Source and conversion history.

Question and rationale

Does a post-return observation depend on source activation, an intervening conversion, their interaction, or the reference trajectory? The preceding experiment did not establish a consistent delay-only relation across acquisition orders. Its largest contrast was +7 counts although both returned values were 744; the two local references differed. This investigation separates the two actions and retains three references for every case.

The wider objective is sovereignty over all available silicon components and their interconnections: SEIGRASM discovery, direct control and observation, replacement of inherited firmware dependencies, and Hyphos/SeigrOS use of discovered electrical capabilities through native Seigsit relations. This experiment characterizes one source and its acquisition path. Source-enable settings and converter commands are not assigned Seigsit states.

Four interventions

Sixteen blocks execute orders A B D C, B C A D, C D B A and D A C B four times each. Successive four-block cycles traverse that order list forwards, backwards, forwards and backwards. Every case occupies every position four times; each directed predecessor pair inside a block occurs four times. Allocation is deterministic.

Each case establishes the disabled source, waits at least 1,000 microseconds and commands three successive references. A guarded departure write either retains the disabled source or enables it. C and D command an intervening conversion; A and B read CS without a command. All four cases then poll readiness, read RESULT and read CS. The untriggered reading is explicitly retained as an observation of the available result, not counted as a new conversion.

An anchored counter wait holds until 50 microseconds after the departure-write report ends. A guarded return write disables the source, followed by zero added wait and two commanded return observations. Both source writes occur in every case. Channel 4, zero divider, disabled empty FIFO and disabled ADC interrupts remain fixed. No SD transaction occurs among the 120 accesses in a four-case block; recording and admission occur between blocks. Current and sticky errors remain visible without automatic retry.

The 1,000-microsecond preparation interval is not proof of complete settling. The disabled source potential is unspecified. Fifty microseconds sets a scheduled hold, not an exact internal switching interval. All actual access brackets remain recorded.

SEIGRASM changes

The authored access stream adds operation kind 7: wait for an offset from an earlier report's end counter. It records the anchor, offset, computed deadline, counter frequency and entry/exit. Invalid anchors, unusable frequency, overflow and deadlines already reached on entry are refused. The wait performs no device access. Each descriptor is bound to its own preceding departure-write report; recovery independently verifies that relation.

The compact executor now admits up to 128 accesses, allowing all 120 accesses of a block to execute before persistence. It clears report storage and stops at the first failed access. Each block retains a 1,210-cell report. Other executable streams are retained.

Entry instructions lead to SEIGRASM route and control instructions, compact execution, senary journal encoding and SD write/readback. PC Python and PowerShell package authored instruction words and descriptors and recover evidence. No foreign target compiler or target OS is used. Existing entry firmware, processor execution, SD transport and retained bridge configuration remain apparatus dependencies. Acquisition calls no firmware service; full replacement of inherited dependencies remains open.

Assessment fixed before return

Verify completeness, source readbacks, commands or their declared absence, errors, deadline anchors and restoration. Retain all cases, including first observations and partial traces. For each completed block, require a common intersection of all four departure-to-return brackets, and separately of all four return-write-to-command brackets, before describing cases as timing-compatible. Overlap cannot prove identical internal physical timing.

Calculate return-minus-reference differences separately from reference 1 and reference 3, for return observations 1 and 2. For those local differences, compare B−A and D−C for source activation; C−A and D−B for the intervening command; and D−C−B+A for interaction. Reference choice, position, drift and all individual readings remain visible. No response sign, magnitude or equality was a pass criterion.

A recurring source or command contrast would support an association under this apparatus. Reference or position dependence would motivate a controlled preparation investigation. Unresolved contrasts would remain unresolved; no preferred result would be selected as a Seigsit state.

Preparation and acquisition

Preparation 23D9BA799F01 froze 31 source inputs and passed 60 image-bound supplied-response checks. Installation verified all 4,096 image chunks and two full image readbacks. The prescribed allowance was three minutes. The returned coordinator interval is 96.315070 seconds, excluding entry loading and the terminal record's write/readback. Powered duration is not independently measured by the capture.

Results

All 2,464 records, 177 operations and 64 cases completed. The journal contains 352 commanded conversions and 32 untriggered observations. All 128 departure/return readbacks matched; no current or sticky conversion errors, missed deadlines or admission failures were reported. ADC and clock restoration passed.

