A high-fidelity simulation method for aging faults using scan chain multiplexing and trigger sampling
By establishing a mapping relationship between the scan chain and gate-level flip-flops in the integrated circuit, inserting activity detection logic, and performing fault injection and status reading, the shortcomings of the DFT scan chain in aging fault assessment are solved, and high-precision fault simulation and reliability assessment are achieved.
Patent Information
- Application Number
- CN202611093345.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-22
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies make it difficult to effectively reuse the DFT scan chain for fault localization and internal state observation during the aging process of integrated circuits, resulting in insufficient accuracy of aging fault assessment and functional safety verification.
By establishing a mapping relationship between scan chain bits and gate-level triggers, activity detection logic is inserted, flip events are collected and compressed trajectory data is generated, and the scan chain is used for fault injection and status reading, combined with a windowed adaptive downsampling mechanism and fault classification.
It improves the accuracy and efficiency of aging failure simulation, can accurately identify internal state contamination and quiescent data corruption, and enhances the completeness of reliability assessment and functional safety verification in the design phase.
Smart Images

Figure CN123070028A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of integrated circuit aging fault simulation technology, and in particular to a high-fidelity simulation method for aging faults using scan chain multiplexing and trigger sampling. Background Technology
[0002] As integrated circuit manufacturing processes continue to evolve to 7nm and below, aging effects such as NBTI and HCI lead to transistor performance degradation and increased critical path delays, posing a severe challenge to the long-term reliability of safety-critical chips such as automotive-grade SoCs and processors. Among related technologies, high-precision aging reliability assessment mainly relies on two schemes: software gate-level fault simulation and FPGA hardware-assisted fault injection. Specifically, software gate-level simulation, by specifying the target node, fault type, and injection cycle in the gate-level netlist, can provide fine spatial and temporal fault coordinates, but it has high execution overhead and is difficult to support rapid iteration of large-scale designs and long program loads. FPGA hardware-assisted methods achieve near-hardware speed fault injection by configuring memory flip-flops or customizing debug logic; however, its injection location usually lacks a definite mapping with the original gate-level netlist, and the observation range is limited to the main output or a small number of debug signals.
[0003] However, while existing DFT scan chains inherently possess the ability to move internal states in, out, and locate them, they are primarily used for production testing in traditional processes and have not been effectively reused as a hardware interface for aging fault injection and internal state observation. Consequently, existing solutions struggle to balance hardware simulation efficiency, gate-level fault location accuracy, and internal state observability. This results in an inability to fully capture potential failure modes such as internal state contamination and silent data corruption caused by aging faults, thus impacting the accuracy of reliability assessments and the completeness of functional safety verification during the design phase. Summary of the Invention
[0004] The present invention aims to at least partially solve one of the technical problems in the related art.
[0005] Therefore, the first objective of this invention is to propose a high-fidelity simulation method for aging faults using scan chain multiplexing and trigger sampling.
[0006] Another objective of this invention is to provide a high-fidelity simulation device for aging faults using scan chain multiplexing and trigger sampling.
[0007] The third objective of this invention is to provide a computer device.
[0008] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.
[0009] To achieve the above objectives, a first aspect of the present invention proposes a high-fidelity simulation method for aging faults using scan chain multiplexing and trigger sampling, comprising:
[0010] S1. Based on the gate-level netlist, static timing analysis report and scan chain configuration file of the design under test, establish a one-to-one mapping relationship between scan chain bits and gate-level triggers, and extract scan units located on the critical path as candidate monitoring units. S2, insert activity detection logic into the candidate monitoring unit, collect the flip events of each monitoring unit within the sampling window during the simulation operation, and generate compressed flip trajectory data through the windowed adaptive downsampling mechanism; S3, reconstruct the activity density of each monitoring unit within the time window based on the flip trajectory data, and generate a pruned fault injection configuration, which includes the target scanning unit, target period and fault type; S4. When the target injection cycle arrives, the multiplexed scan chain pauses the function clock and switches to scan mode. The current register snapshot is read out through a shift operation. The corresponding scan bit is modified according to the fault injection configuration to simulate aging fault. The modified scan frame is then written back through a shift operation. After the function clock is restored, the operation resumes. After the fault propagates, the internal register state is read out again through the scan chain and compared with the golden execution trajectory to classify the fault.
