Linear frequency modulation DFB laser smoothing predistortion iteration nonlinear correction method

Abstract

The invention discloses a linear frequency modulation DFB laser smooth predistortion iteration nonlinear correction method, and relates to the technical field of photoelectron and laser radar, and the method comprises the following steps: generating an initial ideal triangular wave driving voltage signal through an FPGA, and driving a laser through a digital-to-analog converter; an optical signal output by the laser is coupled into the unbalanced interferometer, a beat frequency signal is obtained through the balanced photoelectric detector, and the beat frequency signal is collected back to the FPGA through analog-to-digital conversion of the analog-to-digital converter; performing Hilbert transform on the collected beat frequency signal, and demodulating to obtain the actual instantaneous frequency of the laser; constructing an ideal frequency curve, calculating a frequency error between the actual instantaneous frequency and the smoothed ideal frequency, and converting the frequency error into a voltage error; and updating the driving voltage waveform of the next round of iteration according to the voltage error, and repeating the steps until the nonlinearity reaches a preset threshold. By introducing a smooth ideal frequency model, error abrupt change at an inflection point is eliminated, and high-precision and fast-convergence linear frequency sweeping is realized.

Description

Technical Field

[0001] This invention relates to the field of optoelectronics and lidar technology, specifically to a method for smoothing predistortion iterative nonlinear correction of linear frequency modulated DFB lasers. Background Technology

[0002] Frequency modulated continuous wave (FMCW) lidar has important applications in autonomous driving and remote sensing due to its high resolution and anti-interference capabilities. However, directly modulated semiconductor lasers (such as DFB lasers) have inherent frequency modulation nonlinearity due to thermal and carrier effects. This nonlinearity leads to spectral broadening, which seriously affects ranging accuracy and imaging quality.

[0003] Among existing nonlinear correction techniques, optical phase-locked loops (OPLLs) are complex and costly; resampling techniques are limited by the Nyquist sampling theorem; and iterative predistortion algorithms have attracted much attention due to their high hardware efficiency. However, traditional iterative algorithms based on linear fitting are prone to alternating error signs at the inflection points of the swept frequency signal (the peaks and troughs of the triangular wave) due to the discontinuity between the linear fitting of the ideal frequency and the actual physical process. This leads to sudden changes in the driving voltage, which introduces high-frequency components, causing the algorithm to fail to converge at the inflection points and limiting the final linearity improvement. Therefore, to address the above-mentioned shortcomings of existing technologies, we propose a smooth predistortion iterative nonlinear correction method for linearly frequency-modulated DFB lasers. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a smooth predistortion iterative nonlinear correction method for linearly frequency-modulated DFB lasers, which solves the problems of poor convergence at the sweep frequency inflection point and easy generation of sudden changes in driving voltage in the existing iterative predistortion technology mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for smoothing predistortion iterative nonlinear correction of a linearly frequency-modulated DFB laser, comprising the following steps: S1: The FPGA generates the initial ideal triangular wave drive voltage signal, which drives the DFB laser after passing through a digital-to-analog converter (DAC); S2: The optical signal output from the laser is coupled into an unbalanced Mach-Zehnder interferometer (MZI), the beat frequency signal is obtained through a balanced photodetector (BPD), and then the signal is converted from analog to digital by an analog-to-digital converter and collected back to the FPGA; S3: Perform a Hilbert transform on the acquired beat frequency signal and demodulate it to obtain the actual instantaneous frequency of the laser. ; S4: Construct the ideal frequency curve The ideal frequency curve is directly calculated based on bandwidth and period in the middle linear region of the sweep cycle, and a smoothing function is used in the inflection point region of the sweep cycle to eliminate the polarity abrupt change of the relative error. S5: Calculate the frequency error between the actual instantaneous frequency and the smoothed ideal frequency, and convert the frequency error into a voltage error; S6: Update the driving voltage waveform for the next iteration based on the voltage error; S7: Repeat steps S2 to S6 until the nonlinearity reaches the preset threshold or the number of iterations reaches the preset value.

