Dynamic anti-interference method and system for high-definition display mainboard

Abstract

The invention relates to the technical field of display mainboards, and particularly discloses a dynamic anti-interference method and system for a high-definition display mainboard. An original noise baseline is established in a standby state, test excitation is injected to identify an active response frequency band, a link disturbance weighting index is dynamically generated in image transmission, an interference main source is positioned, and frequency band weight distribution containing time stability is constructed, so that the system can accurately determine the frequency band weight distribution according to the actual physical performance of each frequency band instead of a preset template. Continuously judging which frequency bands are influencing the link integrity; and realizing real-time identification and distinguishing of key interference frequency bands by combining frequency band division filtering, asynchronous re-estimation period setting, coupling strength detection and master-slave cooperative compensation. According to the process, the anti-interference action is always anchored on the current most active, most unstable and most link-harmful frequency band, so that the physical layer stability of image transmission is maintained in the noise frequency point drifting process, and the image anomaly phenomenon caused by frequency point mismatch is effectively relieved.

Description

Technical Field

[0001] This invention relates to the field of display motherboard technology, and more specifically to a dynamic anti-interference method and system for high-definition display motherboards. Background Technology

[0002] Current high-definition display motherboards, when supporting HDMI 2.1 or DP 2.1 ultra-high-definition image transmission, generally face the practical constraint of multiple sources of high-frequency noise coexisting. Typical motherboards integrate multiple functional units such as CPU, GPU, high-speed memory, PCIe expansion devices, and USB controllers. Their operating clock harmonics, switching power supply ripple, and data link reflection signals form a densely overlapping noise spectrum in the 200MHz to 3GHz frequency band. Existing technologies mainly employ static filtering schemes: placing LC notch filters with a fixed center frequency near the HDMI or DP interface, or relying on large-area ground planes and local decoupling capacitor arrays in the PCB stack-up design for broadband suppression. These solutions can meet basic eye diagram requirements during the board-level verification stage, but once the entire system enters the real operating environment, affected by temperature changes, load fluctuations, and component aging, the interference frequency shifts, and the original filtering structure becomes detuned, leading to increased common-mode noise coupling, manifesting as image edge flickering, color banding, or occasional black screens.

[0003] This raises a clear technical problem: under conditions where multiple high-frequency noise sources coexist on the motherboard and the frequency points drift with operating conditions, it is impossible to identify and distinguish the key interference frequency bands that pose a substantial threat to HDMI / DP2.1 ultra-high-definition image transmission in real time, causing anti-interference measures to lag behind the interference evolution process. Summary of the Invention

[0004] The present invention aims to provide a dynamic anti-interference method and system for high-definition display motherboards, which can effectively alleviate image abnormalities caused by frequency mismatch.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a dynamic anti-interference method for a high-definition display motherboard, comprising:

[0006] In the standby state of the video output controller, the original noise signal is synchronously acquired through multiple analog voltage probes distributed on the motherboard, and the original noise baseline is established based on the original noise signal;

[0007] Based on the original noise baseline, a test excitation signal is injected into the HDMI or DP physical layer, the response signal is collected, a differential spectrum is generated by comparing the response signal with the original noise baseline, and the active response frequency band is identified from the differential spectrum.

[0008] Activate image data stream transmission, collect noise signals during image data stream transmission, calculate link disturbance spectrum, and generate link disturbance weighted index by combining protocol layer link status indicators. The link disturbance spectrum is the difference between the current noise signal and the original noise baseline.

[0009] Based on the contribution of each frequency band in the link disturbance weighted index, the dominant source of interference is located, and a frequency band weight distribution including frequency, spatial location and temporal stability is generated.

[0010] Based on the frequency band weight distribution, multiple tunable notch networks are configured for frequency band filtering, and the reestimation period of each notch network is set according to the time stability of each frequency band.

[0011] By perturbing each notch network and observing the response changes in other frequency bands, a coupling strength matrix between frequency bands is constructed to identify the master-slave relationship. Based on the master-slave relationship, the master frequency band is adjusted, and the slave frequency band is compensated for following.

[0012] Preferably, establishing the original noise baseline includes:

[0013] The variance of each signal in the original noise signal is calculated to form a variance vector;

[0014] Power spectrum curves are generated by estimating the power spectral density of each original noise signal based on the variance vector.

[0015] Identify local peak intervals in the power spectrum curve where the energy is continuously higher than the mean by a preset multiple;

[0016] The local peak intervals are marked as candidate frequency bands for the original noise baseline, with the center frequency and bandwidth determined.

[0017] Preferably, the identified active response frequency band includes:

[0018] Frequency points in the differential spectrum where the energy increment exceeds a preset threshold are marked;

[0019] Determine whether the marked frequency point appears synchronously in the monitoring signals of multiple adjacent analog voltage probes;

[0020] Cluster the synchronously occurring marked frequency points into active response frequency bands, and record the center frequency and bandwidth of the active response frequency bands.

[0021] Preferably, the generation of the link disturbance weighted index includes:

[0022] Read the receiver eye diagram height attenuation rate, clock jitter accumulation value, and symbol error count rate from the protocol layer;

[0023] Spatiotemporal alignment of the energy integral value of a specified frequency band in the link disturbance spectrum with the protocol layer link state index;

[0024] Based on the alignment results, weighting coefficients are assigned to the frequency bands corresponding to different monitoring points, and the weighted frequency band energy integral values ​​are summed to generate the link disturbance weighted index.

[0025] Preferably, generating the frequency band weight distribution includes:

[0026] Based on the motherboard hardware resource mapping table, the frequency bands that contribute significantly to the link disturbance weighted index are matched to the physical interference sources;

[0027] The time stability was calculated by counting the number of times each frequency band appeared in ten consecutive sampling windows.

[0028] Construct a three-dimensional weight distribution, with the horizontal axis representing frequency, the vertical axis representing spatial location, and the vertical axis representing temporal stability;

[0029] The frequency band with the highest value in the weight distribution is used as the criterion for determining the correlation between spatial location and temporal stability.