[0011] In one embodiment of the present invention, S1 includes: Information is extracted from DFT specifications, gate-level netlists, static timing analysis reports, and scan chain configuration files to complete scan chain parsing and critical path association, generating a scan chain index table containing scan chain number, scan index, trigger instance name, clock domain, timing margin, critical path marker, and monitoring enable marker. Based on the timing margins in the static timing analysis report, sort them from smallest to largest, extract the endpoint triggers or affected scan units from the critical path as candidate monitoring units, and filter and force them in batches according to a preset ratio T% or a specified module level path.
[0012] In one embodiment of the present invention, it further includes: By parsing the scan chain configuration in the DFT specification, the scan chain number and scan index are associated with the trigger instance name, hierarchical path, and input / output port name in the gate-level netlist; The scan chain index table is generated by combining the timing margin and critical path markers in the static timing analysis report. The scan chain index table also includes a monitoring enable flag field, which identifies whether the scan unit is selected as a candidate monitoring unit.
[0013] In one embodiment of the present invention, S2 includes: For each candidate monitoring unit, an activity detection logic consisting of the original flip-flop output, a shadow flip-flop, and an XOR gate is inserted. After each functional clock edge, the current cycle flip-flop value is compared with the previous cycle flip-flop value through an XOR gate to generate a toggle signal. The simulation time is divided into sampling windows of length W, and each window is set with a sampling resolution K. The first K flip events are accurately recorded, and then the sampling is downsampled step by step at a doubly increasing interval to generate a record item containing the cell identifier, the event period within the window, and the sampling level as compressed flip trajectory data.
[0014] In one embodiment of the present invention, it further includes: The window length W and sampling resolution K are both powers of 2, i.e., W = 2^w, K = 2^k, and satisfy the following conditions: ; Within each sampling window, the sampling tier for the first K events is 0. After that, the tier increases by 1 for every K events filled. The corresponding sampling interval increases by a factor of 2. The record is represented as (cell_id, cycle, tier), where cell_id is determined by the scan index and cycle represents the event occurrence cycle within the window.
[0015] In one embodiment of the present invention, S3 includes: Based on the sampling level of each recorded event in the flip trajectory data, and according to the weight... To compensate for events discarded due to downsampling, the activity estimate is obtained by summing the weighted number of events within the same scanning unit in the window. The activity estimate is converted into an injection probability by combining a global scaling factor Γ. Generate a fault injection configuration table that includes the target scanning unit, target cycle, fault type, observation time, and number of repetitions.
[0016] In one embodiment of the present invention, S4 includes: The following operations are performed via the state machine in the fault injection controller: after waiting for the target cycle to arrive, pause the function clock and switch scan enable, and perform a Shift-Out operation to read the current register snapshot; Based on the position of the target scanning unit in the scanning chain in the fault injection configuration, the corresponding scanning bit is modified by bit flipping or writing the old value of the previous cycle, fixing 0, fixing 1 or random parsing value; Shift-In operation writes the modified scan frame back to the design under test, restores the functional clock, and continues running. After observing the fault propagation for a period of time, the internal register status is read again through the scan chain and compared with the golden execution trajectory by period, scan index and main output port. The fault is classified as being logic masked, internal state contamination, silent data corruption, main output error or system abnormality.
[0017] To achieve the above objectives, a second aspect of the present invention provides a high-fidelity simulation device for aging faults using scan chain multiplexing and trigger sampling, comprising: The scanning unit mapping and filtering module is used to establish a one-to-one mapping relationship between scanning chain bits and gate-level triggers based on the gate-level netlist, static timing analysis report and scanning chain configuration file of the design under test, and to extract scanning units located on the critical path as candidate monitoring units. The unit activity sampling compression module is used to insert activity detection logic into the candidate monitoring unit, collect the flip events of each monitoring unit within the sampling window during the simulation operation, and generate compressed flip trajectory data through a windowed adaptive downsampling mechanism. The fault injection configuration generation module is used to reconstruct the activity density of each monitoring unit within the time window based on the flip trajectory data, and generate the pruned fault injection configuration, which includes the target scanning unit, target period and fault type. The scan chain fault injection and discrimination module is used to reuse the scan chain to pause the function clock and switch to scan mode when the target injection cycle arrives. It reads the current register snapshot through shift operation, modifies the corresponding scan bit according to the fault injection configuration to simulate aging fault, writes back the modified scan frame through shift operation, resumes the function clock and runs, and after the fault propagates, reads the internal register state through the scan chain again and compares it with the golden execution trajectory to classify the fault.
