v0.0.5

Other

• ac4 round 50: section-boundary DP optimiser + SNF emission
• ac4 round 49: HCB1..11 codebook-selection optimiser + parameterised max_sfb
• ac4 round 48: forward MDCT + ASF entropy encoder for arbitrary PCM
• ac4 round 47: IMS bitstream_version=2 TOC parser + mono SIMPLE/ASF tone encoder
• round 46: AC-4 IMS encoder scaffold + ACPL_1 surround Ls/Rs spec audit
• ac4 round 45: stereo-CPE M=2 synced companding for ACPL_3 surround pair
• ac4 round 44: companding sync_flag=1 cross-channel exact synchronisation
• ac4 round 43: companding sync_flag=1/avg branches + ACPL_1 sb0 hookup
• ac4 round 42: cfg0/cfg1/cfg3 trailer-aware ASPX + §5.7.5 companding
• ac4 round 41: 5_X cfg2 ASPX trailers + Table 181 SAP for ACPL_1
• ac4 round 40: SAP a/b/c/d (Pseudocode 59) + Table 183 + Ls/Rs walker
• ac4 round 39: 5_X cfg0/cfg1/cfg3 dispatch + 7_X additional-channel pair
• ac4 round 38: LFE body decoder + cfg2_back_mono end-to-end + ACPL_3 centre
• ac4 round 37: wire 7_X ACPL_1/_2 dispatch + cfg0 centre end-to-end decode
• ac4 round 36: wire 5_X ASPX_ACPL_1 / ACPL_2 Pseudocode 117 into decoder
• Round 35: extend ETSI validation suite to float reference tables
• drop dead linkme dep
• round 35 — EMDF payloads_substream parser + DRC PCM gain application
• cargo fmt pass after round 34
• update round 34 status (SNF + FIXVAR/VARFIX/VARVAR atsg + ACPL_3)
• ac4 round 34: FIXVAR/VARFIX/VARVAR atsg + SNF inject + 5_X ACPL_3 wiring
• auto-register via oxideav_core::register! macro (linkme distributed slice)
• unify entry point on register(&mut RuntimeContext) (#502)

Added

• Round 50 — Section-boundary DP optimiser + Spectral Noise Fill
  (SNF) emission (TS 103 190-1 §5.7.4 section_data + §5.7.6 SNF +
  Pseudocodes 100 + 105 + Table 39/42 + Table SCFB):

  • encoder_asf::dp_optimise_sections(cost_band_cb, max_sections) —
    dynamic-program over scale-factor bands that finds the globally
    cheapest sequence of (start, end, cb) sections, paying the
    per-section header cost (4 + 3 * (floor((L-1)/7)+1) bits per
    Table 39) against each band's per-codebook bit cost. Section
    count capped at 16 per ETSI Table SCFB. Supersedes the round-49
    greedy run-length codebook-merge optimiser.
  • encoder_asf::section_overhead_bits(len) — closed-form per-section
    overhead in bits matching the spec's n_sect_bits=3 / esc=7
    long-frame layout: 7 bits for L ∈ 1..=7, 10 bits for L ∈ 8..=14,
    13 bits for L ∈ 15..=21, etc.
  • encoder_asf::build_band_codebook_cost_table(natural_q_per_band)
    — precomputed per-band per-codebook bit cost (rows of length 12;
    cb=0 cost 0 only for all-zero bands; HCB1..11 costs from
    bit_cost_for_band; u32::MAX for codebooks that can't represent
    the band's natural quant magnitudes). Drives the DP via O(1)
    prefix sums where every band is feasible for the codebook.
  • encoder_asf::build_sections_from_dp(sections, max_sfb) — lowers
    the DP-derived (start, end, cb) triples into an AsfSections
    suitable for the existing write_section_data /
    write_spectral_data_sections emitters.
  • encoder_asf::compute_snf_dpcm_for_zero_quant_bands(coeffs, sfb_offset, max_sfb, sfb_cb, max_quant_idx) — for each band that
    quantises to all-zero (cb == 0 || mqi == 0), estimates the band's
    RMS energy from the original MDCT coefficients and picks the
    HCB_SNF index whose snf_gain = 2^((idx*1.5 - 84)/4) best matches.
    Returns Some(per_band_idx) when at least one zero-quant band
    has measurable energy; None for fully-silent input.
  • encoder_asf::write_snf_data(bw, snf, sfb_cb, max_quant_idx, max_sfb)
    — emits b_snf_data_exists (1 bit) plus per-zero-quant-band
    HCB_SNF Huffman codewords per Table 42 / Pseudocode 105. Round-trips
    cleanly through the existing parse_asf_snf_data decoder path
    (round 36+).
  • encoder_asf::measure_greedy_vs_dp_bits(transform_length, max_sfb, coeffs) — diagnostic helper returning (greedy_bits, dp_bits)
    for any input spectrum so callers can quantify the section-boundary
    optimiser's contribution to total frame size.
  • Ac4ImsEncoder::encode_frame_pcm_with_max_sfb now drives the DP
    optimiser + SNF emission internally; existing call sites get the
    new path automatically with no API change. White-noise spectral
    SNR holds at 27.5 dB (round-49 baseline) with section overhead
    reduced and SNF emission turned on for high-frequency zero-quant
    content.