[0030] Preferably, the configuration of multiple tunable notch networks for frequency band filtering includes:

[0031] Based on the frequency band with the highest weight in the frequency band weight distribution, a tuning command is sent to the corresponding tunable notch network.

[0032] The center frequency of the tunable notch filter network is adjusted using a three-order ramping method: coarse adjustment, medium adjustment, and fine adjustment.

[0033] After each adjustment, wait for the circuit to reach a steady state and collect the output voltage fluctuation variance. If the variance decreases by more than the preset proportion of the previous value, the adjustment is confirmed to be effective and the next adjustment is continued.

[0034] Preferably, the setting of the reestimation period for each notch network includes:

[0035] Based on the basic revaluation interval, the revaluation period of each notch network is calculated by linear mapping according to the time stability of each frequency band.

[0036] The residual fluctuation variance sequence at the output of the notch network is continuously monitored during the revaluation period.

[0037] If the residual variance sequence shows a continuous sudden increase, the reassessment process is triggered in advance. During the reassessment, the image data stream is paused, a quiet period is inserted, and baseline data is reacquired to update the center frequency.

[0038] Preferably, the construction of the inter-band coupling strength matrix includes:

[0039] A small triangular wave modulation was applied sequentially to the center frequency of each tunable notch network, and the variance change rate of the corresponding monitoring points in other frequency bands was recorded simultaneously during the modulation period.

[0040] Fill the normalized slope of the variance change rate relative to the modulation amplitude into the corresponding position in the matrix;

[0041] The unidirectional influence relationship between the main frequency band and the slave frequency band is identified based on the distribution of non-zero elements in the matrix.

[0042] Preferably, the following compensation for the slave frequency band includes:

[0043] When the center frequency of the tunable notch filter network corresponding to the main frequency band is updated, the compensation amount of the slave frequency band is calculated based on the row elements of the main frequency band in the inter-band coupling strength matrix.

[0044] The center frequency of the tunable notch network corresponding to the frequency band is adjusted synchronously according to the compensation amount.

[0045] Ensure that the secondary frequency band compensation is completed within 15 microseconds after the main frequency band adjustment is completed, and verify whether the fluctuation range of the link disturbance weighted index has narrowed after compensation.

[0046] On the other hand, the present invention proposes a dynamic anti-interference system for a high-definition display motherboard, comprising:

[0047] Multiple analog voltage probes are distributed on the motherboard to collect noise signals;

[0048] The synchronous sampling unit connects to multiple analog voltage probes to synchronously acquire the original noise signal, response signal, and noise signal during the transmission of image data stream;

[0049] The processing unit, connected to the synchronous sampling unit, is used to establish the original noise baseline, generate the differential spectrum and link perturbation spectrum, identify active response frequency bands, generate the link perturbation weighting index and frequency band weight distribution, and construct the inter-band coupling strength matrix.

[0050] Multiple tunable notch filters are connected to the processing unit to perform frequency band filtering based on the frequency band weight distribution, and to perform main frequency band adjustment and slave frequency band following compensation according to the instructions of the processing unit.

[0051] The processing unit works in collaboration with multiple tunable notch networks to achieve a dynamic anti-interference process that sets the revaluation period and master-slave relationship based on time stability.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] This invention establishes an initial noise baseline in standby mode, injects test stimuli to identify active response frequency bands, dynamically generates a link disturbance weighting index during image transmission, and then locates the dominant interference source and constructs a time-stable frequency band weight distribution. This enables the system to continuously determine which frequency bands are affecting link integrity based on the actual physical performance of each frequency band rather than a preset template. Combined with frequency band filtering, asynchronous reassessment period setting, coupling strength detection, and master-slave collaborative compensation, it achieves real-time identification and differentiation of key interference frequency bands. This process ensures that anti-interference actions are always anchored to the most active, unstable, and link-hazardous frequency bands, thereby maintaining the physical layer stability of image transmission during noise frequency drift and effectively mitigating image anomalies caused by frequency mismatch. Attached Figure Description

[0054] Figure 1 This is a flowchart of the dynamic anti-interference method for a high-definition display motherboard according to the present invention;

[0055] Figure 2 This is a block diagram of the dynamic anti-interference system for a high-definition display motherboard according to the present invention. Detailed Implementation

[0056] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0057] like Figure 1 As shown, this invention proposes a dynamic anti-interference method for high-definition display motherboards. The core of this method lies in the real-time identification and differentiation of key interference frequency bands affecting the transmission quality of HDMI / DP2.1 ultra-high-definition images, addressing the actual operating conditions of multiple high-frequency noise sources coexisting within the motherboard. The entire operation process does not rely on pre-set spectrum templates or external calibration signals. Instead, it initiates an autonomous sensing path immediately after system power-on, establishing a response mapping relationship between noise sources and the image link step by step, and continuously updating the spatiotemporal weight distribution of this mapping as the system runs. Specifically, it includes the following steps:

[0058] In the standby state of the video output controller, the original noise signal is synchronously acquired through multiple analog voltage probes distributed on the motherboard, and the original noise baseline is established based on the original noise signal;

[0059] Specifically, this includes: calculating the variance of each signal in the original noise signal to form a variance vector; estimating the power spectral density of each original noise signal based on the variance vector to generate a power spectral curve; identifying local peak intervals in the power spectral curve where the energy is continuously higher than a preset multiple of the mean; and marking these local peak intervals as candidate frequency bands for the original noise baseline with a determined center frequency and bandwidth. The resulting original noise baseline accurately reflects the motherboard's inherent electromagnetic state when there is no image data transmission. The local peak intervals it contains directly correspond to actual periodic noise sources, providing an irreplaceable physical reference for subsequent identification of whether interference has entered the image link.