[0018] This invention discloses a high-fidelity simulation method and apparatus for aging faults using scan chain multiplexing and trigger sampling. It reuses the original scan chain of the chip to complete aging fault injection and state acquisition, reduces hardware design costs, accurately targets critical path units for fault simulation, and improves the efficiency and coverage of timing aging fault testing of integrated circuits.
[0019] To achieve the above objectives, a third aspect of this application provides a computer device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, for implementing the method described in the first aspect embodiment.
[0020] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.
[0021] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0022] Figure 1 This is a flowchart of an aging fault high-fidelity simulation method based on scan chain multiplexing and trigger sampling according to an embodiment of the present invention; Figure 2 This is a system overall framework diagram according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the extraction of key scanning units according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the activity detection logic structure according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the sampling trajectory and activity density according to an embodiment of the present invention; Figure 6 This is a schematic diagram of a scan chain fault injection structure according to an embodiment of the present invention; Figure 7 This is a system operation flowchart according to an embodiment of the present invention; Figure 8 This is a flowchart of the fault injection process according to an embodiment of the present invention; Figure 9 This is a result analysis diagram according to an embodiment of the present invention; Figure 10 This is a structural diagram of an aging fault high-fidelity simulation device based on scan chain multiplexing and trigger sampling according to an embodiment of the present invention; Figure 11 It is a computer device according to an embodiment of the present invention. Detailed Implementation
[0023] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0025] The following description, with reference to the accompanying drawings, describes a high-fidelity simulation method and apparatus for aging faults based on scan chain multiplexing and trigger sampling according to an embodiment of the present invention.
[0026] Figure 1 This is a flowchart of a high-fidelity simulation method for aging faults based on scan chain multiplexing and trigger sampling according to an embodiment of the present invention, as follows: Figure 1 As shown, it includes: S1. Based on the gate-level netlist, static timing analysis report and scan chain configuration file of the design under test, establish a one-to-one mapping relationship between scan chain bits and gate-level triggers, and extract scan units located on the critical path as candidate monitoring units. S2, insert activity detection logic into the candidate monitoring unit, collect the flip events of each monitoring unit within the sampling window during the simulation operation, and generate compressed flip trajectory data through the windowed adaptive downsampling mechanism; S3, reconstruct the activity density of each monitoring unit within the time window based on the flip trajectory data, and generate a pruned fault injection configuration, which includes the target scanning unit, target period and fault type; S4. When the target injection cycle arrives, the multiplexed scan chain pauses the function clock and switches to scan mode. The current register snapshot is read out through a shift operation. The corresponding scan bit is modified according to the fault injection configuration to simulate aging fault. The modified scan frame is then written back through a shift operation. After the function clock is restored, the operation resumes. After the fault propagates, the internal register state is read out again through the scan chain and compared with the golden execution trajectory to classify the fault.
[0027] This invention proposes a high-fidelity simulation system for aging faults using scan chain multiplexing and trigger sampling, such as... Figure 2 As shown, this system is based on a gate-level netlist with a DFT scan chain, and combines static timing analysis results with test loads to achieve rapid and high-precision assessment of aging delay faults on an FPGA or hardware simulation platform. The overall system process is divided into two stages: the first stage performs activity-aware adaptive sampling, and the second stage performs scan chain multiplexing fault injection simulation.
[0028] Furthermore, during the netlist preparation phase, the system first inserts a scan chain into the design under test according to the DFT specifications and saves the scan chain configuration file. This configuration file includes at least the scan chain number, the scan index of the scan trigger within the chain, the trigger instance name, the hierarchical path, the input / output port names, the clock domain, and the correspondence with the critical path. The system associates this configuration file with the gate-level netlist and the STA report to form a scan chain index table. The scan chain index table is used to ensure that the scan bits in subsequent hardware simulations can be determined to correspond to the gate-level triggers, rather than being indirectly inferred through FPGA configuration bits.
[0029] Furthermore, such as Figure 3 As shown, the scan chain index mapping first extracts information from the DFT specification, gate-level netlist, STA report, and scan chain configuration file, and then completes scan chain parsing and critical path association respectively. The generated scan chain index table contains fields such as scan_chain_id, scan_index, FF instance, clock domain, timing margin, critical path marker, and monitoring enable marker. Subsequently, the hardware simulation platform only needs to locate the scan bit based on scan_chain_id and scan_index, and then look up the original gate-level trigger instance from the index table, thereby establishing a one-to-one correspondence between hardware scan bits and gate-level fault locations.
