The Tonic Canvas of Consciousness: A Formal Translation from Contemplative Narrative to Intrinsic Neural Rendering

Philosophical and contemplative traditions often describe awareness as a relatively formless field within which differentiated contents arise. Psychology and neuroscience instead describe background state, perceptual differentiation, prediction error, recurrent access, action selection, and learning. Self-Aware Networks (SAN) has proposed since 2022 that tonic oscillatory activity functions as an active "canvas of consciousness," while phase-wave differentials (PWDs) constitute changing content within that canvas. This paper asks whether the narrative architecture of "ground, arising form, discernment, action, and return" can be translated into a mechanistic architecture without claiming that a neural model proves a religious metaphysics. The proposed translation maps a history-conditioned tonic reference state to the field-like continuity of awareness; typed, receiver-relative deviations from that reference to arising forms; learned neural tuning to discernment; coherence-weighted recruitment to valuation and action; and reafferent/plastic updating to the return of consequences into the next reference state. Information is defined conditionally: a tonic state is not intrinsically information-free, but may have low surprisal and low entropy rate under the system's current model, whereas a discriminable deviation may carry high information for a particular receiver. Synchrony is treated as coordination rather than consciousness itself. Both excessive synchrony and unstructured dispersion can destroy usable differentiation; conscious content is hypothesized to occupy a metastable regime in which structured differences are locally readable, recurrently available, embodied, and capable of changing future processing. To discipline the proposed bridge, the paper combines four formal tools: a Causal Interface Sequence, a category-theoretic partial functor, a process-calculus encoding, and Gorla's five validity criteria for encodings. A machine-checkable five-role interface model establishes an exact labeled-graph isomorphism and passes finite-interface analogues of the five criteria. Draft 2 adds a bounded source grammar, a target grammar with silent transitions, weak behavioral equivalence, semantic name substitutions, and four independent parallel contexts. The validator exhaustively checks 970 generated source terms across 3,880 term-context pairs and rejects eight adversarial mutations. The resulting claim remains deliberately bounded. The certificate proves weak operational equivalence only for the declared finite languages and contexts; a whole-tradition, full pi-calculus, empirical-neural, or ontological isomorphism is neither established nor required. The framework yields falsifiable experiments comparing predictable background activity, informative residuals, hallucinated content, minimal-content awareness, action, and memory. It also distinguishes an external computer display from intrinsic rendering: in the latter, the rendering state is part of the same embodied causal system that detects, acts, remembers, and updates itself. Draft 3 executes the comparator and ablation benchmark that Draft 2 left open. A deterministic 768-trial synthetic design crosses two tonic contexts, two receiver signs, two content labels, and controlled noise. Tonic-only, phasic-only, pure-synchrony, and unstructured-dispersion implementations remain at chance on content; a feedforward typed classifier recovers content but fails return and memory; and complete abstract predictive and active-inference-like controls match the generic behavioral chain. The complete declared SAN chain also passes that chain, while a phase-sign intervention distinguishes its phase-typed implementation from the abstract controls. Seven single-feature ablations each break at least one preregistered observable. This is a test of declared toy implementations, not evidence of neural implementation, phenomenal identity, or SAN's uniqueness over every predictive architecture. Draft 4 closes the remaining local source and design gaps without changing that result. It separates six philosophical and contemplative source families instead of treating them as one doctrine; binds predictive coding and active inference to independently authored definitions while identifying the synthetic controls as local test doubles; specifies a two-stage, receiver-relative phase-measurement and closed-loop intervention protocol; and records explicit nonduplication boundaries against adjacent SAN papers. Closed-loop phase-locked human stimulation establishes feasibility for a phase-conditioned intervention, not evidence for TCT. The empirical claim remains prospective: SAN gains support only if receiver- and context-specific phase relations add held-out explanatory value and survive