A nanometer-scale coating thickness on-line quantitative detection device and method based on spectral analysis

Abstract

This invention provides an online measurement device and method for nanoscale film thickness based on spectral analysis. The method includes: acquiring absolute phase-encoded data generated by the rotation of the workpiece mounting frame; dividing the spectral acquisition time sequence into a baseline extraction time window and a target thickness measurement time window based on the absolute phase-encoded data; within the baseline extraction time window, controlling the probe to acquire ambient background scattered light beams to generate a real-time background spectrum and updating it to obtain the target reference spectrum; within the target thickness measurement time window, controlling the probe to receive the mixed reflection spectrum returned from the surface of the workpiece; performing differential alignment processing between the mixed reflection spectrum and the target reference spectrum to extract the target interference spectrum and analyze the nanoscale film thickness. This invention effectively eliminates measurement interference caused by dynamic ambient background luminescence, improving the accuracy of online film thickness detection.

Description

Technical Field

[0001] This invention relates to the field of optical thin film inspection technology, and in particular to an online measurement device and method for nanoscale coating thickness based on spectral analysis. Background Technology

[0002] Currently, in industrial coating production, vacuum coating chambers are typically operated under high vacuum and continuous heating, and contain plasma radiation sources for depositing film-forming materials. To monitor film quality, a spectral probe is usually used to project a probe beam onto the surface of the workpiece and receive the reflected beam that interferes after penetrating the thin film. By analyzing the interference spectral characteristics of the reflected beam, the thickness scalar of the nanoscale coating can be calculated.

[0003] In long-cycle continuous thin film deposition processes, plasma discharge density and luminescence intensity exhibit continuous dynamic drift. Plasma glow and ambient thermal radiation emitted by the heating source generate complex ambient background scattered light within the vacuum coating chamber. The intensity and spectral energy distribution of this ambient background radiation are affected by fluctuations in evaporation source current parameters and vacuum level adjustments, resulting in dynamic changes. Fixed calibration reference data are typically insufficient to accommodate real-time ambient luminescence drift.

[0004] When a spectral probe receives interference reflection signals from a workpiece surface, it often simultaneously receives ambient background scattered light from inside the chamber. This aliasing phenomenon causes a dynamic shift in the analytical interference spectral intensity baseline. This shift in the baseline can easily lead to positioning errors at interference extrema such as peaks and troughs, resulting in deviations in the final calculated nanoscale coating thickness values ​​and affecting the reliability of high-precision online measurement and process closed-loop control. Summary of the Invention

[0005] The purpose of this invention is to provide an online measurement device and method for nanoscale coating thickness based on spectral analysis, so as to solve the problems mentioned in the background art.

[0006] In a first aspect, the present invention provides an online measurement device for nanoscale coating thickness based on spectral analysis, comprising a vacuum coating chamber, a workpiece mounting frame disposed inside the vacuum coating chamber and supporting the workpiece to be measured for continuous rotation, a spectral detection probe disposed outside the vacuum coating chamber and emitting a detection beam to the workpiece to be measured and receiving the reflected beam, and a spectral data processing host electrically connected to the spectral detection probe; the spectral data processing host analyzes the nanoscale coating thickness based on the reflected beam.

[0007] The spectral data processing host is equipped with a rotating phase-locked loop module and a dynamic baseline follow-up calibration module.

[0008] The rotating phase-locked loop module acquires the absolute phase encoding data generated when the workpiece mounting frame rotates in real time, and divides the spectral acquisition time sequence into a baseline extraction time window and a target thickness measurement time window in real time based on the absolute phase encoding data; the baseline extraction time window corresponds to the physical gap area between adjacent workpieces to be measured; the target thickness measurement time window corresponds to the area where the workpiece to be measured faces the spectral detection probe.

[0009] The dynamic baseline follow-up calibration module controls the spectral detection probe to collect the ambient background scattered beam in the vacuum coating chamber and generate a real-time background spectrum within the baseline extraction time window. The real-time background spectrum is then used to replace and update the initial reference spectrum to obtain the target reference spectrum.

[0010] The dynamic baseline follow-up calibration module controls the spectral detection probe to receive the mixed reflection spectrum returned from the surface of the workpiece under test within the target thickness measurement time window, and performs differential alignment processing on the mixed reflection spectrum and the target reference spectrum to extract the target interference spectrum;

[0011] The spectral data processing host analyzes the nanoscale coating thickness based on the target interference spectrum.

[0012] Optionally, the spectral detection probe includes a broadband light source emitting component, an optical fiber transmission component, and a spectral signal conversion component;

[0013] The detection beam emitted by the broadband light source emitting component is transmitted through the optical fiber transmission component to the optical window opened on the side wall of the vacuum coating chamber, and then projected onto the surface of the workpiece to be tested through the optical window.

[0014] The reflected light beam from the surface of the workpiece under test returns via the optical window and is received by the beam splitting signal conversion component.

[0015] Optionally, the spectral data processing host has a pre-stored library of refractive index dispersion models corresponding to different materials;

[0016] When the spectral data processing host analyzes the nanoscale coating thickness based on the target interference spectrum, it includes:

[0017] The extreme points of the peak wavelength and the extreme points of the trough wavelength within the target interference spectrum are located, and the refractive index of the target material is matched in the refractive index dispersion model library to calculate the thickness of the nanoscale coating.

[0018] Optionally, the spectral data processing host is equipped with a digital filtering module;

[0019] Before locating the extreme points of the peak wavelength and the extreme points of the trough wavelength, the digital filtering module uses moving average smoothing logic to filter out the high-frequency noise extremes in the target interference spectrum, thereby obtaining a smooth interference spectrum curve.

[0020] Optionally, the spectral data processing host is bidirectionally connected to the evaporation source controller installed in the vacuum coating chamber;

[0021] The spectral data processing host has a thickness standard lower limit threshold set in it. When the calculated thickness of the nanoscale coating is lower than the thickness standard lower limit threshold, the spectral data processing host sends an acceleration and flow increase command to the evaporation source controller.

[0022] Optionally, a servo drive motor is connected to the bottom drive shaft of the workpiece mounting frame, and a rotary encoder is installed at the tail end of the servo drive motor;

[0023] The rotating phase-locked loop module reads the pulse count value output by the rotary encoder in real time and converts the pulse count value into absolute angular coordinates as the absolute phase encoding data.

[0024] Optionally, the rotating phase lock module has a preset gap angle start and end range and a facing angle start and end range;

[0025] When the absolute angular coordinates fall within the start and end range of the gap angle, the rotating phase lock module determines that it has entered the baseline extraction time window;

[0026] When the absolute angular coordinates fall within the range of the starting and ending angles of the opposing angle, the rotating phase-locked loop module determines that it has entered the target thickness measurement time window.

[0027] Optionally, the dynamic baseline follow-up calibration module is internally configured with a sliding weighted fusion unit;

[0028] The sliding weighted fusion unit acquires multiple historical background spectra generated within each rotation cycle, and performs a weighted average calculation on the real-time background spectrum and the multiple historical background spectra, using the calculated weighted average background spectrum as the target reference spectrum.

[0029] Optionally, the dynamic baseline follow-up calibration module is equipped with a safe upper limit threshold for ambient scattered light intensity;

[0030] After acquiring the real-time background spectrum, the dynamic baseline follow-up calibration module calculates the total global light intensity integral value of the real-time background spectrum.

[0031] When the total global light intensity integral exceeds the safe upper limit threshold of the ambient scattered light intensity, the dynamic baseline follow-up calibration module discards the currently acquired real-time background spectrum and continues to use the historical reference spectrum.