[0030] Furthermore, the STA report allows for the selection of critical paths based on timing margins from smallest to largest, and the extraction of critical path endpoint triggers or affected scan units along the path as candidate monitoring units. For designs of different scales, candidate monitoring units can be processed in batches according to a percentage (T%); for safety-critical modules requiring higher coverage, module-level paths can be specified for forced inclusion. This screening step aims to limit the spatial scope of aging delay faults to timing-sensitive areas, avoiding indiscriminate monitoring of all registers.
[0031] Further, in the first stage, the system first reads the HDL design, standard cell library, design constraints, and DFT specifications to generate a gate-level netlist with scan chains. Based on the STA results, it extracts scan cells located on the critical path or near timing violation boundaries. For selected scan cells, the system inserts lightweight XOR gates and shadow triggers, comparing the trigger values of the current cycle and the previous cycle to detect whether the scan cell has flipped in the current cycle. Since aging-induced delay faults typically only manifest as sampling errors when the path flips, this module can directly filter out highly relevant fault time windows under real workloads.
[0032] Furthermore, such as Figure 4 As shown, the activity detection logic consists of the original flip-flop output Q, the shadow flip-flop shadow_Q, and an XOR gate. After each functional clock edge, shadow_Q stores the Q value of the previous cycle, and XOR(Q, shadow_Q) outputs a toggle signal. The toggle signal, along with scan_index, the current cycle counter, and the window number, is written to the sampling buffer. For scan units not in the critical path set, the system does not insert this monitoring logic, thereby reducing area overhead. For cross-clock domain designs, each clock domain maintains its own cycle count or is aligned using a unified timestamp.
[0033] Furthermore, to reduce the storage overhead of flip trajectories, the system employs a windowed adaptive sampling mechanism. The simulation time is divided into sampling windows of length W, with each window having a sampling resolution of K. At the beginning of each window, the first K flip events are recorded precisely; as the number of flips continues to increase, the sampling density is gradually reduced, for example, recording once every 2, 4, and 8 events. This mechanism can retain periodic-level accuracy for control signals with low activity but potentially high safety impacts, while compressing the recording of high-activity data paths.
[0034] Furthermore, both the window length W and the sampling resolution K are powers of 2, i.e., W = 2^w and K = 2^k, so that the hardware can determine the level through shifting and counters. The system preferably satisfies K^2 ≥ W, ensuring that the sampling interval remains within a controllable range even in the worst-case scenario. Each record item within a sampling window can be represented as (cell_id, cycle, tier), where cell_id is determined by scan_index, cycle represents the event occurrence cycle within the window, and tier represents the sampling level of the event. The tier for the first K events is 0; thereafter, tier increments by 1 for every K events filled, corresponding to a doubling of the sampling interval.
[0035] Furthermore, such as Figure 5 As shown, after sampling, the host computer assigns weights to the recorded events according to different sampling levels, reconstructs the activity density of each scanning unit in different time windows, and generates a pruned fault injection configuration. The fault injection configuration includes information such as the target scanning unit, target period, fault type, observation time, and number of repetitions. In this way, subsequent fault injection is concentrated on the critical path and the true flip window, reducing the number of invalid experiments.
[0036] Furthermore, during the host computer reconstruction, for the i-th recorded event, a weight w(i) can be assigned according to its sampling level, preferably w(i) = 2^tier, to compensate for events dropped by downsampling on the hardware side. The system sums the weighted event counts of the same scanning unit within the window to obtain the activity estimate within that window, and combines it with the global scaling factor Γ to generate the injection probability P(FI). The higher P(FI), the more likely the scanning unit is to expose aging delay faults within that window. Combined with the global scaling factor Γ, the injection probability is obtained, and finally, a fault injection configuration table that can be directly read by the FI controller is generated.
[0037] Furthermore, in the second stage, the system reuses the DFT scan chain as both a fault injection interface and an internal state observation interface, such as... Figure 6As shown, when the target injection cycle arrives, the controller pauses the function clock and switches to scan mode, reading the current register snapshot via Shift-Out; then, it modifies the corresponding scan bit according to the position of the target scan unit in the scan chain, for example, flipping the bit to simulate a late aging fault; then, it writes the modified scan frame back to the DUT via Shift-In, resumes operation after the function clock is restored, and allows the fault to propagate in subsequent logic cones.