• Round 49 — Codebook-selection optimiser (HCB1..11) + parameterised
  max_sfb (TS 103 190-1 §5.7 + Pseudocodes 17 + 19 + 20 + Annex A.0
  huff_codes + Table SCFB):

  • encoder_asf::pick_best_codebook_for_band — per-band codebook
    optimiser sweeping HCB1..11 and choosing the lowest-bit-cost
    codebook whose q_max covers the band's natural quantised range.
    Anchor scalefactor targets peak quant ≈ 12 (HCB9/10's q_max → 3×
    more quantisation levels per band than the round-48 HCB5-only
    baseline). HCB11 always qualifies via its Pseudocode 20 escape so
    very-high-energy bands don't clip.
  • encoder_asf::bit_cost_for_band — precise bit-counter modelling
    HCB1..11 codeword length + sign bits for unsigned codebooks +
    per-Pseudocode-20 n_ext extension bits for HCB11 escapes.
    Mirrors the encoder emitter's exact bit shape (inline magnitude
    saturates at 16 for HCB11, sign bit per non-zero post-saturation
    line for unsigned codebooks).
  • encoder_asf::build_sections_from_per_band_cb — collapses runs
    of consecutive same-codebook bands into a single AsfSections
    entry so the emitted asf_section_data() honours the spec's
    grouping pseudocode without spurious cb-switch overhead.
  • encoder_asf::write_section_data + write_spectral_data_sections
    — multi-section asf_section_data + asf_spectral_data emitters.
    Per-section emission walks sect_start..sect_end bins with the
    section's codebook, handles cb == 0 silent bands, and writes
    Pseudocode 20 escape bits for HCB11 outliers.
  • Ac4ImsEncoder::encode_frame_pcm_with_max_sfb(frame, max_sfb) —
    new public entry point parameterising max_sfb (round-48 default
    was hard-coded to 40 → ~6.4 kHz at tl=1920 / 48 kHz). Pad target
    scales with max_sfb (2KB / 4KB / 8KB tiers). encode_frame_pcm
    keeps the round-48 default of max_sfb=40 for backwards
    compatibility.
  • White-noise round-trip SNR jumps from 13.6 dB (round-48 HCB5-only
    baseline) to 27.5 dB (round-49 HCB1..11 optimiser, q_target=12,
    max_sfb=50) — measured spectrally against the encoder's own MDCT
    coefficients pre/post quantisation. 1 kHz tone reconstruction at
    max_sfb=55 preserves >100% of input energy (vs ~40% at the
    round-48 max_sfb=40 default).

• Round 48 — Forward MDCT analysis + ASF entropy encoder for arbitrary
  PCM input (TS 103 190-1 §5.5 MDCT + §5.7 SIMPLE + §5.8 ASF +
  Pseudocodes 17-19 + Annex A.0 huff_codes):

