Acquisition spacing.

Question and rationale

Does additional spacing between acquisitions change six-state median recognition, and where do elapsed intervals occur within the ADC observation sequence?

The second ordered timing return recognized 863/864 candidate medians. At an unchanged state-1 request, readings 2880, 2847 and 2880 produced a rejected median in the gap above its 2830–2869 acceptance interval. The next triple was correct; its ordered end followed 104.913944 nominal ms later. Each call in the rejected triple lasted about 34.9 ms. The overall timing record contains no discontinuity, but the existing whole-call bracket does not distinguish trigger, status and result-access contributions.

This experiment varies spacing while retaining the current reference and recognition rules. It adds ordered observations at the existing ADC trace points. A lower unresolved count at longer spacing would support further investigation of acquisition cadence; it would not identify an electrical cause or independently prove that a fresh conversion supplied each value.

Method

Retain six rounds, six reference holds of 36 readings and six validation holds of 72 readings per round: 3888 readings and 864 non-overlapping validation medians. Voltage request codes, trigger mode, reference renewal, 1 ms programmed settling, one-third-gap acceptance, operating guards and restoration are unchanged. Validation observations cannot alter calibration.

Add a wait of 0, 5000 or 20000 microseconds before each validation acquisition. The condition is (triple position within the held request + zero-based round + state) modulo three. All three readings in a median share a condition. Each held request supplies eight medians per condition; each condition therefore has 288 medians and 864 validation readings across the boot. References receive no added wait. The first reading of a triple uses that triple's condition even when the preceding triple used a different one.

The existing ordered whole-call interval starts after the added wait. Actual spacing is measured from the preceding call's ordered end to the current call's ordered start; it includes ordinary intervening work as well as the added wait. Tick requirements round upward. These are minimum added waits, not a fixed sampling frequency or a claim about the physical conversion instant.

The schedule balances the condition across states and rounds. It is deterministic, and consecutive observations share the same apparatus and held request. Carryover, time trends and phase relationships remain possible; samples are not independent trials.

SEIGRASM implementation and recording

The cadence operation performs the prescribed wait and calls the existing timed acquisition. The ADC operation now places an ISB before and after each retained counter observation at the trigger-slot start and end, the completed status read and the four completed result-octet reads. Seven existing instruction sites call two appended observation routines. Bus operations, values, trace layout and recognition inputs are retained.

These points permit trigger-slot, trigger-end-to-status and successive completed-read intervals to be reconstructed. Differences include the intervening instructions and transport. They are not isolated ADC conversion durations. Existing timeout/reset counters are unchanged and are not promoted to ordered measurement endpoints. The ordering instructions add work; this experiment's zero-added-wait condition supplies a concurrent comparison with the same instrumentation.

Configuration schema 11 retains the schema-10 row and section layout. Configuration words 60–62 carry the three additional waits; word 63 retains the gap divisor. Each row still contains 64 words and each whole-call record six words. The journal remains bounded by 32151 records within 32704 reservations. No additional raw-record volume is introduced.

Target execution consists of the existing literal SEIGRASM entry, cadence, timed electrical acquisition, reference calibration, recognition and checked raw-sector recording. PC tooling copies the literal opcode words and initializes image contexts; it does not implement target operations or generate instructions. The existing EEPROM image-loading dependency remains. No new firmware service, target language, target compiler or filesystem operation is introduced.

Assessment

Require a matching run identity, unchanged boot payload, complete journal, all 3888 attempts and all prescribed restorations. Reconstruct each native recognition outcome from its recorded inputs. Require 3888 valid whole-call intervals, monotonically ordered available ADC observations inside each call, and every prescribed minimum inter-call spacing met. The configuration and recorded endpoints must agree with the frozen method.

Report correct, wrong, unresolved and unavailable medians separately for each spacing condition, including state and round when examining exceptions. Compare all 288 scheduled medians per condition; unavailable outcomes cannot be dropped. Longer spacing supports the recognition hypothesis only if it yields more correct and fewer unresolved or unavailable medians without increasing wrong assignments relative to the zero-added-wait condition. Complete recognition remains a separate criterion. A single boot cannot establish a reliable error probability.