[0060] Based on the original noise baseline, test excitation signals are injected into the HDMI or DP physical layer to collect response signals. A differential spectrum is generated by comparing the response signals with the original noise baseline, and active response frequency bands are identified from the differential spectrum. Specifically, this includes: marking frequency points in the differential spectrum where the energy increment exceeds a preset threshold; determining whether the marked frequency points appear synchronously in the monitoring signals of multiple adjacent analog voltage probes; clustering the synchronously appearing marked frequency points into active response frequency bands, and recording the center frequency and bandwidth of the active response frequency bands. The identified active response frequency bands include both the enhanced response of existing frequency bands in the original noise baseline under excitation and new frequency bands that are not apparent in the baseline but have a strong coupling effect on the transmission channel, thus revealing potential threat points that may truly interfere with the image link.

[0061] Activate image data stream transmission, collect noise signals during image data stream transmission, calculate link disturbance spectrum, and generate link disturbance weighted index by combining protocol layer link status indicators. The link disturbance spectrum is the difference between the current noise signal and the original noise baseline.

[0062] The generation of the link disturbance weighted index specifically includes: reading the receiver eye diagram height attenuation rate, clock jitter accumulation value, and symbol error count rate from the protocol layer; spatiotemporally aligning the energy integral values ​​of specified frequency bands in the link disturbance spectrum with the protocol layer link status indicators; assigning weight coefficients to the frequency bands corresponding to different monitoring points based on the alignment results; and summing the weighted frequency band energy integral values ​​to generate the link disturbance weighted index. The generated link disturbance weighted index directly correlates the abstract frequency domain energy distribution with measurable image link physical layer degradation phenomena, enabling the system to quantitatively assess the impact of different frequency bands on actual image quality and providing an operable criterion for subsequently distinguishing key interference frequency bands.

[0063] Based on the contribution of each frequency band in the link disturbance weighted index, the dominant interference source is located, and a frequency band weight distribution including frequency, spatial location, and temporal stability is generated. Specifically, this includes: matching the frequency bands that contribute significantly in the link disturbance weighted index to the physical interference source according to the motherboard hardware resource mapping table; calculating the temporal stability by counting the number of times each frequency band appears in ten consecutive sampling windows; constructing a three-dimensional weight distribution with frequency on the horizontal axis, spatial location on the vertical axis, and temporal stability on the vertical axis; and using the frequency band with the highest value in the weight distribution as the basis for determining the correlation between spatial location and temporal stability.

[0064] By restoring the interference phenomenon to its specific physical source and spatial location, and introducing a time dimension to characterize its evolution trend, the system can distinguish which frequency bands are persistent and stable threats, and which are occasional disturbances, thus providing a reliable basis for differentiated intervention.

[0065] Based on the frequency band weight distribution, multiple tunable notch filters are configured for frequency band filtering. Specifically, this includes: issuing tuning commands to the corresponding tunable notch filter based on the frequency band with the highest weight in the frequency band weight distribution; adjusting the center frequency of the tunable notch filter using a three-order ramp-up method (coarse tuning, medium tuning, and fine tuning); after each order adjustment, waiting for the circuit to reach a steady state and collecting the output voltage fluctuation variance; if the variance decrease exceeds a preset proportion of the previous value, the adjustment is confirmed as effective, and the process continues to the next order. The reassessment period for each notch filter is set based on the time stability of each frequency band. Specifically, this includes: using the base reassessment interval as a benchmark, calculating the reassessment period of each notch filter based on the time stability of each frequency band through linear mapping; continuously monitoring the residual fluctuation variance sequence at the output of the notch filter during the reassessment period; if the residual fluctuation variance sequence shows a continuous sudden increase, the reassessment process is triggered in advance, during which the image data stream is paused, a silent period is inserted, and baseline data is reacquired to update the center frequency.

[0066] This ensures that each tunable notch network always operates at the most effective resonant point under the current operating conditions, avoiding transient disturbances caused by global uniform refresh and preventing low-stability frequency bands from losing their suppression effect due to update lag, thus ensuring that the filtering action is consistent with the rhythm of noise evolution.

[0067] By perturbing each notch network and observing the response changes in other frequency bands, an inter-band coupling strength matrix is ​​constructed. Specifically, this involves: sequentially applying small triangular wave modulation to the center frequency of each tunable notch network, simultaneously recording the variance change rate of corresponding monitoring points in other frequency bands during modulation; filling the normalized slope of the variance change rate relative to the modulation amplitude into the corresponding position in the matrix; and identifying the unidirectional influence relationship between the master and slave frequency bands based on the distribution of non-zero elements in the matrix. After identifying the master-slave relationship and adjusting the master frequency band accordingly, compensation is performed on the slave frequency band. Specifically, when the center frequency of the tunable notch network corresponding to the master frequency band is updated, the compensation amount for the slave frequency band is calculated based on the row elements of the master frequency band in the inter-band coupling strength matrix; the center frequency of the tunable notch network corresponding to the slave frequency band is adjusted synchronously according to the compensation amount; and slave frequency band compensation is completed within fifteen microseconds after the master frequency band adjustment is completed, verifying whether the fluctuation amplitude of the link disturbance weighted index has narrowed after compensation.

[0068] This transforms the physical coupling relationship between multiple interference frequency bands into executable adjustment logic, avoiding cascading mismatches caused by single-point tuning, and significantly improving the convergence stability and response consistency of the link disturbance weighted index.

[0069] On the other hand, this invention proposes a dynamic anti-interference system for high-definition display motherboards, such as... Figure 2 As shown, it includes:

[0070] Multiple analog voltage probes are distributed on the motherboard to collect noise signals;

[0071] The synchronous sampling unit connects to multiple analog voltage probes to synchronously acquire the original noise signal, response signal, and noise signal during the transmission of image data stream;

[0072] The processing unit, connected to the synchronous sampling unit, is used to establish the original noise baseline, generate the differential spectrum and link perturbation spectrum, identify active response frequency bands, generate the link perturbation weighting index and frequency band weight distribution, and construct the inter-band coupling strength matrix.

[0073] Multiple tunable notch filters are connected to the processing unit to perform frequency band filtering based on the frequency band weight distribution, and to perform main frequency band adjustment and slave frequency band following compensation according to the instructions of the processing unit.