[0038] Furthermore, scan chain multiplexing fault injection is performed by a state machine in the FI controller. This state machine includes states such as waiting for the target cycle, pausing the function clock, switching scan enable, Shift-Out reading the scan frame, modifying the target bit, Shift-In writing back the scan frame, resuming the function clock, and observing the propagation results. The position of the target scan bit is determined by scan_index. If the same design contains multiple scan chains, it is located by both scan_chain_id and scan_index. When modifying the target bit, bit flipping can be performed, or the old value from the previous cycle can be written, fixed at 0, fixed at 1, or a random parsed value can be written according to the aging fault model.
[0039] Furthermore, after observing the fault propagation for a period of time, the system can again read the internal register state through the scan chain and compare it with the golden execution trajectory. Based on the comparison results, the system classifies the fault into types such as logic masking, internal state contamination, silent data corruption, main output error, or system anomaly. Compared to black-box methods that only monitor the main output, this invention can discover potential internal errors that are difficult to identify using traditional hardware fault injection methods.
[0040] Furthermore, the golden execution trajectory can be pre-collected during fault-free operation or provided by a parallel Golden DUT. The comparison process uses cycles as the time alignment unit, scan_index as the internal state alignment unit, and the main output port or program checkpoint as the external result alignment unit. If the internal snapshot during faulty operation differs from the golden snapshot but the main output remains the same, it indicates that the fault has entered the architectural state but has not yet become apparent; if the main output or program checkpoint differs, it indicates that the fault has propagated into a visible error. This classification result is used to generate a list of fault-sensitive scan cells and hardening priorities.
[0041] Furthermore, this invention primarily addresses aging-induced delay faults, supporting models such as late-arrival faults, early-sampling faults, and metastable-induced faults. In a preferred embodiment, late-arrival fault is used as the primary model, where the target trigger samples the old value from the previous cycle after the critical path delay exceeds a clock cycle. This model is directly related to timing degradation caused by aging and is suitable for early chip reliability assessment and functional safety verification.
[0042] Furthermore, such as Figure 7As shown, the system first receives the HDL design, standard cell library, design constraints, and DFT specifications. After DFT netlist synthesis and STA analysis, a gate-level netlist with scan chains and a set of key scan cells are obtained. The system does not directly perform fault traversal on all triggers and all cycles. Instead, in Phase 1, it first collects the workload-related activities of the key scan cells, and then the host computer reconstructs the activity density and generates a pruned fault injection configuration. In Phase 2, the scan chain is reused as a bidirectional interface that can write faults and read internal states. Finally, a fault classification and reliability report are generated through the state comparison module.
[0043] Furthermore, after obtaining the pruned fault injection configuration, the present invention enters stage two. Stage two does not rely on indirect inference between FPGA configuration bits and gate-level flip-flops, but directly utilizes the Shift-Out and Shift-In capabilities of the DFT scan chain to complete accurate fault injection. The following figure shows the scan chain multiplexing fault injection process.
[0044] Furthermore, such as Figure 8 As shown, the FI controller pauses the functional clock during the target injection cycle to stabilize the DUT architecture state; it then switches to scan mode and reads the pre-injection register snapshot via Shift-Out; the controller locates the target trigger based on the scan_index and modifies the corresponding scan bit; the modified scan frame is written back to the DUT via Shift-In, achieving gate-level trigger-level fixed-point fault injection; finally, the functional clock is restored, allowing the fault to propagate in subsequent logic cones. This Pause-Shift-Resume protocol enables the scan chain to simultaneously serve as both a fault injection interface and an internal state probe interface.
[0045] Furthermore, after fault injection is complete, simply observing the main output cannot cover all fault effects. For example, some faults may have already corrupted internal registers but have not yet propagated to the main output; some faults may cause silent data corruption, while the external interface may still function normally for a short period. Therefore, this invention rereads the internal snapshot after the observation window ends and analyzes the results according to the flowchart below.
[0046] Furthermore, such as Figure 9 As shown, the system aligns the faulty main output, faulty internal snapshot, golden main output, and golden internal snapshot according to cycle, scan_index, and output port, and calculates the difference set. If there is no difference between the internal state and the main output, it indicates that the fault is masked by logical or timing behavior. If the system continues to run and the main output has not yet reported an error but the internal state is already unacceptable, it is classified as silent data corruption. If there are differences in the main output, abnormal signals, or program results, it is classified as visible output error. The analysis results are further used to generate a list of fault-sensitive scanning units, fault propagation paths, and hardening priorities.
