{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","cell_id":"activation_decay_measurement__computer_science","arm":"CONSTRAINED_HIGH","candidate_id":"activation_decay_measurement__computer_science__CONSTRAINED_HIGH","decision":"PROPOSAL","abstention_reason":null,"proposal":{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","cell_id":"activation_decay_measurement__computer_science","arm":"CONSTRAINED_HIGH","candidate_id":"activation_decay_measurement__computer_science__CONSTRAINED_HIGH","proposal_index":1,"version":0,"title":"Matched Charge-Decay Sentinel for DRAM Refresh Expiry","problem":"A DRAM bank stores bits as electrical charge that decays with time, temperature, leakage, and device variation. A fixed refresh interval can become stale relative to the actual retention margin: if charge reaches the sensing limit first, stored bits can be corrupted; if the interval is much shorter than necessary, refresh consumes avoidable energy and memory availability.","actors":["DRAM storage capacitors and access transistors","Matched sacrificial sentinel cells","Sense amplifiers and voltage comparators","Hardware refresh pulse generator","Memory hardware validation engineer"],"observable_state":"The observable state is elapsed time from charge injection to the point at which a sentinel cell's differential read voltage approaches a preset minimum sensing margin. Immediate sentinel voltage establishes A(0); destructive reads of parallel sentinels after staggered hardware delays provide A(t), while temperature and supply voltage define the tested physical context.","consequence":"Reliance on an expired charge state can produce incorrect memory reads. Refreshing without reference to measured decay can instead impose unnecessary electrical activity and block memory access. The candidate is intended to align refresh timing with a directly observed physical retention boundary.","affected_objective":"Maintain memory-state integrity and availability within an electrical-energy budget.","intervention":"Place a small bank of sacrificial one-transistor/one-capacitor sentinel cells beside each protected memory region. A normal write or refresh pulse charges both the data cells and sentinels, but the sentinels are then isolated from refresh. Hardware delay taps destructively probe separate sentinels at increasing intervals, and analog comparators test their remaining differential voltage against a guarded minimum sense margin. When the threshold is approached, a hardwired pulse circuit electrically refreshes the data region and recharges the sentinel bank. If the sentinel indication is absent, inconsistent, or outside its calibrated envelope, the circuit falls back to a conservative fixed refresh oscillator rather than treating the charge state as current.","structural_mapping":[{"archetype_element":"Activation Target Definition","domain_realization":"The target activation is the stored electrical charge differential required for a DRAM sense amplifier to distinguish a one from a zero."},{"archetype_element":"Baseline Activation Capture","domain_realization":"A sample-and-hold measurement records sentinel differential voltage immediately after charge injection."},{"archetype_element":"Observable Activation Proxy","domain_realization":"Voltage and destructive-read margin in physically matched sacrificial cells proxy the otherwise disruptive measurement of data-cell charge."},{"archetype_element":"Delayed Probe Schedule","domain_realization":"Parallel sentinels are read through fixed hardware delay taps spanning the candidate retention interval."},{"archetype_element":"Decay Curve Model","domain_realization":"The ordered pattern of comparator crossings forms a piecewise physical retention profile; bench traces establish whether that profile bounds data-cell decay."},{"archetype_element":"Usable Activation Threshold","domain_realization":"A comparator reference represents the minimum guarded voltage margin allowed before ordinary sensing becomes unreliable."},{"archetype_element":"Intervention Window Rule","domain_realization":"Refresh is electrically initiated before the sentinel reaches the guarded comparator threshold."},{"archetype_element":"Refresh or Reprime Protocol","domain_realization":"A wordline refresh pulse restores data-cell charge and simultaneously recharges fresh sentinel cells."},{"archetype_element":"Context and Load Register","domain_realization":"Qualification repeats physical measurements across bounded temperature, supply-voltage, access-load, and aging conditions rather than assuming one retention window transfers unchanged."},{"archetype_element":"Expiry Boundary and Stop Rule","domain_realization":"A threshold crossing, missing sentinel response, or out-of-envelope condition invalidates reliance on the measured window and selects the conservative oscillator."},{"archetype_element":"Activation Half-Life Estimator","domain_realization":"The delay tap at which the sentinel falls through half of its initial measured margin supplies a hardware-observable decay-time estimate."