[0074] The processing unit works in collaboration with multiple tunable notch networks to achieve a dynamic anti-interference process that sets the revaluation period and master-slave relationship based on time stability.

[0075] Furthermore, the components in the above system also implement other steps of the aforementioned dynamic anti-interference method for a high-definition display motherboard, as follows:

[0076] Step 1: Start the time-domain synchronous sampling channel and establish the original noise baseline.

[0077] After the motherboard completes power stabilization, the video output controller enters standby mode. At this time, the HDMI / DP physical layer has not yet activated the data stream, but the transceiver power supply is ready, and each reference clock oscillator is in free oscillation mode. In this state, the system first activates a set of analog voltage probes distributed on the motherboard, including those near the two ends of the power supply filter capacitor of the PCIe slot, near the VDDIO pin of the southbridge chip, the ground terminal of the shielding layer of the DP main link differential pair, and the common-mode voltage monitoring point of the first differential pair of the HDMITMDS channel.

[0078] These four probes are not evenly spaced, but rather selected based on the location where high-frequency current loops are most easily coupled in the actual motherboard wiring topology. Their outputs are buffered with low gain and then sent to a unified 16-bit synchronous sampling unit. This sampling unit continuously acquires four voltage signals at a fixed rate of 1.2 GHz, with each acquisition lasting 256 microseconds, generating a time series set of 32,768 points, denoted as . The subscript 0 indicates the initial stage, and t is the discrete-time index in nanoseconds. This set undergoes no pre-filtering or peak clipping, preserving all original fluctuation characteristics.

[0079] In obtaining Then, the system calculates the variance value of each signal within a 256-microsecond window, obtaining a four-dimensional variance vector. This vector reflects the background disturbance intensity distribution at each monitoring point when there is no image data transmission, serving as the benchmark for all subsequent comparisons. Due to the inherent ripple and switching noise in the motherboard power supply network, It is not zero, but rather exhibits a stable but non-uniform combination of values, for example... It is usually higher than the other three because it is close to the PLL power supply loop inside the Southbridge chip, which still maintains periodic charging and discharging behavior in the idle state.

[0080] based on The system further extracts the power spectral density estimate for each signal. Using the modified periodogram method, each 32768-point sequence is divided into 128 segments of 256 points each, with a 50% overlap between segments. After applying a Hanning window, a 1024-point Fast Fourier Transform is performed, and the results from the 128 segments are averaged to obtain four power spectral density curves with a frequency resolution of 1.17MHz, denoted as... The value of f ranges from 0 to 500MHz, covering all key frequency bands involved in the HDMI / DP2.1 protocol. This set of spectral lines constitutes the original noise baseline, and its shape directly reflects the electromagnetic radiation fingerprint of the motherboard in a silent state, serving as the starting point for all subsequent frequency band discrimination.

[0081] Finally, the system from Three local peak intervals with energy continuously exceeding twice the mean were identified and marked as follows: Its corresponding center frequency is denoted as The bandwidth is 12MHz for all three ranges. These three ranges are not fixed but drift slowly with motherboard temperature, but are considered quasi-static within a single power-on cycle. Their existence indicates that the motherboard has three independent periodic noise sources, such as... This likely corresponds to the switching frequency of the CPU voltage regulation module. Corresponding to the harmonic clusters after frequency multiplication and division of the memory PHY layer clock, This corresponds to the leakage component of the internal oscillator of the USB 3.2 host chip. These intervals will become the initial candidate set for determining whether interference has entered the image link in the second step.

[0082] Step 2: Inject test stimulus and observe link response offset

[0083] After the initial noise baseline is established, the system enters the active detection phase. At this time, the video output controller still does not send valid pixel data, but injects a set of low-amplitude, wide-bandwidth, non-periodic pseudo-random voltage excitation signals into the DP main link or HDMI™DS channel. This excitation does not carry image information; it serves only as a detection carrier, and its amplitude is limited to a differential swing of less than 20mVpp to ensure that it does not trigger receiver mis-locking or adaptive equalizer activation. The excitation signal is generated by an on-chip direct digital frequency synthesizer, covering a continuous frequency band from 100MHz to 12GHz. However, the energy distribution is not uniform; instead, a main energy peak is set at logarithmic intervals every 100MHz, with the peak width controlled within 8MHz, thus forming sparse but well-covered detection pins in the frequency domain.

[0084] After the excitation is sent into the transmission channel via the physical layer driving circuit, it will generate a superimposed response at the four monitoring points. Since the excitation itself is not periodic, the voltage changes it causes at each probe are mainly due to the combined effects of channel reflection, impedance mismatch, and near-field coupling. The system restarts the synchronous sampling unit to collect a new round of 256-microsecond window data with the same parameters, denoted as... Here, subscript 1 indicates the response state after the initial stimulus injection, which is related to... The only difference is that it incorporates a secondary disturbance caused by the stimulus.

[0085] It is worth noting that, The time-domain waveforms of the various signals are no longer solely dominated by the motherboard's inherent noise floor, but exhibit transient rises and phase jitter at the millivolt level, especially in... This is even more pronounced in the middle, because the shielding layer and common-mode node of these two paths, which are directly related to the physical channels of DP / HDMI, are most sensitive to abnormalities in the traveling wave on the transmission line.

[0086] Subsequently, the system... Perform the same spectral estimation procedure as in the first step to obtain the power spectrum set under the response state. .Will By subtracting point by point, we obtain the difference spectrum. This difference spectrum no longer reflects absolute energy, but rather characterizes the energy increment at each frequency point after excitation injection. Observations show that... Instead of a uniform rise across the entire excitation frequency band, sharp positive peaks appear near several discrete frequency points, with the peak width remaining around 8MHz, consistent with the width of the excitation pin points; simultaneously, The presence of a negative dip nearby indicates that the original noise source underwent phase cancellation or energy redistribution under excitation. This coexistence of positive and negative spectral shifts suggests that the motherboard's electromagnetic environment exhibits strong nonlinear response characteristics and cannot be simply regarded as a linear superposition system.

