Laser phase noise measurement method based on time-delay self-heterodyne and digital demodulation

Abstract

The application discloses a laser phase noise measurement method based on delay self-heterodyne and digital demodulation, and relates to the technical field of laser measurement, and comprises the following contents: after being adjusted by a variable optical attenuator, to-be-measured laser is divided into a frequency shift arm and a delay arm, the frequency shift arm is frequency-shifted by an acousto-optic modulator, the delay arm is delayed by a delay optical fiber, and original electric beat frequency signals are obtained by a balanced photoelectric detector after the two arms are coupled; homologous and synchronous quadrature demodulation is carried out on the original electric beat frequency signals, envelope credibility judgment and abnormal sample repair are combined, and a continuous differential phase sequence is obtained; actual optical time delay is determined according to the beat frequency spectrum characteristics and the continuous differential phase sequence, and a differential phase power spectral density is formed; then, transmission relationship conversion is carried out according to the actual optical time delay, bounded spectrum compensation is implemented on a zero point neighborhood and a low credibility frequency point, and an intrinsic phase noise power spectral density is obtained. The application can improve the stability, continuity and result credibility of phase noise measurement under the condition of short delay.

Description

Technical Field

[0001] This invention relates to the field of laser measurement technology, specifically to a laser phase noise measurement method based on time-delay self-heterodyne and digital demodulation. Background Technology

[0002] Narrow-linewidth, low-phase-noise lasers are primarily used in fiber optic sensing, coherent optical communication, lidar, microwave photonics, precision spectroscopy, and cold atom detection. System performance is closely related to linewidth, specifically the power spectral density of phase noise at various Fourier frequencies. In current narrow-linewidth laser measurements, spectrometers and Fabry-Perot type equipment struggle to meet the requirements for high-coherence laser measurements. Two-beam heterodyne methods require an additional reference laser and its matching. Time-delay self-heterodyne methods, because they do not require an additional reference laser and have a simpler structure, typically acquire beat frequency signals through frequency shifting using an acousto-optic modulator, time-delay fiber, and detectors, and then further analyze them using an oscilloscope, spectrum analyzer, or digital acquisition platform.

[0003] To reduce system noise floor, dual-detector, dual-IQ demodulation, or cross-correlation spectrum estimation structures have been proposed. However, this increases the consistency of optical paths, circuits, and channels, making it difficult to achieve stable system calibration and long-term stable operation. Recently, real-time oscilloscopes or FPGAs have been used to replace traditional spectrum analyzers for online implementation, directly reconstructing frequency noise or phase noise from time-domain beat waveforms. However, the results still require delay settings, operating points, and back-end digital processing. This contradiction is mainly reflected in online detection and miniaturized instruments. With the development of kilohertz or lower linewidth lasers and low-noise lasers, the measurement conditions that typically require delays much greater than coherence time lead to longer optical fibers, higher insertion loss, and stronger fiber ingress noise. Short-delay schemes can greatly reduce size, but they introduce interference fringes and envelopes due to correlated beat frequencies, resulting in a spectrum shape that does not conform to the true noise distribution.

[0004] The existing technology has at least the following technical problems: In single-probe, single-link delay self-heterodyne digital measurement, the power spectral density of the intrinsic phase noise of the laser under test after synchronous demodulation is affected by the actual time delay parameters, the zero point of the transfer function, the beat frequency envelope amplitude, and the sampling noise. When the delay is comparable, the received optical power is reduced, or the beat frequency envelope is locally fading, the additive noise will greatly aggravate the instantaneous phase extraction error and cause discontinuity in phase expansion. Furthermore, spurious peaks, spurious valleys, and noise floor rise are easily generated near the zero point of the transfer function, making the output spectrum unable to reflect the true phase noise of the laser, which affects the selection of devices, the tuning of control parameters, and the evaluation of the overall performance. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a laser phase noise measurement method based on time-delay self-heterodyne and digital demodulation. The method involves performing synchronous orthogonal demodulation on the original electronic beat frequency signal, combined with envelope reliability determination and outlier sample repair, to obtain a continuous differential phase sequence. Based on the beat frequency characteristics and the continuous differential phase sequence, the actual optical delay is determined, and a differential phase power spectral density is formed. Then, based on the actual optical delay, a transfer relationship conversion is performed, and bounded spectral compensation is applied to the zero-neighborhood and low-reliability frequency points to obtain the intrinsic phase noise power spectral density. This method improves the stability, continuity, and reliability of phase noise measurements under short delay conditions, and solves the technical problems described in the background art.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A laser phase noise measurement method based on delayed self-heterodyne and digital demodulation includes: dividing the continuous optical signal output by the laser under test into a frequency-shifting arm and a delay arm after being adjusted by a variable optical attenuator; the frequency-shifting arm is frequency-shifted by an acousto-optic modulator; the delay arm is delayed by a delay fiber; the two arms are coupled by a 2×2 fiber coupler and the original electronic beat frequency signal is obtained by a balanced photodetector; the method includes: performing analog-to-digital conversion on the original electronic beat frequency signal; synchronously orthogonally demodulating the signal using a digital local oscillator that is of the same origin as the acousto-optic modulator driving signal; and obtaining a continuous differential phase sequence by combining envelope confidence determination and abnormal sample point repair. The actual optical time delay is determined based on the beat spectrum characteristics of the continuous differential phase sequence and the original electronic beat frequency signal, and the differential phase power spectral density is formed. Based on the actual optical time delay, the differential phase power spectral density is converted by the transfer relationship, and the zero-neighborhood and low-confidence frequency points are screened and bounded spectrum compensation is performed to obtain the intrinsic phase noise power spectral density of the laser under test.

[0007] Furthermore, a time-delayed self-heterodyne beat frequency measurement link is established and the system operating point is preset, including: adjusting the beat frequency envelope amplitude entering the balanced photodetector through a variable optical attenuator, so that the balanced photodetector and subsequent acquisition circuits operate within the linear range, and sending the output of the balanced photodetector as the raw electrical beat frequency signal into the acquisition module.

[0008] Furthermore, synchronous quadrature demodulation using a digital local oscillator that is from the same source as the acousto-optic modulator driving signal includes: performing analog-to-digital conversion on the original electronic beat frequency signal to obtain a discrete beat frequency sequence, and using the digital local oscillator for synchronous mixing to obtain in-phase and quadrature components, and then performing low-pass filtering, decimation downsampling, and amplitude normalization on the in-phase and quadrature components.

[0009] Furthermore, a continuous differential phase sequence is obtained by combining envelope confidence determination and abnormal sample point repair, including: determining the envelope sequence and envelope reference of the beat frequency signal based on the in-phase component and the quadrature component, generating valid sample point markers accordingly, and performing freezing, removal, interpolation repair or neighborhood smoothing repair on invalid samples.

[0010] Furthermore, a continuous differential phase sequence is obtained, including: extracting the instantaneous differential phase from the baseband differential information corresponding to the effective sample points, and performing phase expansion, linear trend elimination and continuous accumulation processing on the instantaneous differential phase to obtain a continuous differential phase sequence for subsequent determination of actual optical time delay and calculation of differential phase power spectral density.

[0011] Furthermore, the actual optical delay is determined based on the beat spectrum characteristics of the continuous differential phase sequence and the original electronic beat frequency signal, including: combining the nominal length of the delay fiber, the fiber refractive index, the additional optical path of the system connection, and the interference fringe spacing, envelope zero position and first extremum position in the beat spectrum to determine the actual optical delay.

[0012] Furthermore, the differential phase power spectral density is formed, including: setting the window length, window overlap rate, averaging number, decimation rate and frequency resolution of the spectrum analysis according to the actual optical time delay; segmenting the continuous differential phase sequence; and sequentially performing DC removal, detrending, windowing and average spectrum estimation on each segment.

[0013] Furthermore, the zero-neighborhood and low-confidence frequency points are screened, including: based on the zero-point position of the transfer relationship corresponding to the actual optical time delay, identifying the zero-neighborhood frequency points, the frequency points near the bottom of the system noise, and the frequency points with insufficient effective samples in the differential phase power spectral density, and using the identification results as the basis for frequency point screening.

[0014] Furthermore, bounded spectral compensation includes: directly performing transfer relationship conversion on the selected reliable frequency points; performing local smoothing, interpolation fitting, or regression equalization on the selected low reliable frequency points based on the spectral values ​​of adjacent reliable frequency points; and forming the intrinsic phase noise power spectral density together with the direct conversion results.

[0015] Furthermore, the intrinsic phase noise power spectral density, time-domain phase jitter sequence, and their statistical results are transmitted to the user's computer via network cable and router. The host computer software then displays the intrinsic phase noise power spectral density, stores the data, performs historical comparisons, and processes and outputs the test results.