},{"archetype_element":"Cohort-Specific Decay Profile","domain_realization":"Sentinel banks are matched to bounded physical regions rather than applying one measured decay profile to the entire device."},{"archetype_element":"Competing Cue Noise Monitor","domain_realization":"Reference capacitors and differential sensing expose supply noise, read disturbance, and comparator drift that could masquerade as leakage decay."},{"archetype_element":"Ethical Timing Guardrail","domain_realization":"The physical analogue is a safety guard band and mandatory conservative fallback; no adaptive threshold may be used to conceal loss of data-retention margin."}],"mechanism_mapping":[{"mechanism_slug":"time-lagged_activation_probe","role":"Staggered destructive reads measure sentinel charge at increasing physical delays after charging.","counterfactual_removal":"Without delayed electrical probes, the device observes only initial charge and cannot distinguish fresh activation from later decay."},{"mechanism_slug":"decay_curve_fitting","role":"Comparator crossings and bench voltage traces establish a bounded piecewise relation between elapsed time and retained charge margin.","counterfactual_removal":"Without relating voltage margin to elapsed time, no measured refresh window can be derived from the sentinel."},{"mechanism_slug":"activation_window_thresholding","role":"An analog comparator converts the decaying voltage margin into a physical expiry boundary.","counterfactual_removal":"Without the threshold, the sentinel provides a trace but no defensible boundary before sensing margin is exhausted."},{"mechanism_slug":"refresh_cadence_adaptation","role":"The next hardware refresh pulse occurs when the matched sentinel approaches its guarded expiry boundary.","counterfactual_removal":"Without cadence response, measurement would remain diagnostic and would not preserve the data-cell charge state."},{"mechanism_slug":"contextual_reactivation","role":"The refresh pulse restores charge in the same bank and electrical context whose sentinel decayed.","counterfactual_removal":"Refreshing an unrelated region would not restore the threatened charge state."},{"mechanism_slug":"decay_segment_comparison","role":"Multiple delay taps distinguish early rapid loss, stable retention, and late threshold approach without requiring software inference.","counterfactual_removal":"A single endpoint could not reveal whether a misleading transient or comparator disturbance caused the apparent boundary."},{"mechanism_slug":"staleness_boundary_enforcement","role":"Threshold, fault, or calibration-envelope violations physically select conservative refresh timing.","counterfactual_removal":"Without an enforced fallback, expired or invalid sentinel evidence could continue to govern retention."}],"causal_chain":["A write or refresh pulse deposits electrical charge representing data and initializes matched sacrificial sentinels.","Leakage physically reduces charge in both data and sentinel capacitors over elapsed time.","Staggered sense operations expose the sentinel's declining differential voltage without disturbing stored data.","Analog comparison identifies when the measured proxy approaches a guarded minimum usable margin.","A hardwired pulse path refreshes the associated data region before the measured expiry boundary.","Refresh replenishes the data-cell charge and starts a new sentinel decay interval.","A missing, inconsistent, or out-of-envelope measurement transfers control to a conservative fixed oscillator."],"baseline":"A fixed hardware refresh interval selected for an assumed worst-case retention condition, optionally shortened by a temperature proxy, without direct observation of local stored-charge decay.","nearest_rivals":["A uniformly shorter fixed refresh interval, which protects margin by spending more refresh activity rather than measuring decay.","Temperature-compensated refresh, which measures a correlate of leakage rather than retained charge itself.","One-time retention characterization and device binning, which measures decay during qualification but does not observe subsequent physical drift.","Error-correcting memory and scrubbing, which repair or detect some consequences after charge loss rather than setting the pre-loss refresh boundary."],"remaining_contrastive_claim":"Relative to a fixed timer or temperature proxy, the candidate's causal distinction is direct, local measurement of a matched charge-decay state followed by electrical replenishment before its guarded threshold. This is a testable mechanism claim, not a claim of novelty or superior effect.","authority_safety":{"decision_authority":"The memory hardware validation lead may authorize only a nonproduction test coupon and may approve later integration only after the sentinel is shown to conservatively bound data-cell retention across the declared operating envelope.","authorized_first_step":"Construct and test one isolated coupon containing data cells, matched sentinel cells, analog comparators, fixed delay taps, and a conservative fallback oscillator.","excluded_actions":["Reducing an existing production refresh rate","Using the circuit for sole protection of irreplaceable data","Disabling error detection or conservative fallback","Extrapolating one coupon's calibration to other processes, layouts, temperatures, voltages, or aging states","Allowing software to override the guarded minimum threshold during the first test"],"halt_rollback":"Halt if any sentinel crosses later than the earliest protected data cell, if probing disturbs adjacent data, or if comparator/reference faults fail to select fallback. Roll back by electrically disconnecting the sentinel trigger and operating the coupon solely from the fixed conservative oscillator."