[0006] Preferably, the smoothing process of the ideal frequency curve in step S4 specifically includes: For the beginning and end phases of up-sweep or down-sweep, i.e., the inflection point region, a quadratic function is used to smoothly connect the ideal linear frequency curve, resulting in the smoothed ideal frequency. Represented as:

[0007] in, The length of the smooth region, For the ideal sweep slope, For half a modulation cycle, and The fitting coefficients should be set according to the specific experiment.

[0008] Preferably, the driving voltage update formula in step S6 is:

[0009] in For the first The driving voltage of the next iteration. As the error factor, The first number calculated based on the frequency error Secondary voltage correction amount.

[0010] A smoothing predistortion iterative nonlinear correction system for a linear frequency modulated DFB laser includes an FPGA control module, a drive circuit module, an optical path detection module, and a data acquisition module.

[0011] Preferably, the FPGA control module is used to generate drive waveforms, receive feedback signals, perform Hilbert transforms, construct smooth ideal frequency curves, and update drive voltage algorithms.

[0012] Preferably, the driving circuit module includes a DAC and an adder, used to superimpose the modulation voltage generated by the FPGA control module with the DC bias voltage to drive the DFB laser.

[0013] Preferably, the optical path detection module consists of an MZI interferometer composed of an optical fiber coupler and a time-delay optical fiber, and a balanced photodetector (BPD), used to convert laser frequency changes into electrical signals.

[0014] Preferably, the data acquisition module includes an analog-to-digital converter for digitizing and transmitting the analog voltage signal output by the balanced photodetector (BPD) to the FPGA control module.

[0015] This invention provides a method for smoothing predistortion iterative nonlinear correction of linearly frequency-modulated DFB lasers, which has the following advantages: 1. This linear frequency modulated DFB laser smoothing predistortion iterative nonlinear correction method uses FPGA to control laser driving and data acquisition, and obtains the laser's beat frequency signal and calculates the instantaneous frequency through an auxiliary interferometric optical path. Unlike traditional linear fitting methods, this scheme constructs an ideal reference frequency curve that has been smoothed. In particular, it introduces quadratic function smoothing at the inflection point of the sweep frequency signal, eliminating the sudden change in driving voltage caused by the alternation of relative error signs. It solves the problem of non-convergence of traditional iterative predistortion at the sweep frequency transition point, realizes high-precision linear frequency modulation, can significantly reduce sweep frequency nonlinearity, and improve the ranging accuracy and imaging quality of FMCW LiDAR. Attached Figure Description

