is converted into an electric signal by an optical sensor such as a photodiode and is displayed on an oscillograph. Since the response speed of the optical sensor is slow, the optical pulse waveform cannot be displayed.

From such a point, conventionally, an optical pulse is incident on a Michelson interferometer and the interference light is subjected to fast Fourier transform to obtain an optical spectrum of a fundamental wave, and an intensity correlation spectrum is also obtained. To display the waveform of the optical pulse by extracting the doubled harmonic, fast Fourier transforming it to obtain the optical spectrum of the second harmonic, and finding the phase where the mutual relationship of these three spectra is established. Has been proposed (IEEE
J. of Q. E. Vol25 No. 6 June
1989 General Method for U
ltrashort Light Pulse Chi
rp Measurement).

[0004]

In the method of finding the spectrum of an optical pulse by utilizing the proposed interferometer and reproducing the waveform of the optical pulse, the fundamental wave optical spectrum, intensity correlation spectrum and double harmonic wave are used. Since the relationship with the spectrum is non-linear, it cannot be solved by simple simultaneous equations. Therefore, give an appropriate phase to the optical spectrum and calculate whether or not the mutual relationship of the three spectra is satisfied.If not satisfied, the phase must be changed slightly and the same thing must be repeated. It has a drawback that it becomes huge and the calculation time becomes remarkably long.

[0005]

According to the present invention, an input optical pulse is incident on an interferometer, the fundamental wave optical spectrum of the input optical pulse is obtained from the interference light, and the frequency of the interference light is doubled. The optical spectrum of the double harmonic is obtained, the unknown phase is given to the fundamental optical spectrum, the inverse discrete Fourier transform is performed, the result of the inverse Fourier transform is squared, and the squared result is discretized. Fourier transform is performed, the result of this Fourier transform is compared with the optical spectrum of the second harmonic, the unknown phase is corrected according to the comparison result, and the above is repeated to obtain the fundamental optical spectrum. The true phase is obtained and the waveform of the optical pulse is reproduced from the optical spectrum having the true phase.
In that case, input the unknown phase to the neural network,
The output of the neural network is used as the correction phase, and each weight of the neural network is corrected according to the comparison result.

In the above description, an unknown phase is given to the 2nd harmonic optical spectrum instead of the fundamental optical spectrum, its inverse discrete Fourier transform is performed, and the inverse Fourier transform result is square root calculated. The phase may be corrected by performing a discrete Fourier transform and comparing the result of the Fourier transform with the fundamental wave optical spectrum.

[0007]

1 shows a measuring device to which the method of the present invention is applied. The input light pulse 11 is incident on a Michelson interferometer 12, is split by the beam splitter 13 thereof, and is incident on a fixed mirror 14 and a movable mirror 15. Both reflected lights interfere with each other at the beam splitter 13, and the interference light 16
Is converted into an electric signal by the light receiving element 17. The electric signal is converted into a digital signal by the A / D converter 18, and this digital signal is supplied to the signal processor 19.

The interference light 16 is branched and supplied to the second harmonic generator 21, the output thereof is supplied to the filter 22 and the frequency component of twice the frequency component is taken out, and the output of the filter 22 is the photomultiplier 23. Is amplified by the photon counter 24 and converted into an electric signal by the photon counter 24. The electric signal is converted into a digital signal by the A / D converter 30,
The digital signal is supplied to the signal processor 19.

The reference light from the helium neon laser 25 is also incident on the Michelson interferometer 12, the interference light is converted into an electric signal by the light receiving element 25, and the converted output has a frequency doubled by the doubler 27. Doubled and this output is A
The signals are given to the / D converters 18 and 30 as sampling pulses, respectively, and are converted into digital signals every unit movement of the movable mirror 15.

