A method and system for bus interface data skew convergence of a co-packaged optical module

Abstract

The application discloses a bus interface data deviation convergence method and system of a co-encapsulation optical module, relates to the technical field of data deviation convergence, and comprises the following steps: at the bus interface layer, selecting a channel of a physical wire crossing a junction of different material layers as a first channel group, and selecting a channel of the physical wire completely located in a same material layer as a second channel group; identifying a relative sliding direction between the different material layers; and when a difference mode changes from systematic differentiation to irregular dispersion, maintaining a current compensation offset amount. The application groups and compares the channel of the physical wire crossing the junction of the different material layers and the channel of the physical wire completely located in the same material layer, identifies the relative sliding direction between the heterogeneous metal layers based on a response difference mode of the two groups of channels, matches the compensation direction with the actual direction of the interlayer sliding, and helps to reduce the problem of mismatching between the compensation direction and the sliding direction caused by the unified temperature compensation in the prior art.

Description

Technical Field

[0001] This invention relates to the field of data deviation convergence technology, and in particular to a method and system for data deviation convergence of a bus interface of a co-packaged optical module. Background Technology

[0002] Co-Packaged Optics (CPO) technology integrates the optical engine and switching chip into the same package, achieving electrical interconnection between the chips through a high-density bus interface. To meet the requirements of high bandwidth and low latency transmission, the electrical interconnection layout within the CPO package is becoming increasingly dense. Its physical traces often need to cross multiple heterogeneous metal layers such as copper redistribution layers, aluminum bonding pads, and through-silicon via copper pillars, forming a complex cross-layer channel network.

[0003] In the actual operating environment of CPO packages, chip power dissipation and ambient temperature changes cause the package to undergo continuous temperature cycling. Due to the significant differences in the thermal expansion coefficients of heterogeneous materials such as copper, aluminum, and silicon, each metal layer expands or contracts to varying degrees when the temperature changes. For channels where physical traces cross the interfaces of different metal layers, the asynchronous deformation of each metal layer will cause relative slippage between layers, resulting in an asymmetric change in the equivalent trace length and interlayer contact state of the cross-layer channel, which in turn causes timing drift in bus interface data transmission.

[0004] Existing technologies typically attribute timing drift in bus interfaces to a uniform temperature effect, employing global temperature compensation strategies, such as applying uniform timing offset compensation to all channels based on overall package temperature measurements. However, this compensation method fails to distinguish the relative displacement direction between metal layers: when different metal layers slip relative to each other during temperature cycling, the drift characteristics of cross-layer channels exhibit systematic differences from those of channels entirely within the same metal layer, and may even show directional differentiation. Equating cross-layer drift with intra-layer drift leads to a mismatch between the compensation direction and the actual direction of inter-layer slip, which not only fails to effectively converge cross-layer channel data deviations but may also exacerbate timing degradation. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for converging bus interface data deviations of co-packaged optical modules, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for converging bus interface data deviation of a co-packaged optical module, comprising the following steps:

[0007] In the bus interface layer, channels where physical traces cross the interface of different material layers are selected as the first channel group, and channels where physical traces are completely located within the same material layer are selected as the second channel group.

[0008] The same probe sequence is sent to the first channel group and the second channel group. Based on the response difference pattern of the two channels, the relative slip direction between different material layers is identified.

[0009] For the first channel group, a compensation offset opposite to the relative sliding direction is applied at the bus interface transmitting end;

[0010] Continuously monitor the response difference patterns between the first channel group and the second channel group. When the difference pattern changes from systematic differentiation to irregular discreteness, maintain the current compensation offset.

[0011] Preferably, the step of selecting channels that cross the interfaces of different material layers as the first channel group includes:

[0012] Traverse all channels of the bus interface layer to obtain the crossing depth and crossing angle of each trace across the material layer interface;

[0013] Based on the crossing depth and crossing angle, the interlayer slip sensitivity coefficient of each channel is calculated. The interlayer slip sensitivity coefficient is positively correlated with the crossing depth and negatively correlated with the crossing angle and the degree of deviation in the vertical direction.

[0014] The channels whose interlayer slip sensitivity coefficient exceeds a preset sensitivity threshold are selected as the first channel group.

[0015] Preferably, the selection of channels whose physical traces are entirely located within the same material layer as the second channel group includes:

[0016] Analyze the material layers that each channel in the first channel group crosses to determine its source material layer and target material layer;

[0017] In the source material layer and the target material layer, channels whose physical traces are completely located within the material layer and whose horizontal projections at the interlayer interface are spatially adjacent to the corresponding channels in the first channel group are selected as candidate reference channels.