[0016] (III) Beneficial Effects This invention provides a laser phase noise measurement method based on time-delay self-heterodyne and digital demodulation, which has the following advantages: The input power operating point is set using an adjustable optical attenuator. The frequency shift arm and delay arm form a beat frequency input in front of the photoelectric balanced detector, which reduces the impact of front-end overdrive, low envelope, and link mismatch on subsequent measurements from the source, enabling the original electronic beat frequency signal to be continuously sampled. Synchronous quadrature demodulation of the digital local oscillator, which is of the same origin as the acousto-optic modulator drive signal, restores the beat frequency signal to the same phase reference as the frequency shift reference, avoiding additional reference drift into the measurement results. At the same time, differential phase information is stably obtained from a single measurement link. Combined with envelope reliability, abnormal sample point repair, phase expansion, and linear trend elimination, the impact of envelope collapse, instantaneous changes, and phase cycle slip on the results is limited to a local range, continuously forming a differential phase sequence.

[0017] By linking the calculation of actual optical delay with the formation of differential phase power spectral density, the front-end delay fiber is not merely a physical delay device, but also the basis for back-end modeling and spectrum construction, reducing spectral estimation mismatch caused by the deviation between nominal and actual delay. The transfer relationship is converted using actual optical delay, and the zero-neighborhood and low-confidence frequency points are screened and bounded spectral compensation is performed, thereby limiting the impact of fringe ripples, local minima amplification, and spectral line divergence on the inversion of intrinsic phase noise power spectral density under short delays. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of the laser phase noise measurement device of the present invention; Figure 2 This is a schematic diagram of the self-heterodyne beat frequency establishment and front-end power balancing in step one of the present invention; Figure 3 This is a schematic diagram of the overall process of the laser phase noise measurement method of the present invention; Figure 4 This is a schematic diagram of the process of generating a continuous differential phase sequence from the original electronic beat frequency sequence by synchronous demodulation from the same source in this invention. Figure 5 This is a schematic diagram of the actual time delay calibration process provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the differential phase spectral density construction process provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figures 1-6This invention provides a laser phase noise measurement method based on time-delay self-heterodyne and digital demodulation. Step 1 is performed by the laser phase noise measurement device. The fiber optic link establishment is completed by the fiber optic connection unit, the acousto-optic modulator drive establishment is completed by the FPGA and amplifier, the balanced photodetector output observation and pre-sampling judgment are completed by the acquisition front end, and the host computer only receives the status record formed in step 1 and does not intervene in the judgment process.

[0021] In a preferred embodiment, the variable optical attenuator is a fiber optic pigtail adjustable attenuator, the 2×2 fiber optic coupler is a polarization-maintaining fiber device, the delay fiber is a coiled and fixed polarization-maintaining single-mode fiber, the balanced photodetector is a dual photodiode differential receiver structure, and the acousto-optic modulator is an acousto-optic crystal assembly matched to the wavelength band of the laser under test. In a parallel embodiment, the variable optical attenuator can be replaced with a free-space continuous attenuation assembly, the delay fiber can be replaced with an on-chip waveguide delay chain, and the acousto-optic modulator can be replaced with a fiber-coupled acousto-optic modulator module. As long as the same original beat frequency signal is still output and the terminological correspondence remains unchanged, it can be included in the implementation scope of this step.

[0022] Step 1: Establish a continuous, traceable, and linearly operating raw beat frequency input environment between the same laser under test, self-heterodyne dual arms, acousto-optic modulator drive link, balanced photodetector differential input, and acquisition front end. Simultaneously generate the raw electronic beat frequency sequence, drive frequency shift marker, front end baseline offset, and initial time delay marker for subsequent steps.

[0023] In the time-delay autoheterodyne measurement, the beat frequency signal is not generated out of thin air, but is formed by the coherent superposition of the optical fields of the frequency-shifting arm and the time-delay arm at the differential port of the same balanced photodetector. If the power of the two arms is unbalanced, or if the signal deviates from the linear operating region of the acquisition front end before entering the balanced photodetector, the beat frequency envelope will exhibit local collapse, bias increase, or decreased differential port utilization in subsequent sampling. Consequently, the in-phase and quadrature components obtained in step two will no longer reflect only phase information, but will be mixed with significant amplitude errors.

[0024] The device first connects the output of the laser under test to the input of a variable optical attenuator via an optical fiber connection unit, and then connects the output of the variable optical attenuator to a first-stage optical splitter coupler. The first-stage optical splitter couples split the same laser under test into a frequency-shifting arm beam and a time-delaying arm beam. The frequency-shifting arm beam sequentially enters an acousto-optic modulator and a section of compensation fiber, while the time-delaying arm beam sequentially enters a time-delaying fiber and another section of compensation fiber. Both arms are then connected together to the input of a 2×2 fiber coupler, whose two outputs are then connected to the two optical inputs of a balanced photodetector. The compensation fiber here is not a new measurement principle, but rather serves to absorb differences in connector pigtail length and wiring bends, ensuring a controllable path relationship between the frequency-shifting and time-delaying arm beams before they enter the 2×2 fiber coupler. Subsequently, the FPGA outputs an RF driver word, which, after digital-to-analog conversion and amplification, is sent to the acousto-optic modulator. This causes the frequency-shifting arm beam to acquire a defined frequency shift, while the time-delaying arm beam maintains its original center frequency. This creates a beat frequency pair at the input of the balanced photodetector that is of the same origin but at different times and frequencies.

[0025] In this system, the variable optical attenuator is not treated as a simple attenuation component, but rather defined as a front-end power balancing actuator. After the device starts, the fiber optic connection unit first sets the variable optical attenuator to the intermediate attenuation position, then acquires the envelope change output from the balanced photodetector monitoring port. If the envelope peak caused by the frequency-shifting arm light and the delay arm light is found to be biased to one side, the variable optical attenuator and the splitting ratio selection are used together to adjust the power balance, ensuring that the power of the frequency-shifting arm light and the power of the delay arm light entering the balanced photodetector are maintained in a balancing relationship conducive to differential beat frequency formation. This action does not aim for absolute equality between the two arms, but rather aims to maintain a readable ratio of the beat frequency component to the total incident power even when acousto-optic modulator insertion loss, delay fiber loss, and connection loss are all present. In the formula, the power balancing coefficient is... The degree of synergy between the optical power of the frequency shift arm and the optical power of the delay arm in the formation of the beat frequency envelope, with a value range of [value range missing]. The closer the value is to the upper limit, the higher the degree of balancing of the two arms at the input end of the balanced photodetector, and the easier it is for the differential beat frequency component to be prominently read by the front end. Frequency shift arm optical power The effective optical power entering the 2×2 fiber coupler after passing through the acousto-optic modulator and frequency shifter path, with a value range of [value missing]. Delay arm optical power The effective optical power entering the 2×2 fiber coupler after passing through the delay fiber and delay arm path, with a value range of [value missing]. ; Frequency shift arm optical power and delay arm optical power This is for front-end balancing mechanism analysis. During online measurement, the system does not require separate real-time measurement of the frequency-shift arm optical power. and delay arm optical power Instead, it obtains the peak value of the beat frequency envelope from the front end within the observation window. Front-end baseline offset and sampling occupancy coefficient Determine the operating point; power balancing factor. The installation calibration results can be written into the configuration table and used as a reference state quantity when constructing the envelope credibility later.

[0026] Taking the desktop package implementation as an example, after the operator connects the pigtail of the laser under test to the input of the variable optical attenuator, they first observe the envelope strength on the external monitoring interface of the balanced photodetector. Then, they slowly adjust the variable optical attenuator knob or send a control word to the variable optical attenuator to gradually raise the envelope from a state submerged by background noise to a continuously discernible state. Subsequently, keeping the acousto-optic modulator drive on, they only change the splitting ratio of the first-stage beam splitter or switch the position of the pre-made jumper to allow the frequency-shifting arm light and the delay arm light to form a more balanced beat frequency envelope at both ends of the balanced photodetector. The result that can be directly seen on site is that the envelope on the monitoring port of the balanced photodetector no longer fluctuates intermittently, but forms a continuous beat frequency profile, creating input conditions for the differential phase load. In use, power balancing logic is simultaneously incorporated into both the variable optical attenuator and the beam splitter. The beat frequency input output from step one no longer relies on empirical connections and can serve as a reproducible balancing basis. This reduces the impact of asymmetric insertion loss and delay arm loss of the acousto-optic modulator on the differential port of the balanced photodetector. Simultaneously, it ensures that the in-phase and quadrature components extracted in step two have a stable amplitude base. Furthermore, the power balancing coefficient... It will be written as a working point marker in the output of step one, and subsequent judgments will be directly related to the front-end state.