[0032] Secondly, the present invention provides a method for online measurement of nanoscale coating thickness based on spectral analysis, which is executed using the online measurement device for nanoscale coating thickness based on spectral analysis as described in the first aspect, and includes the following control steps:

[0033] Receive absolute phase encoded data generated when the workpiece mounting frame rotates;

[0034] The spectral acquisition time sequence is divided into a baseline extraction time window and a target thickness measurement time window in real time based on the absolute phase encoding data; the baseline extraction time window corresponds to the physical gap area between adjacent workpieces to be measured; the target thickness measurement time window corresponds to the area of ​​the workpiece to be measured facing the spectral detection probe.

[0035] Within the baseline extraction time window, the control spectral probe collects the ambient background scattered beam in the vacuum coating chamber and generates a real-time background spectrum. The real-time background spectrum is then used to replace and update the initial reference spectrum to obtain the target reference spectrum.

[0036] Within the target thickness measurement time window, the spectral detection probe is controlled to receive the mixed reflection spectrum returned from the surface of the workpiece to be measured;

[0037] The mixed reflectance spectrum is differentially aligned with the target reference spectrum to extract the target interference spectrum;

[0038] The nanoscale coating thickness was determined based on the target interference spectrum analysis.

[0039] The present invention has achieved the following beneficial effects:

[0040] This invention divides the spectral acquisition time sequence into a baseline extraction time window and a target thickness measurement time window in real time by acquiring the absolute phase encoded data generated when the workpiece mounting frame rotates. Within the baseline extraction time window corresponding to the physical gap between adjacent workpieces under test, the system acquires the ambient background scattered beam within the vacuum coating chamber to generate a real-time background spectrum, which is then used to replace and update the target reference spectrum. Within the target thickness measurement time window corresponding to the area of ​​the workpiece directly facing the probe, the system receives the mixed reflection spectrum and performs differential alignment with the updated target reference spectrum to extract the target interference spectrum. This design can dynamically adapt to the drift of plasma discharge density and luminescence intensity during long-cycle coating production, eliminates the superposition interference of ambient background radiation on the interference spectrum intensity baseline, avoids extreme point positioning errors, improves the accuracy of online measurement and calculation of nanoscale coating thickness, and ensures the reliability of closed-loop control of the coating deposition process.

[0041] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0042] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0043] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0044] Figure 1 As described in the embodiments of the present invention Figure 1 This is a schematic diagram of the structure of the online measurement device for nanoscale coating thickness based on spectral analysis in an embodiment of the present invention;

[0045] Figure 2 This is a flowchart of the control execution steps of the online measurement method for nanoscale coating thickness based on spectral analysis in an embodiment of the present invention. Detailed Implementation

[0046] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0047] This application provides an online measurement device for nanoscale coating thickness based on spectral analysis, specifically addressing this topic. (Refer to...) Figure 1 As shown, the online measurement device for nanoscale coating thickness based on spectral analysis includes a vacuum coating chamber, a workpiece mounting frame disposed inside the vacuum coating chamber and continuously rotating to support the workpiece to be measured, a spectral detection probe disposed outside the vacuum coating chamber and emitting a detection beam to the workpiece to be measured and receiving the reflected beam, and a spectral data processing host electrically connected to the spectral detection probe. The spectral data processing host analyzes the nanoscale coating thickness based on the reflected beam.

[0048] Understandably, in an industrial coating production environment, the vacuum coating chamber operates under high vacuum and heating conditions, and contains a plasma radiation source for depositing film-forming materials. The workpiece mounting frame employs a planetary rotating frame structure, carrying multiple workpieces to be tested within the vacuum coating chamber, allowing for continuous revolution and rotation, thus ensuring uniform deposition of physical thin film materials on the surface of the workpieces. The spectral data processing host, acting as the processing and calculation control node, is responsible for coordinating the transmission and reception timing of external optical components and analyzing the interference fringe characteristic data contained within the reflected beam.

[0049] Specifically, the vacuum coating chamber is a sealed outer shell structure made of stainless steel, with light-transmitting holes on the side walls. The spectral detection probe, through an optical window mounted at one of the light-transmitting holes, forms an optical transmission channel with the interior space of the vacuum coating chamber, projecting a broadband detection beam onto the surface of the workpiece under test. When the detection beam penetrates the nanoscale thin film on the surface of the workpiece, optical physical reflections occur on the upper and lower surfaces of the film, forming multiple reflected beams with interference phase differences. These multiple reflected beams superimpose to produce optical interference, which is then received and converged by the spectral detection probe into the reflected beam. After receiving the photoelectric conversion value of the reflected beam, the spectral data processing host analyzes the nanoscale coating thickness scalar based on the optical thin film interference principle calculation model. This non-contact optical detection method avoids the mechanical friction loss caused by physical contact devices in vacuum and high-temperature environments, enabling real-time online measurement of the nanoscale thin film thickness.

[0050] Furthermore, the spectral detection probe includes a broadband light source emitting component, an optical fiber transmission component, and a beam splitting signal conversion component. The broadband light source emitting component uses a light-emitting diode array or tungsten halogen LED as the core light source, outputting a continuous broadband signal covering a set wavelength band. The detection beam emitted by the broadband light source emitting component is coupled into the internal channel of the optical fiber transmission component through a collimating lens group. The optical fiber transmission component uses a multi-core quartz fiber bundle and is responsible for guiding the low-loss transmission of optical signals between the separately arranged hardware modules.

[0051] The probe beam emitted by the broadband light source emitting component is transmitted through the optical fiber transmission component to the optical viewing window on the side wall of the vacuum coating chamber, and then projected onto the surface of the workpiece under test through the optical viewing window. The optical viewing window includes a quartz glass substrate and a metal sealing ring flange. The quartz glass substrate is embedded inside the metal sealing ring flange and fixed in the mounting hole on the side wall of the vacuum coating chamber by a fastening nut structure. A fluororubber vacuum sealing gasket is provided between the quartz glass substrate and the metal sealing ring flange to maintain the vacuum pressure state inside the vacuum coating chamber. The emitting end face of the optical fiber transmission component is vertically aligned with the center point of the outer surface of the optical viewing window through a fixed mechanical bracket. After leaving the optical fiber transmission component, the probe beam passes through the optical viewing window and enters the physical space inside the vacuum chamber, and illuminates the surface of the workpiece under test along a preset straight light path as it rotates with the drive shaft.

[0052] The reflected light beam from the surface of the workpiece under test returns via the optical window and is received by the beam splitter signal conversion component. The reflected light path transmission structure and the incident light path transmission structure adopt a coaxial confocal optical physical architecture design. The reflected light beam returning from the thin-film interference on the surface of the workpiece under test penetrates the optical window, is focused by the lens assembly, and is guided into the receiving fiber channel distributed inside the fiber optic transmission component. The fiber optic transmission component guides the reflected light beam to the physical entrance slit at the front end of the beam splitter signal conversion component. The beam splitter signal conversion component is equipped with a diffraction grating array element, which disperses the incident multi-wavelength composite interference beam in space, converting it into a monochromatic spatial spectral band linearly arranged according to wavelength nodes. A linear array of photoelectric image sensors, positioned behind the beam splitter signal conversion component, performs synchronous photoelectric integration and acquisition, converting the photon radiation energy into an analog level signal, and then outputting a digital sequence matrix to the spectral data processing host through an analog-to-digital sampling conversion circuit.