  • encoder_mdct::mdct_naive + EncoderMdctState — forward MDCT
    direction complementing the decoder's mdct::imdct. Naive
    O(N²) direct-summation cosine basis (correctness-first; encoder
    isn't on a hot path). Sign convention + scaling matched against
    the decoder's IMDCT through a Princen-Bradley TDAC round-trip
    test (constant-signal recovery in steady-state middle frame
    within 1% error). EncoderMdctState carries the previous-frame
    N PCM samples for cross-frame 50% TDAC overlap.
  • encoder_asf::quantise_coeff + pick_scalefactor_for_band +
    encode_pair + write_sect_len_incr + write_scalefac_data +
    build_mono_simple_asf_body_from_pcm_spectrum — closed-form
    forward ASF entropy encoder for the long-frame, single-window-
    group, mono SIMPLE channel case. Per-band scalefactor selection
    via the closed-form solve sf_min = ceil(100 + 4*log2(max_abs/q_max^(4/3)))
    keeping every quantised line within HCB5's ±4 magnitude bound
    after q = round(sign(c)*|c/sf_gain|^(3/4)). Single-section
    HCB5 emission across 0..max_sfb; reference scalefactor
    • DPCM-coded per-band deltas via HCB_SCALEFAC; b_snf_data_exists = 0.
  • Ac4ImsEncoder::encode_frame_pcm(input: &[f32]) — public entry
    point taking arbitrary float PCM input (range [-1.0, 1.0])
    and emitting a structurally-valid IMS v2 frame end-to-end:
    forward MDCT analysis → per-band scalefactor + quantisation →
    HCB5 entropy coding → wrap in v2 IMS TOC + audio_size header.
    Lazily initialises a per-encoder EncoderMdctState on first
    call; bumps sequence_counter modulo 1024. Mono / 48 kHz / 24 fps
    by default (frame_len = 1920 samples, max_sfb = 40 covering
    bins 0..600 ≈ 7.5 kHz).
  • 13 new unit tests across the three modules cover: forward MDCT
    zero-in-zero-out + linearity + Princen-Bradley constant-signal
    • sine-wave SNR > 40 dB; quantise/dequantise round-trip;
      pick_scalefactor q_max bound; encode_pair signed/unsigned
      round-trip via huff_decode + split_qspec;
      build_mono_simple_asf_body_from_pcm_spectrum end-to-end parse
      via the existing ASF decoder; full encode → decode round-trip
      for 1 kHz tone (peak amplitude > 1000 i16), multi-tone
      250+500+1000 Hz (SNR > 10 dB on steady-state frame), and
      silence (peak < 50 i16); encode_frame_pcm bumps
      sequence_counter per call.
  • Codebook-selection optimiser (try HCB1..11 per section, pick
    min-bits), section-boundary optimiser (split bands by
    codebook), spectral noise fill, and stereo / multichannel
    forward analysis remain deferred for round 49+.

• Round 46 — AC-4 IMS encoder scaffold + ACPL_1 surround Ls/Rs
  ASPX-extension spec audit (TS 103 190-2 §6.2.1.1 / §6.3.2.5,
  TS 103 190-1 §4.2.6.6 Table 25):

  • encoder_ims::Ac4ImsEncoder — Auditor-mode AC-4 IMS encoder
    skeleton. Emits a structurally-valid raw_ac4_frame() payload
    with the IMS-flavoured ac4_toc() (bitstream_version = 2 +
    ac4_presentation_v1_info() + ac4_substream_group_info()) per
    TS 103 190-2 §6.2.1.1. Public API: Ac4ImsEncoder::new() (defaults
    to mono 48 kHz 24 fps single-presentation single-substream-group
    iframe), with_v0() / with_stereo() / with_5_1() builders,
    encode_frame(body_padding_bytes) (bumps the 10-bit
    sequence_counter modulo 1024), and encode_frame_v0(...) that
    forces the TS 103 190-1 v0 layout for round-trip-with-decoder
    tests. Audio body is zero-byte placeholder bits — full encoder
    pipeline (MDCT analysis, scalefactor selection, entropy coding,
    A-SPX envelope coding, A-CPL parameter extraction) deferred to
    future rounds. Eight new unit tests cover sequence_counter wrap,
    parse_ac4_toc round-trip for mono / stereo / 5.1, full
    Ac4Decoder round-trip emitting silent audio, and the leading
    bitstream_version bit-pattern invariant for both v0 (0b00)
    and v2 (0b10) frames.
  • decoder.rs::dispatch_acpl_5x_pair-driving block: spec-confirms
    NOT-ASPX finding for the ACPL_1 surround Ls/Rs carriers per ETSI
    TS 103 190-1 §4.2.6.6 Table 25 row case ASPX_ACPL_1:. The
    trailer order is aspx_data_2ch() (L/R primary pair) +
    aspx_data_1ch() (centre mono) + two acpl_data_1ch() parameter
    sets — there is NO third ASPX trailer for the surround pair. The
    Ls/Rs sSMP,3 / sSMP,4 carriers are joint-MDCT residuals that feed
    Pseudocode 117 raw, with the post-Pseudocode-117 surround-output
    bandwidth shape coming from the L/R-carrier ACPL synthesis
    (alpha/beta/decorrelator), not from independent surround-pair
    extension. Same finding for the M=2 surround-pair synced
    companding cohort: with no surround carriers, no companding to
    sync. The existing raw-PCM dispatch path is correct per spec —
    no behavioural change in this round, only a documentation note
    closing out the round-45 follow-up flagged in dispatch_acpl_5x_pair.