[0047] The embodiments of the present invention also have the following technical effects: This invention reuses existing DFT scan chains to achieve targeted injection of aging faults and internal state reading, establishing a one-to-one correspondence between scan chain bits and gate-level flip-flops, thus improving the accuracy of hardware fault injection. This invention reduces invalid injection through key scan unit screening, activity-aware sampling, and Bayesian reconstruction, concentrating fault injection within the critical path window where flip-flops occur under real workloads, resulting in higher efficiency compared to full gate-level simulation and traditional hardware injection. This invention can read internal register snapshots before and after fault injection, discovering internal state contamination and silent data corruption that cannot be identified by main output monitoring alone, providing more accurate data for subsequent reliability assessment, functional safety verification, and circuit hardening.
[0048] To achieve the above embodiments, such as Figure 10 As shown, this embodiment also provides a high-fidelity simulation device 10 for aging faults using scan chain multiplexing and trigger sampling, comprising: The scanning unit mapping and filtering module 100 is used to establish a one-to-one mapping relationship between scanning chain bits and gate-level triggers based on the gate-level netlist, static timing analysis report and scanning chain configuration file of the design under test, and to extract scanning units located on the critical path as candidate monitoring units. The unit activity sampling compression module 200 is used to insert activity detection logic into the candidate monitoring unit, collect the flip events of each monitoring unit within the sampling window during the simulation operation, and generate compressed flip trajectory data through a windowed adaptive downsampling mechanism. The fault injection configuration generation module 300 is used to reconstruct the activity density of each monitoring unit within the time window based on the flip trajectory data, and generate a pruned fault injection configuration, which includes the target scanning unit, target period and fault type. The scan chain fault injection and discrimination module 400 is used to, when the target injection cycle arrives, reuse the scan chain to pause the function clock and switch to scan mode, read the current register snapshot through shift operation, modify the corresponding scan bit according to the fault injection configuration to simulate aging fault, write back the modified scan frame through shift operation, resume operation after the function clock is restored, and after the fault propagates, read the internal register state through the scan chain again and compare it with the golden execution trajectory to classify the fault.
[0049] This invention provides a high-fidelity simulation device for aging faults using scan chain multiplexing and trigger sampling. It reuses the original scan chain of the chip to complete aging fault injection and status acquisition, reducing hardware design costs, accurately simulating faults for critical path units, and improving the efficiency and coverage of timing aging fault testing for integrated circuits.
[0050] To implement the methods of the above embodiments, the present invention also provides a computer device, such as... Figure 11 As shown, the computer device 600 includes a memory 601 and a processor 602; wherein, the processor 602 reads executable program code stored in the memory 601 to run a program corresponding to the executable program code, so as to implement the various steps of the method described above.
[0051] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.
[0052] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0053] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
Claims
1. A high-fidelity simulation method for aging faults using scan chain multiplexing and trigger sampling, characterized in that, include: S1. Based on the gate-level netlist, static timing analysis report and scan chain configuration file of the design under test, establish a one-to-one mapping relationship between scan chain bits and gate-level triggers, and extract scan units located on the critical path as candidate monitoring units. S2, insert activity detection logic into the candidate monitoring unit, collect the flip events of each monitoring unit within the sampling window during the simulation operation, and generate compressed flip trajectory data through the windowed adaptive downsampling mechanism; S3, reconstruct the activity density of each monitoring unit within the time window based on the flip trajectory data, and generate a pruned fault injection configuration, which includes the target scanning unit, target period and fault type; S4. When the target injection cycle arrives, the multiplexed scan chain pauses the function clock and switches to scan mode. The current register snapshot is read out through a shift operation. The corresponding scan bit is modified according to the fault injection configuration to simulate aging fault. The modified scan frame is then written back through a shift operation. After the function clock is restored, the operation resumes. After the fault propagates, the internal register state is read out again through the scan chain and compared with the golden execution trajectory to classify the fault.