[0027] Once the operating point of the variable optical attenuator is locked, step one does not end immediately. Instead, it continues to transform the phase fluctuations of the laser under test into a differential phase load that can be read by the balanced photodetector through an acousto-optic modulator and a delay fiber. The key is not simply shifting one optical path and delaying the other, but ensuring that both paths originate from the same laser under test, thus ensuring that the phase difference in the output of the balanced photodetector strictly corresponds to the phase state of the same laser at two different moments. Only a beat frequency input established in this way is suitable for the homogeneous demodulation logic in step two. In the formula, the original beat frequency signal Balanced photodetector at time The range of the output analog electrical signal is constrained by the linear range of the balanced photodetector and the acquisition front end; the photoelectric conversion coefficient The electrical conversion capability of the balanced photodetector and the pre-amplifier link for the superposition of the optical fields of the two arms, with a value range of [value range missing]. Frequency shift arm optical power With delay arm optical power The meaning remains the same as before and will not be changed. instantaneous phase of the laser under test The laser under test at time The phase state, with values ​​in the real number domain, provides a phase sample for the current moment; initial delay marker. Step 1: The initial optical delay obtained based on the physical path of the delay arm, with a value range of [value missing]. ; Drive frequency shift marker The radio frequency drive frequency corresponding to the acousto-optic modulator has a value range of [value range missing]. The beat frequency center is moved to a frequency band that is easy for the balanced photodetector and the acquisition front end to read; phase constant : A constant term formed by the connection phase, the fixed phase of the device, and the relationship between the coupler output ports, with a value range of . Front-end baseline offset The DC baseline component of the balanced photodetector and the acquisition front end is limited in value range by the front end bias circuit and is used for zero-point correction before subsequent sampling. In the formula This is a direct result of the time-delayed self-heterodyne physical process: the frequency-shifting arm carries the current phase, while the time-delaying arm carries an initial time delay marker earlier than the current time. The phase of the two fibers is recoupled by a 2×2 fiber coupler and then enters the differential port of the balanced photodetector. The acousto-optic modulator drives the frequency. Then the phase difference is overlaid onto the readable beat frequency center.

[0028] In a preferred embodiment, the acousto-optic modulator is mounted on an independent heat dissipation base plate, the delay fiber is coiled on a fixed fiber spool and local stress is released by pressure strips, and the 2×2 fiber coupler and the balanced photodetector are arranged in adjacent areas of the same metal mounting base plate to shorten the length of the suspended jumper from the coupler to the balanced photodetector; with this arrangement, the beat frequency center position is more stable, and the phase constant caused by connection disturbances is reduced. Drifting is easier to control.

[0029] In parallel implementations, the acousto-optic modulator can employ a free-space crystal assembly with a collimator, or it can use an optical fiber-coupled acousto-optic modulator module; the delay path can be formed by either a coiled delay fiber or an on-chip waveguide delay chain, as long as the corresponding initial delay marker is used. Able to be recorded separately and maintain the same original electronic beat frequency signal One-to-one correspondence.

[0030] During use, the original electro-taper frequency signal The input is no longer a coarse-grained one that only requires beat frequency, but a singular input object with a clear source of phase, time delay, and frequency shift. In step two, during synchronous quadrature demodulation, it becomes clear that the recovered signal is the same laser beam spanning the initial time delay marker. The differential phase, rather than unknowns mixed with multi-source drift.

[0031] Even if a stable beat frequency light input has been obtained, if there are still problems such as excessively high or low occupancy, excessive baseline offset, or missing time delay records when the output of the balanced photodetector enters the acquisition front end, the analog-to-digital conversion and synchronous quadrature demodulation in step two will still lose a unified reference.

[0032] In this system, the output of the balanced photodetector and the input of the acquisition front-end are connected via a short-distance coaxial connection. After receiving the original beat frequency signal, the acquisition front-end first performs baseline observation, then amplitude judgment, and finally sends the qualified analog signal to the analog-to-digital converter to form the original beat frequency sequence. Simultaneously, the fiber optic connection unit writes the delay fiber type, delay fiber wiring path, connector additional length, and current acousto-optic modulator drive frequency shift into the step-one state register area on the FPGA side. This state register area is not a generalized software cache, but a front-end state record area shared by subsequent steps, storing at least the original beat frequency sequence index and drive frequency shift marker. Power balancing coefficient Front-end baseline offset and initial delay mark .

[0033] Furthermore, if the peak value of the envelope of the original electronic beat frequency signal is too high, clipping will occur at the acquisition front end, and the in-phase and quadrature components in step two will carry obvious harmonic distortion; if the peak value of the envelope is too low, the effective bits after analog-to-digital conversion will be swallowed up by the noise floor. Although synchronous quadrature demodulation can continue in form, the recovered phase sequence will be frequently disturbed by envelope fluctuations.

[0034] Therefore, step one needs to provide a duty cycle criterion with the same caliber as the acquisition front end before analog-to-digital conversion, to link the operating point of the variable optical attenuator, the amplifier output amplitude, and the full-scale range of the acquisition front end, where: In the formula, the sampling amplitude coefficient The percentage of the original beat frequency signal relative to the full-scale range of the acquisition front end is taken as: Peak frequency envelope The peak amplitude of the raw beat frequency signal output by the balanced photodetector and observed by the acquisition front end within the current observation window, with a value range of [value missing]. Front-end baseline offset The meaning remains the same as before, and its absolute value is used to deduct the interference of static zero-point drift on the amplitude judgment; front-end full-scale voltage : The maximum allowable linear swing amplitude at the front-end input terminal, with a value of Its function is to define the safety boundary before the analog-to-digital conversion; In a preferred embodiment, the acquisition front end continuously reads the peak and valley values ​​output by the balanced photodetector within a fixed observation window, and calculates the sampling amplitude coefficient. Return to FPGA; if the sampling occupancy coefficient If the sampling duty cycle is too high, the FPGA should first reduce the transmitted light power of the variable optical attenuator, and then check whether the amplifier output still meets the frequency shift condition of the acousto-optic modulator; if the sampling duty cycle is too high... If the signal is too low, first increase the transmitted light power of the variable optical attenuator, then check whether the monitoring port of the balanced photodetector still maintains a continuous beat frequency envelope. The adjustment sequence here is intentionally set to first the variable optical attenuator, then the amplifier. This is because the variable optical attenuator directly affects the total optical power of both arms, while the amplifier affects the drive intensity of the acousto-optic modulator. Changing both simultaneously and erratically would make it difficult to determine in step one which path the amplitude change originates from. Taking a rack-mounted implementation as an example, the operator can directly see the level indicator of the acquisition front end on the front panel. When the level indicator continuously exceeds the limit, the variable optical attenuator is reverted first; when the level indicator stabilizes but the beat frequency center disappears, the acousto-optic modulator drive link is then checked. There is a direct correspondence between the on-site actions and the results, which can be reproduced by those skilled in the art.

[0035] When using it, the sampling amplitude coefficient By unifying optical power balancing and electrical sampling boundaries into a single continuous criterion, the scenario where the optical path appears to have a beat frequency but the sampling end is unusable is avoided. This eliminates the need for subsequent synchronous quadrature demodulation to perform front-end clipping or noise floor enhancement. Furthermore, the sampling occupancy factor... Will be related to the power balance factor Paired data is saved for later differentiation between envelope anomalies caused by power imbalance and those caused by sampling boundary offset.

[0036] In step one, the end point is not when the beat frequency has been established, but when the set of objects that can be directly called by subsequent steps has been encapsulated.

[0037] Specifically, before the analog-to-digital conversion begins, the device records the effective optical path of the delay fiber, the additional optical path of the connector, the additional optical path of the compensation fiber, and the current acousto-optic modulator drive frequency shift in the front-end status recording area. After the analog-to-digital conversion begins, the sampled original beat frequency sequence is bound to the above records and output. The result of step one thus includes both the signal itself and the context of why the signal is formed in this way. This eliminates the need to re-estimate the acousto-optic modulator frequency shift center during synchronous quadrature demodulation in step two, and allows step three, when performing time delay self-calibration, to start from the physical path record rather than from zero. In the formula, the initial time delay mark Step 1: The initial optical delay calculated based on the physical information of the delay path is set to a value of [value missing]. Group refractive index The group refractive index of the delay fiber or on-chip waveguide delay chain in the wavelength band of the laser under test, with a value range of [value range missing]. ; Delay path length The effective delay length of the delay fiber or on-chip waveguide delay chain, with a value range of [value missing]. ; Connect additional optical path The additional optical path length introduced by the difference in length between the connector pigtail, compensating fiber, and necessary patch cord can be positive, zero, or negative. Its function is to correct for errors caused by relying solely on the delay path length. The resulting time delay estimation error; speed of light in vacuum This is a constant for light propagation, with a fixed value. Its function is to convert optical path length to time. In a preferred embodiment, after the delay fiber is coiled, the path is registered first, and then the start and end points of the coil and the connector number are written into the device configuration file. When the operator replaces the delay fiber coil or changes the jumper position, the fiber connection unit requires re-registration of the additional optical path. Then, the acquisition front-end is allowed to enter the sampling state. For the parallel implementation of the on-chip waveguide delay chain, the layout length and layout corner are incorporated into the delay path length. The calculations include taking the input-output coupling segment into account for the additional optical path of the connection. The results visible on-site are as follows: regardless of whether a coiled delay fiber or an on-chip waveguide delay chain is used, the same set of output objects will be generated on the device interface at the end of step one, namely the original beat frequency sequence and the driving frequency shift marker. Power balancing coefficient Sampling amplitude coefficient Front-end baseline offset and initial delay mark .

[0038] When in use, the initial delay flag Instead of merely stating the approximate length of the delay fiber, it is now fixed as an engineering parameter sequentially bound to the original beat frequency sequence. This avoids terminological drift caused by the independent interpretation of the front-end state in steps two and three, and also ensures that the actual delay self-calibration in step three converges from an existing physical starting point.