Retain every rejected or wrong triple, its ordered endpoints, request/readback, ADC trace, calibration interval and next decision in the same held request. Compare trigger-slot and successive read-endpoint duration distributions across conditions. Timing differences alone cannot establish analog continuity, conversion freshness or causal electrical mechanisms. The independent reader retains timing, spacing and recognition outcomes separately.

Powered duration

Use one uninterrupted 26-minute powered session for this preparation, overriding the shared 30-minute window. Other handling remains unchanged.

The preceding timing boots reached the final recording section 1402.098862 and 1402.166795 nominal seconds after trial entry. Boot 2 records 1.043355 seconds before acquisition, 136.213763 seconds in acquisition and 1264.909677 seconds between acquisition end and final-section start. The current recorder checks each empty reservation, writes the senary record and reads back the full sector before advancing. The post-acquisition recording path dominates the measured duration; there is no prescribed 30-minute measurement loop.

The new spacing schedule adds exactly 21.6 nominal seconds of requested waits across validation. The previous later final-section timestamp plus that budget is 1423.766795 seconds (about 23 minutes 44 seconds), leaving about 136 seconds within the 26-minute window. This is a preparation estimate, not a measured completion time for the new image. Ordering overhead, card variation and the final section's own write/readback remain outside that estimate's demonstrated bound. The returned journal will determine completeness; an incomplete return must remain incomplete evidence.

Preparation

Candidate 4E92B31DB139 passed 258 supplied-device checks. Relative to timing boot 2, 32 instruction regions retain their bytes and addresses. The ADC observation region changes at seven counter sites and two appended helpers; the cadence region is added. The boot configuration, electrical parameters, calibration and recognition instructions, and recording layout are unchanged. Preparation manifest; validation record; instruction and apparatus comparison.

Second boot

Repeat the same acquisition and recognition procedure to test whether complete candidate recognition and the absence of a spacing advantage recur across boots. Retain all instructions, electrical settings, the balanced 0/5/20 ms schedule and the 26-minute powered window. Assign a new recording identity and empty journal. Preparation 127AAB5030AE retains all 34 instruction regions byte-for-byte; the repeat comparison records the unchanged code and acquisition settings. Both returns are now available below.

Results across both boots

Complete candidate median recognition recurred at every spacing. Each boot recovered 3888 readings, 864 validation medians and all 32151 journal records. Every prescribed minimum wait and every retained ADC ordering check passed. Neither boot showed a recognition advantage from adding 5 or 20 ms.

All remaining baseline outcomes were unresolved: four in boot 1 and five in boot 2. Neither rule produced a wrong or unavailable median. The fixed candidate rule correctly accepted all nine baseline rejections. These are two boot records with correlated observations; complete observed counts do not establish an error probability or equivalence between spacing conditions.

Boot 1 measurements · Boot 2 measurements · Implications for the next experiment

Boot 1 results

The return contains all 3888 scheduled readings, 864 validation medians and 32151 journal records. All 3888 whole-call intervals and 27216 retained ADC observations satisfy the ordering checks. Every prescribed minimum spacing was met. All six candidate calibrations retained separated intervals; the smallest intervening rejection region contained 14 integer ADC codes.

The renewed baseline had no wrong or unavailable medians. Its four rejections were correctly accepted by the unchanged separation-based candidate. No candidate rejection occurred, including at state 1, where the preceding timing return retained an exception. Complete candidate recognition occurred in this boot; the declared hypothesis of improved recognition with longer spacing was not supported because the zero-added-wait condition was also complete.

Request and readback agree throughout each listed triple: code 90 for state 0, 93 for state 1 and 96 for state 2. The candidate acceptance rule was fixed before acquisition; no interval was adjusted using these validation observations.

Boot 1 intervals

All 7776 whole-call endpoint frequency observations report 54000000 Hz. Durations below are nominal milliseconds derived from that configuration. Each condition contains 864 validation readings. The whole-call median across all 3888 readings is 34.911991 ms, with a range of 34.839833–39.172685 ms.

The recorded gaps distinguish all three spacing conditions. Median ADC access intervals are similar across them. These measurements locate elapsed work between ordered observations, including transport and instructions; they do not isolate the physical conversion instant or its duration.