[0018] A pre-detection sequence is sent to the candidate reference channel to filter out channels whose response drift direction is consistent with the overall thermal expansion trend of the package, thus forming the second channel group.

[0019] Preferably, sending the same probe sequence to the first channel group and the second channel group includes:

[0020] The detection sequence is arranged to include a rising edge dominant segment and a falling edge dominant segment. The rising edge dominant segment is composed of consecutive alternating transitions from a first logic level to a second logic level, and the falling edge dominant segment is composed of consecutive alternating transitions from a second logic level to a first logic level. Both segments have the same symbol length and bit timing period.

[0021] The detection sequence is sent to the first channel group and the second channel group, and the response timing of each channel group to the rising edge dominant segment and the falling edge dominant segment is recorded respectively to construct the response difference pattern.

[0022] Preferably, identifying the relative slip direction between different material layers based on the response difference patterns of the two sets of channels includes:

[0023] During the temperature cycling process of the package, the detection sequence is sent to the first channel group and the second channel group at the heating phase, the holding phase and the cooling phase respectively, and the response timing offset of the two channels at each phase is recorded to construct the phase-offset difference mode.

[0024] Extract the response timing offset under the heating phase and the cooling phase, compare the directional polarity of the two, and when the directional polarity is opposite and symmetrical and the response timing offset under the heat preservation phase is significantly reduced relative to the heating phase and the cooling phase, determine the offset direction under the heating phase as the preliminary result of the relative sliding direction.

[0025] Preferably, after determining the preliminary judgment result, the method further includes:

[0026] Calculate the difference in response timing offset of each channel in the first channel group under the heating phase and the cooling phase, and construct an offset difference sequence;

[0027] The depth of the physical traces of each channel in the first channel group crossing the material layer interface is associated with the offset difference sequence one by one according to the channel to construct a depth-difference response sequence.

[0028] Determine the monotonicity of the depth-difference response sequence: when the sequence monotonically increases with increasing crossing depth, confirm the initial judgment result; when the sequence monotonically decreases with increasing crossing depth, reverse the initial judgment result to the relative slip direction.

[0029] Preferably, applying a compensation offset opposite to the relative slip direction at the bus interface transmitting end to the first channel group includes:

[0030] From the response difference pattern, extract the response timing offset of the first channel group to the rising edge dominant segment and the falling edge dominant segment, and calculate the difference between the two as the edge asymmetric differentiation amount.

[0031] Based on the edge asymmetry differentiation amount, the compensation offset is decomposed into rising edge compensation component and falling edge compensation component. The rising edge compensation component and falling edge compensation component are opposite to the relative slip direction, and their absolute values ​​are differentiated according to the magnitude of the edge asymmetry differentiation amount.

[0032] At the bus interface transmitting end, the data transmission timing of each channel in the first channel group is loaded according to the effective logic transition type: the rising edge compensation component is loaded for logic transitions with the same polarity as the rising edge dominant segment, and the falling edge compensation component is loaded for logic transitions with the same polarity as the falling edge dominant segment.

[0033] Preferably, the continuous monitoring of the response difference pattern between the first channel group and the second channel group includes:

[0034] In each monitoring cycle, the response offset polarity of each channel in the first channel group relative to the corresponding reference channel in the second channel group is extracted to construct a polarity sequence;

[0035] Calculate the polarity consistency index of the polarity sequence, which is used to characterize the degree to which the response offset polarity of each channel in the first channel group points in the same direction;

[0036] When the polarity consistency index decreases from above the preset consistency threshold to below the preset consistency threshold, it is determined that the difference pattern changes from systematic differentiation to irregular dispersion.

[0037] Preferably, maintaining the current compensation offset when the difference pattern changes from systematic differentiation to irregular discreteness includes:

[0038] After determining that the difference pattern has turned into irregular discreteness, the current compensation offset is used as a candidate maintenance value, and the verification maintenance period is started.

[0039] During the verification maintenance period, the probe sequence is sent to the first channel group and the second channel group with the candidate maintenance value as a fixed compensation amount to continuously monitor the polarity consistency index;

[0040] When the polarity consistency index remains below the preset consistency threshold during the verification maintenance period, the candidate maintenance value is confirmed as a valid maintenance value.

[0041] Based on the edge asymmetry differentiation amount, maintenance tolerances are set for the rising edge compensation component and the falling edge compensation component, respectively. The maintenance tolerance is related to the response fluctuation amplitude of the corresponding edge component during the verification maintenance period.

[0042] In subsequent monitoring, when the fluctuations of both the rising edge compensation component and the falling edge compensation component are within the corresponding maintenance tolerance, the effective maintenance value is maintained.