[0039] Step two below is executed collaboratively by the acquisition front-end and the FPGA. The acquisition front-end is responsible for converting the raw beat frequency signal into discrete sampling objects, while the FPGA is responsible for performing synchronous demodulation, baseband shaping, envelope confidence gating, abnormal sample point repair, and continuous accumulation. In a preferred embodiment, the sampling clock and the acousto-optic modulator driving clock are from the same source, and the low-pass filter can adopt a finite impulse response structure or a combination of a cascaded integrator comb filter and a compensation filter.

[0040] Step 2: Using the original beat frequency sequence output from Step 1, drive the frequency shift marker. Front-end baseline offset Power balancing coefficient Sampling amplitude coefficient and initial delay mark To ensure consistent input, a continuous differential phase sequence is generated through synchronous demodulation from the same source, confidence gating, and phase increment repair. The continuous differential phase sequence, along with the envelope sequence and valid sample markers required in step three, is then output.

[0041] The original beat frequency signal output in step one includes the initial time delay marker of the laser fly-through. The differential phase information, but it is still superimposed on the driving frequency shift marker. At the beat frequency center, if zero-crossing interpretation or simple peak tracking is performed directly on the original electronic beat frequency sequence, envelope fluctuations and front-end baseline offset will occur. Sampling amplitude coefficient Fluctuations will be superimposed on the phase estimation. If the demodulation reference and the acousto-optic modulator drive are not from the same source, a reference drift unrelated to the laser under test will also be introduced. The acquisition front end is biased according to the front end baseline. After zero-position shifting of the input analog waveform, the original beat frequency signal is discretized using the sampling period to obtain the discrete beat frequency sequence. .

[0042] Subsequently, the FPGA calls a numerically controlled oscillator that shares a phase accumulator with the acousto-optic modulator drive link to process the discrete beat frequency sequence. Perform synchronous quadrature downconversion in complex form, followed by low-pass filter suppression of the sum and frequency terms and high-frequency spurious signals to obtain the complex baseband sequence. And from the complex baseband sequence Extract the envelope sequence and envelope reference, and combine them with the power balancing coefficient. With sampling occupancy coefficient Constructing envelope confidence .

[0043] Before starting step two, the FPGA reads the drive frequency shift flag stored in step one. and front-end baseline offset The acousto-optic modulator-driven phase accumulator and the demodulator-driven phase accumulator are cleared on the same reset edge, and then the sampling gate is closed, discrete beat frequency sequence. Each sampling point corresponds to a traceable demodulation phase position, avoiding the drift of the reference phase starting point due to the absence of beat frequency.

[0044] In the formula, the mixing complex sequence Discrete beat frequency sequence The complex mixing result before being transferred to baseband under the same demodulation reference takes values ​​in the complex domain; discrete beat frequency sequence. The data acquisition front end is based on the sampling period. The sequence obtained by discretely sampling the original beat frequency signal output from step one has a value range constrained by the full-scale voltage of the sampling front end; front end baseline bias. The meaning follows step one, changing the DC bias from the discrete beat frequency sequence. Mid-stripping; driving frequency shift marker The meaning follows the same as in step one, and its function is to determine the center frequency of the complex down-conversion. Sampling period The time interval between two adjacent discrete sampling points, with a value range of [value missing]. Its function is to map the sampling sequence number to the actual time; the starting point of the same phase. The demodulation phase accumulator and the acousto-optic modulator drive the phase accumulator to share the reference phase at the start time of step two after the reset edge, and the value range is: Its function is to lock the driving link in step one and the demodulation link in step two to the same phase starting point. In the formula, the imaginary unit : Orthogonal unit in the complex plane, satisfying The cosine and sine references are combined into a single complex exponential form; where the sampling number is... : The position number of the discrete sequence, which is a non-negative integer, and is used to uniformly number the discrete objects in step two; Preferably, the numerically controlled oscillator employs a combination of a lookup table and a phase accumulator; in parallel embodiments, a coordinate rotation digital calculation structure can also be used. (Same-source phase starting point) Unifying the acousto-optic modulator drive and synchronous quadrature demodulation within the same reference frame, discrete beat frequency sequence The baseband relocation no longer introduces additional reference drift; front-end baseline offset It is subtracted before mixing.

[0045] Furthermore, only mixed complex sequences This is still insufficient for phase recovery because it simultaneously contains difference frequency components, sum frequency residues, and broadband noise. Therefore, the mixing complex sequence is first... The sequence is fed into a low-pass filter to form a complex baseband sequence. Then from the complex baseband sequence Extract the envelope sequence and envelope reference, and then use the power balancing coefficient. With sampling occupancy coefficient Construct envelope confidence quantity This allows the front-end state established in step one to continue participating in the numerical determination in step two.

[0046] In the formula, the complex baseband sequence Mixing complex sequences The complex baseband result after low-pass filtering takes values ​​in the complex domain; filter coefficients The low-pass filter in the first... The coefficients at each tap position take values ​​in the real number domain; the filter length... The number of taps in the low-pass filter, taking integer values. Tap number : Filter coefficient position number, with values ​​being integers and Its function is to iterate through each tap of the low-pass filter; Envelope confidence : No. Complex baseband sequence at each sampling location The reliability of subsequent phase retrieval is determined by a value ranging from [value range missing]. envelope sequence Complex base band sequence The modulus value is specifically taken as The value is Its function is to reflect the baseband envelope strength of the current sampling point; Envelope Reference In the first Within a local window centered on a sampling point, the envelope sequence The reference envelope obtained by performing median arithmetic has a range of values. Weighting factors Envelope reference to envelope confidence level The degree of participation, with a value range of Power balancing coefficient With sampling occupancy coefficient The meaning of these terms follows that of step one, and their functions correspond to the front-end dual-arm balancing state and the sampling level amplitude occupancy state, respectively. Preferably, envelope reference The low-pass filter is generated using an odd-length median window and employs a two-stage structure consisting of a cascaded integrator comb filter and a finite impulse response compensated filter.

[0047] Taking the benchtop embodiment as an example, when the operator lightly touches the external jumper, a dip will appear in the beat frequency envelope on the monitoring port of the balanced photodetector. At this time, the envelope sequence... Consequently, the envelope reference decreases, while Maintaining relative stability, envelope confidence level The synchronous decrease indicates that this segment is not suitable for direct participation in phase recovery. Complex baseband sequence After shaping, its envelope sequence With envelope reference It is now possible to distinguish whether the current sampling point is experiencing normal fluctuations or local mismatches; simultaneously, the power balancing coefficient... With sampling occupancy coefficient It was reintroduced.

[0048] The envelope strength mentioned in step two is within a preset effective range, specifically determined by the envelope confidence value. AND gate limited edition The comparison result is achieved; when When, the corresponding sampling point is determined as a valid sampling point, when When this happens, the corresponding sampling point is determined to be an invalid sample point or a sample point to be repaired. Therefore, the preset valid interval no longer has a separate criterion, but is consistent with the envelope confidence criterion.

[0049] In use, the device outputs no longer a simple mixing result, but rather a result carrying an envelope sequence. Envelope Reference and envelope confidence Complex baseband sequence Since the demodulation reference and the acousto-optic modulator drive are of the same origin, the complex baseband sequence The phase term in the equation retains the same physical source as in step one; due to the envelope confidence... The power balancing factor has been set. With sampling occupancy coefficient By introducing a digital processing chain, subsequent phase recovery can be carried out with the front-end state known.

[0050] Even if the complex baseband sequence has been obtained and envelope confidence If the arctangent phase is calculated and directly expanded for all sampling points, phase wrapping, low envelope points, and local interruptions will still cause artificial reversals in the phase sequence. Under the conditions of a single balanced photodetector and a single measurement link, a continuous chain needs to be constructed within this step, consisting of discrimination, extraction, repair, and accumulation. The FPGA first uses the envelope confidence value... Valid sample point markers are generated, and phase extraction is performed only on these valid samples. Subsequently, the phase increments between adjacent valid samples are wrapped and their continuity is determined. Phase increments within invalid sample point intervals are then repaired using predicted increments. Finally, the repaired phase increments are used as accumulation objects, and a detrending agent is superimposed to obtain a continuous difference phase sequence. In this process, step three requires more than just a continuous differential phase sequence. It also includes valid sample point markings and repair section markings.

[0051] Furthermore, the arctangent operation only applies to complex baseband sequences. The envelope confidence factor only has physical meaning when both its real and imaginary parts are reliable. The current envelope sequence has been indicated. Deviating from local envelope reference If phase extraction is still performed, it will only convert the envelope collapse into a pseudo-phase. Therefore, the FPGA first converts the envelope confidence value... AND gate limited edition Comparison to form valid sample point markers Only when valid sample points are marked The complex baseband sequence is only processed when the current point is valid. The real and imaginary parts are called using the two-parameter arctangent function to obtain the wrapped differential phase.