Acquisition lasted 157.869650 nominal seconds (about 2 minutes 38 seconds). Trial entry preceded acquisition by 0.988726 seconds; a further 1264.786437 seconds elapsed from acquisition end to final-section entry. The latter occurred at 1423.644813 seconds after trial entry, about 23 minutes 44 seconds, compared with the preparation estimate of 1423.766795 seconds. The full journal was recovered under the prescribed 26-minute procedure. External powered duration was not separately measured, and final-section entry precedes that section's write and readback. This return supports the shorter procedure for this image and card; it is not a general completion bound.

Boot 2 results

The return contains all 3888 scheduled readings and 864 medians. All 3888 whole-call intervals and 27216 ADC observations are ordered; every minimum spacing was met. All six candidate calibrations retained separated acceptance intervals, with a smallest intervening rejection region of 15 integer ADC codes. The candidate recognized 864/864 medians; the renewed baseline recognized 859/864 and rejected five.

The median at ordinal 464 equals the candidate's upper acceptance endpoint. Its distance to that boundary is zero; successful classification does not establish tolerance to further upward movement. All acceptance intervals were fixed from the reference observations before their validation holds.

Individual raw readings still require separate interpretation. The first round's raw reference intervals overlap between states 2 and 3, leaving 432 renewed raw classifications unavailable. Later renewed raw classification has six unresolved readings. Complete three-reading median recognition therefore does not qualify every individual acquisition as a distinguishable state.

All 7776 endpoint frequency observations report 54000000 Hz. Whole-call duration has a median of 34.912139 ms and a range of 34.829500–39.176500 ms. Acquisition lasted 157.873332 nominal seconds. Final-section entry occurred 1423.791605 seconds after trial entry, about 23 minutes 44 seconds. The complete persisted journal was recovered; its final target readback remains unproven.

Interpretation and next comparison

The two planned boots establish recurrence of complete candidate median recognition and ordered recording under this procedure. Both give no observed recognition benefit from the added waits. Zero added wait remains the supported reference for the next comparison; these results give no basis for requiring a longer delay or widening acceptance intervals.

The preceding timing boots recognized 864/864 and 863/864 candidate medians. They lacked the current ADC ordering instructions and mixed spacing schedule. Their comparison with these returns cannot attribute the absent rejection to either change. Complete counts also leave little room to detect a spacing benefit under the sampled conditions.

The next useful investigation should distinguish conversion freshness from repeated access to a retained result. It requires an observable relation to a distinct conversion and a control capable of contradicting that relation; ordered timestamps and repeated values alone cannot supply it. The state-0 boundary observation and raw-reference overlap remain concrete challenges for recognition tests. Further repetitions of the same spacing schedule would need an additional, explicit question.

Acquisition integrity

Two read-only acquisitions agree by SHA-256: 0a7575a68514346101de85fc51d97a61606a44d5862900301feffbcfc9e6175e. The preparation identity matches and both boot files are unchanged. All 32151 journal records are valid and consecutive, with 553 unchanged empty reservations. All six block restorations completed; final restoration and pin release report success. The complete persisted journal does not prove the target's final readback. Seven sectors outside the journal differ from the prepared image; their writer and timing remain unidentified. No procedure deviations were reported.

Boot 2 acquisition integrity

Two read-only acquisitions agree by SHA-256: 8276605bc88c4febd090c6cd884da6568b38652fc2dc9a29d70cd06b674aefe7. The preparation identity matches and both boot files are unchanged. All 32151 journal records are valid and consecutive; 553 empty reservations remain unchanged. All six block restorations completed, and final restoration and pin release report success. Seven sectors outside the journal differ from preparation, with writer and timing unidentified. No deviations were reported; external powered duration was not separately measured.

Interpretation limits

The measuring frequency is configured rather than independently calibrated. ERA anchoring is absent. Ordered timestamps do not independently qualify ADC conversion freshness, analog voltage accuracy or firmware exclusion. A result is specific to the tested apparatus and conditions; other instruction architectures and silicon clusters are not qualified by this run.