[0016] Figure 1 A schematic diagram of the structure of a nonlinear correction system based on a smooth predistortion iterative algorithm provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the principle and process of the smoothing predistortion iterative algorithm in an embodiment of the present invention; Figure 3 This is a schematic diagram of a smoothed ideal frequency curve constructed in an embodiment of the present invention. Figure 4 The figures show the spectrum of the beat frequency signal before correction and the residual diagram in this embodiment of the invention; (a) is a comparison diagram of the initial driving voltage and beat frequency signal waveform before correction, (b) is the spectrum of the beat frequency signal before correction, (c) is the upper sweep light before correction, and (d) is the instantaneous frequency and residual error of the lower sweep light; Figure 5 The figures show the spectrum and residual diagram after correction using the traditional linear fitting method in this embodiment of the invention; (a) is a comparison diagram of the waveforms of the initial driving voltage and beat frequency signal after correction, (b) is the spectrum of the beat frequency signal after correction, (c) is the upper sweep frequency light after correction, and (d) is the instantaneous frequency and residual error of the lower sweep frequency light; Figure 6The figure shows the spectrum and residual diagram after correction using the smoothing algorithm of the present invention in this embodiment of the invention; (a) is a comparison diagram of the waveforms of the initial driving voltage and beat frequency signal after correction, (b) is the spectrum of the beat frequency signal after correction, (c) is the upper sweep frequency light after correction, and (d) is the instantaneous frequency and residual error of the lower sweep frequency light; Figure 7 This is a comparison chart of the nonlinearity convergence curves after different algorithms are corrected and the number of iterations varies in this embodiment of the invention. Figure 8 The image shows the effect of FMCW LiDAR imaging using the light source corrected by the present invention; (a) in the figure shows the imaging effect after correction using the method of the present invention, and (b) shows the imaging effect after correction using the conventional method. Figure 9 The image entropy comparison diagram is shown for FMCW LiDAR imaging using the light source corrected by the present invention; in the figure, (a) is the original data image generated by the smooth predistortion iterative algorithm, (b) is the grayscale converted image generated by the smooth predistortion iterative algorithm, (c) is the original data image generated by the traditional linear fitting predistortion iterative algorithm, and (d) is the grayscale converted image generated by the traditional linear fitting predistortion iterative algorithm. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Please see Figures 1 to 9 This invention provides a technical solution: a nonlinear correction method for FMCW lasers based on a smooth predistortion iterative algorithm, mainly comprising the following steps: S1. Use FPGA to generate an initial ideal triangular wave drive voltage signal, which drives the DFB semiconductor laser after digital-to-analog conversion; S2. Couple the optical signal output from the laser to an unbalanced Mach-Zehnder interferometer (MZI), convert the optical signal into a beat frequency electrical signal, and collect it back to the FPGA; S3. Demodulate the acquired beat frequency signal to obtain the actual instantaneous frequency of the laser as time changes; S4. Based on the sweep bandwidth and modulation period of the laser, construct an ideal frequency curve after smoothing; the ideal frequency curve uses a smooth function transition in the sweep inflection point region to eliminate the polarity abrupt change of the relative error; S5. Calculate the frequency error between the actual instantaneous frequency and the smoothed ideal frequency curve, and convert the frequency error into a correction amount for the driving voltage; S6. Based on the iterative learning control strategy, update the driving voltage waveform of the next cycle using the voltage correction amount, and repeat the above process until the sweep frequency nonlinearity meets the preset requirements.

[0019] S1 specifically includes the following steps: The initial linear frequency modulation parameters are preset in the FPGA control module, and the generation period is... Peak-to-peak value is Standard triangular wave digital signal; The digital signal is converted into an analog voltage signal by a high-speed digital-to-analog converter (DAC); the analog voltage signal is superimposed with a DC bias current source using an adder circuit to generate the final driving current, which is injected into the DFB semiconductor laser to enable the laser to generate a preliminary swept frequency optical signal.

[0020] S2 specifically includes the following steps: The optical signal output from the DFB laser is split proportionally (e.g., 99:1) using an optical fiber coupler, and a small portion of the optical signal is introduced into the auxiliary measurement optical path. The auxiliary measurement optical path adopts an unbalanced Mach-Zehnder interferometer (MZI) structure, which includes two 50:50 couplers and a fixed-length delay single-mode fiber to generate interference beat frequencies. A balanced photodetector (BPD) is used to receive the interference optical signal and convert it into a beat frequency signal in the form of an analog voltage. The analog beat frequency signal is converted into a digital sequence by an analog-to-digital converter (ADC) at a preset sampling rate and transmitted in real time to the random access memory (RAM) inside the FPGA for caching.

[0021] S3 specifically includes the following steps: The digital signal processing module inside the FPGA is invoked to perform a Hilbert transform on the buffered beat frequency signal sequence; the instantaneous phase information of the transformed signal is extracted; the instantaneous phase is differentiated, and the actual instantaneous frequency of the laser under the current driving voltage is obtained by demodulation. .

[0022] S4 specifically includes the following steps: Abandoning the traditional full-segment linear fitting method, a piecewise defined ideal frequency model is constructed in the FPGA. In the middle linear region of the sweep frequency period ( The ideal frequency is set as a linear function. ,in The target sweep slope; in the initial segment of the sweep cycle ( ) and the concluding paragraph ( (i.e., the frequency sweep inflection point region, using a quadratic function) Smooth the fit to the ideal frequency; use the smooth fit coefficients The solution ensures that the ideal frequency curve is continuous and smooth at the inflection point, thereby avoiding the alternation of positive and negative error signs caused by frequency steps in subsequent error calculations.