[0043] A bus interface data deviation convergence system for a co-packaged optical module includes:

[0044] The channel grouping selection module is used to select, at the bus interface layer, channels that physical traces cross the interface of different material layers as the first channel group, and channels that physical traces are completely located within the same material layer as the second channel group.

[0045] A probe sequence sending module is used to send the same probe sequence to the first channel group and the second channel group;

[0046] The response acquisition and interlayer slip recognition module is used to acquire the responses of the first channel group and the second channel group to the detection sequence, and to identify the relative slip direction between different material layers based on the response difference patterns of the two channels.

[0047] The compensation offset loading module is used to apply a compensation offset opposite to the relative sliding direction to the first channel group at the bus interface transmitting end;

[0048] The monitoring and maintenance module is used to continuously monitor the response difference pattern between the first channel group and the second channel group. When the difference pattern changes from systematic differentiation to irregular dispersion, the current compensation offset is maintained.

[0049] The technical effects and advantages of this invention are as follows:

[0050] (1) The present invention compares channels that cross the interface of different material layers with channels that are completely located in the same material layer in groups. Based on the response difference pattern of the two groups of channels, the relative slip direction between heterogeneous metal layers is identified, and a compensation offset opposite to the direction is applied so that the compensation direction matches the actual slip direction between the layers. This helps to reduce the mismatch between the compensation direction and the slip direction caused by uniform temperature compensation in the prior art, thereby suppressing the further deterioration of the timing deviation of cross-layer channels.

[0051] (2) This invention sends detection sequences in three phases of temperature cycling of the package body: heating, holding and cooling. It uses the reverse symmetry of the phase response polarity of heating and cooling and the attenuation of the phase shift of holding to construct the initial judgment result. It also verifies the result by combining the monotonicity of the response sequence of the crossing depth and the shift difference. This provides a thermodynamic consistency verification basis for the determination of the sliding direction and helps to reduce the probability of misjudgment of direction.

[0052] (3) By extracting the edge asymmetric differentiation amount, the present invention decomposes the compensation offset into rising edge compensation component and falling edge compensation component, and loads them respectively at the bus interface transmitting end according to the effective logic transition type, so that the compensation configuration is adapted to the asymmetric influence characteristics of interlayer slip on the charging path and the discharging path, which helps to improve the synchronization of rising edge and falling edge timing convergence.

[0053] (4) This invention monitors the change from systematic differentiation to irregular dispersion of the response difference pattern between the first channel group and the second channel group through the polarity consistency index, and confirms the effectiveness of the compensation amount by verifying the maintenance period. It provides a basis for judging the maintenance of the compensation amount after the interlayer slip enters the stable period, which helps to reduce the situation where the transient compensation amount is solidified into overcompensation. Attached Figure Description

[0054] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings:

[0055] Figure 1 This is a flowchart illustrating the data deviation convergence method of the present invention;

[0056] Figure 2 This is a schematic diagram of the data deviation convergence system of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Example 1: As Figure 1 The present invention provides a method for converging bus interface data deviation of a co-packaged optical module, comprising the following steps:

[0059] S1, Channel Grouping Selection

[0060] In this example, at the bus interface layer, channels where physical traces cross the interface between different material layers are selected as the first channel group, and channels where physical traces are entirely within the same material layer are selected as the second channel group. Specifically:

[0061] Iterate through all channels of the bus interface layer, extract the stack-up information of the physical traces for each channel from the package layout database (such as GDSII or LEF / DEF format), and obtain the crossing depth and crossing angle of each trace across the material layer interface. The crossing depth is defined as the vertical distance of the trace crossing the material layer interface; the crossing angle is defined as the angle between the trace extension direction and the normal direction of the interlayer interface.

[0062] The inter-layer slip sensitivity coefficient for each channel is calculated based on the span depth and span angle. In this example, the formula is used: K = d × cosθ;

[0063] In the formula, d represents the crossing depth, and θ represents the crossing angle. When θ = 0° (vertical crossing), cosθ = 1, and the sensitivity coefficient is at its maximum; when θ = 90° (horizontal crossing), cosθ = 0, and the sensitivity coefficient is at its minimum. This formula reflects the characteristic that the sensitivity coefficient is positively correlated with the crossing depth and negatively correlated with the crossing angle and the degree of deviation in the vertical direction.

[0064] Channels whose interlayer slip sensitivity coefficient exceeds a preset sensitivity threshold are selected as the first channel group. In this example, the preset sensitivity threshold Kth is set based on the magnitude of the difference in thermal expansion coefficients between the heterogeneous metal layers within the package. For example, when the copper layer (with a thermal expansion coefficient of approximately...)... / °C) and aluminum layer (coefficient of thermal expansion approximately At the boundary of / °C, It is set to 50% of the maximum allowable equivalent trace drift under the maximum thermal cycling temperature range of the package, with a typical value range of 0.5μm to 2.0μm.