[0053] In this embodiment, the spectral data processing host is equipped with a rotating phase-locked loop module and a dynamic baseline follow-up calibration module. The plasma radiation source generates ambient background radiation within the vacuum coating chamber. The intensity and spectral energy distribution of this ambient background radiation change dynamically with fluctuations in the evaporation source current parameters, vacuum level adjustments, and localized arc discharges. If the spectral probe directly acquires the reflected signal containing the ambient background radiation, the resolved interference spectral intensity baseline will drift, leading to errors in interference extremum positioning and deviations in thickness calculation. To eliminate dynamic background stray light interference, a control algorithm is configured to independently extract and numerically cancel the ambient background radiation.

[0054] Specifically, the rotating phase-locked loop module establishes a low-level communication connection with the servo drive system at the bottom of the workpiece mounting frame to acquire the absolute phase encoding data generated during the rotation of the workpiece mounting frame in real time. A servo drive motor is connected to the bottom drive shaft of the workpiece mounting frame, and a rotary encoder is installed at the tail end of the output shaft of the servo drive motor. The rotating phase-locked loop module reads the pulse count value output by the rotary encoder in real time and converts the pulse count value into absolute angular coordinates as the absolute phase encoding data. The servo drive motor serves as the mechanical power output source, driving the workpiece mounting frame to perform constant-speed circular revolution. The grating engraving disk inside the rotary encoder cuts the infrared light path of the internal photoelectric pair, outputting an orthogonal square wave pulse signal stream containing position node information. The orthogonal square wave pulse signal stream contains a reference zero-position pulse representing the revolution period and incremental pulses representing the subdivision deflection phase.

[0055] The rotating phase-locked loop (PLL) module is internally equipped with a digital counting circuit that captures the rising and falling edges of the square wave pulses transmitted from the rotary encoder and performs accumulation or decrement operations based on the orthogonal phase relationship. When a reference zero-position pulse is received, the digital counting circuit triggers a reset logic to perform a zeroing operation. The PLL module calculates the rotor's deflection angle relative to the zero point by multiplying the accumulated pulse count value by an angle conversion ratio coefficient according to a calibrated formula mapping the total pulse count constant to the mechanical circumferential angle equivalent. The data processor encapsulates the deflection angle value into absolute angle coordinates and pushes it as absolute phase encoded data to the system shared data bus.

[0056] Based on the absolute phase encoding data, the rotating phase-locked loop module divides the spectral acquisition time sequence into a baseline extraction time window and a target thickness measurement time window in real time. The baseline extraction time window corresponds to the physical gap region between adjacent workpieces under test; the target thickness measurement time window corresponds to the region where the workpiece under test faces the spectral probe. It is understood that multiple workpieces under test are arrayed on the outer surface of the workpiece mount, with physical gap regions between adjacent workpieces to avoid physical collisions and to accommodate mechanical fixtures. When the workpiece mount rotates the physical gap region to the front of the spectral probe, the probe beam passes through the physical gap region. At this time, the optical return signal received by the spectral probe mainly consists of plasma glow within the vacuum coating chamber and ambient thermal radiation emitted by the heating source. The dynamic scanning projection of the physical gap region on the time axis, as defined by the timing segmentation command issued by the rotating phase-locked loop module, is defined as the baseline extraction time window. When the solid film surface of the workpiece under test rotates into the effective illumination field of view of the spectral probe, the state machine control logic reverses, and the system defines the current control period as the target thickness measurement time window. By extracting the radiation signal corresponding to the physical gap region, the ambient background scattered beam inside the vacuum coating chamber is obtained.

[0057] The rotating phase-locked loop module has preset gap angle start and end ranges and facing angle start and end ranges. When the absolute angle coordinate falls within the gap angle start and end ranges, the rotating phase-locked loop module determines to enter the baseline extraction time window; when the absolute angle coordinate falls within the facing angle start and end ranges, the rotating phase-locked loop module determines to enter the target thickness measurement time window. Angle threshold parameters defining interval boundaries are pre-written in the system's non-volatile storage area. The gap angle start and end range parameters define the starting and ending boundary scalars of the physical gap area when it enters and leaves the effective projection measurement path of the probe beam. The facing angle start and end range parameters define the effective working start and end angle area of ​​the workpiece intercepting the probe beam. The rotating phase-locked loop module cyclically reads the absolute angle coordinate parameters on the current data bus at a set scan refresh frequency. The logic value comparator verifies the magnitude relationship between the acquired absolute angle coordinates and the pre-stored gap angle start and end range parameters. If the verification shows that the absolute angle coordinate value is greater than the gap start boundary scalar and less than the gap end boundary scalar, the data processor outputs an extraction action flag, instructing the control program to enter the baseline extraction time window. As the motor rotates, the coordinate values ​​are continuously updated. When the logic value comparator determines that the current absolute angle coordinates fall within the preset angle start and end range parameter interval, the data processor cancels the extraction action flag and outputs the thickness measurement action flag, instructing the system to close the baseline extraction channel and activate the target thickness measurement time window.

[0058] Within the baseline extraction time window, the dynamic baseline follow-up calibration module controls the spectral probe to acquire the ambient background scattered light beam within the vacuum coating chamber and generate a real-time background spectrum. This real-time background spectrum is then used to replace and update the initial reference spectrum to obtain the target reference spectrum. While the extraction action flag maintains its trigger level, the photoelectric conversion element inside the spectral probe performs a short-time integration exposure operation. Since the probe's field of view is directly facing the physical gap region at this time, the acquired optical signal reflects the background light emission energy state within the vacuum coating chamber at a specific time point. The spectral probe converts the received light radiation energy into an analog electrical signal, which is then discretized and quantized by the analog-to-digital sampling channel into a two-dimensional digital array containing a wavelength index sequence and corresponding light intensity response values, generating the real-time background spectrum. The dynamic baseline follow-up calibration module retrieves the data transmission channel and stores the real-time background spectrum in a high-speed memory buffer block.

[0059] In long-cycle continuous thin-film deposition processes, plasma discharge density and luminescence intensity exhibit a continuous dynamic drift trend. A fixed calibration reference data array cannot adapt to real-time ambient luminescence drift. After acquiring the real-time background spectrum digital array generated by the latest exposure, the dynamic baseline follow-up calibration module initiates an internal memory overwrite control routine. This routine uses a memory address overwrite instruction to replace the numerical content of the address segment corresponding to the initial reference spectrum pre-stored in the system memory module with the data parameters corresponding to the real-time background spectrum. Through this data overwrite update operation, the target reference spectrum aligned with the system's physical luminescence state is generated.

[0060] The dynamic baseline follow-up calibration module is internally configured with a sliding weighted fusion unit. This unit acquires multiple historical background spectra generated within each rotation cycle and performs a weighted average calculation on the real-time background spectrum and the multiple historical background spectra. The calculated weighted average background spectrum is used as the target reference spectrum. The acquired real-time background spectrum is affected by local plasma scintillation and may contain high-frequency transient noise components. To improve the numerical stability of the reference data array, the sliding weighted fusion unit constructs a fixed-depth first-in-first-out (FIFO) circular queue in system memory. As the mechanical operation cycle progresses, the FIFO circular queue temporarily stores multiple historical background spectrum data frames acquired in adjacent operation cycles. When the latest acquired real-time background spectrum data frame is pushed to the head of the FIFO circular queue, the earliest enqueued historical data frame stored at the end of the queue is popped from memory and its address space is released. The multiplication and accumulation calculation module retrieves the internal weight allocation constant table and matches specific weight coefficient scalars to each spectral matrix in the queue. The closer the time attribute of the spectral matrix is ​​to the current physical moment, the higher the value of the matched weight coefficient scalar. The multiplication and accumulation module multiplies the light intensity values ​​at the same wavelength nodes of each spectral matrix with the corresponding weighting coefficient scalars, and then sums all the multiplication products vertically to synthesize the weighted average background spectrum. Specifically, the mathematical model for synthesizing the weighted average background spectrum is as follows: ;in, The calculated weighted average background spectrum at a specific wavelength node The light intensity response value at that location; The total number of background spectral data frames contained in the first-in-first-out circular queue is a constant. The first item in the queue, ordered chronologically. Frame background spectrum at wavelength node The original value of the light intensity at that location is set. Corresponding to the latest real-time background spectrum currently acquired; To match up to the The weighting coefficient scalar of the frame background spectrum. To ensure that the spectral matrix with a time attribute closer to the current physical moment has a higher weight, this weighting coefficient scalar is set to follow an exponential function distribution: ,in The preset smoothing attenuation constant is greater than zero. Through time-domain weighted smoothing operations, computational disturbances caused by short-term abrupt changes in background radiation are suppressed, and the calculated weighted average background spectrum is overwritten and updated to the target reference spectrum.