[0052] In the formula, the wrapping differential phase : Composed of complex base band sequences The extracted phase values ​​range from 1 to 10. Operator : A two-parameter arctangent function whose inputs are, in turn, complex baseband sequences. The imaginary and real parts are used to output the quadrant-consistent phase angle of the corresponding complex vector in the complex plane; the imaginary part operator Operations that take the imaginary part of a complex number; real part operator. The operation of extracting the real part of a complex number has the following effects: (1) Extracting the real part of a complex number from the complex baseband sequence. Separate orthogonal components and in-phase components in the middle; Valid sample point marking : No. Whether a sampling point is allowed to enter the phase increment recovery chain is determined by a binary set of values. Threshold Envelope confidence The decision threshold is set to a value of , to convert continuous envelope confidence values Convert to explicit valid sample point labels Envelope confidence With complex base band sequence The meaning remains the same as in the previous text; In the preferred embodiment, the threshold quantity The device is based on the power balancing factor With sampling occupancy coefficient The corresponding front-end status table was retrieved. The differential phase of the package is included. It no longer comes from the indifference evaluation of all samples, but only from the already enveloped confidence value. Valid sample points for review; valid sample point marking A clear boundary was established for incremental repair.

[0053] Only the differential phase of the package is obtained. This is still insufficient to proceed directly to step three because it involves wrapping the differential phase. Naturally existing Periodic reversal, while effective sample point marking. The segments with a value of zero need to be crossed consecutively.

[0054] Therefore, instead of directly performing large-span interpolation in the phase value domain, we first work in the phase increment domain: wrapping the differential phase between adjacent valid samples. The difference is calculated and wrapping is removed to obtain the phase increment under the constraint of the current sampling point; when a valid sample point at a certain position is marked... When the value is zero, the original increment at that position is not used; instead, the predicted increment is invoked. As an alternative.

[0055] Predicted increment The effective samples before and after the repair section are obtained using piecewise linear interpolation or piecewise cubic Hermite interpolation. For cases with long, continuous invalid sections, a processing path is adopted that freezes the local slope and then reconnects at the recovery point. After the phase increment to be repaired is formed, a local trend increment is applied to it. The elimination of these phases eventually leads to a continuous accumulation of differential phase sequences. ,in: In the formula, the repair phase increment : No. Each sampling location is marked with a valid sample point. The phase increment after gating and abnormal section repair takes values ​​in the real number domain; where, the operator The phase wrapper operator, in its specific form, maps the input phase difference to an interval. Internally, its function is to eliminate the differential phase of the package. Full circle reversal between adjacent sampling points; Where: Input quantity The phase difference to be normalized; the output falls within the interval Its function is to reflect the phase difference across integer cycles back to the unified principal value interval.

[0056] Package differential phase With valid sample point marking The meaning remains the same as before; predicting increments Mark valid sample points Within the zero-valued segment, the alternative phase increment is given by interpolation or local slope extension of the neighboring valid samples, and its value range is in the real number domain; when the number of consecutive invalid samples does not exceed a preset length threshold. At that time, predict the increment The phase increments corresponding to the nearest valid samples on both sides of the notch are generated using piecewise cubic Hermite interpolation; when the number of consecutive invalid samples exceeds a preset length threshold... At that time, predict the increment The segment is generated using a local slope preservation method based on the most recent effective phase increment, and this segment is simultaneously written into the repair segment marker.

[0057] Continuous differential phase sequence The discrete continuous phase object output in step two has a real number value and serves as the direct input for step three, which performs actual time delay self-calibration and differential phase spectrum estimation. Local trend increment The sliding fitter adjusts the phase increment within a local window. The trend component obtained after linear fitting, with values ​​in the real number domain, is used to remove residual bias and slow drift from the continuous differential phase sequence. The effect; in length of Within the symmetrical sliding window, the repair phase increment Perform a linear fit, and use the slope of the fitted line as the local trend increment. Under symmetrical window conditions, the local trend increment It can be written as: Among them, window half length It is a positive integer. Local trend increment. : in the Local linear trend components calculated at each sampling location; sampling location : The number of the discrete sampling point currently being processed; the half-length of the window. The number of sampling points expanded to the left and right sides of the local fitting window. Therefore, the total window length is... The range of sampling points covered is from arrive Relative position index : relative to the current sampling point The offset. When When, it corresponds to the current point; when When, it corresponds to the sample points behind; when At that time, corresponding to the sample points in front; repair phase increment. : in position The phase increment value at that point; Preferably, when the repair section is short, the predicted increment is... Piecewise cubic Hermite interpolation is used for generation; when the repair segment is long, the predicted increment is used. The slope is generated by freezing local slopes. Taking the generalized implementation in the field as an example, when the operator re-plugs the short jumper cable near the balanced photodetector, an envelope trough will appear on the acquisition interface; at this time, the valid sample point is marked. Within this trough range, the phase shifts to zero, and the repairer no longer uses the original package differential phase. Instead of directly taking the increment, the predicted increment is constructed along the effective sample points before and after the trough. After the envelope is recovered, it is smoothly connected to subsequent valid samples, and the continuous differential phase sequence displayed on the interface is displayed. There will be no obvious breakage.

[0058] Phase increment repair changes the focus of abnormal segment processing from the phase value itself to the phase increment, which is a local trend increment. The elimination further separates the residual beat frequency center drift from the phase fluctuations of the laser under test. After processing, the continuous differential phase sequence output in step two... Valid sample point marking Together with the repair section markers, they constitute the unified input for step three.

[0059] Step 3: Based on the continuous differential phase sequence The observation results mark the initial time delay. Corrected to actual delay mark and in actual delay marking For continuous differential phase sequences under constraints The process involves segmentation, windowing, weighting, and fusion to output the differential phase spectral density, which can be directly used in step four. And its corresponding zero-point neighborhood marker.

[0060] The initial delay marker given in step one Originating from the delay fiber path registration and connection additional optical path accounting, it has clear engineering significance, but this initial delay marker... This still represents the physical starting point of the installation process and is not equivalent to the actual time delay required for the measurement of the relationship in step four. In scenarios involving re-fixing of the polarization-maintaining fiber optic tray, changes in the bending radius of the pigtail, re-insertion of connectors, uneven distribution of ambient temperature along the fiber tray, or an increase in the proportion of the repaired section in step two, the continuous differential phase sequence... The stripe spacing shown will deviate from the initial time delay marking. The expected theoretical position. If the initial time delay marker is used directly at this point... The zero point of the transfer relation in step four will then be related to the actual differential phase spectral density. The misalignment of zero and valley frequencies causes confusion at the boundaries between reliable and unreliable frequencies.

[0061] In this process, the FPGA first marks the valid sample points output in step two. and envelope sequence For continuous differential phase sequences The observation segment is extracted to form a two-level object: a coarse observation segment and a check observation segment. The coarse observation segment is used to quickly locate the approximate position of the stripe valley values, while the check observation segment is used to remove false valley values ​​caused by locally repaired sections. Subsequently, the host computer compares the stripe valley value sequence obtained from the coarse observation segment with the initial time delay marker. By comparing the corresponding theoretical zero-point intervals and combining them with the allowable range given by the physical accounting of the delay fiber, the actual delay mark is calculated. In this chain, the continuous differential phase sequence output in step two... The subject of observation is the initial time delay marker output in step one. It is a physical prior, and the two together constitute the time delay self-calibration input in step three.

[0062] If the continuous difference phase sequence If all segments are fed into the stripe observation process at once, then the repaired segments and envelope sequences from step two will be... Sections with large fluctuations and effective sample point markings Sparse segments will form local depressions on the coarse spectrum, which will be mistaken for time delay autoheterodyne fringe valleys.

[0063] To avoid this problem, the FPGA first scans the continuous differential phase sequence along the time axis in a circular buffer. Only when valid sample points are marked within a certain candidate segment Continuous dense, envelope sequence A segment is only included in the coarse observation segment set when the local undulations do not show sharp depressions and the repair segment marker from step two does not cover the main body of the segment. Subsequently, envelope sequences are selected from the coarse observation segment set. The set of verification observation segments consists of segments that are more stable, have shorter repair sections, and have more coherent phase connections.

[0064] Preferably, the time span of the coarse observation segment is taken as a preset observation window, and the preset observation window does not exceed the time span marked by the initial time delay. The calculated fringe observation length is doubled; the time span of the verification observation segment is taken as one-half to two-thirds of the time span of the coarse observation segment, so as to enhance local discriminative power without losing fringe interval information. The Tukey window is preferably used as the window function, with a shrinkage coefficient preferably between 0.1 and 0.35, because this range can suppress leakage at the window edges without excessively weakening the local amplitude near the fringe valleys. In parallel implementations, the window function can also be replaced with a Blackman-Harris window or a discrete Prolate window. The replaced fringe observation objects still maintain a two-level structure of coarse and verification observation segments, without changing the overall processing logic of step three.