[0065] Analyze the material layers crossed by each channel in the first channel group to determine its source and target material layers. Within the source and target material layers, select physical traces that are entirely within their respective material layers and whose horizontal projection distance at the interlayer interface with the corresponding channel in the first channel group is less than a preset proximity threshold. The channel is used as a candidate baseline channel. In this example, The value is set to 50μm to ensure that the candidate reference channel and the corresponding cross-layer channel have the same thermal field distribution and overall deformation characteristics of the package at the interlayer interface.

[0066] A pre-probe sequence is sent to the candidate reference channels. The pre-probe sequence uses the same code structure as the probe sequence in subsequent step S2, but the symbol length is shortened to 1 / 4 of the data symbol length to reduce the pre-probe's impact on normal bus data transmission. The timing offset polarity of each candidate reference channel's response to the pre-probe sequence is recorded. Based on the transmission time of the pre-probe sequence, the phase advance of the receiver's recovery clock relative to the transmission clock is recorded as the positive drift direction, and the phase lag is recorded as the negative drift direction. When the polarity direction deviates from the expected drift direction caused by the theoretical overall deformation direction by less than 15°, it is determined to be consistent, and consistent channels are selected to form the second channel group.

[0067] It should be noted that the overall thermal expansion trend of the package is calculated by real-time acquisition of temperature gradients using an array of temperature sensors located at the four corners and center of the package substrate. This data, combined with the nominal thermal expansion coefficients of each material layer, determines the theoretical overall deformation direction of the package at the current temperature. This screening mechanism ensures that the second channel group is driven only by the overall thermal expansion of the package and is not affected by local interlayer slippage, thus forming a high signal-to-noise ratio reference system with the first channel group.

[0068] S2, Probe sequence transmission

[0069] In this example, the same probe sequence is sent to both the first channel group and the second channel group, specifically:

[0070] The probe sequence is arranged to include rising-edge dominant segments and falling-edge dominant segments. The rising-edge dominant segment consists of consecutive alternating transitions from a first logic level to a second logic level, such as the code pattern "010101…"; the falling-edge dominant segment consists of consecutive alternating transitions from a second logic level to a first logic level, such as the code pattern "101010…". Symbol length The standard timing period for the bus interface Integer multiples (e.g.) =8 ), ensuring that the probe sequence is compatible with the normal data timing, and that the rising edge-dominant segment and the falling edge-dominant segment and They are all the same.

[0071] The probe sequences are sent to the first channel group and the second channel group, and the response timing of each channel group to the rising edge dominant segment and the falling edge dominant segment is recorded respectively to construct the response difference pattern.

[0072] It should be noted that the first logic level and the second logic level correspond to the standard differential signal levels of the bus interface. For example, the first logic level is -400mV and the second logic level is +400mV. (Bit timing period) The settings are based on the bus data rate, for example, in a 56Gbps PAM4 bus. The value is 142.86 ps. By recording the response timing of the rising edge-dominant segment and the falling edge-dominant segment respectively, the asymmetric influence of interlayer slip on the charging and discharging paths can be made explicit, providing a data basis for subsequent orientation identification and edge differentiation compensation.

[0073] S3, Interlayer slip direction recognition

[0074] In this example, the relative slip direction between different material layers is identified based on the response difference patterns of the two sets of channels, specifically:

[0075] During the temperature cycling process of the package, the temperature change rate of the package substrate is monitored in real time by an array of temperature sensors. Detection sequences are sent to the first and second channel groups during the heating phase, the holding phase, and the cooling phase, respectively.

[0076] The criterion for determining the heating phase is: the rate of temperature change over three consecutive sampling periods. / All are greater than +0.5°C / min; the criterion for determining the heat preservation phase is: the absolute value of the temperature change rate. The temperature change rate is less than 0.2°C / min and the temperature fluctuation range is less than ±2°C, with a sustained stable time exceeding 30 seconds; the criterion for determining the cooling phase is the temperature change rate over three consecutive sampling periods. / All are less than -0.5°C / min.

[0077] Record the timing offset of the response of the first and second channel groups to the probe sequence at each phase. A phase-shift difference model is constructed. The response time shifts under the heating and cooling phases are extracted, and their directional polarities are compared. When the heating phase shift... Phase offset with cooling The signs are opposite (i.e.) × <0), and the ratio of their absolute values ​​is in the range of 0.5 to 2.0, while the thermal insulation phase offset When the absolute value is less than 30% of the absolute value of the heating phase offset and the absolute value of the cooling phase offset, the polarity of the determination direction is reversed and symmetrical, and the heat preservation phase is significantly attenuated. The offset direction under the heating phase is determined as the preliminary result of the relative sliding direction.