[0061] The dynamic baseline follow-up calibration module has a set safe upper limit threshold for ambient scattered light intensity. After acquiring the real-time background spectrum, the module calculates the total global light intensity integral of the real-time background spectrum. When the total global light intensity integral exceeds the safe upper limit threshold for ambient scattered light intensity, the module discards the currently acquired real-time background spectrum and maintains the historical reference spectrum. Under abnormal equipment operating conditions, sudden changes in vacuum can trigger abnormal internal arc discharge, causing an unexpected nonlinear surge in ambient background light radiation intensity, resulting in an uncontrollable and rapid increase in the ambient background light radiation intensity. If the real-time background spectrum containing the characteristics of numerical surge is pushed into the first-in-first-out circular queue, it will cause the target reference spectrum data to overflow and become distorted, leading to the failure of the differential subtraction operation. The operation terminal configuration program inputs an energy scalar constant containing a set tolerance margin as the safe upper limit threshold for ambient scattered light intensity. After generating the digital matrix file of the real-time background spectrum, the anomaly interception verification module performs a pre-logic check on the digital matrix file. The arithmetic unit scans the light intensity response values ​​of all wavelength nodes in the matrix file and performs a summation series accumulation operation, outputting the global light intensity integral value representing the total light radiation energy level. A logical judgment operator verifies whether the global light intensity integral value is greater than the preset safe upper limit threshold for environmental scattered light intensity. If the verification result meets the condition, the interrupt management controller triggers an anomaly interception command. The anomaly interception command clears the buffer memory address temporarily storing the real-time background spectrum, removes the abnormal data matrix from memory, and prevents the data array from entering the sliding weighted fusion calculation execution process. The control system suspends the current baseline replacement flag, instructing the differential logic operator to continue reading the safe-verified historical reference spectrum and performing differential subtraction compensation in the upcoming computation task.

[0062] Within the target thickness measurement time window, the dynamic baseline follow-up calibration module controls the spectral probe to receive the mixed reflection spectrum returned from the surface of the workpiece under test. As the servo drive motor continues to rotate, the workpiece under test rotates from the circumferential side into front of the spectral probe, and the state machine controller enters the target thickness measurement time window. The nanoscale thin film on the surface of the workpiece under test produces interference reflection of the continuously projected probe beam. Simultaneously, the ionized plasma glow background radiation in the vacuum coating chamber converges and aliased within the interference reflection optical path signal. The photoelectric conversion photosensitive array inside the spectral probe is activated by a trigger signal, performing a broadband integration exposure operation. The photoelectric conversion photosensitive array performs full-wavelength channel integration photoelectric sampling on the total incident photon energy after aliasing and converts it into a digital signal array, i.e., the mixed reflection spectrum. The spectral data processing host transfers the mixed reflection spectrum to the internal thickness measurement calculation memory block through a communication transmission channel.

[0063] The dynamic baseline follow-up calibration module performs differential alignment processing on the mixed reflectance spectrum and the target reference spectrum to extract the target interference spectrum. The target reference spectrum is a data matrix sequence updated and acquired within the baseline extraction time window immediately preceding the target thickness measurement time window. The time interval between the two sets of spectral acquisition actions is on the millisecond timescale. The plasma discharge environment exhibits steady-state luminescence characteristics on the millisecond timescale. After extracting the two sets of spectral data matrix sequences from system memory, the dynamic baseline follow-up calibration module calls the index matching logic to execute the wavelength dimension start and end address alignment verification instruction, ensuring that the light intensity response values ​​of the two sets of matrix sequences at the same wavelength horizontal axis nodes are mutually mapped and correspond. After successful verification, the arithmetic subtraction unit executes the wavelength-by-wavelength differential subtraction operation instruction, subtracting the environmental noise light intensity value of the corresponding wavelength node in the target reference spectrum from the light intensity values ​​of each wavelength in the mixed reflectance spectrum. To prevent the thermal noise fluctuation of the photoelectric circuit from causing invalid negative scalars less than zero in the subtraction calculation result, a lower limit clamping protection node is connected in series at the output of the arithmetic subtraction unit. When the subtraction difference scalar value is below zero, the lower limit clamping protection node forces the light intensity calculation result of the corresponding wavelength node to be rewritten as a zero constant. After the differential subtraction operation of all band channel nodes has been traversed and executed, the system will reassemble and splice the generated difference data sequence to generate a data matrix that filters out environmental superimposed interference components, and construct the target interference spectrum stripped of the environmental substrate.

[0064] The spectral data processing host analyzes the nanoscale coating thickness based on the target interference spectrum. The spectral data processing host is equipped with a digital filtering module; before locating the peak and trough wavelength extreme points, the digital filtering module uses moving average smoothing logic to filter out high-frequency noise extremes in the target interference spectrum, obtaining a smooth interference spectrum curve. In the photoelectric signal conversion link, there is a superposition phenomenon between dark current random fluctuations and amplifier thermal noise. This superimposed noise manifests as high-frequency transient interference on the spectral array, causing false zero-crossings in the discrete derivative calculation step, interfering with the extreme value location engine's judgment. The digital filtering module allocates a sliding operation window of a set length in memory space. The sliding operation window has a set wavelength node coverage width parameter. The data processor issues instructions to control the sliding operation window to perform point-by-point translation along the wavelength coordinate axis of the target interference spectrum. At each translation dwell point, the arithmetic accumulation unit extracts the light intensity response values ​​of all wavelength nodes within the coverage area of ​​the sliding operation window and performs summation calculations. The arithmetic division unit divides the summation result by the total number of nodes covered by the window, outputting the local interval arithmetic mean. The system then updates the wavelength node address mapped to the center of the sliding operation window with this local interval arithmetic mean. After traversing the entire wavelength coordinate axis, high-frequency oscillation noise is suppressed by the sliding smoothing algorithm, and the discrete intensity matrix is ​​reconstructed into the smoothed interference spectrum curve that satisfies the derivative smoothing continuity condition.