[0023] S5 specifically includes the following steps: The actual instantaneous frequency obtained in step S3 With the smooth ideal frequency constructed in step S4 By subtracting point by point, the frequency error curve is obtained. Using the frequency modulation response efficiency parameter of the laser, the frequency error curve is... Linear mapping to voltage error curve The voltage error curve eliminates high-frequency oscillation components at the inflection point due to the use of a smooth reference.

[0024] S6 specifically includes the following steps: Iterative update formula Calculate the next generation of driving waveforms, where The driving voltage for the current cycle. Let [the error convergence factor] be the updated [factor]. Write the waveform into the FPGA's waveform memory to replace the original drive waveform; Determine whether the current sweep frequency nonlinearity or residual error is lower than a preset threshold (e.g., nonlinearity < 0.03%). If not, re-execute steps S1 to S5 using a new driving waveform until the convergence condition is met.

[0025] like Figure 1 As shown, an FPGA-based FMCW laser nonlinear correction system includes an FPGA control module, a laser driver module, a DFB laser assembly, and an auxiliary optical interferometry measurement module.

[0026] The FPGA control module serves as the core processing unit of the system, used to generate initial digital drive signals and process feedback signals. The output of the FPGA is connected to the laser drive module, which includes a high-speed digital-to-analog converter (DAC) and an adder circuit. The digital waveform generated by the FPGA is converted into an analog voltage signal by the DAC and then superimposed with the DC bias current (DCBias) in the adder circuit to jointly drive the DFB semiconductor laser.

[0027] The output optical path of the DFB laser is connected to a 99:1 fiber coupler, where 99% of the optical power output is used for subsequent LiDAR detection missions, and 1% of the optical power is coupled into the auxiliary optical interferometry module as monitoring light.

[0028] The auxiliary optical interferometry module adopts an unbalanced Mach-Zehnder interferometer (MZI) structure, including a first 50:50 coupler, a delay fiber, a second 50:50 coupler, and a balanced photodetector (BPD). In this embodiment, the length of the delay fiber is preferably 4 meters, providing a relative delay of approximately 19.47 ns. After the two interference optical signals are converted into beat frequency voltage signals by the BPD, they are input to the analog-to-digital converter (ADC). The digital output of the ADC is connected back to the FPGA control module to form a closed-loop feedback control circuit.

[0029] Example 2: A nonlinear correction method for FMCW lasers based on a smooth predistortion iterative algorithm, such as... Figure 2 As shown, the nonlinear correction method is executed in the internal logic of the FPGA, and the specific steps are as follows: Step 1: Initial Excitation and Signal Acquisition FPGA generation cycle during system startup 50 μs (corresponding to a repetition frequency of 20 kHz), peak-to-peak value A standard triangular wave voltage signal of 1.75V is used as the initial driving waveform. The ADC acquires the beat frequency signal $I(t)$ output by the MZI in real time and caches it in the RAM of the FPGA.

[0030] Step 2: Instantaneous frequency demodulation For the acquired beat frequency signal The Hilbert Transform is performed; specifically, the FPGA extracts the instantaneous phase of the analytic signal. The actual instantaneous frequency of the laser is obtained by differentiating it with respect to time. :

[0031] This step is achieved through a digital differentiation algorithm, which can accurately obtain the frequency response characteristics of the laser under the current driving voltage.

[0032] Step 3: Construct a smooth ideal frequency reference curve This step is the core of this solution. In order to eliminate the abrupt error changes at the inflection points of the triangular wave in traditional algorithms, this embodiment does not use full-segment linear fitting, but instead constructs a piecewise defined smooth ideal frequency model. Let the time for a single up-scan or down-scan be . The sweep bandwidth is (Approximately 8.6 GHz in this embodiment), the time length of the inflection point smoothing region is defined as... .

[0033] Intermediate linear region ( ); In this region, the ideal frequency strictly follows a linear relationship. , Inflection point smoothing region ( and ): In this region, a quadratic function is used. Instead of linear functions, the system solves for coefficients using the least squares method or boundary condition constraints. , making exist and The function values ​​and first derivatives at the connection points remain continuous; this smoothing process ensures the reference curve remains intact. The smoothness avoids errors in frequency calculation. At that time, due to the conflict between the inertia of the actual physical frequency and the ideal linear angle, a violent jump between positive and negative signs occurs.