[0078] After determining the initial judgment result, the difference in response timing offset of each channel in the first channel group under the heating phase and the cooling phase is calculated to construct an offset difference sequence. In this example, signed difference is used. The depth at which the physical traces of each channel in the first channel group cross the material layer interface. By associating the offset difference sequence with each channel, a depth-difference response sequence is constructed. .

[0079] Determining the monotonicity of the depth-difference response sequence: A linear fitting method is used to calculate the fitting slope k and the coefficient of determination of the sequence. When k> (like =0.1ns / μm) and When k > 0.75, the judgment sequence monotonically increases with the crossing depth, confirming the initial judgment result; when k < - and When the value is greater than 0.75, the judgment sequence decreases monotonically with the increase of the crossing depth, and the initial judgment result is reversed to the relative slip direction.

[0080] It should be noted that the direction polarity of the response timing offset is referenced to the second channel group: when the response timing of the first channel group is ahead of the second channel group, the offset polarity is recorded as positive; when it is behind, it is recorded as negative. The 30% threshold for significant attenuation of the insulation phase offset is based on the physical characteristic that interlayer slip tends to stabilize under thermal equilibrium. If the insulation phase offset remains at a high level, it indicates that the drift may be dominated by non-thermal cycling factors (such as permanent contact degradation caused by process defects) rather than interlayer slip. In this case, the system marks an anomaly and triggers an alarm. Threshold and The calibration is based on theoretical calculations and measured statistical results of the difference in thermal expansion coefficients between heterogeneous metal layers within the package. For example, when the theoretical equivalent trace change rate caused by the maximum thermal expansion difference between the copper and aluminum layers is 0.15 ns / μm, the calibration is set... =0.1ns / μm to retain a decision margin; A value >0.75 ensures that the linear correlation between depth and difference is statistically significant, eliminating interference from local process noise. The monotonicity check of the depth-difference response sequence utilizes the physical law that the greater the span depth, the more significant the impact of interlayer slip on the equivalent trace length. If the initial judgment is correct, depth and difference should be positively correlated; if they are negatively correlated, it indicates that the initial judgment has reversed the polarity of the response to slip exacerbation and slip mitigation, requiring reverse correction to block the source of cascading failure.

[0081] S4, Compensation Offset Loading

[0082] In this example, for the first channel group, a compensation offset opposite to the relative sliding direction is applied at the bus interface transmitting end, specifically as follows:

[0083] From the response difference pattern, the response timing offsets of the first channel group for the rising edge-dominant segment and the falling edge-dominant segment are extracted and denoted as follows: and Calculate the edge asymmetric differentiation, which is relative to the bit time period. The normalized dimensionless value of is: ;

[0084] Based on the asymmetric differentiation at the edge The compensation offset is decomposed into rising edge compensation components. With falling edge compensation component In this example, a basic compensation amount is set. Opposite to the direction of relative slip, and using the following decomposition rules: ; ;

[0085] In the formula, γ is a dimensionless differentiation weight coefficient, which is calibrated according to the asymmetric sensitivity of the channel to the interlayer contact state, and typically ranges from 0.1 to 0.5. The rising edge compensation component and the falling edge compensation component are both opposite to the relative slip direction, and their absolute values ​​are differentiated according to the magnitude of the edge asymmetric differentiation.

[0086] At the bus interface transmitter, the data transmission timing of each channel in the first channel group is loaded according to the effective logic transition type: a rising edge compensation component is loaded for logic transitions with the same polarity as the rising edge dominant segment (i.e., 0→1 transition). Add a falling edge compensation component to logic transitions (i.e., 1→0 transitions) that have the same polarity as the dominant falling edge segment. .

[0087] It should be noted that the physical application of the compensation offset at the transmitting end is achieved through an adjustable delay line (TDL) array or a delay phase-locked loop (DLL) phase adjuster. The transmitting end controller, based on the current logic transition type, retrieves the corresponding compensation component value from a lookup table (LUT), converts it into a delay control code, and drives the TDL to perform picosecond-level precision phase adjustment on the output data edges. Because... To normalize the dimensionless value, γ is also a dimensionless coefficient; the above formulas have consistent dimensions. and All with Both are time-based quantities. Marginal asymmetric differentiation quantity. The introduction of this feature makes the compensation action isomorphic to the physical asymmetric characteristics of interlayer slip, avoiding the conventional deviation of using symmetric compensation to deal with asymmetric drift, and ensuring that the timing convergence of the rising and falling edges is restored synchronously after compensation.