In the signal processor 19, the digital signal from the A / D converter 18 is fast Fourier transformed to obtain the fundamental wave optical spectrum E ₀ (ω) of the input optical pulse,
In addition, the signal from the photon counter 24 is subjected to fast Fourier transform to obtain an optical spectrum U ₀ of the second harmonic of the input optical pulse.
Find (ω). These optical spectra have only amplitude and have no phase information. Therefore, as shown in FIG. 2, an appropriate unknown phase φ _e (ω) is first assumed for each angular frequency ω (S ₁ ), and this is given to each fundamental wave optical spectrum E _o (ω) (S ₂ ). Next, the fundamental spectrum with such a phase is subjected to inverse fast Fourier transform to E (t) φ
Obtain _e (t) (S ₃ ). The result of the inverse fast Fourier transform is squared (S ₄ ). The squared result is subjected to fast Fourier transform to obtain a doubled harmonic light spectrum U (ω) (S ₅ ).

Doubled harmonic spectrum obtained by the calculation
The tram U (ω) and the harmonic light double of the double obtained by the previous measurement
Pectrum U₀Compare with (ω) (S₆). The comparison
Check whether the result (error) is a predetermined value or less
If not set, the phase φ previously set _eCorrect (ω) and fix it
Corrected phase φ_e′ (Ω) is φ_eStep S as (ω)
₂Return to (S₇). Repeat the above and calculate twice
Harmonic spectrum U (ω) is measured value U₀Incorrect with (ω)
If the difference is less than the specified value, the phase φ at that time
_e(Ω) is the true phase of each spectrum of the input optical pulse
This phase φ _e(Ω) and the fundamental optical spectrum measured
Tram E₀Inverse fast Fourier transform using (ω) and
The optical pulse waveform is obtained and displayed on the display 28 (S₉).

By the way, the phase correction in step S ₈ is performed using a neural network. For example, as shown in FIG. 3, when the frequency components decomposed by the fast Fourier transform are 256 points, the phases φ _e (ω ₀ ) to φ _e (ω ₂₅₅ ) for each frequency component are input to the neural network 31. . As this neural network, a Boltzmann machine type multi-layer structure is preferable, and the respective weights are compared with each other in the comparison result (error) in step S _6.
And the corrected phases φ _e ′ (ω _o ) to φ _e ′ (ω ₂₅₅ ) are obtained as the output of the neural network 31. By doing so, fast convergence, that is, a true phase can be obtained. In the learning process, an inverse propagation algorithm is used, and an annealing process is introduced according to the degree of learning. In this way, the number of repetitions is 2500-
It can be reduced to about 3.

In step S ₂ in FIG. 2, not the fundamental wave optical spectrum but the second harmonic optical spectrum U
₀ (ω) is given a phase, square root calculation is performed in step S ₄ instead of square calculation, and the calculated fundamental wave optical spectrum E (ω) and measured fundamental wave spectrum E ₀ (ω) are calculated in step S ₆ . Comparison may be made, and other points may be the same as those described above.

[0014]

As described above, according to the present invention, since the phase correction is performed using the neural network, the true phase is quickly converged, that is, the calculation time is short and the calculation amount is small, and the short optical pulse The waveform can be played.

[Brief description of drawings]

FIG. 1 is a block diagram showing an example of an apparatus for carrying out the method of the present invention.

FIG. 2 is a flowchart showing a part of the operational flow of the method of the present invention.

FIG. 3 is a diagram showing a phase correction neural network.

Claims (1)

[Claims] 1. An input optical pulse is incident on an interferometer, a spectrum of a fundamental wave of the input optical pulse is obtained from the interference light, and the frequency of the interference light is doubled to double the input optical pulse. Obtain the spectrum of the harmonics, give an unknown phase to each of the obtained fundamental wave spectrum or the double harmonic spectrum, perform the inverse discrete Fourier transform, and perform the square or square root calculation of the conversion result. Discrete Fourier transform is performed on the above, and the result of the Fourier transform is compared with the 2nd harmonic spectrum or the fundamental wave spectrum, and the phase is modified according to the comparison result and the above is repeated to repeat the above basic Waveform spectrum or 2nd harmonic spectrum, obtain the true phase, and reproduce the waveform of the above input optical pulse from the spectrum with the true phase. An optical pulse waveform measuring method according to claim 1, wherein the unknown phase is input to a neural network, the output of the neural network is used as the modified phase, and each weight of the neural network is modified according to the comparison result. Optical pulse waveform measurement method.