[0065] After the operator replaces the delay fiber reel in the equipment cabinet, the initial delay marker obtained in step one is recorded. The physical path has been updated, but the tension of the fiber reel fixing strip is different from the previous one. After the device enters step three, several continuous coarse observation strips first appear on the interface, among which the section near the fiber reel connector is due to the envelope sequence. There were obvious fluctuations, and the sample only entered the coarse observation set, not the verification observation set; while the middle section of the fiber disc had a gentle envelope and effective sample point markings. Continuous segments were simultaneously included in both the coarse observation segment set and the check observation segment set. The visible result was that subsequent extraction of stripe valley values ​​revolved solely around the latter; while the former was not deleted, it no longer served as the primary determinant of time delay.

[0066] In practice, fringe observations are no longer based on the assumption that all segments with phase data are equally valid, but rather on continuous differential phase sequences. Above the stable section screening. This involves marking the valid sample points formed in step two. and envelope sequence Continuing the propagation further reduces the interference of repaired sections and local envelope depressions on fringe spacing recognition, thereby improving the actual time delay marking from the source. The reliability of the solution.

[0067] After completing the double-layer sieving, step three proceeds to the time delay calculation stage. The basic mechanism is that in a time-delay autoheterodyne system, the interval between adjacent fringe valleys is inversely related to the actual optical time delay. Therefore, the time delay can be inferred from the fringe valley interval extracted from the calibration observation segment. However, relying solely on the fringe valley interval is easily affected by local pseudo-valleys and the observation window boundary, and relying solely on the initial time delay marker is also problematic. Furthermore, it cannot cover additional optical path lengths and environmental changes, therefore it is necessary to integrate the two into a unified actual time delay marker. .

[0068] To this end, the system first extracts the adjacent valley interval sequence from the coarse spectrum of the verification observation segment, then performs median suppression on the valley interval sequence, and finally compares it with the initial time delay marker. Weighted fusion is performed, where: In the formula, the actual time delay is marked The optical time delay after correction in step three has a range of values. ; Fusion weights The weighting coefficient between physical accounting values ​​and fringe observation values, with a range of values ​​of [value range missing]. Used to balance the initial delay marker Stability and field adaptability of stripe valley interval observation; initial time delay marking The meaning follows step one, providing the delay prior obtained from the physical path registration; median operator The operation of taking the median of the input sequence is used to suppress the bias caused by local spurious valleys; the stripe valley interval. : Verify the frequency difference between two adjacent effective fringe valley values ​​in the coarse spectrum of the observation segment, with a range of values ​​of . Its function is to characterize the fringe spacing caused by actual time delay; valley index : The position number of the stripe valley interval sequence, which is a positive integer, used to traverse different stripe intervals; The extraction of stripe valley values ​​employs a combination of five-point single-valley judgment and consistency verification of adjacent valley intervals: first, candidate valley values ​​with a center point lower than the two points before and after are found on the coarse spectrum of the verification observation segment; then, it is determined whether the interval between adjacent valley values ​​falls within the range defined by the initial time delay marker. Within the derived allowable band; if two consecutive candidate valley values ​​simultaneously exceed the allowable band, then the envelope sequence within them is... The segment with greater fluctuations is reverted to the observation reference segment and is not included in the actual time delay marking of the current round. Calculate the value.

[0069] In parallel implementations, the stripe valley value can also be obtained by combining the first-order difference sign change with local quadratic curve fitting. When the implementation method is changed, the stripe valley value interval in the formula is adjusted. The definition remains unchanged.

[0070] When in use, the actual delay marker No longer simply inheriting the initial delay flag The static value is no longer a simple instantaneous estimate from a certain observation window, but a fusion result of the combined effects of physical path accounting and spectral fringe observation.

[0071] Among them, the actual delay mark is obtained. Then, if we still directly apply the entire continuous differential phase sequence Performing a single long-window Fast Fourier Transform will result in two problems: first, if the duration of the analysis segment exceeds the actual time delay marker... The corresponding relevant boundaries, the local correlation in the differential phase will be mixed into the single-segment spectral estimation, resulting in the coupling of spectral fluctuations and stripe valleys; secondly, the effective sample point markings in different segments. Density, envelope sequence The degree of stability and the proportion of repaired segments are not consistent. If the spectrum of each segment is averaged equally, the suppression of abnormal segments in step two will be diluted again in step three.

[0072] The system first marks the actual time delay. With sampling period Calculate the segment length, and then mark the valid sample points. The continuous density and repair segment distribution determine the segment step interval; then, the continuous differential phase sequence within each analysis segment is analyzed. Perform mean removal, intra-segment trend removal, window loading, and weighted spectral estimation to generate piecewise differential phase spectral density. Finally, segment weights are generated based on the effective sample density, envelope consistency, and repair ratio of each segment, and the differential phase spectral density of all segments is then analyzed. Perform weighted fusion to obtain the final differential phase spectral density. Step 3 outputs the differential phase spectral density. Will be consistent with actual delay marker Write them together into the input area of ​​step four.

[0073] Pay attention to the length of each segment and the distance between adjacent segments. If the segment length is too short, the differential phase spectral density... The frequency resolution is insufficient, making it difficult to distinguish between fringe valleys and spectral background details; if the analysis segment is too long, the same analysis segment will span beyond the actual time delay mark. The time frame allows related items within a segment to be re-included.

[0074] To avoid issues at both ends, step three establishes the analysis segment length directly on the actual time delay marker. With sampling period The ratio is based on, rather than on, a fixed number of points, where: In the formula, the length of the analysis segment is... The number of discrete sampling points contained in each analysis segment, taking a positive integer value, serves to limit the duration of each analysis segment on the time axis; the rounding operator. : Performs rounding operations on the input real number, the purpose of which is to round the actual delay value. and sampling period The calculated real number length is converted into the number of sampling points; Segmentation scaling factor : Analysis segment length relative to actual time delay marker The proportion, with a range of values ​​being Actual delay marker The meaning remains the same as before; its function is to provide physical boundaries for the length of the analysis segment; sampling period The meaning follows from step two, and its function is to convert between the number of discrete points and the actual time.

[0075] Preferably, the step interval of the analysis segment is equal to the length of the analysis segment. The interval is one-quarter to one-half because it preserves continuous coverage between adjacent analysis segments without allowing them to completely overlap. When the repair segments are densely distributed, the analysis segment step interval is taken as the length of the analysis segment. The window function is half the length of the analysis segment to reduce the probability of adjacent analysis segments simultaneously crossing the same repair section. Tukey windows or Blackman-Harris windows are preferred; in FPGA-restricted implementations, the window function is pre-stored in on-chip memory as a lookup table, with the window index determined by the analysis segment length. Corresponding switching. Using a purely textual example, when the operator lightly presses the edge of the delay fiber optic disk, the continuous differential phase sequence... A period of slow fluctuation will occur. If a fixed long window is used, multiple analysis segments will cross the same disturbed area. Using a time delay marker... Constraint analysis segment length Afterward, the disturbed area is confined to fewer analysis segments, and the boundary between the accepted analysis segments and the analysis segments with reduced weights on the interface becomes clear.

[0076] When using it, analyze the segment length. The step interval is no longer set independently of the physical delay, but rather revolves around the actual delay marker. The changes are consistent. This ensures that the segmented spectral estimation in step three maintains the same scale as the delayed link in step one, while also reducing the impact of two types of distortions—long-segment correlation and insufficient short-segment resolution—on the differential phase spectral density. The impact.

[0077] In the analysis segment length Once determined, it is also necessary to answer the question of how much each segment contributes to the final result. If all analysis segments are equally weighted and merged, then the valid sample points are marked. If densely populated analysis segments and analysis segments with a high proportion of repair segments are treated equally, the quality boundary established in step two will become ineffective.

[0078] Therefore, the segmented differential phase spectral density is first calculated for each analysis segment. Then, segment weights are generated based on effective sample density, envelope consistency, and repair ratio. Finally, the spectra of each segment are fused, and the frequency points near the spectral valleys are written into the zero-neighborhood marker for use in step four. In the formula, segment weight : No. Each analysis segment at the final differential phase spectral density The fusion weight it occupies is a value of Effective sample density : No. Valid sample point markings within each analysis segment The percentage of sample points with a value of 1, taking values ​​of ;No. within paragraph The proportion of sample points to total sample points; envelope consistency : No. Envelope sequence within each analysis segment The ratio of the local median to the local peak value is taken as . Used to characterize the envelope sequence within this analysis segment. The degree of stability; the first Intra-segment envelope sequence The ratio of local median to local peak value; the percentage of repaired area. : No. The percentage of sample points within an analysis segment covered by the repaired segment markers from step two is taken as a value. ;No. The percentage of sample points covered by the repair section markers within the segment; stability constant. To prevent small positive numbers with a denominator of zero, the range of values ​​is: And it is much less than 1, used to ensure segment weights. The computation is stable; In the formula, the piecewise differential phase spectral density : No. The analysis segment is in the first The differential phase spectral density on each frequency grid takes the value of a non-negative real number; window sequence Loading to the 1st The time-domain window function over each analysis segment takes non-negative real numbers; segmented valid sample point markings. Marking global valid sample points Mapped to the The binary sequence within each analysis segment takes the value of ; Piecewise continuous differential phase sequence From global continuous differential phase sequence Extract and complete the segment after removing the mean and linear trend. A sequence of analysis segments, with values ​​ranging from the real number domain; frequency grid number. : The discrete frequency position number where the differential phase spectral density is located, with a value that is a non-negative integer; the total number of analysis segments. : The number of analysis segments entering the fusion phase, with a positive integer value; segment weight. The meaning follows the previous formula, in the final differential phase spectral density High-quality analysis sections are highlighted in the composition; Final differential phase spectral density The differential phase spectral density output in step three, with non-negative real values, is used as the direct input for the transfer relation inversion and bounded spectral compensation in step four; stability constant. The meaning remains the same as the previous formula, and its function is to maintain the piecewise differential phase spectral density when the effective sample points in the analysis segment are scarce. The computation is stable; In this process, the FPGA first generates a segmented continuous differential phase sequence for each analysis segment. and segmented effective sample point marking Then, the piecewise differential phase spectral density is calculated using an on-chip multiplier array. The host computer then determines the segment weights. Perform fusion and analyze the final differential phase spectral density. A local valley scan is performed, and segments with three or more consecutive frequency grids simultaneously below the adjacent background are marked as zero-point neighborhood markers. These zero-point neighborhood markers are not involved in the spectral correction in this step but are passed unprocessed to step four as prior information for reliable frequency point selection. A textual example illustrates the envelope consistency of individual analysis segments when someone near the device touches the short jumper wire at the front end of the balanced photodetector. Decline and Repair Ratio Rise, segment weight Consequently, the value decreases; the analysis segment is not deleted from the interface, but its contribution is reduced in the fusion result, resulting in the final differential phase spectral density. It is still dominated by the majority of high-quality analysis segments.