[0088] S5, Continuous Monitoring and Maintenance

[0089] In this example, the response difference pattern between the first channel group and the second channel group is continuously monitored. When the difference pattern changes from systematic differentiation to irregular dispersion, the current compensation offset is maintained, specifically:

[0090] In each monitoring cycle, the response offset polarity of each channel in the first channel group relative to the corresponding reference channel in the second channel group is extracted, and a polarity sequence P={p1,p2,…,pN} is constructed, where pi∈{+1,-1}, +1 represents positive offset (advance), and -1 represents negative offset (lag).

[0091] Calculate the polarity consistency index of the polarity sequence. In this example, the formula is used: ;

[0092] The polarity consistency index is used to characterize the degree to which the response offset polarity of each channel in the first channel group points in the same direction. When the polarity consistency index... From above the preset consistency threshold The state decreased to The following condition indicates that the pattern of difference has shifted from systematic differentiation to irregular dispersion. In this example, Set to 0.85.

[0093] After the differential pattern is determined to transform into irregular discreteness, the current compensation offset is used as a candidate maintenance value, and a verification maintenance period is initiated. In this example, the verification maintenance period is set to 50 complete monitoring cycles (or corresponding to 10 temperature cycle cycles). During the verification maintenance period, probe sequences are sent to the first and second channel groups using the candidate maintenance value as a fixed compensation amount to continuously monitor the polarity consistency index. .

[0094] When the polarity consistency index The value remains below the preset consistency threshold throughout the validation maintenance period. When the candidate maintenance value is confirmed, it is considered a valid maintenance value.

[0095] Based on the asymmetric differentiation at the edge The sustain tolerances for the rising edge compensation component and the falling edge compensation component are set separately. In this example, the formula is used: ; ;

[0096] In the formula, and These represent the standard deviation of the response fluctuation of the corresponding edge components during the verification maintenance period; β is a dimensionless asymmetric sensitivity coefficient, calibrated according to the asymmetric tolerance of the channel process. For example, when the asymmetric change rate of interlayer contact resistance exceeds 10%, β is set to 0.2 to match the tolerance amplification requirements of this asymmetry, with a typical value range of 0.1 to 0.3. Since... To normalize dimensionless values, the above formulas have consistent dimensions. and , and All are time units. The maintenance tolerance is correlated with the response volatility of the corresponding edge component during the validation maintenance period, and is expressed through edge asymmetric differentiation. Perform adaptive adjustment: when When the value is large, it indicates that the effect of interlayer slip on the edge itself has a large asymmetry, the micro-tremor base during the stable period increases accordingly, and the maintenance tolerance is adaptively relaxed; when When =0, it degenerates to the standard 1.5σ tolerance.

[0097] In subsequent monitoring, the deviation between the actual applied rising edge compensation component and the corresponding component in the effective maintenance value of each channel is calculated in real time. And the deviation between the falling edge compensation component and the corresponding component in the effective maintenance value. When | |≤ And | |≤ At that time, maintain the effective maintenance value.

[0098] It should be noted that the verification maintenance period is designed to eliminate residual thermal inertia interference during the transition of interlayer slip from the driving phase to the stable phase. If during the verification maintenance period... A rebound to The above situation indicates that the interlayer slip has not yet truly entered the stable period. The system terminates the verification and re-executes step S3 for orientation identification and compensation adjustment to avoid solidifying the transient instantaneous compensation amount into overcompensation. The maintenance tolerance adopts... The rules allow random fluctuations in the micro-contact state during the stable period to adapt within their respective tolerances, while ensuring that the tolerance setting matches the inherent asymmetric characteristics of the channel through the edge asymmetric differentiation amount. This prevents the compensation amount from becoming ineffective due to excessive leniency or frequently triggering readjustment due to excessive strictness, thus achieving a balance between macro-compensation benchmark locking and micro-fluctuation adaptation.

[0099] Example 2: Figure 2 The present invention provides a bus interface data deviation convergence system for a co-packaged optical module, including a channel group selection module, a probe sequence transmission module, a response acquisition and interlayer slip recognition module, a compensation offset loading module, and a monitoring and maintenance module.

[0100] In this example, the channel grouping selection module is deployed in the programmable logic unit of the FPGA. It reads the pre-stored package layout database through the SPI interface, parses the stack-up attributes and geometric parameters of each channel, and performs extraction of crossing depth and crossing angle, sensitivity coefficient calculation and grouping screening. It is used to select channels that cross the interface of different material layers as the first channel group and channels that are completely located in the same material layer as the second channel group at the bus interface layer.