[0065] The spectral data processing host pre-stores a refractive index dispersion model library corresponding to different materials. When the spectral data processing host analyzes the nanoscale coating thickness based on the target interference spectrum, it includes: locating the extreme wavelength points of the peak and trough wavelengths within the target interference spectrum, and calculating the nanoscale coating thickness by combining the refractive index of the target material matched in the refractive index dispersion model library. The optical refractive index of the film-forming material exhibits nonlinear dispersion physical characteristics as the incident light wavelength changes. The refractive index dispersion model library, in the form of a two-dimensional data table, pre-records the discrete refractive index mapping parameters of conventional coating material targets at various wavelength nodes. The extreme value location logic unit performs differential analytical operations on the aforementioned smoothed interference spectrum curve. The operation unit extracts the light intensity correction parameters of adjacent wavelength nodes, performs parameter subtraction to obtain the difference, divides the difference by the wavelength step constant, and obtains the sequence of real numbers of the first-order derivatives characterizing the physical characteristics of the waveform slope. The logic judgment unit retrieves the positive and negative sign bit logic states of the values ​​of adjacent nodes in the sequence of real numbers of the first-order derivatives. When a symbol transition node is located where the symbol state transitions from positive to negative or from negative to positive, it is determined that a mathematical zero-crossing point exists in the current region. The logic judgment unit maps the mathematical zero-crossing point back to the original wavelength coordinate system, locking in the peak wavelength extreme point parameters corresponding to the top of the spur peak of the spectral envelope curve, and the valley wavelength extreme point parameters corresponding to the bottom of the concave valley of the spectral envelope curve.

[0066] The processing core sends an addressing command to the refractive index dispersion model library based on the deposition material process code entered externally by the operating terminal. The refractive index dispersion model library retrieves the real scalar value of the target material's refractive index that matches the wavelength range of the peak wavelength extrema and the trough wavelength extrema. The processing core constructs a system of simultaneous equations for solving the thin film interference order based on the relationship between the phase difference constants of adjacent extrema points. The data processor inputs the acquired peak wavelength extrema, trough wavelength extrema, and the real scalar value of the target material's refractive index into the system of simultaneous equations for solving the thin film interference order. Specifically, the algebraic expression of the system of simultaneous equations for solving the thin film interference order is: ;in, and These are the adjacent wavelength extrema of the peaks and the troughs, respectively, and their arrangement satisfies the wavelength condition based on physical dispersion. ; and These are the values ​​retrieved from the refractive index dispersion model library at the corresponding wavelengths. and The target material's refractive index is a real scalar value. The thickness of the nanoscale coating to be measured and solved; To detect the optical refraction angle of a light beam inside a nanoscale thin film, Constructs an optical path difference correction operator for non-perpendicular incident light; The integer unknown parameters of the interference order to be analyzed and stripped are defined as follows: The computation module uses a numerical approximation iterative algorithm to analyze and strip the integer unknown parameters of the interference order, and performs multiple rounds of root-finding calculations on the simultaneous equations of the thin film interference order. The computation module derives and outputs a thickness physical constant with convergent numerical characteristics. After numerical rounding, truncation, and formatting, the real-time nanoscale coating thickness is output.

[0067] The spectral data processing host is bidirectionally connected to the evaporation source controller located inside the vacuum coating chamber. The spectral data processing host has a set lower limit threshold for thickness. When the calculated nanoscale coating thickness is lower than this lower limit threshold, the spectral data processing host sends an acceleration command to the evaporation source controller. The evaporation source controller is a hardware adjustment device for heating power located inside the control cabinet of the vacuum coating chamber. It is responsible for providing power to the electron beam generator assembly or the resistance-heated evaporation boat assembly, and adjusting the evaporation and vaporization physical rate of the film-forming material. During long-cycle thin film deposition, physical loss of the material loaded in the crucible can cause a decrease in the actual deposition vaporization rate. The control terminal pre-enters the lower limit threshold parameter corresponding to the current process node in the system configuration setting area. The closed-loop communication comparison circuit continuously extracts the nanoscale coating thickness output by the thickness calculation engine and performs an arithmetic comparison verification with the lower limit threshold parameter. When the nanoscale coating thickness scalar is less than the lower limit threshold parameter scalar, the control system determines that the actual film accumulation physical progress lags behind the predetermined process control trajectory.

[0068] The network communication driver module of the spectral data processing host encapsulates and generates a speed-up and current-increase command data packet. The speed-up and current-increase command data packet contains a set device operation control code and a scalar value for the heating power increment step, and includes a cyclic redundancy check (CRC) verification field at the end of the data frame. The network communication driver module transmits the encapsulated speed-up and current-increase command data packet to the network receiving port of the evaporation source controller via an industrial field serial bus hardware link. After the internal microprocessor of the evaporation source controller decodes and verifies the data packet check code, it drives the internally mounted analog-to-digital converter (ADC) isolation output circuit. The ADC isolation output circuit outputs a control reference analog voltage parameter with increased voltage amplitude to the external high-power power supply module. The control reference analog voltage parameter is commandatically increased to provide the beam current intensity operating parameter to the electron beam deflection electrode, or to increase the constant heating current operating parameter of the resistive heating evaporation boat. The increase in the operating power parameter causes the physical temperature of the molten pool surface of the evaporation source target to rise, increasing the vaporization physical rate of the film-forming material molecules ejected and diffused into the external vacuum space, thus accelerating the accumulation of the film deposition thickness on the surface of the workpiece under test. By using real-time feedback of online dynamic measurement data, combined with the underlying heating power hardware to execute closed-loop communication adjustment and coordination logic, the cumulative deviation of insufficient coating thickness caused by material consumption in long-cycle continuous production conditions is corrected.

[0069] To address abnormal communication conditions caused by electromagnetic interference in industrial environments leading to interruptions in the underlying encoder communication link, the spectral data processing host firmware runtime layer embeds a software phase-locked loop (PLL) follow-up takeover algorithm. During periods of uninterrupted hardware communication network operation, this algorithm continuously records the hardware timer clock timestamp parameters for each acquisition of the zero-position reference synchronization pulse. The calculation module evaluates and outputs the real-time average angular velocity constant scalar of the workpiece mounting frame. When the hardware communication monitoring pin detects an interruption in the acquisition of the absolute phase encoded data, the microprocessor interrupt response service switches to open-loop timing prediction mode. Under open-loop timing prediction mode, the microprocessor utilizes the real-time average angular velocity constant scalar, combined with the hardware timer within the system motherboard, to perform linear spatiotemporal parameter interpolation extrapolation calculations. The extrapolation calculation generates an estimated angle coordinate data stream to supplement and replace the hardware encoded pulses missing due to communication interruption. The control system relies on the estimated angle coordinate data stream to maintain the generation of segmentation extraction instructions and the timing segmentation operation scheduling logic between the baseline extraction time window and the target thickness measurement time window. Until the physical encoder communication cable link is restored to the handshake connection configuration, the online thickness measurement function is guaranteed to maintain fault-tolerant and safe processing capabilities in the communication disturbance space.

[0070] Based on the aforementioned device hardware topology and functional module control logic configuration, this application provides an online measurement method for nanoscale coating thickness based on spectral analysis, executed using the aforementioned online measurement device for nanoscale coating thickness based on spectral analysis. The measurement method encompasses a measurement and control digital execution flow chain, from underlying pulse level acquisition, timing logic segmentation and mapping, optical matrix differential smoothing and cleaning to iterative solution of the interferometric extremum algorithm equation. (Refer to...) Figure 2 As shown, the specific control execution steps are as follows:

[0071] Step S1: Receive the absolute phase encoded data generated when the workpiece mounting frame rotates.