[0087] Furthermore, the system performs cluster analysis on all positive peaks in the difference spectrum: if a certain frequency point place If the peak value appears simultaneously in at least two of the three adjacent probes, then it is included in the active response frequency band set. Seven segments have been identified so far, with center frequencies of [missing information]. to ,in Spacing less than 15MHz, The spacing is 9 MHz, while the other four bands are far from the original baseline peak. This indicates that the excitation not only stimulates new resonances but also disturbs the radiation patterns of existing noise sources. The new frequency bands that do not overlap with the original baseline are more likely to become unpredictable interference sources in subsequent image transmissions.

[0088] Step 3: Activate the image data stream and capture link integrity metrics

[0089] When the active response frequency band set Once confirmed, the system exits pure probe mode and switches to real-world operation. The video output controller begins sending standard UHD 60Hz image frames to the DP main link or HDMI interface, with a resolution of 3840×2160, a color depth of 10bit, and an RGB 4:4:4 format. The link rate reaches 32.4Gbps (DP2.1UHBR13.5) or 48Gbps (HDMI2.1VRR+DSC). At this point, the physical layer transceiver enters full-load operation, the equalizer automatically adjusts the tap coefficients, the clock recovery circuit locks the center of the data eye diagram, link training is complete, and the image is stably displayed on the external monitor.

[0090] During the continuous transmission of image data, the system does not interrupt sampling at the four monitoring points. Instead, it extends the sampling window length from 256 microseconds to 10 milliseconds to balance high-frequency noise details with low-frequency drift trends. The new round of acquired data is denoted as... Its time span covers approximately 600 frames, sufficient to reflect the electromagnetic behavior of the link under steady-state load. Unlike the previous two steps, The signals no longer exhibit a clear periodic or pseudo-random structure, but are instead a mixture of broadband transients caused by high-speed serial data transitions, slow drifts caused by power rail collapse, and microvolt-level DC offsets caused by thermal expansion. These components together constitute the real noise field during image transmission.

[0091] based on The system recalculates the power spectral density to obtain the service state spectrum. Subsequently, the link perturbation spectrum was constructed. This refers to the net increment of the current noise field relative to the original baseline. This spectrum includes contributions from the active response bands identified in the second step, as well as new components such as the spectral sidelobes of the image data itself, residual oscillations from the link training process, and reverse coupling of the receiver feedback signal.

[0092] In particular, Overlapping areas A significant increase was observed, while The value remains only slightly positive, indicating that some active frequency bands are suppressed under real load, while others are amplified. This phenomenon suggests that the image data stream itself constitutes a dynamic modulation source, which changes the local electromagnetic boundary conditions of the motherboard.

[0093] To quantify the degree of interference in the image link, the system reads three physical layer metrics from the link status register of the DP or HDMI protocol layer: receiver eye diagram height attenuation rate. Clock jitter cumulative value and symbol error count rate All three are returned as dimensionless ratios, ranging from 0 to 1, with higher values ​​indicating poorer link integrity. These three metrics are then compared with... Perform spatiotemporal alignment and find that when When a certain threshold is exceeded, Rising in sync, and The integral value only begins to increase after reaching a higher level. This indicates the common-mode voltage monitoring point. The frequency band is most sensitive to energy changes, and its impact on eye diagram quality precedes its impact on bit error rate, thus it can serve as an early warning signal.

[0094] Therefore, the system defines a link disturbance weighted index:

[0095] ;

[0096] That These are the 12MHz bandwidth ranges corresponding to the second peak range in the original baseline, the third frequency band in the active response, and the sixth frequency band in the active response, respectively. The disturbance spectra corresponding to the third, fourth, and second monitoring points, respectively; coefficients Its value is determined by the average correlation of each monitoring point with the three agreement indicators during historical operation, and does not change with the current operating conditions. This index This provides a unified benchmark for decision-making regarding intervention in the fourth step, comprehensively reflecting the overall pressure exerted on the physical layer of the link by the three most threatening frequency bands in current image transmission.

[0097] Step 4: Locate the dominant source of interference and generate frequency band weight distribution.

[0098] Link Disturbance Weighted Index After the calculations are complete, the system enters the interference source tracing phase. Previous steps did not categorize noise sources; they only recorded their spatial and frequency domain characteristics. The goal of this step is to... Each contribution is mapped to its specific physical location and driving behavior, thus providing a basis for subsequent dynamic adjustments. The system calls the hardware resource mapping table pre-stored in the motherboard management controller. This table records the functional units adjacent to each monitoring point and their typical operating frequency range. For example, monitoring point 3 (DP shield ground terminal) is adjacent to the GPU memory controller and the PCIe 5.0 root complex. When fully loaded, the two generate harmonic clusters around 3.2GHz and 8GHz respectively. However, after attenuation by PCB traces, their third and fifth harmonic components can still be detected in the frequency band below 500MHz.

[0099] Based on the mapping table, the system performs... The three intervals are matched one by one with possible sources. For Its center frequency The measured clock speed was 492.3MHz, which does not coincide with the 1 / 4 divider clock (400MHz) and its second harmonic (800MHz) of the DDR5-6400 memory controller. However, it closely matches the doubled frequency of the 246MHz reference clock of the USB3.2 Gen2x2 PHY within the southbridge chip, with a deviation of only ±0.5MHz. This deviation is exactly equal to the temperature drift of the motherboard crystal oscillator at the current temperature. Therefore, it is determined that... The main cause is clock leakage from the USB host controller chip, whose energy is coupled to the DP shield layer through the southbridge power supply plane, thereby modulating the common-mode voltage.