[0101] The probe sequence sending module is integrated into the code generator of the transmitting end SerDes. According to the control instructions issued by the FPGA, it generates and sends the same probe sequence to the first channel group and the second channel group.

[0102] The response acquisition and interlayer slip recognition module consists of a phase detector in the receiver's CDR (Clock Data Recovery) circuit and an embedded processor. The phase detector samples the arrival time of the response edges of each channel, while the embedded processor runs a temperature phase determination algorithm, a direction polarity comparison algorithm, and a depth-difference sequence monotonicity verification algorithm to acquire the responses of the first and second channel groups to the probe sequence. Based on the response difference patterns of the two groups of channels, the relative slip direction between different material layers is identified.

[0103] The compensation offset loading module consists of an independent adjustable delay line (TDL) array for each channel at the transmitting end and a delay control code lookup table (LUT). The TDL array receives the compensation component values ​​from the response acquisition and inter-layer slip recognition module, selects the corresponding delay control code according to the valid logic transition type, and applies a compensation offset opposite to the relative slip direction to the first channel group at the bus interface transmitting end.

[0104] The monitoring and maintenance module consists of a state machine within the FPGA and an interface with the temperature sensor array. The state machine polls the polarity consistency index according to the monitoring cycle, performs verification maintenance period timing and maintenance tolerance comparison, and is used to continuously monitor the response difference patterns between the first channel group and the second channel group. When the difference pattern changes from systematic differentiation to irregular dispersion, the current compensation offset is maintained.

[0105] It should be noted that the above-mentioned functional modules can be implemented, in whole or in part, through software, hardware, firmware, or other arbitrary combinations. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for converging bus interface data deviation of a co-packaged optical module, characterized in that, Includes the following steps: In the bus interface layer, channels where physical traces cross the interface of different material layers are selected as the first channel group, and channels where physical traces are completely located within the same material layer are selected as the second channel group. The same probe sequence is sent to the first channel group and the second channel group. Based on the response difference pattern of the two channels, the relative slip direction between different material layers is identified. For the first channel group, a compensation offset opposite to the relative sliding direction is applied at the bus interface transmitting end; Continuously monitor the response difference patterns between the first channel group and the second channel group. When the difference pattern changes from systematic differentiation to irregular discreteness, maintain the current compensation offset.

2. The method for converging bus interface data deviation of a co-packaged optical module according to claim 1, characterized in that, The selection of channels that cross the interfaces of different material layers as the first channel group includes: Traverse all channels of the bus interface layer to obtain the crossing depth and crossing angle of each trace across the material layer interface; Based on the crossing depth and crossing angle, the interlayer slip sensitivity coefficient of each channel is calculated. The interlayer slip sensitivity coefficient is positively correlated with the crossing depth and negatively correlated with the crossing angle and the degree of deviation in the vertical direction. The channels whose interlayer slip sensitivity coefficient exceeds a preset sensitivity threshold are selected as the first channel group.

3. The method for converging bus interface data deviation of a co-packaged optical module according to claim 2, characterized in that, The selection of channels whose physical traces are entirely located within the same material layer as the second channel group includes: Analyze the material layers that each channel in the first channel group crosses to determine its source material layer and target material layer; In the source material layer and the target material layer, channels whose physical traces are completely located within the material layer and whose horizontal projections at the interlayer interface are spatially adjacent to the corresponding channels in the first channel group are selected as candidate reference channels. A pre-detection sequence is sent to the candidate reference channel to filter out channels whose response drift direction is consistent with the overall thermal expansion trend of the package, thus forming the second channel group.

4. The method for converging bus interface data deviation of a co-packaged optical module according to claim 3, characterized in that, Sending the same probe sequence to the first channel group and the second channel group includes: The detection sequence is arranged to include a rising edge dominant segment and a falling edge dominant segment. The rising edge dominant segment is composed of consecutive alternating transitions from a first logic level to a second logic level, and the falling edge dominant segment is composed of consecutive alternating transitions from a second logic level to a first logic level. Both segments have the same symbol length and bit timing period. The detection sequence is sent to the first channel group and the second channel group, and the response timing of each channel group to the rising edge dominant segment and the falling edge dominant segment is recorded respectively to construct the response difference pattern.