[0072] The hardware input capture circuit of the spectral data processing host monitors in real time the dual-channel physically orthogonal pulse signals emitted by the rotary encoder connected to the servo drive shaft. The signal conditioning front-end trigger circuit performs hardware low-pass filtering, shaping, and anti-jitter delay processing on the received orthogonal pulse signals. The digital counter performs channel phase detection on the edge level flip state of the shaped pulses and performs hardware step-by-step accumulation calculation. When the reference zero-position alignment pulse flag is detected on the pin, the digital counter triggers the internal clearing circuit to reset and clear the constant value of the count accumulation register. The acquired real-time pulse count value is transmitted to the scaling multiplier. The scaling multiplier multiplies the real-time pulse count value by a pre-entered angle equivalent mapping conversion coefficient to calculate the absolute angular coordinate scalar of the workpiece mounting frame relative to the mechanical home zero point. The communication coprocessor packages the extracted absolute angular coordinate scalar into the communication protocol payload byte segment and encapsulates it to generate the absolute phase encoded data. The communication coprocessor pushes the absolute phase encoded data into the shared memory data bus of the system motherboard.

[0073] Step S2: Based on the absolute phase encoding data, the spectral acquisition time sequence is divided into a baseline extraction time window and a target thickness measurement time window in real time; the baseline extraction time window corresponds to the physical gap area between adjacent workpieces to be measured; the target thickness measurement time window corresponds to the area of ​​the workpiece to be measured facing the spectral detection probe.

[0074] The main control scheduling management program periodically reads the absolute phase encoded data refreshed in the shared memory data bus. The interval boundary value comparison logic module extracts the angle coordinate values ​​carried in the data packet and performs interval verification boundary calculations with the gap start and end boundary parameter arrays and the facing start and end boundary parameter arrays stored in the system configuration flash memory area. When the angle coordinate value is detected to meet the logical boundary condition of being greater than the gap start parameter and less than the gap end parameter, the system state machine module switches to the baseline extraction measurement and control operation state. The system state machine module outputs a time window action valid flag level to the external hardware, starting the baseline extraction time window sequence. When the angle coordinate value continues to increase with the rotation of the rotating machinery, meeting the logical boundary condition of being greater than the facing start parameter and less than the facing end parameter, the system state machine module switches to the thickness measurement sampling measurement and control operation state. The system state machine module cancels the baseline extraction action flag level and synchronously outputs the thickness measurement action flag level, starting the target thickness measurement time window sequence.

[0075] Step S3: Within the baseline extraction time window, the spectral probe is controlled to collect the ambient background scattered beam in the vacuum coating chamber and generate a real-time background spectrum. The real-time background spectrum is then used to replace and update the initial reference spectrum to obtain the target reference spectrum.

[0076] During the period when the baseline extraction action flag level remains high at the trigger threshold, the exposure trigger master controller sends a hardware trigger drive pulse sequence to the spectral probe. The photoelectric conversion sensing target inside the spectral probe initiates a photon collection physical process with a set integration time parameter. The light radiation energy source entering the probe's optical window is primarily plasma-mixed background stray light. The analog-to-digital conversion isolation card quantizes the analog voltage array output from the photoelectric conversion sensing target into a digital signal matrix data structure with a specific sampling resolution, thus forming the real-time background spectrum. The microprocessor performs a full-wavelength channel light intensity numerical summation arithmetic operation on the real-time background spectrum digital signal matrix to calculate the global light intensity integral total scalar value for a specific capture period. After the numerical comparison module verifies that the global light intensity integral total scalar value does not exceed the preset safe upper limit threshold for ambient scattered light intensity, the memory direct access control channel moves the real-time background spectrum to a dynamic buffer circular stack queue. The weighted summation filter module reads multiple frames of historical background spectrum data accumulated in the buffer circular stack queue, as well as the latest real-time background spectrum data. The weighted summation filter module applies a decreasing exponential weight distribution sequence, performing node-wise multiplication and summation operations on each array in the queue, and outputting a smoothed reconstructed weighted average background spectrum. The storage management allocation controller directly overwrites and stores the smoothed reconstructed weighted average background spectrum in a dedicated reference calibration memory sector. The storage overwrite operation completes the erasure and parameter overwriting of the initial reference spectrum, and the system permanently generates and outputs the target reference spectrum that needs to be subtracted for the current cycle's difference operation.

[0077] Step S4: Within the target thickness measurement time window, control the spectral detection probe to receive the mixed reflection spectrum returned by the surface of the workpiece to be measured.

[0078] During the period when the thickness measurement action flag level remains high at the trigger threshold, the exposure trigger main controller sends parallel hardware trigger drive pulses to the spectral probe. The projected broadband probe beam undergoes optical interference and reflection at the nanoscale thin film interface on the surface of the workpiece. The returned interference and reflection beam physically aliased and superimposed with the plasma background beam distributed throughout the vacuum chamber. The spectral probe performs grating dispersion, photoelectric photosensitive integration conversion, and digital-to-digital conversion quantization encoding processing steps on the aliased and superimposed composite beam. After the processing steps are completed, a digital matrix array with a row-column wavelength intensity mapping network structure is generated, which is the mixed reflection spectrum. The memory allocation system pushes the mixed reflection spectrum into the thickness measurement operation task buffer address memory segment.

[0079] Step S5: Perform differential alignment processing on the mixed reflection spectrum and the target reference spectrum to extract the target interference spectrum.

[0080] The differential calculation engine controller synchronously locks the address pointers of the physical addresses storing the mixed reflectance spectrum and the target reference spectrum. The arithmetic logic control unit utilizes a cyclically incrementing addressing hardware mechanism to sequentially extract the aliased light intensity response value and the noise floor reference value of node addresses with the same wavelength mapping offset. The subtractor hardware module within the arithmetic logic control unit executes real-number subtraction logic instructions to obtain the wavelength node difference constant. A boundary constraint lower limit checker verifies whether the difference constant falls below the preset digital zero lower limit reference. If the verification result meets the fall condition, the lower limit clamping protection logic circuit forcibly overwrites and clears the memory area of ​​the current node's output difference constant to zero. If the verification result does not meet the fall condition, the system retains the original output difference constant. After the logic traversal operation covers all wavelength mapping node addresses, the combined output data array, filtered to remove background light intensity bias interference, is reconstructed to reconstruct the target interference spectrum.

[0081] Step S6: Analyze the nanoscale coating thickness based on the target interference spectrum.

[0082] The microprocessor's central control core retrieves the digital smoothing filter operator program and sets a sliding operation window with a specific span width. The sliding operation window performs a moving average denoising and smoothing algorithm on the target interferometric spectral data array, outputting a smoothed interferometric spectral curve array model free of high-frequency glitches. The extreme value coordinate positioning operator performs initial-order difference quotient differentiation on the smoothed interferometric spectral curve array model, generating a derivative slope mapping array. The logic symbol processor scans the positive and negative polarity sign transition nodes contained within the derivative slope mapping array. The logic symbol processor extracts the peak wavelength extreme point coordinate parameters and trough wavelength extreme point coordinate parameters corresponding to the extreme physical positions of the spectral interferometric curves. The interferometric solution processor receives the pre-set process film material identifier code and performs a search command matching in the refractive index dispersion model library lookup table stored in non-volatile flash memory. It retrieves the real physical scalar of the target material's refractive index for the corresponding measurement band. The interferometric solution processor constructs an algebraic model of the interferometric extreme value order equations, including the extreme wavelength parameter constant and the real physical scalar of the refractive index. The microprocessor central control core uses a numerical approximation algorithm to solve for the order-order integer unknowns in the algebraic model of the equation. It derives and calculates the nanoscale physical thickness numerical parameters that meet the set mathematical tolerance convergence accuracy, completes data recording, and outputs the nanoscale coating thickness constant to the peripheral bus. The parallel-operation feedback closed-loop control task executes logic to compare the nanoscale coating thickness constant with a set lower limit threshold for thickness. If the comparison result shows that the thickness constant is lower than the lower limit threshold, the control task encapsulates a speed-up and current-increasing network instruction data packet containing heating physical power increment parameters. The control task sends the speed-up and current-increasing network instruction data packet to the evaporation source controller network interaction node via the communication bus, intervening to adjust the underlying evaporation vaporization current amplitude output parameters.