All 32 untriggered results in A and B exactly equal reference 3, including the 16 observations with the source enabled. The intervening commanded readings in D exceed reference 3 by 114–117 counts (median +115); those in C differ by −1 to +1 (median 0). This separates observed result retention from the response following a command. It does not independently timestamp or verify an internal analog conversion.

Post-return differences are much smaller. The first return in A has a median of −1 relative to reference 3; B and C have medians of 0, and D +0.5. Every case has a second-return median of 0 against reference 3.

Reference trajectory and first observations

Reference 2 exceeds reference 1 in 58/64 cases, equals it in four, and is lower in two. The median increment is +1 count in each of A, B, C and D. Reference 3 minus reference 2 has median 0, with 41 equal, ten positive and thirteen negative increments. These observations precede each case's intervention. They expose a repeatable association with reference position; they do not identify whether elapsed time, conversion activity, preparation or another coupled influence causes it.

The first case records references 746, 743, 744 and returns 743, 743. The next case, B, records references 742, 743, 744 and returns 744, 744. For both returns, B−A is +5 using reference 1 and +1 using reference 3. The first case is retained throughout. Reference-1 block medians fall from 742 to 728; reference-3 medians fall from 744 to 729.5.

Source and conversion contrasts

All 16 blocks satisfy the timing criterion, so the all-block and timing-compatible summaries contain the same observations. Each contrast subtracts local return-minus-reference differences; it is not a comparison of unmatched absolute readings. Values below are descriptive summaries of this boot.

For the first return against reference 3, B−A is positive in 11 blocks and zero in five; C−A is positive in nine and zero in seven. D−C is positive in eight, zero in seven and negative in one; D−B is positive in six, zero in nine and negative in one. Thus activation and an intervening command each show a small upward contrast relative to A, but neither produces a uniform additional effect in the presence of the other. All five second-return contrast medians are zero with both references. A zero median does not mean every contrast is zero or that an effect has been excluded.

Recorded timing

All accesses report a 54 MHz counter. Departure-to-return write brackets span 51.796–56.204 microseconds; return-write-to-first-command brackets span 3.704–8.407 microseconds. Zero added wait therefore does not mean simultaneous actions. All 64 anchored waits entered before their deadline and exited at or after it.

The intervals overlap within every block as required. This limits gross scheduling mismatch; it does not resolve exact internal source switching, sampling or conversion times.

Individual cases and contrasts

All 64 cases appear below in acquisition order. The first row in each block is its first-position case. References and returns are commanded. I is the intervening observation: U denotes no command in A/B; C denotes a command in C/D. Every row is complete and error-clear. Units are raw counts.

Interpretation and next question

The experiment separates a retained RESULT observation from the response after a command, and shows that three nominally identical reference acquisitions are not interchangeable at the one-count scale. Source activation and intervening conversion both associate with small first-return differences relative to the unchanged/no-command case. The second observation reduces the pooled differences, while individual deviations remain. These are bounded associations from one boot, not an identified internal mechanism.

The next discriminating question is whether the reference-position shift follows elapsed time after preparation or conversion activity. A proposed comparison would keep the source disabled, match the total preparation-to-observation interval, and vary the number of intervening conversion commands while retaining a wait-only control. Record the first and later results, deadlines, readiness and source controls without discarding an early reading. This tests the action history behind the shift instead of selecting a delay or omitting data to suppress it. The subsequent preparation time and conversion activity experiment conducts this comparison with fixed source control and matched probe deadlines.

The reusable result for wider silicon discovery is a recorded comparison between an intentional action and the same interval without that action, with local references and explicit result retention. It remains one component-level capability. Independent analog freshness, six maintained electrical states, native physical Seigsit creation and transfer, replacement firmware and sovereignty over every component in the cluster remain unestablished by this run.

Integrity, restoration and provenance

Two read-only captures agree. Independent raw reconstruction verifies all senary record fields, integrity checks, grouped commands or their absence, source readbacks, deadline anchors and arithmetic, observations and interval bounds. Unused journal reservations match preparation. Restored ADC control and divider are zero, interrupts disabled and FIFO empty; clock control is zero with divider/select retained at 1.

The partition table, geometry and both boot files match preparation; FAT copies agree. Seven sectors outside the journal differ, including System Volume Information. Their writer and timing are unestablished.

Capture SHA-256: 53d6ef84434717f1e6807c4993a75159c85561b4159e103d95b603efb520f8dd. Image SHA-256: 91b3da6bd4eadfe028062b0c036c2e1447ed46f554db6d9c98e0faa9f0099ca6.