• Round 44 — companding sync_flag == 1 cross-channel synchronisation
  (Pseudocode 121's exact g_synch(ts) = (∏ g_ch)^(1/M)) (TS 103
  190-1 §5.7.5.2 + Pseudocode 121):

  • aspx::compute_companding_levels(q, sb0, sb1) exposes the
    per-slot energy L_ch(ts) = 0.9105 * mean E_ch(sb,ts) without
    applying gain — the building block the synced path collects
    across channels.
  • aspx::levels_to_scales_per_slot(levels) and
    aspx::levels_to_scale_averaged(levels) produce the per-slot
    array (PerSlot / SyncPerSlot) and single constant (Averaged /
    SyncAveraged) scales g(ts) * G from a level vector.
  • aspx::apply_companding_scales_on_qmf(q, sb0, sb1, scales)
    applies pre-computed scales — used by both the single-channel
    legacy path and the synced multi-channel path.
  • aspx::apply_synchronised_companding_across_channels(channels, mode)
    implements Pseudocode 121's sync_flag == 1 branch directly:
    collects every channel's L_ch(ts), computes g_synch(ts) as
    the geometric mean across M channels (SyncPerSlot) or
    g_avg,synch from per-channel averaged levels (SyncAveraged),
    then applies the synced gain g_synch(ts) * G UNIFORMLY to
    every contributing channel. Geometric mean is computed via
    log-sum / exp for numerical stability across many channels.
  • Ac4Decoder::aspx_extend_to_qmf(...) is the new phase-1 split:
    runs the QMF analysis + HF generation + envelope adjustment +
    noise / tone injection, returns (qmf_matrix, sbx, sbz) BEFORE
    companding + synthesis. Returns None when the QMF
    preconditions trip (length / table / patch derivation).
  • Ac4Decoder::qmf_synthesise_pcm(q, out_len) is the phase-2
    split: runs the inverse QMF synthesis and returns PCM. Caller
    is responsible for having applied any §5.7.5 gain.
  • Ac4Decoder::extend_5x_channels_with_sync_companding(...) is
    the integration glue: drives every entry through phase-1,
    collects QMF matrices, calls the synced apply, drives every
    entry through phase-2. Channels whose phase-1 returned None
    fall through to the unmodified PCM (matching the per-channel
    aspx_extend_pcm contract).
  • Ac4Decoder::extend_5x_entries(...) is the shared front-end
    used by dispatch_5x_cfg{0,1,2,3}_simple_aspx: routes through
    the synced cross-channel path when companding.sync_flag == 1,
    otherwise through the per-channel path.
  • Ac4Decoder::five_x_synced_mode(cc) resolves the synced mode
    once for the whole 5_X frame (Pseudocode 121 broadcasts
    compand_on[0] to every channel under sync_flag == 1).
  • All four dispatch_5x_cfg{0,1,2,3}_simple_aspx routes
    refactored to build a per-channel entries list then route
    through extend_5x_entries. Cfg2 now also folds the centre
    channel into the synced cohort when back_mono is present, so
    M == 5 for the geometric mean across all five 5_X channels
    (was M == 4 in the round-43 approximation).
  • The round-43 single-channel approximation (SyncPerSlot /
    SyncAveraged collapsing to per-channel g_ch / g_avg,ch)
    is RETIRED — every 5_X SIMPLE/ASPX dispatcher now applies the
    exact synchronised gain across all contributing channels.