[0083] In specific application scenarios, the online measurement device for nanoscale coating thickness based on spectral analysis can be configured as a multi-channel synchronous detection topology. When the workpiece mounting rack has multiple independent workpiece carrying rings distributed along the vertical axis, the sidewall of the vacuum coating chamber has multiple sets of optical viewing windows arranged in an array along the vertical axis. Each set of optical viewing windows has an independent spectral probe mounted on its outer side. These multiple spectral probes emit detection beams towards the corresponding height level of the workpiece carrying rings and receive reflected beams. The mainboard of the spectral data processing host integrates a multi-channel high-speed parallel data acquisition interface to simultaneously receive the mixed reflectance spectral data streams output by the multiple spectral probes.

[0084] The data processor utilizes a multi-threaded parallel computing architecture, allocating independent memory address spaces, independent sliding weighted fusion unit state machines, and independent differential calculation engines to each physical detection channel. Each detection channel shares absolute phase-coded data output from a specific rotating phase-locked loop module to ensure hardware synchronization of the timing segmentation logic between the baseline extraction time window and the target thickness measurement time window for each channel. Within a single mechanical revolution cycle, the spectral data processing host can parallelly analyze and output multiple nanoscale coating thickness constants covering different physical levels at different vertical heights of the workpiece mount. Multi-channel thickness data is pushed to the host computer manufacturing execution system via the industrial Ethernet protocol. The host computer manufacturing execution system compares the numerical deviations of the multi-channel thickness data, quantitatively evaluates the uniformity distribution parameters of the physical jet flow field of the evaporation source in vertical space, and generates a flow field uniformity quality analysis report.

[0085] To address the physical problem of optical transmission attenuation caused by thermal aging of fiber optic transmission components under high-temperature environments, the system is equipped with optical path attenuation compensation logic. A standard broadband reflective reference plate, independent of the workpiece mounting frame, is installed inside the vacuum coating chamber. The surface reflectivity parameters of the standard broadband reflective reference plate are fixed. During the no-load operation period before production begins, the spectral data processing host controls the workpiece mounting frame to stop rotating and adjusts the illumination angle of the spectral probe so that the probe beam is directly projected onto the surface of the standard broadband reflective reference plate. The spectral probe collects the reference reflection spectrum, which is stored in the memory of the spectral data processing host as the initial lossless optical path reference curve. At subsequent specific equipment maintenance times, the spectral data processing host again executes the no-load alignment sampling process to acquire the reflection spectrum data of the standard broadband reflective reference plate at the current physical moment. The data processor calls the division operation logic to divide the original light intensity values ​​of each wavelength node in the reflection spectrum data acquired at the current physical moment by the light intensity values ​​of the same wavelength nodes in the initial lossless optical path reference curve, generating an optical path intensity attenuation ratio coefficient matrix. After acquiring the mixed reflection spectrum in the subsequent thickness measurement step, and before performing differential alignment processing, the data processor performs an inverse gain amplification operation on the mixed reflection spectrum. The measured light intensity values ​​for each wavelength node are extracted from the mixed reflection spectrum data sequence and divided by the proportionality constant of the same wavelength nodes in the optical path light intensity attenuation proportionality coefficient matrix. The quotient is then used to overwrite the original measured light intensity values. Through this arithmetic calculation, optical transmission loss caused by fiber aging is compensated, ensuring that the absolute amplitude distribution ratio of the target interference spectrum entering the interferometric extreme value positioning module remains constant, thus maintaining the stability of the thickness equation solution values.

[0086] In the data structure implementation of the spectral data processing host, the real-time background spectrum, mixed reflectance spectrum, and target reference spectrum are all defined as one-dimensional vector arrays containing double-precision floating-point data. The data processor accesses array elements in a contiguous memory address space by using pointer base addresses plus byte offsets. During the differential alignment process in step S5, the data processor calls the vector operation instruction set to divide the one-dimensional vector array into multiple data blocks according to the cache line length. Within a single processor clock cycle, the light intensity values ​​of multiple adjacent wavelength nodes are read in parallel, and subtraction arithmetic instructions are executed in batches. Through underlying vector operation hardware acceleration, the iteration calculation latency of large-scale spectral node data is reduced, and the computational throughput of differential alignment processing is improved.

[0087] For scenarios involving spectral thickness calculation of thicker films or complex multilayer film systems, the system can optionally be configured with a thickness extraction engine based on frequency domain transformation. For thin film structures with thicknesses exceeding the conventional interference period, the target interference spectrum exhibits a dense high-frequency oscillating envelope. In frequency domain transformation calculation mode, the data processor performs a nonlinear transformation of the coordinate axes of the target interference spectrum, converting the equally spaced wavelength abscissas into reciprocal wavenumber coordinates. The light intensity sequence in the wavenumber coordinates is resampled using a spline interpolation function to generate an interference sequence with equal wavenumber spacing. The data processor applies a Fast Fourier Transform algorithm to the equally spaced wavenumber interference sequence, mapping the interference signal from the spectral wavenumber domain to the spatial frequency domain, generating a spectral amplitude function matrix characterizing the optical thickness distribution. The data processor performs array traversal sorting and searching to locate the frequency domain coordinate points corresponding to the energy peaks in the spectral amplitude function matrix. Combining the pre-stored target material refractive index scalar parameters and the incident angle trigonometric function factor, the nanometer-scale coating thickness is directly calculated using the spatial frequency conversion formula. By switching to the frequency domain transformation extraction path, the local pseudo-stationary point interference phenomenon that is prone to occur when locating the extreme points of dense interference in the wavelength domain is avoided. Specifically, the specific mathematical expression of the spatial frequency conversion formula is as follows: ;in, The thickness of the nanoscale coating is a value obtained through direct calculation; To determine the frequency domain coordinates corresponding to the energy peaks in the locked spectral amplitude function matrix, since the Fast Fourier Transform directly inverts and maps the equally spaced wavenumber coordinates (in the form of the wavelength reciprocal) to the spatial distance domain, this... The physical dimensions of the parameter are converted into the length dimension; To obtain the pre-stored average refractive index scalar parameter of the target material in the measurement band; The angle of refraction of the probe beam as it penetrates the thin film under test. That is, the trigonometric function factor of the incident angle.

[0088] To prevent analog-to-digital converter (ADC) numerical saturation overflow caused by the spectral probe on highly reflective workpiece surfaces, the system is equipped with an adaptive integral time adjustment module. During continuous acquisition, the adaptive integral time adjustment module extracts the mixed reflectance spectrum acquired in the previous running cycle and retrieves the maximum light intensity amplitude response constant from the spectral array. This maximum light intensity amplitude response constant is compared with the full-scale saturation threshold and the low signal-to-noise ratio (SNR) lower limit threshold calibrated by the photoelectric conversion circuit. When the maximum light intensity amplitude response constant exceeds a specific percentage of the full-scale saturation threshold, the adaptive integral time adjustment module calculates the amplitude attenuation ratio and generates a reduction configuration command, which is sent to the spectral probe to simultaneously shorten the exposure integral time parameter of the photoelectric sensor array. When the maximum light intensity amplitude response constant is below the low SNR lower limit threshold, the adaptive integral time adjustment module generates an amplification configuration command, extending the exposure integral time parameter according to the calculated ratio. After adjusting the integration time parameter, the spectral data processing host synchronously adjusts the light intensity normalization coefficient in the memory. Before performing the differential calculation between the mixed reflectance spectrum and the target reference spectrum, the two sets of arrays are pre-normalized and calibrated at the numerical level according to the proportional difference of the integration time to ensure that the operands of the differential subtraction operation have the same energy response dimensions.