[0079] When using it, mark the valid sample points from step two. and envelope sequence Continuing to retain the data at the frequency domain level, the actual time delay obtained in step three is then marked. The entire process of segmenting the analysis is introduced. The final output is the differential phase spectral density. Its zero-point neighborhood markers possess both physical traceability and digital processing feasibility.

[0080] Step 4: Mark the actual delay and differential phase spectral density A zero-constrained inversion link is constructed, and the intrinsic phase noise power spectral density is output in a single-chain processing process that first determines the reliable frequency point, then performs bounded spectral compensation, and finally completes the result assembly. .

[0081] Among them, the differential phase spectral density obtained in step three It only delays the differential response of the self-heterodyne link to the phase fluctuations of the laser under test. Only by marking the actual time delay... After introducing the transitivity, the differential phase spectral density Only then can it be mapped to the intrinsic phase domain; simultaneously, because this transitivity is marked in the zero neighborhood. The frequency range covered will have an amplification effect where the denominator tends to be smaller. Step four cannot be used to invert all frequency points. Instead, it is necessary to first distinguish which frequency points are suitable for direct conversion and which frequency points can only be compensated by nearby reliable anchor points.

[0082] The host computer first reads the actual delay marker in the spectrum result frame. Frequency step and differential phase spectral density A transfer relation inversion kernel is constructed on the frequency grid; subsequently, based on the zero-neighborhood markers... A confidence coefficient is generated using a local noise floor reference, and then the confidence coefficient is compared with a threshold to obtain the frequency point state label. Frequency status marker The value can be either the direct inversion frequency point or the compensation frequency point.

[0083] For zero-point neighborhood labeling Let the theoretical zero frequency be ,in It is a positive integer; when a certain frequency is the grid frequency satisfy: At that time, the frequency grid is marked as the zero-neighbor frequency point, that is... ;otherwise .in The number of grid cells for zero-point expansion is preferably 1 to 3. The system explicitly calculates the actual time delay for each frequency grid. The inversion denominator is determined, and a regularization base is added at the minimum point of the denominator to avoid numerical divergence near the zero point. This yields the candidate intrinsic phase noise spectrum. Although it has not undergone frequency filtering, it already maintains a one-to-one correspondence with the physical transmission relationship, including: In the formula, the candidate intrinsic phase noise spectrum Marked according to actual delay For differential phase spectral density The frequency domain result after the initial inversion is taken as a non-negative real number; differential phase spectral density. The meaning follows from step three, and its function is to provide the first Differential phase energy on each frequency grid; frequency grid frequency : No. The actual frequency position corresponding to each frequency grid is taken as a value. ; Actual delay marker The meaning follows from step three; its function is to determine the zero-point distribution and amplification degree of the transfer relationship at different frequency positions; regularization base quantity. : Add a non-zero lower bound to the denominator of the transitive relation, with a range of values ​​of 100. Used in Preserving the candidate intrinsic phase noise spectrum near zero The values ​​are stable; frequency grid number : Discrete frequency position number, with a value of a non-negative integer, used to traverse the entire frequency axis; Preferably, the host computer checks the actual delay marker after receiving the spectrum result frame. Check if it matches the current batch, then proceed by frequency step. Reconstructed frequency grid frequency List the candidates, then calculate the candidate intrinsic phase noise spectrum. When the operator replaces the delay fiber optic tray and re-executes the measurement procedure, the actual delay marker on the interface is updated first. Subsequently, the candidate intrinsic phase noise spectrum The stripe positions shifted as a whole, and the host computer no longer used the inversion kernel corresponding to the old fiber disk.

[0084] When used, the inversion starting point in step four is directly affected by the actual time delay marker. Constraints prevent the initial delay from being marked. Misused in the final inversion. Add regularization to the baseline. Subsequently, even in frequency ranges where the denominator of the transitivity relation approaches zero, the candidate intrinsic phase noise spectrum... It will also not exhibit unbounded divergence.

[0085] Only candidate intrinsic phase noise spectrum This cannot be directly taken as the final result because the frequency grid near the zero point of the transitivity has been regularized. While it restricts unbounded growth, its spectral values ​​still contain significant uncertainties.

[0086] Frequency point reliability is constructed from two aspects: first, the current frequency grid distance to the zero-point neighborhood marker. The farther the boundary, the weaker the influence of the zero point of the transitivity relationship; secondly, the candidate intrinsic phase noise spectrum. A higher frequency relative to the local noise floor reference indicates a lower likelihood that the frequency point is dominated by the measurement noise floor. By incorporating both factors into the confidence coefficient and then comparing it with a threshold, the frequency point status label can be obtained. ,in: In the formula, the credibility coefficient : No. Each frequency grid can directly employ the candidate intrinsic phase noise spectrum. The degree, the range of values ​​is Candidate intrinsic phase noise spectrum The meaning remains the same as before; local noise floor reference : At the current frequency grid frequency Centered on and removing the zero-point neighborhood marker Within the local frequency window after the frequency coverage point, from the candidate intrinsic phase noise spectrum The extracted low quantile reference value, a non-negative real number, is used to characterize the instrument noise floor and remaining fringe baseline near the current frequency grid; zero-point distance. Current frequency grid frequency The distance to the nearest zero-neighborhood boundary frequency is taken as a value. Suppression constant Zero Distance For the credibility coefficient The adjustment constant is set to a value of ; Frequency point status marker By comparing credibility coefficients With confidence threshold Generate, when The time is recorded as the direct inversion frequency point, when The time is denoted as the compensation frequency point; where, the confidence threshold is... The range of values ​​is Its function is to delineate the direct inversion region and the compensation region; Preferably, local noise floor reference The value is obtained using the 20th or 25th percentile value within a local frequency window. This is illustrated in a purely textual example when the measurement object is a low-noise laser and multiple zero-neighborhood markers are output in step three. At this time, the host computer interface will display status bars of different colors on both sides of these frequency bands; the frequency point status markers near the zero boundary It is set as the compensation frequency, far from zero and above the local noise floor reference. Frequency point status marker It was set as the direct inversion frequency point.

[0087] When using it, utilize the credibility coefficient. At zero distance and local noise floor reference A continuous transition is established between them. At the same time, it can retain a sufficient number of direct inversion frequency points as compensation anchor points, while avoiding the direct output of pseudo-valley values ​​near the zero point as the final result.

[0088] Step four has already distinguished between the direct inversion frequency and the compensated frequency. However, if only simple linear interpolation is performed on the compensated frequency, although it can cross the zero-neighborhood, it will destroy the candidate intrinsic phase noise spectrum. The local slope information is already preserved in the reliable section; if a high-order polynomial is used for global fitting, it is easy to introduce additional oscillations between anchor points.

[0089] The host computer first searches for continuous compensated frequency segments on the frequency axis. For each compensated frequency segment, at least one directly inverted frequency point is found on both the left and right sides as the left and right anchor points, respectively. If there are insufficient directly inverted frequency points on one side, the search range is expanded until an anchor point is found, or the compensated frequency segment is marked as a caution segment. Subsequently, anchor point spectral values ​​are constructed between the left and right anchor points for each compensated frequency segment. Then take the anchor point spectral value With candidate intrinsic phase noise spectrum According to credibility coefficient The fusion process generates the final intrinsic phase noise power spectral density. Finally, the host computer outputs the intrinsic phase noise power spectral density. Convert to output phase noise level and mark the frequency point status. Zero-point neighborhood marker The cautious section markers are written to the results file.