• Round 43 — companding b_compand_avg + sync_flag == 1 branches +
  ASPX_ACPL_1 sb0 = acpl_qmf_band hookup (TS 103 190-1 §5.7.5.2 +
  Pseudocode 121):

  • aspx::CompandingMode enum captures the (sync_flag,
    b_compand_on[ch], b_compand_avg) product per Pseudocode 121:
    Off, PerSlot, Averaged, SyncPerSlot, SyncAveraged.
  • aspx::CompandingMode::from_control(cc, slot) resolves the
    active branch for a single channel from a parsed
    CompandingControl, honouring sync_flag == true (slot 0
    broadcasts) and the b_compand_avg presence-rule.
  • aspx::apply_companding_on_qmf_with_mode(q, sb0, sb1, mode)
    extends the round-42 single-channel path to all four active
    sub-branches: PerSlot / SyncPerSlot apply g_ch(ts) * G
    per timeslot; Averaged / SyncAveraged average L_ch(ts)
    over the full A-SPX interval, derive a single constant
    g_avg,ch * G and broadcast it across all timeslots. (Off
    is a strict no-op.) apply_companding_on_qmf retained as a
    PerSlot thin wrapper for backward compatibility.
  • Ac4Decoder::aspx_extend_pcm signature now takes
    (compand_mode: CompandingMode, compand_sb0_override: Option<u32>)
    instead of a raw compand_on: bool. The override implements
    §5.7.5.2's sb0 selection rule: for the ASPX_ACPL_1 codec mode
    sb0 = acpl_qmf_band (from acpl_config_1ch_partial.qmf_band);
    otherwise sb0 falls back to tables.sbx (= aspx_xover_band).
  • Stereo CPE path (receive_frame) lifts the override from
    tools.acpl_config_1ch_partial.qmf_band whenever the active
    stereo or 5_X mode is AspxAcpl1. Cfg0/Cfg1/Cfg2/Cfg3 SIMPLE/ASPX
    dispatchers pass None (those paths never run on ACPL_1).
  • Ac4Decoder::five_x_compand_mode_for_slot resolves a
    CompandingMode per output slot for the 5_X dispatchers; the
    legacy five_x_compand_on_for_slot is retained as a thin
    mode != Off wrapper for round-42 unit-test compatibility.
  • sync_flag == 1 cross-channel synchronisation: the per-channel
    pipeline computes g_ch(ts) per channel; for M = 1 (the
    dominant case in our pipeline) the geometric mean across
    channels reduces to the local gain (exact). For M > 1 channels
    processed independently the synchronisation is approximated —
    documented as a known limitation in
    apply_companding_on_qmf_with_mode.
  • 5_X ACPL_3 path companding wiring: already in place from round
    42 via the stereo CPE path's compand_mode_pri /
    compand_mode_sec (the ACPL_3 walker populates tools.companding
    from companding_control(2); the L/R carriers go through the
    standard stereo CPE primary/secondary path which now lifts
    CompandingMode and applies it before the §5.7.7.6.2
    Pseudocode 118 multichannel synthesis).
  • 7 new tests cover: CompandingMode::from_control resolves all
    six sync/on/avg product states; apply_companding_on_qmf_with_mode
Off strict no-op + Averaged/SyncAveraged constant-scale
    invariant + sb0-override band shift; five_x_compand_mode_for_slot
    branch resolution; aspx_extend_pcm with sb0 override + with
    Averaged mode diverges from baseline.

• Round 42 — 5_X SIMPLE/ASPX cfg0/cfg1/cfg3 trailer-aware ASPX
  dispatch + §5.7.5 companding tool (TS 103 190-1 §4.2.6.6 / §5.7.5):