[0100] for Its center frequency MHz, with an error of 0.2% compared to the GPU memory controller's 1024MHz base clock, and this frequency band is within... The largest increase was observed at monitoring point 4, which is located near the power decoupling capacitor bank of the GDDR6X memory chip array, confirming that its source is the sudden change in switching current generated by the memory controller during high-bandwidth read / write operations. Similarly, The center frequency is 2187.4MHz, which is highly consistent with the 2200MHz reference clock harmonic group of the PCIe 5.0 link, and in The strongest performance was observed at point 2, which is located at both ends of the power supply filter capacitor of the PCIe slot. Therefore, it was determined that the disturbance originated from the power rail disturbance caused by the high-speed communication of the PCIe device.

[0101] After locating the three source points, the system no longer treats each frequency band as an isolated entity, but instead constructs a three-dimensional weighted distribution: the horizontal axis represents frequency (in MHz), the vertical axis represents spatial location (represented by monitoring point numbers 1–4), and the vertical axis represents temporal stability (represented by the probability of the frequency band appearing within ten consecutive 10-millisecond windows). This distribution is denoted as... Where p∈{1,2,3,4}, τ∈[0,1]. For example... This indicates that the 492.3MHz frequency point has 98% time stability at monitoring point 3; This indicates that the stability of 1042.6MHz at monitoring point 4 is 83%. This distribution is not a static table, but is updated in real time with each new window of sampling. The weight of each item is equal to the number of times that item is true in the most recent ten observations divided by ten.

[0102] Step 5: Activate the frequency band adaptive filtering path based on the weight distribution.

[0103] The three-dimensional weight distribution generated in the fourth step The three key frequency bands have been mapped one-to-one with their spatial response locations and temporal stability, but this distribution itself is not operational and needs to be converted into practically functional electrical paths. The system then activates a set of pre-installed analog tunable notch filters on the motherboard, with four independent channels, each connected to the power supply plane or shielding loop of the sensitive area corresponding to the four monitoring points. These notch filters are not fixed-center-frequency LC structures, but rather consist of a voltage-controlled capacitor array and a fine-tuning inductor. Their resonant frequency can be continuously adjusted within the range of 200MHz to 3.5GHz, with an adjustment step of 0.8MHz and a response delay of less than 80 nanoseconds.

[0104] in accordance with The three terms with the highest median values—that is The system sequentially sends tuning commands to the third, fourth, and second notch filter channels, locking their center frequencies at 492.3MHz, 1042.6MHz, and 2187.4MHz, respectively. The first channel is temporarily disabled because its corresponding weight is not among the top three. The tuning process is not a one-step process, but rather a three-stage ramp-up: first, a coarse tuning step (4MHz) is used to approach the target frequency; then, a medium tuning step (0.4MHz) is used for calibration; and finally, a fine tuning step (0.08MHz) is used for micro-scanning. After each adjustment, a 12-microsecond wait is allowed until the circuit enters a new steady state before the output voltage fluctuation variance of that channel is collected. If the variance decreases by more than 17% of the previous value, the step is considered valid, and the next step is initiated; otherwise, the process reverts to the previous step and is retried. This process ensures that the notch filter always operates at the actual impedance matching state, rather than the theoretical resonant point.

[0105] After the three notch filters are put into operation, the system immediately starts a new round of 10-millisecond synchronous sampling to obtain the dataset. .and compared to, The fluctuations around 492MHz converged significantly. The envelope narrows around 1042MHz. The peak density decreased near 2187MHz, while the waveforms of the other frequency bands and the other three signals remained largely unchanged. This indicates that the filtering action was highly localized and did not induce cross-band energy transfer or crosstalk from adjacent paths. It is worth noting that... DC offset is relatively The voltage increased by about 1.3mV. This is because the notch filter changed the current distribution path of the local grounding reference point after it was connected. This is an expected parasitic effect and does not affect the common-mode noise suppression effect.

[0106] based on The system recalculates the perturbation spectrum. And reassess the link disturbance weighted index:

[0107] ;

[0108] The subscript 3 indicates the third round state. These are the disturbance spectra of the third, fourth, and second monitoring points under the current filtering state, respectively. (Measured data...) The decrease was 39.2%, with the first item decreasing by 51.6%, the second by 33.8%, and the third by 28.1%. This non-uniform attenuation ratio indicates that different sources respond differently to filtering: USB clock leakage is suppressed most quickly due to its short path and strong coupling; memory controller switching noise requires a longer time to stabilize due to its high energy and wide bandwidth; and PCIe power supply disturbances are affected by the inertia of the entire board's decoupling network, resulting in the slowest response. This difference will serve as the basis for setting the dynamic update rhythm in step six.

[0109] Step 6: Set the frequency band update period according to the time stability gradient

[0110] The fifth step verified the effectiveness of the frequency band filtering, but it also revealed a problem: the stability of the three frequency bands. The two approaches are not identical. If all notch filter parameters are refreshed using the shortest cycle, frequent re-adjustments in high-stability frequency bands will occur, causing unnecessary phase jitter. Conversely, if the longest cycle is used, low-stability frequency bands may miss critical windows for interference pattern changes. Therefore, the system no longer uses a global timer but instead configures an independent re-estimation cycle for each enabled notch filter channel. This cycle is not set to a fixed number of milliseconds but is determined by the time stability of the corresponding frequency band. It is obtained through linear mapping.

[0111] Specifically, the system defines the basic recalculation interval. Milliseconds, corresponding to stability In the case of; when As the temperature rises, the cycle lengthens. When the value decreases, the period shortens. The mapping relationship is as follows:

[0112] ;

[0113] in The reestimation period for the p-th enabled channel is in milliseconds. The frequency band corresponding to this channel The stability value in the code; the coefficient 1.2 is derived from the joint calibration of the dielectric constant of the motherboard PCB material with temperature change rate and the temperature drift coefficient of the copper foil resistance, ensuring that the periodic adjustment matches the actual physical drift rate. After substituting the values, the third channel (corresponding to...) )of Milliseconds, fourth channel ( )of Milliseconds, second channel ( )of Milliseconds. The three differ by about 300 milliseconds, which is sufficient to stagger re-estimation actions and avoid transient superposition caused by simultaneous adjustments from multiple paths.