},"negative_tests":{"strongest_counterevidence":"Matched sentinels may not represent the weakest data cells; layout, process tails, access history, temperature gradients, or aging could make protected cells decay faster than the sentinel.","problem_falsifier":"The stale-window problem is falsified for the scoped device if direct retention measurements show that every tested data cell preserves the required sensing margin beyond the fixed baseline interval throughout the declared operating envelope.","intervention_falsifier":"The intervention is falsified if the guarded sentinel crossing fails to precede the earliest loss of required margin in any tested data cell, lacks a stable relation to data-cell decay across contexts, or if its refresh pulse does not restore the required charge margin.","risks":["A slow-decaying sentinel could create false confidence and data corruption.","A fast-decaying sentinel could cause excessive refresh activity.","Destructive sentinel reads or switching transients could disturb nearby data cells.","Temperature gradients, voltage noise, fabrication variation, and aging could invalidate physical matching.","Comparator or reference drift could shift the expiry boundary.","Additional cells, routing, and refresh pulses could consume area, energy, or memory availability.","Repeated probing could interact with access-pattern-dependent disturbance mechanisms."]},"next_evidence_step":"On one nonproduction coupon, charge 64 data cells and 8 adjacent sentinels with alternating bit patterns; measure immediate voltage margin and delayed margin at fixed taps under three bounded temperatures and two supply points, both idle and under a fixed access load. Predefine success as every guarded sentinel crossing occurring before the earliest data-cell margin failure, correct fallback under injected open-sentinel and comparator faults, and demonstrated charge restoration by the resulting refresh pulse. Do not change any production refresh setting from this test.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Not assessed against other proposals because runtime isolation forbids inspecting them; this candidate is derived solely from the supplied archetype and computer-science domain card.","revision_record":{"parent_version":null,"progress_targets_addressed":["Construct a complete physical-substrate candidate","Preserve activation-decay measurement, threshold, refresh, and expiry structure","Make the essential effect independent of software, analytics, reporting, incentives, authorization, and procedural enforcement"],"conceptual_changes":["Initial version maps transient cognitive activation to transient electrical charge in volatile memory.","The action window becomes a guarded retention interval and reprime becomes electrical refresh."],"operational_changes":["Specified matched sacrificial cells, staggered hardware probes, analog thresholding, direct refresh pulses, and conservative fallback.","Bounded the first test to a nonproduction coupon."],"evidence_changes":["No external evidence or prior-art search was used.","Added explicit physical falsifiers and a coupon-level measurement step."],"claim_changes":["Omitted novelty, prevalence, demand, and effect-size claims.","Limited the contrastive claim to the proposed causal mechanism."]},"substrate_contract":{"primary_allowed_process":"MEASUREMENT_INSTRUMENTATION","counterfactual_independence":"Removing all software, algorithmic inference, databases, dashboards, reporting, incentives, authorization rules, and procedural enforcement leaves the essential effect intact: capacitor leakage changes a directly sensed voltage; analog comparators identify a physical threshold; and a hardwired electrical pulse replenishes charge. Human authority governs whether the prototype may be deployed but does not produce its retention effect.","forbidden_channel_audit":"No software model, analytics pipeline, information-routing system, or human report determines the refresh event. The sensor does not merely generate advice: its analog output directly gates an energetic charge-restoration pulse. Calibration records and safety authorization are optional support wrappers; deleting them would reduce assurance but would not remove the physical sense-threshold-refresh causal chain."}}}