[0034] Step 4: Error Calculation and Voltage Update The frequency error is obtained by calculating the difference between the actual instantaneous frequency and the smoothed ideal frequency. :

[0035] By utilizing the laser's frequency modulation efficiency (FM Efficiency) parameter, the frequency error is linearly mapped to a correction amount for the driving voltage. , Update the control law using iterative learning Drive voltage of the next cycle :

[0036] in, The error factor is used to control the iteration speed and stability. In this embodiment, The value of is set empirically based on the system noise level to ensure smooth convergence.

[0037] Step 5: Iterate Repeat steps 2 to 4 above to determine whether the current sweep frequency nonlinearity is lower than a preset threshold (e.g., 0.03%) or whether the number of iterations has reached a preset value.

[0038] Regarding the verification of nonlinear correction performance, based on the hardware system constructed in Embodiment 1 and the correction method described in Embodiment 2, a verification experiment was conducted on the frequency sweep linearity of the DFB semiconductor laser. The experimental conditions were set as follows: driving triangular wave period The frequency response time is 50 μs (corresponding to a repetition frequency of 20 kHz), the peak-to-peak voltage is 1.75 V, and the laser sweep bandwidth is approximately 8.6 GHz. The experiment first compared the spectral characteristics of the beat frequency signal before and after correction, such as... Figure 4 As shown, in the uncorrected state, due to the inherent frequency modulation nonlinearity of the laser, the beat frequency signal spectrum is severely broadened, with a full width at half maximum (FWHM) of approximately 2.1 MHz and a peak intensity of only -9.87 dBm; at this time, the nonlinearities of the up-sweep and down-sweep frequencies are 3.3% and 1.3%, respectively. Subsequently, iterative corrections were performed using both the traditional linear fitting predistortion method and the smoothing predistortion method proposed in this invention. After correction using traditional linear fitting methods, such as Figure 5 As shown, although the nonlinearity is reduced, the driving voltage changes abruptly due to the discontinuity of the inflection point processing frequency, and there is still obvious high-frequency sidelobe noise in the beat frequency signal spectrum, with a peak intensity of about 1.6 dBm. After correction using the smoothing predistortion method of this invention, as shown Figure 6 As shown, the FWHM of the beat frequency signal spectrum is significantly compressed to below 0.12MHz, the peak intensity is increased to 5.17dBm, the spectral noise floor is flat, and the nonlinearity of both the up-sweep and down-sweep frequencies is reduced to 0.022%, corresponding to a residual nonlinearity coefficient 1-r 2 They reached 8.458×10 -7 and 8.4208×10 -7 .

[0039] Further comparison of the convergence stability of the two algorithms, such as Figure 7 As shown, during the iteration process, the nonlinearity of the traditional linear fitting algorithm will rebound (diverge) after decreasing; while the nonlinearity of the method of this invention shows a monotonically decreasing trend with the increase of the number of iterations, and converges stably after about 2700 iterations, verifying the significant improvement effect of smoothing on the stability of the algorithm.

[0040] Finally, an FMCW ranging system was built based on the corrected light source to measure the distance to targets in the range of 0.5m to 5m. Experimental data show that after correction using the method of this invention, the maximum distance measurement error of the system is only 9mm, which is better than the 16mm after correction by the traditional method, thus meeting the requirements of high-precision measurement.

[0041] FMCW LiDAR Imaging Applications Using a coaxial scanning architecture, a two-dimensional scanning galvanometer was employed to scan a diffuse reflection target (a cardboard cutout of the letter "NJUPT") located 1.1 meters away point by point, achieving an imaging resolution of 128×128 pixels. This embodiment introduces image entropy as an evaluation index of imaging quality. The lower the image entropy, the more concentrated the gray-level distribution of the image, the higher the contrast between the target and the background, and the better the imaging quality.

[0042] Figure 8 (a) shows the imaging results using the smooth pre-distortion corrected light source of the present invention. The image background is clean and the target outline is clear and sharp. Figure 8 (b) shows the imaging results using traditional linear fitting to correct the light source. The image background has obvious noise and jitter, and the target edge is relatively blurry.