[0114] Before the first revaluation period arrives, the system continuously monitors the residual fluctuation variance at the output of each notch channel. Taking the third channel as an example, its variance sequence is denoted as... , k is the number of sampling rounds since activation (each round is 10 milliseconds). When five consecutive rounds... The standard deviation is less than If the channel has reached its optimal suppression state under the current operating conditions, subsequent sampling will only be performed for maintenance purposes, without triggering a reassessment. If a sudden increase occurs in a certain round (the increase exceeds the average by 2.3 times), a reassessment will be initiated in advance, regardless of the cycle. This criterion is based on the typical jump threshold of the contact resistance of copper interconnects under thermal stress, which is an inherent characteristic of the hardware and does not require external calibration.

[0115] When the third channel first reaches the 1376 millisecond node, the system pauses the image data stream for 120 microseconds, inserts a silent period, and then resumes acquisition. Using similar data, a new baseline spectrum was obtained. and with the current In comparison, it was found that The center frequency has drifted from 492.3MHz to 492.7MHz, an offset of 0.4MHz, which is due to the current junction temperature increase of 1.8℃. Therefore, the system updates the center frequency of the third channel to 492.7MHz and records the drift rate as 0.29MHz / ℃, which will be used to correct the subsequent temperature compensation model.

[0116] The other two channels execute the same process in sequence according to their respective cycles, but because the start times are different, the re-evaluation actions are scattered on the time axis. As a result, the parameter maintenance of the entire motherboard electromagnetic environment is transformed into a loosely coupled and asynchronous evolution state, which ensures timely response and avoids system-level oscillation.

[0117] Step 7: Construct the inter-band coupling strength matrix and identify master-slave relationships.

[0118] Step six achieved independent maintenance of each frequency band. However, in actual operation, it was found that when the frequency of the third channel was increased by 0.4MHz, the suppression effect of the fourth channel at 1042.6MHz actually improved by 2.1%, while the residual variance of the second channel at 2187.4MHz increased by 1.8%. This cross-frequency band linkage phenomenon cannot be explained by linear superposition, indicating that there is an indirect electrical coupling path between the three interference sources. To clarify this relationship, the system activated the coupling detection mode in step seven: while maintaining the normal operation of the three notch filters, a small perturbation was applied to each channel in sequence, that is, a ±0.15MHz triangular wave modulation was superimposed on its center frequency, with a modulation period of 200 milliseconds and an amplitude far lower than the bandwidth of the notch filter (12MHz), to ensure that its basic function was not damaged.

[0119] The first disturbance was applied to the third channel, and the system simultaneously recorded the variance change rate of the other two monitoring points during the disturbance period. The results showed that when the frequency of the third channel increased, the slope of the variance decrease at the fourth monitoring point increased, while the slope of the variance increase at the second monitoring point accelerated simultaneously. When the second disturbance was applied to the fourth channel, the variance change at the third monitoring point was slight (<0.3%), while the variance increase at the second monitoring point intensified. When the third disturbance was applied to the second channel, neither the third nor the fourth monitoring points showed a significant response in variance. Therefore, it can be determined that the third frequency band is the driving source, the fourth frequency band is its strong follower, and the second frequency band is its weak follower, forming a unidirectional chain effect structure.

[0120] Based on the above observations, the system constructs a third-order coupling strength matrix. Where row i represents the disturbed frequency band, column j represents the observed frequency band, and element Defined as the normalized slope of the variance change rate of band j relative to the perturbation amplitude of band i. After averaging over ten rounds of perturbations, we get:

[0121] ;

[0122] in This indicates that disturbances in the third frequency band have the strongest impact on the fourth frequency band. This indicates that its impact on the second frequency band is relatively weak. This indicates that the disturbance in the fourth frequency band has a moderate impact on the second frequency band, and the rest are zero. This matrix does not contain self-feedback terms (the diagonal is zero) because the modulation of the notch filter itself is not considered internal coupling. The matrix structure confirms the existence of a master-slave hierarchy and provides a topological basis for designing the coordinated adjustment strategy in step eight.

[0123] Step 8: Implement primary frequency band priority adjustment and secondary frequency band follow-up compensation

[0124] The coupling matrix obtained in step seven This reveals a clear master-slave structure, upon which the system reconstructs its intervention logic: instead of treating the three frequency bands as equal entities, the third frequency band (USB clock leakage) is established as the adjustment master, with the other two frequency bands used as derivative responses for compensation. Specifically, whenever the third channel needs to update its center frequency due to temperature drift or load changes, the system does not adjust that channel individually, but rather... For the corresponding row element, compensation commands are simultaneously sent to the fourth and second channels.

[0125] For example, when the center frequency of the third channel is increased from 492.7MHz to 492.9MHz (+0.2MHz), the system calculates the compensation proportionally: the fourth channel should be decreased. MHz, that is, adjusted from 1042.6MHz to 1042.576MHz; the second channel should be down-adjusted. The frequency was adjusted from 2187.4MHz to 2187.392MHz. Both compensations were completed within 15 microseconds after the third channel update, utilizing the inherent response margin of the notch filter's voltage-controlled capacitor.

[0126] The effect of this coordinated adjustment is that the image link perturbation weighted index... Within 10 milliseconds of the main frequency band update, it fell back to below the level before the update, and the fluctuation range narrowed by 42%, which is a decrease of 39.2% compared to the single-point adjustment in the fifth step, and the overall stability was further improved.

[0127] More importantly, the changing trends of the three indicators at the protocol layer are converging: After master-slave collaboration, synchronous convergence is achieved, and the previous issues no longer occur. The phenomenon of delayed rise indicates that, by respecting the physical coupling nature between frequency bands, the system has successfully integrated the originally discrete anti-interference actions into an organic process, enabling the motherboard to truly possess the ability to identify and distinguish key interference frequency bands for HDMI / DP2.1 ultra-high-definition image transmission in real time, even in environments with coexisting multi-source high-frequency noise.