[0090] Furthermore, the host computer searches for the left and right anchor points for each compensated frequency segment, both of which are derived from the frequency status flags. This is a frequency grid for direct frequency inversion. Subsequently, the anchor point spectral values... Instead of global fitting, piecewise cubic Hermite conformal interpolation or monotonic rational interpolation is used to generate the slope point by point between the left and right anchor points. This approach inherits the local slope trend at the left and right anchor points while avoiding additional oscillations in the middle of the compensation section caused by higher-order polynomials.

[0091] Anchor point spectral value After generation, it is compared with the candidate intrinsic phase noise spectrum. According to credibility coefficient Mixed, in which: In the formula, the final intrinsic phase noise power spectral density The intrinsic phase noise spectrum result after confidence constraint and compensation processing is a non-negative real number; the confidence coefficient... The meaning remains the same as before, and its function is to control the candidate intrinsic phase noise spectrum. With anchor point spectral values The fusion ratio between them; candidate intrinsic phase noise spectrum The meaning remains the same as before; anchor point spectral value : The frequency domain reference value generated by conformal interpolation from the directly inverted frequency points on both sides of the current compensated frequency segment, is a non-negative real number; frequency grid number The meaning remains the same as before; Preferably, if there are a sufficient number of directly inverted frequency points on both the left and right sides of a certain compensated frequency segment, then the anchor point spectrum value Piecewise cubic Hermite conformal interpolation is used for generation; if anchor points are sparse on one side, monotonically rational interpolation is used to prevent edge overshoot. A plain text example illustrates this: when a continuous segment of zero-point neighborhood markers appears on the interface... When covering a segment, the host computer will select a set of direct inversion frequency points on the left and right sides of the segment as anchor points. Then the compensation curve will smoothly transition from the left anchor point to the right anchor point, and no sharp dips or protrusions will appear in the middle of the segment.

[0092] In practice, the compensation frequency points no longer cross the zero-point segment via a simple straight line connection, but instead form anchor point spectrum values ​​that are coordinated with the slope of the reliable segment under the constraints of left and right anchor points. Then, through the credibility coefficient... With candidate intrinsic phase noise spectrum The fusion and compensation results will neither completely deviate from the physical inversion nor output the pseudo-valley values ​​near zero as is.

[0093] After completing the final intrinsic phase noise power spectral density After construction, the results still need to be delivered to the user. If only the intrinsic phase noise power spectral density is output... If the linear value is not specified, it is difficult for users to directly compare it with historical measurement records; if only graphs are output without frequency point status markers... and zero-point neighborhood label If this is not the case, subsequent reviews will be unable to distinguish which frequency points are obtained directly from the inversion and which frequency points are obtained after compensation.

[0094] Therefore, the results are compiled into three types of objects—linear spectrum, logarithmic level spectrum, and state flags—and output uniformly, where: In the formula, the output phase noise level The final intrinsic phase noise power spectral density The displayed quantity obtained after transformation to logarithmic dimensions takes the value in the real number domain; the final intrinsic phase noise power spectral density The meaning remains the same as before; The power spectral density is defined using a two-sided phase aperture; for external display, it is converted to a one-sided phase noise level. If a single-sided spectral aperture is used, the logarithmic conversion formula no longer includes the half coefficient, and this remains consistent throughout the text.

[0095] logarithmic function The base-10 logarithmic operation is used to compress the intrinsic phase noise power spectral density. The dynamic range; coefficients Used to convert the bilateral phase spectrum to a single-sided display aperture, its function is to reduce the output phase noise level. Consistent with commonly used phase noise spectrum aperture; frequency grid number The meaning remains the same as before; In practical implementation, the host computer calculates the eigenphase noise power spectral density. Output phase noise level Frequency point status marker Zero-point neighborhood marker Actual delay marker The batch number is written to the same result file. The result file preferably includes five parts: a file header, a frequency axis, a linear spectrum array, a logarithmic spectrum array, and a status flag array. As illustrated in the plain text example, after the measurement is completed, a continuous output phase noise level is displayed on the interface. The curve, with frequency status markers displayed below it. The resulting state band; when the operator drags the cursor over a certain frequency position, the interface will simultaneously display the intrinsic phase noise power spectral density corresponding to that point. Output phase noise level Frequency point status marker and zero-point neighborhood label .

[0096] When used, it ensures the intrinsic phase noise power spectral density. It can be called upon by subsequent algorithms, which in turn affects the output phase noise level. It allows for direct manual viewing and comparison with historical documents, while also using frequency point status markers. and zero-point neighborhood label The engineering semantics of the compensation process are preserved.

[0097] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0098] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0099] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0100] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A laser phase noise measurement method based on time-delay self-heterodyne and digital demodulation, including: The continuous optical signal output from the laser under test is divided into a frequency-shifting arm and a delay arm after being adjusted by a variable optical attenuator. The frequency-shifting arm is frequency-shifted by an acousto-optic modulator, and the delay arm is delayed by a delay fiber. The two arms are coupled by a 2×2 fiber coupler and the original electronic beat frequency signal is obtained by a balanced photodetector. The feature is that it includes: The original electronic beat frequency signal is converted from analog to digital, and synchronous quadrature demodulation is performed using a digital local oscillator that is from the same source as the acousto-optic modulator driving signal. Combined with envelope confidence determination and abnormal sample point repair, a continuous differential phase sequence is obtained. The actual optical time delay is determined based on the beat spectrum characteristics of the continuous differential phase sequence and the original electronic beat frequency signal, and the differential phase power spectral density is formed. Based on the actual optical time delay, the differential phase power spectral density is converted by the transfer relationship, and the zero-neighborhood and low-confidence frequency points are screened and bounded spectrum compensation is performed to obtain the intrinsic phase noise power spectral density of the laser under test.

2. The laser phase noise measurement method according to claim 1, characterized in that: Establish a time-delayed self-heterodyne beat frequency measurement link and preset the system operating point, including: adjusting the beat frequency envelope amplitude entering the balanced photodetector through a variable optical attenuator, so that the balanced photodetector and subsequent acquisition circuits operate within the linear range, and sending the output of the balanced photodetector as the raw electrical beat frequency signal into the acquisition module.

3. The laser phase noise measurement method according to claim 2, characterized in that: Synchronous quadrature demodulation using a digital local oscillator that is from the same source as the acousto-optic modulator driving signal includes: performing analog-to-digital conversion on the original electronic beat frequency signal to obtain a discrete beat frequency sequence, and using the digital local oscillator for synchronous mixing to obtain in-phase and quadrature components, and then performing low-pass filtering, decimation downsampling, and amplitude normalization on the in-phase and quadrature components.

4. The laser phase noise measurement method according to claim 3, characterized in that: A continuous differential phase sequence is obtained by combining envelope reliability determination and abnormal sample point repair, including: determining the envelope sequence and envelope reference of the beat frequency signal based on the in-phase component and the quadrature component, generating valid sample point markers accordingly, and performing freezing, removal, interpolation repair or neighborhood smoothing repair on invalid samples.

5. The laser phase noise measurement method according to claim 4, characterized in that: The continuous differential phase sequence is obtained by: extracting the instantaneous differential phase from the baseband differential information corresponding to the effective sample points, and performing phase expansion, linear trend elimination and continuous accumulation processing on the instantaneous differential phase to obtain the continuous differential phase sequence for subsequent determination of actual optical time delay and calculation of differential phase power spectral density.

6. The laser phase noise measurement method according to claim 5, characterized in that: The actual optical delay is determined based on the beat spectrum characteristics of the continuous differential phase sequence and the original electronic beat frequency signal, including: combining the nominal length of the delay fiber, the fiber refractive index, the additional optical path of the system connection, and the interference fringe spacing, envelope zero position and first extremum position in the beat spectrum to determine the actual optical delay.

7. The laser phase noise measurement method according to claim 6, characterized in that: The differential phase power spectral density is formed by: setting the window length, window overlap rate, averaging number, decimation rate and frequency resolution of the spectrum analysis according to the actual optical time delay; segmenting the continuous differential phase sequence; and performing DC removal, detrending, windowing and average spectrum estimation on each segment in sequence.

8. The laser phase noise measurement method according to claim 7, characterized in that: The screening of zero-neighborhood and low-confidence frequency points includes: identifying zero-neighborhood frequency points, frequency points near the system noise floor, and frequency points with insufficient effective samples in the differential phase power spectral density based on the zero-point position of the transfer relationship corresponding to the actual optical time delay, and using the identification results as the basis for frequency point screening.

9. The laser phase noise measurement method according to claim 8, characterized in that: Bounded spectral compensation includes: directly performing transfer relationship conversion on the selected reliable frequency points; performing local smoothing, interpolation fitting, or regression equalization on the selected low reliable frequency points based on the spectral values ​​of adjacent reliable frequency points; and forming the intrinsic phase noise power spectral density together with the direct conversion results.

10. The laser phase noise measurement method according to claim 9, characterized in that: The intrinsic phase noise power spectral density, time-domain phase jitter sequence, and their statistical results are transmitted to the user's computer via network cable and router. The host computer software displays the intrinsic phase noise power spectral density, stores the data, performs historical comparisons, and processes and outputs the test results.

Images