2. The method as described in claim 1, characterized in that, S1 includes: Information is extracted from DFT specifications, gate-level netlists, static timing analysis reports, and scan chain configuration files to complete scan chain parsing and critical path association, generating a scan chain index table containing scan chain number, scan index, trigger instance name, clock domain, timing margin, critical path marker, and monitoring enable marker. Based on the timing margins in the static timing analysis report, sort them from smallest to largest, extract the endpoint triggers or affected scan units from the critical path as candidate monitoring units, and filter and force them in batches according to a preset ratio T% or a specified module level path.
3. The method as described in claim 2, characterized in that, The method further includes: By parsing the scan chain configuration in the DFT specification, the scan chain number and scan index are associated with the trigger instance name, hierarchical path, and input / output port name in the gate-level netlist; The scan chain index table is generated by combining the timing margin and critical path markers in the static timing analysis report. The scan chain index table also includes a monitoring enable flag field, which identifies whether the scan unit is selected as a candidate monitoring unit.
4. The method as described in claim 1, characterized in that, The S2 includes: For each candidate monitoring unit, an activity detection logic consisting of the original flip-flop output, a shadow flip-flop, and an XOR gate is inserted. After each functional clock edge, the current cycle flip-flop value is compared with the previous cycle flip-flop value through an XOR gate to generate a toggle signal. The simulation time is divided into sampling windows of length W, and each window is set with a sampling resolution K. The first K flip events are accurately recorded, and then the sampling is downsampled step by step at a doubly increasing interval to generate a record item containing the cell identifier, the event period within the window, and the sampling level as compressed flip trajectory data.
5. The method as described in claim 4, characterized in that, The method further includes: The window length W and sampling resolution K are both powers of 2, i.e., W = 2^w, K = 2^k, and satisfy the following conditions: ; Within each sampling window, the sampling tier for the first K events is 0. After that, the tier increases by 1 for every K events filled. The corresponding sampling interval increases by a factor of 2. The record is represented as (cell_id, cycle, tier), where cell_id is determined by the scan index and cycle represents the event occurrence cycle within the window.
6. The method as described in claim 1, characterized in that, The S3 includes: Based on the sampling level of each recorded event in the flip trajectory data, and according to the weight... To compensate for events discarded due to downsampling, the activity estimate is obtained by summing the weighted number of events within the same scanning unit in the window. The activity estimate is converted into an injection probability by combining a global scaling factor Γ. Generate a fault injection configuration table that includes the target scanning unit, target cycle, fault type, observation time, and number of repetitions.
7. The method as described in claim 1, characterized in that, The S4 includes: The following operations are performed via the state machine in the fault injection controller: after waiting for the target cycle to arrive, pause the function clock and switch scan enable, and perform a Shift-Out operation to read the current register snapshot; Based on the position of the target scanning unit in the scanning chain in the fault injection configuration, the corresponding scanning bit is modified by bit flipping or writing the old value of the previous cycle, fixing 0, fixing 1 or random parsing value; Shift-In operation writes the modified scan frame back to the design under test, restores the functional clock, and continues running. After observing the fault propagation for a period of time, the internal register status is read again through the scan chain and compared with the golden execution trajectory by period, scan index and main output port. The fault is classified as being logic masked, internal state contamination, silent data corruption, main output error or system abnormality.
8. A high-fidelity simulation device for aging faults using scan chain multiplexing and trigger sampling, characterized in that, include: The scanning unit mapping and filtering module is used to establish a one-to-one mapping relationship between scanning chain bits and gate-level triggers based on the gate-level netlist, static timing analysis report and scanning chain configuration file of the design under test, and to extract scanning units located on the critical path as candidate monitoring units. The unit activity sampling compression module is used to insert activity detection logic into the candidate monitoring unit, collect the flip events of each monitoring unit within the sampling window during the simulation operation, and generate compressed flip trajectory data through a windowed adaptive downsampling mechanism. The fault injection configuration generation module is used to reconstruct the activity density of each monitoring unit within the time window based on the flip trajectory data, and generate the pruned fault injection configuration, which includes the target scanning unit, target period and fault type. The scan chain fault injection and discrimination module is used to reuse the scan chain to pause the function clock and switch to scan mode when the target injection cycle arrives. It reads the current register snapshot through shift operation, modifies the corresponding scan bit according to the fault injection configuration to simulate aging fault, writes back the modified scan frame through shift operation, resumes the function clock and runs, and after the fault propagates, reads the internal register state through the scan chain again and compares it with the golden execution trajectory to classify the fault.
9. A computer device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.