  • asf::SubstreamTools gained cfg0_aspx_{lr,ls_rs,centre},
    cfg1_aspx_{lr,ls_rs,centre}, cfg3_aspx_{lr,ls_rs,centre} —
    mirrors of the round-41 cfg2 trailer slots. The 5_X SIMPLE/ASPX
    outer walker now stores the captured Table-25 case ASPX:
    trailer triplet (aspx_data_2ch + aspx_data_2ch + aspx_data_1ch)
    into the slot matching the active coding_config instead of
    discarding round-41's cfg0/1/3 captures.
  • Ac4Decoder::dispatch_5x_cfg{0,1,3}_simple_aspx now apply the
    A-SPX bandwidth-extension per output channel using the captured
    trailers, with the canonical Table-25 trailer-to-slot mapping
    (1st-2ch -> L/R, 2nd-2ch -> Ls/Rs, 1ch -> C). Independent of
    cfg0's b_2ch_mode (ASPX is applied after the channel-element
    decode produces PCM, so the L,Ls / R,Rs inner stereo coding
    doesn't shuffle the trailer assignment).
  • Ac4Decoder::maybe_extend_5x_slot — internal helper that picks
    the right (trailer, primary/secondary) for an output slot and
    runs aspx_extend_with_trailer with the resolved companding-on
    flag for that slot.
  • aspx::apply_companding_on_qmf(q, sbx, sbz) — §5.7.5.2
    companding tool decoder side, applied per-channel on the QMF
    matrix in [sbx, sbz) for all timeslots. Implements the
    sync_flag == 0, b_compand_on == true branch (the dominant
    case): per-slot mean-absolute level L_ch(ts) per Pseudocode
    equations, then g_ch(ts) = L^((1-α)/α) with α = 0.65 and
    Q_out = g * G * Q_in with G = 2^α. Zero-signal slot uses
    unit gain (avoids 0^negative_exp = inf). The b_compand_avg
    averaging and the sync_flag == 1 per-channel-product synched
    gain branches are scaffolded for a later round.
  • Ac4Decoder::aspx_extend_pcm extended with a compand_on: bool
    parameter — applies apply_companding_on_qmf between envelope
    adjustment and inverse QMF synthesis when set. All call sites
    updated: stereo CPE primary/secondary (uses
    tools.companding[0] / [1]), and every 5_X cfg0/1/2/3 slot
    (uses the companding_control(5) slot lookup via
    five_x_compand_on_for_slot).
  • Ac4Decoder::five_x_compand_on_for_slot(cc, slot) — resolves
    b_compand_on[slot] from a CompandingControl, honouring
    sync_flag == true (slot 0 broadcasts), per-channel, and absent
    flags (returns false).
  • 5 new tests cover: per-cfg trailer-aware dispatch produces
    output that diverges from the round-39 low-band-only path;
    companding flag resolver across mono / per-channel / sync
    branches; aspx_extend_pcm with companding diverges from
    baseline; and edge-case no-ops for apply_companding_on_qmf
    (degenerate band + zero signal).

• Round 41 — 5_X SIMPLE/ASPX cfg2 ASPX bandwidth-extension trailers +
  Table 181 first-stage SAP matrix for 5_X / 7_X ASPX_ACPL_1 (TS
  103 190-1 §4.2.6.6 / §5.3.4.3.2):