[0089] In industrial IoT integrated environments, the spectral data processing host is equipped with a network routing module that interfaces with the upper-level manufacturing execution system. This network routing module embeds a message queue telemetry transmission communication protocol stack and a structured query language database connection client. After a single nanometer-level coating thickness calculation is completed and a closed-loop intervention is executed, the data processor extracts the absolute angular coordinates, thickness value, total integral value of the ambient reference light intensity, and hardware error code status words for the current operation cycle, assembling them into a data record in Extensible Markup Language (XML) format. The network routing module pushes this data record as a message payload to the message broker node configured on the plant server, or executes a database insert statement to serialize the record and store it in a relational database table. The data routing action and the underlying differential calculation belong to different independent processes, and data is transferred through an inter-process communication mailbox mechanism to ensure the archiving of production traceability data.

[0090] It is understandable that the matrix operations, logic state control, and closed-loop instruction encapsulation steps in the aforementioned online measurement method for nanoscale coating thickness based on spectral analysis are all implemented using computing device hardware. The computing device includes a processor module and a storage medium module. The storage medium module stores the computer's executable program code instructions. The processor module reads and processes these program code instructions, schedules various peripheral interface circuits, and executes the aforementioned steps of spectral extraction, differential cleaning, extreme value calculation, and feedback processing. The hardware architecture provides computational processing and data routing support through a standard bus.

[0091] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An online measurement device for nanoscale coating thickness based on spectral analysis, comprising a vacuum coating chamber, a workpiece mounting frame disposed inside the vacuum coating chamber and supporting the workpiece to be measured for continuous rotation, a spectral detection probe disposed outside the vacuum coating chamber and emitting a detection beam to the workpiece to be measured and receiving the reflected beam, and a spectral data processing host electrically connected to the spectral detection probe; the spectral data processing host analyzes the nanoscale coating thickness based on the reflected beam; Its features are, The spectral data processing host is equipped with a rotating phase-locked loop module and a dynamic baseline follow-up calibration module. The rotating phase-locked loop module acquires the absolute phase encoding data generated when the workpiece mounting frame rotates in real time, and divides the spectral acquisition time sequence into a baseline extraction time window and a target thickness measurement time window in real time based on the absolute phase encoding data; the baseline extraction time window corresponds to the physical gap area between adjacent workpieces to be measured; the target thickness measurement time window corresponds to the area where the workpiece to be measured faces the spectral detection probe. The dynamic baseline follow-up calibration module controls the spectral detection probe to collect the ambient background scattered beam in the vacuum coating chamber and generate a real-time background spectrum within the baseline extraction time window. The real-time background spectrum is then used to replace and update the initial reference spectrum to obtain the target reference spectrum. The dynamic baseline follow-up calibration module controls the spectral detection probe to receive the mixed reflection spectrum returned from the surface of the workpiece under test within the target thickness measurement time window, and performs differential alignment processing on the mixed reflection spectrum and the target reference spectrum to extract the target interference spectrum; The spectral data processing host analyzes the nanoscale coating thickness based on the target interference spectrum.

2. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 1, characterized in that, The spectral detection probe includes a broadband light source emitting component, an optical fiber transmission component, and a spectral signal conversion component; The detection beam emitted by the broadband light source emitting component is transmitted through the optical fiber transmission component to the optical window opened on the side wall of the vacuum coating chamber, and then projected onto the surface of the workpiece to be tested through the optical window. The reflected light beam from the surface of the workpiece under test returns via the optical window and is received by the beam splitting signal conversion component.

3. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 2, characterized in that, The spectral data processing host has a pre-stored library of refractive index dispersion models for different materials. When the spectral data processing host analyzes the nanoscale coating thickness based on the target interference spectrum, it includes: The extreme points of the peak wavelength and the extreme points of the trough wavelength within the target interference spectrum are located, and the refractive index of the target material is matched in the refractive index dispersion model library to calculate the thickness of the nanoscale coating.

4. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 3, characterized in that, The spectral data processing host is equipped with a digital filtering module; Before locating the extreme points of the peak wavelength and the extreme points of the trough wavelength, the digital filtering module uses moving average smoothing logic to filter out the high-frequency noise extremes in the target interference spectrum, thereby obtaining a smooth interference spectrum curve.

5. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 4, characterized in that, The spectral data processing host is bidirectionally connected to the evaporation source controller installed in the vacuum coating chamber. The spectral data processing host has a thickness standard lower limit threshold set in it. When the calculated thickness of the nanoscale coating is lower than the thickness standard lower limit threshold, the spectral data processing host sends an acceleration and flow increase command to the evaporation source controller.

6. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 1, characterized in that, A servo drive motor is connected to the bottom drive shaft of the workpiece mounting frame, and a rotary encoder is installed at the tail end of the servo drive motor. The rotating phase-locked loop module reads the pulse count value output by the rotary encoder in real time and converts the pulse count value into absolute angular coordinates as the absolute phase encoding data.

7. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 6, characterized in that, The rotating phase lock module has a preset gap angle start and end range and a direct angle start and end range. When the absolute angular coordinates fall within the start and end range of the gap angle, the rotating phase lock module determines that it has entered the baseline extraction time window; When the absolute angular coordinates fall within the range of the starting and ending angles of the opposing angle, the rotating phase-locked loop module determines that it has entered the target thickness measurement time window.

8. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 7, characterized in that, The dynamic baseline follow-up calibration module is internally configured with a sliding weighted fusion unit; The sliding weighted fusion unit acquires multiple historical background spectra generated within each rotation cycle, and performs a weighted average calculation on the real-time background spectrum and the multiple historical background spectra, using the calculated weighted average background spectrum as the target reference spectrum.

9. The online measurement device for nanoscale coating thickness based on spectral analysis according to claim 8, characterized in that, The dynamic baseline follow-up calibration module is set with a safe upper limit threshold for ambient scattered light intensity. After acquiring the real-time background spectrum, the dynamic baseline follow-up calibration module calculates the total global light intensity integral value of the real-time background spectrum. When the total global light intensity integral exceeds the safe upper limit threshold of the ambient scattered light intensity, the dynamic baseline follow-up calibration module discards the currently acquired real-time background spectrum and continues to use the historical reference spectrum.

10. A method for online measurement of nanoscale coating thickness based on spectral analysis, performed using the online measurement equipment for nanoscale coating thickness based on spectral analysis as described in any one of claims 1 to 9, characterized in that, Includes the following control steps: Receive absolute phase encoded data generated when the workpiece mounting frame rotates; The spectral acquisition time sequence is divided into a baseline extraction time window and a target thickness measurement time window in real time based on the absolute phase encoding data; the baseline extraction time window corresponds to the physical gap area between adjacent workpieces to be measured; the target thickness measurement time window corresponds to the area of ​​the workpiece to be measured facing the spectral detection probe. Within the baseline extraction time window, the control spectral probe collects the ambient background scattered beam in the vacuum coating chamber and generates a real-time background spectrum. The real-time background spectrum is then used to replace and update the initial reference spectrum to obtain the target reference spectrum. Within the target thickness measurement time window, the spectral detection probe is controlled to receive the mixed reflection spectrum returned from the surface of the workpiece to be measured; The mixed reflectance spectrum is differentially aligned with the target reference spectrum to extract the target interference spectrum; The nanoscale coating thickness was determined based on the target interference spectrum analysis.

Images