[0043] Figure 9 The image entropy of the imaging results using the smooth pre-distortion correction light source of this invention is shown. Calculations show that the image entropy is reduced to 1.4063 bits. This result demonstrates that the method of this invention can effectively suppress phase noise and significantly improve the 3D imaging sharpness and signal-to-noise ratio of FMCW LiDAR, verifying the application effect of the FMCW LiDAR system based on the correction method of this invention in the field of 3D imaging. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for smoothing predistortion iterative nonlinear correction of a linearly frequency-modulated DFB laser, characterized in that: Includes the following steps: S1: The initial ideal triangular wave drive voltage signal is generated by the FPGA and then driven by the DFB laser after passing through the digital-to-analog converter (DAC); S2: The optical signal output from the laser is coupled into an unbalanced Mach-Zehnder interferometer (MZI), the beat frequency signal is obtained through a balanced photodetector (BPD), and then the signal is converted from analog to digital by an analog-to-digital converter and collected back to the FPGA; S3: Perform a Hilbert transform on the acquired beat frequency signal and demodulate it to obtain the actual instantaneous frequency of the laser. ; S4: Constructing the ideal frequency curve The ideal frequency curve is directly calculated based on bandwidth and period in the middle linear region of the sweep cycle, and a smoothing function is used in the inflection point region of the sweep cycle to eliminate the polarity abrupt change of the relative error. S5: Calculate the frequency error between the actual instantaneous frequency and the smoothed ideal frequency, and convert the frequency error into a voltage error; S6: Update the driving voltage waveform for the next iteration based on the voltage error; S7: Repeat steps S2 to S6 until the nonlinearity reaches the preset threshold or the number of iterations reaches the preset value.

2. The method for smoothing predistortion iterative nonlinear correction of a linearly frequency-modulated DFB laser according to claim 1, characterized in that: The smoothing process for the ideal frequency curve in step S4 is as follows: For the beginning and end phases of up-sweep or down-sweep, i.e., the inflection point region, a quadratic function is used to smoothly connect the ideal linear frequency curve, resulting in the smoothed ideal frequency. Represented as: ； in The length of the smooth region, For the ideal sweep slope, For half a modulation cycle, and The fitting coefficients should be set according to the specific experimental conditions to ensure smoothness.

3. The method for smoothing predistortion iterative nonlinear correction of a linearly frequency-modulated DFB laser according to claim 1, characterized in that: The driving voltage update formula in step S6 is: ； in For the first The driving voltage of the next iteration. As the error factor, The first number calculated based on the frequency error Secondary voltage correction amount.

4. A smooth predistortion iterative nonlinear correction system for a linearly frequency-modulated DFB laser, used to implement the smooth predistortion iterative nonlinear correction method for a linearly frequency-modulated DFB laser as described in claim 1, characterized in that: It includes an FPGA control module, a driver circuit module, an optical path detection module, and a data acquisition module.

5. The linear frequency modulated DFB laser smoothing predistortion iterative nonlinear correction system according to claim 4, characterized in that: The FPGA control module is used to generate drive waveforms, receive feedback signals, perform Hilbert transforms, construct smooth ideal frequency curves, and update drive voltage algorithms.

6. The linear frequency modulated DFB laser smoothing predistortion iterative nonlinear correction system according to claim 4, characterized in that: The driving circuit module includes a DAC and an adder, which are used to superimpose the modulation voltage generated by the FPGA control module with the DC bias voltage to drive the DFB laser.

7. The linear frequency modulated DFB laser smoothing predistortion iterative nonlinear correction system according to claim 4, characterized in that: The optical path detection module consists of an MZI interferometer composed of an optical fiber coupler and a time-delay optical fiber, as well as a balanced photodetector (BPD), used to convert laser frequency changes into electrical signals.

8. The linear frequency modulated DFB laser smoothing predistortion iterative nonlinear correction system according to claim 4, characterized in that: The data acquisition module includes an analog-to-digital converter for digitizing and transmitting the analog voltage signal output by the balanced photodetector (BPD) to the FPGA control module.

Images