5. The method for bus interface data deviation convergence of a co-packaged optical module according to claim 4, characterized in that, The method for identifying the relative slip direction between different material layers based on the response difference patterns of two sets of channels includes: During the temperature cycling process of the package, the detection sequence is sent to the first channel group and the second channel group at the heating phase, the holding phase and the cooling phase respectively, and the response timing offset of the two channels at each phase is recorded to construct the phase-offset difference mode. Extract the response timing offset under the heating phase and the cooling phase, compare the directional polarity of the two, and when the directional polarity is opposite and symmetrical and the response timing offset under the heat preservation phase is significantly reduced relative to the heating phase and the cooling phase, determine the offset direction under the heating phase as the preliminary result of the relative sliding direction.

6. The method for converging bus interface data deviation of a co-packaged optical module according to claim 5, characterized in that, After determining the preliminary judgment result, the following is also included: Calculate the difference in response timing offset of each channel in the first channel group under the heating phase and the cooling phase, and construct an offset difference sequence; The depth of the physical traces of each channel in the first channel group crossing the material layer interface is associated with the offset difference sequence one by one according to the channel to construct a depth-difference response sequence. Determine the monotonicity of the depth-difference response sequence: when the sequence monotonically increases with increasing crossing depth, confirm the initial judgment result; when the sequence monotonically decreases with increasing crossing depth, reverse the initial judgment result to the relative slip direction.

7. The method for bus interface data deviation convergence of a co-packaged optical module according to claim 6, characterized in that, Applying a compensation offset opposite to the relative sliding direction at the bus interface transmitting end to the first channel group includes: From the response difference pattern, extract the response timing offset of the first channel group to the rising edge dominant segment and the falling edge dominant segment, and calculate the difference between the two as the edge asymmetric differentiation amount. Based on the edge asymmetry differentiation amount, the compensation offset is decomposed into rising edge compensation component and falling edge compensation component. The rising edge compensation component and falling edge compensation component are opposite to the relative slip direction, and their absolute values ​​are differentiated according to the magnitude of the edge asymmetry differentiation amount. At the bus interface transmitting end, the data transmission timing of each channel in the first channel group is loaded according to the effective logic transition type: the rising edge compensation component is loaded for logic transitions with the same polarity as the rising edge dominant segment, and the falling edge compensation component is loaded for logic transitions with the same polarity as the falling edge dominant segment.

8. The method for bus interface data deviation convergence of a co-packaged optical module according to claim 7, characterized in that, The continuous monitoring of the response difference patterns between the first channel group and the second channel group includes: In each monitoring cycle, the response offset polarity of each channel in the first channel group relative to the corresponding reference channel in the second channel group is extracted to construct a polarity sequence; Calculate the polarity consistency index of the polarity sequence, which is used to characterize the degree to which the response offset polarity of each channel in the first channel group points in the same direction; When the polarity consistency index decreases from above the preset consistency threshold to below the preset consistency threshold, it is determined that the difference pattern changes from systematic differentiation to irregular dispersion.

9. The method for converging bus interface data deviation of a co-packaged optical module according to claim 8, characterized in that, When the difference pattern changes from systematic differentiation to irregular discreteness, maintaining the current compensation offset includes: After determining that the difference pattern has turned into irregular discreteness, the current compensation offset is used as a candidate maintenance value, and the verification maintenance period is started. During the verification maintenance period, the probe sequence is sent to the first channel group and the second channel group with the candidate maintenance value as a fixed compensation amount to continuously monitor the polarity consistency index; When the polarity consistency index remains below the preset consistency threshold during the verification maintenance period, the candidate maintenance value is confirmed as a valid maintenance value. Based on the edge asymmetry differentiation amount, maintenance tolerances are set for the rising edge compensation component and the falling edge compensation component, respectively. The maintenance tolerance is related to the response fluctuation amplitude of the corresponding edge component during the verification maintenance period. In subsequent monitoring, when the fluctuations of the rising edge compensation component and the falling edge compensation component are both within the corresponding maintenance tolerance, the effective maintenance value is maintained.

10. A bus interface data deviation convergence system for a co-packaged optical module, characterized in that, include: The channel grouping selection module is used to select, at the bus interface layer, channels that physical traces cross the interface of different material layers as the first channel group, and channels that physical traces are completely located within the same material layer as the second channel group. A probe sequence sending module is used to send the same probe sequence to the first channel group and the second channel group; The response acquisition and interlayer slip recognition module is used to acquire the responses of the first channel group and the second channel group to the detection sequence, and to identify the relative slip direction between different material layers based on the response difference patterns of the two channels. The compensation offset loading module is used to apply a compensation offset to the first channel group at the bus interface transmitting end, which is opposite to the relative sliding direction. The monitoring and maintenance module is used to continuously monitor the response difference pattern between the first channel group and the second channel group. When the difference pattern changes from systematic differentiation to irregular dispersion, the current compensation offset is maintained.

Images