  • aspx::FiveXAspxTrailer + aspx::FiveXAspxChannelTrailer —
    captured per-trailer bitstream state (xover, frequency tables,
    framing, qmode, delta-dir, sig/noise envelopes, hfgen
    add_harmonic / tna_mode) for one aspx_data_2ch() or
    aspx_data_1ch() payload. Wraps everything aspx_extend_pcm
    needs for one or two channels.
  • asf::capture_aspx_data_2ch_trailer /
    asf::capture_aspx_data_1ch_trailer — wrap parse_aspx_data_*_body
    with snapshot/restore over the per-substream ASPX-trailer slots
    so multiple sequential trailers can be parsed without corrupting
    each other's state.
  • parse_5x_audio_data_outer SIMPLE/ASPX branch now walks the
    Table 25 row case ASPX: trailer triplet (aspx_data_2ch + aspx_data_2ch + aspx_data_1ch) and stores the captured trailers
    on tools.cfg2_aspx_lr / cfg2_aspx_ls_rs / cfg2_aspx_centre.
    Cfg0/Cfg1/Cfg3 still parse the bits (so the bitreader lands at
    the right offset) but don't yet wire to dispatch.
  • Ac4Decoder::dispatch_5x_cfg2_simple_aspx extended to apply
    A-SPX bandwidth-extension per-channel: L/R use aspx_lr (primary
    / secondary); Ls/Rs use aspx_ls_rs; centre uses aspx_centre.
    SIMPLE-mode (no trailers) and trailer-parse-miss paths fall
    through to round-38 low-band-only PCM.
  • asf::apply_sap_table_181(a, b, s3, s4, chparam_pair, max_sfb, tl) -> (l, r, ls, rs) — Table 181 first-stage matrix for
    5_X_codec_mode == ASPX_ACPL_1. Mixes (sSMP_A, sSMP_B) with
    (sSMP_3, sSMP_4) per-sfb using the (a, b, c, d) coefficients
    extracted from each chparam_info() payload (Pseudocode 59) into
    preliminary (L, R, Ls, Rs) spectra. Bands past max_sfb_master
    pull L/R from A/B and zero Ls/Rs.
  • parse_aspx_acpl_1_2_inner_body (5_X) +
    parse_7x_audio_data_outer (7_X ASPX_ACPL_1 branch) now persist
    the parsed chparam_info() pair on
    tools.acpl_1_residual_chparam plus max_sfb_master on
    tools.acpl_1_residual_max_sfb_master — round 40 already
    persisted the residual spectra; round 41 closes the loop with
    the SAP coefficients + bound.
  • Ac4Decoder::receive_frame 5_X ACPL_1 dispatch now applies
    Table 181's SAP matrix in the spectral domain when all four
    inputs (sSMP_A, sSMP_B, sSMP_3, sSMP_4 + chparam pair +
    max_sfb_master) are available, IMDCTs the resulting (L, R, Ls,
    Rs) preliminary spectra, and feeds them into Pseudocode 117
    (run_acpl_5x_pair_pcm) as the already-mixed PCM inputs the
    spec expects. ACPL_2 mode (no residual pair) falls through to
    round-40 silence placeholders.

• Round 40 — SAP a/b/c/d coefficient extraction (Pseudocode 59) +
  Table 183 7_X SIMPLE/ASPX final channel mapping + standalone Ls/Rs
  surround mono walker for ACPL_1 Mode 1 (TS 103 190-1 §5.3.2 /
  §5.3.4.4.1 / §5.7.7.6.1):

  • asf::SapCoeffs + asf::extract_sap_abcd(info, max_sfb_per_group)
    — Pseudocode 59 implementation. Walks chparam_info.sap_mode ∈
    {0, 1, 3} and emits per-(g, sfb) (a, b, c, d) quartets:
    • sap_mode == 0 → identity (a=d=1, b=c=0).
    • sap_mode == 1, ms_used → M/S inverse (a=b=c=1, d=-1) per-sfb;
      identity where ms_used == false.
    • sap_mode == 3 → alpha-driven SAP. Pair-major DPCM differential
      decode of dpcm_alpha_q → alpha_q[g][sfb] (odd sfbs inherit
      the even pair-mate; cross-group delta_code_time folds in when
      g != 0 and max_sfb_g == max_sfb_prev). `sap_gain = alpha_q
      • 0.1drives(a, b, c, d) = (1 + sap_gain, 1, 1 - sap_gain, -1)for SAP-coded bands;(1, 0, 0, 1)` for skipped bands.
  • Ac4Decoder::dispatch_7x_additional_channel_pair extended:
    accepts partner_pair_spectra + partner_slots + chparam. With
    b_use_sap_add_ch == true AND partner spectra of matching
    transform length, applies Table 183's SAP matrix per-sfb in the
    spectral domain — [out_high; out_low] = [a b; c d] · [partner; add_ch] — then IMDCTs both halves independently. With identity
    SAP the partner slot is left untouched (its independent 5_X-core
    IMDCT renders elsewhere) and only F/G land at slots 5/6 per Table
    182 — round-39 behaviour preserved.
  • Ac4Decoder::receive_frame 7_X branch resolves the partner pair
    from the active five_x_coding_config: cfg2 picks
    four_channel_data[2,3] (Ls/Rs); cfg3 picks
    five_channel_data[3,4]; cfg1 picks the trailing
    two_channel_data[0,1]. cfg0 has no natural surround partner
    inside the 5_X core — falls through to identity passthrough.
  • Standalone Ls/Rs surround mono walker for ACPL_1's Mode 1
    surround-driven path: parse_aspx_acpl_1_2_inner_body (5_X)
    and parse_7x_audio_data_outer (7_X) now persist the joint-MDCT
    residual pair (sSMP,3 / sSMP,4 per Table 181) on
    tools.acpl_1_residual_pair: [Option<(u32, Vec<f32>)>; 2]. The
    decoder's 5_X / 7_X ACPL_1 dispatch IMDCTs them into Ls / Rs PCM
    carriers and feeds them as the x3 / x4 inputs of Pseudocode
    117 — replacing the round-37 silence placeholder for slots 3 / 4.
    ASPX_ACPL_2 mode never emits a residual pair (no max_sfb_master
    in its walker), so the detach is None and surround stays at
    silence as before.
  • New tests (+8): extract_sap_abcd_mode_zero_returns_identity,
    extract_sap_abcd_mode_one_swaps_per_sfb_on_ms_used,
    extract_sap_abcd_mode_three_pair_dpcm_decode,
    extract_sap_abcd_mode_three_unused_bands_pass_through,
    extract_sap_abcd_mode_three_delta_code_time_cross_group,
    sap_coeffs_identity_helper,
    dispatch_7x_additional_pair_sap_identity_routes_partner_and_additional,
    dispatch_acpl_5x_pair_with_real_ls_rs_carriers_emits_surround_energy.
    The existing
    parse_5x_aspx_acpl_1_non_iframe_walks_three_channel_data grew an
    assertion that tools.acpl_1_residual_pair[0/1].is_some() after
    the inner walker runs. 568 tests total (was 560).