[0128] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A dynamic anti-interference method for a high-definition display motherboard, characterized in that, include: In the standby state of the video output controller, the original noise signal is synchronously acquired through multiple analog voltage probes distributed on the motherboard, and the original noise baseline is established based on the original noise signal; Based on the original noise baseline, a test excitation signal is injected into the HDMI or DP physical layer, the response signal is collected, a differential spectrum is generated by comparing the response signal with the original noise baseline, and the active response frequency band is identified from the differential spectrum. Activate image data stream transmission, collect noise signals during image data stream transmission, calculate link disturbance spectrum, and generate link disturbance weighted index by combining protocol layer link status indicators. The link disturbance spectrum is the difference between the current noise signal and the original noise baseline. Based on the contribution of each frequency band in the link disturbance weighted index, the dominant source of interference is located, and a frequency band weight distribution including frequency, spatial location and temporal stability is generated. Based on the frequency band weight distribution, multiple tunable notch networks are configured for frequency band filtering, and the reestimation period of each notch network is set according to the time stability of each frequency band. By perturbing each notch network and observing the response changes in other frequency bands, a coupling strength matrix between frequency bands is constructed to identify the master-slave relationship. Based on the master-slave relationship, the master frequency band is adjusted, and the slave frequency band is compensated for following.

2. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The establishment of the original noise baseline includes: The variance of each signal in the original noise signal is calculated to form a variance vector; Power spectrum curves are generated by estimating the power spectral density of each original noise signal based on the variance vector. Identify local peak intervals in the power spectrum curve where the energy is continuously higher than the mean by a preset multiple; The local peak intervals are marked as candidate frequency bands for the original noise baseline, with the center frequency and bandwidth determined.

3. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The identified active response frequency bands include: Frequency points in the differential spectrum where the energy increment exceeds a preset threshold are marked; Determine whether the marked frequency point appears synchronously in the monitoring signals of multiple adjacent analog voltage probes; Cluster the synchronously occurring marked frequency points into active response frequency bands, and record the center frequency and bandwidth of the active response frequency bands.

4. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The generated link disturbance weighted index includes: Read the receiver eye diagram height attenuation rate, clock jitter accumulation value, and symbol error count rate from the protocol layer; Spatiotemporal alignment of the energy integral value of a specified frequency band in the link disturbance spectrum with the protocol layer link state index; Based on the alignment results, weighting coefficients are assigned to the frequency bands corresponding to different monitoring points, and the weighted frequency band energy integral values ​​are summed to generate the link disturbance weighted index.

5. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, Generating the frequency band weight distribution includes: Based on the motherboard hardware resource mapping table, the frequency bands that contribute significantly to the link disturbance weighted index are matched to the physical interference sources; The time stability was calculated by counting the number of times each frequency band appeared in ten consecutive sampling windows. Construct a three-dimensional weight distribution, with the horizontal axis representing frequency, the vertical axis representing spatial location, and the vertical axis representing temporal stability; The frequency band with the highest value in the weight distribution is used as the criterion for determining the correlation between spatial location and temporal stability.

6. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The configuration of multiple tunable notch networks for frequency band filtering includes: Based on the frequency band with the highest weight in the frequency band weight distribution, a tuning command is sent to the corresponding tunable notch network. The center frequency of the tunable notch filter network is adjusted using a three-order ramping method: coarse adjustment, medium adjustment, and fine adjustment. After each adjustment, wait for the circuit to reach a steady state and collect the output voltage fluctuation variance. If the variance decreases by more than the preset proportion of the previous value, the adjustment is confirmed to be effective and the next adjustment is continued.

7. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The recalculation period for each notch network is defined as follows: Based on the basic revaluation interval, the revaluation period of each notch network is calculated by linear mapping according to the time stability of each frequency band. The residual fluctuation variance sequence at the output of the notch network is continuously monitored during the revaluation period. If the residual variance sequence shows a continuous sudden increase, the reassessment process is triggered in advance. During the reassessment, the image data stream is paused, a quiet period is inserted, and baseline data is reacquired to update the center frequency.

8. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The constructed inter-band coupling strength matrix includes: A small triangular wave modulation was applied sequentially to the center frequency of each tunable notch network, and the variance change rate of the corresponding monitoring points in other frequency bands was recorded simultaneously during the modulation period. Fill the normalized slope of the variance change rate relative to the modulation amplitude into the corresponding position in the matrix; The unidirectional influence relationship between the main frequency band and the slave frequency band is identified based on the distribution of non-zero elements in the matrix.

9. The dynamic anti-interference method for a high-definition display motherboard according to claim 1, characterized in that, The following compensation for the frequency band includes: When the center frequency of the tunable notch filter network corresponding to the main frequency band is updated, the compensation amount of the slave frequency band is calculated based on the row elements of the main frequency band in the inter-band coupling strength matrix. The center frequency of the tunable notch network corresponding to the frequency band is adjusted synchronously according to the compensation amount. Ensure that the secondary frequency band compensation is completed within 15 microseconds after the main frequency band adjustment is completed, and verify whether the fluctuation range of the link disturbance weighted index has narrowed after compensation.

10. A dynamic anti-interference system for a high-definition display motherboard for implementing the method as described in any one of claims 1-9, characterized in that, include: Multiple analog voltage probes are distributed on the motherboard to collect noise signals; The synchronous sampling unit connects to multiple analog voltage probes to synchronously acquire the original noise signal, response signal, and noise signal during the transmission of image data stream; The processing unit, connected to the synchronous sampling unit, is used to establish the original noise baseline, generate the differential spectrum and link perturbation spectrum, identify active response frequency bands, generate the link perturbation weighting index and frequency band weight distribution, and construct the inter-band coupling strength matrix. Multiple tunable notch filters are connected to the processing unit to perform frequency band filtering based on the frequency band weight distribution, and to perform main frequency band adjustment and slave frequency band following compensation according to the instructions of the processing unit. The processing unit works in collaboration with multiple tunable notch networks to achieve a dynamic anti-interference process that sets the revaluation period and master-slave relationship based on time stability.

Images