• Round 39 — 5_X SIMPLE/ASPX cfg0/cfg1/cfg3 dispatch helpers +
  7_X SIMPLE/ASPX additional-channel pair render (TS 103 190-1
  §5.3.4.3.1 / Table 180 columns 0/1/3 + §5.3.4.4.1 / Table 182):

  • Ac4Decoder::dispatch_5x_cfg0_simple_aspx — IMDCTs each
    two_channel_data.scaled_spec_per_channel[0..2] into PCM slots
    per the 1-bit b_2ch_mode: false -> tcd_a→[0,1] (L,R) /
    tcd_b→[3,4] (Ls,Rs); true -> tcd_a→[0,3] (L,Ls) /
    tcd_b→[1,4] (R,Rs). cfg0_centre_mono.scaled_spec (the trailing
    mono_data(0)) lands on slot 2 (C).
  • Ac4Decoder::dispatch_5x_cfg1_simple_aspx — IMDCTs
    three_channel_data[0..3] into slots 0/1/2 (L/R/C) and
    two_channel_data[0..2] into slots 3/4 (Ls/Rs).
  • Ac4Decoder::dispatch_5x_cfg3_simple_aspx — IMDCTs
    five_channel_data[0..5] straight into slots 0..4 (L/R/C/Ls/Rs).
  • Ac4Decoder::dispatch_7x_additional_channel_pair — IMDCTs the
    seven_x_additional_channel_data.scaled_spec_per_channel[0..2]
    into PCM slots 5 / 6 (the F / G preliminary outputs in Table 182).
    SAP companding (Table 183 a,b,c,d) is the identity for now —
    b_use_sap_add_ch == false collapses the matrix; the explicit
    SAP coefficient extraction lands in a future round.
  • Ac4Decoder::receive_frame switch: round 38's cfg2-only branch
    now selects across Cfg0/Cfg1/Cfg2/Cfg3 based on the parsed
    five_x_coding_config. Mutually exclusive with the ACPL_3 / pair
    paths via the existing five_x_simple_aspx_active gate. The
    7_X SIMPLE/ASPX path additionally runs the additional-channel
    pair dispatch on top of the core 5-channel mapping.
  • New tests (+9): dispatch_5x_cfg0_populates_l_r_c_ls_rs_default_2ch_mode,
    dispatch_5x_cfg0_alternate_2ch_mode_maps_to_l_ls_r_rs,
    dispatch_5x_cfg1_populates_l_r_c_ls_rs,
    dispatch_5x_cfg3_populates_l_r_c_ls_rs,
    dispatch_5x_cfg013_noop_on_length_mismatch,
    dispatch_7x_additional_pair_populates_slots_5_and_6,
    dispatch_7x_additional_pair_noop_on_length_mismatch,
    plus integration tests
    five_x_simple_cfg0_walker_populates_two_two_plus_centre,
    five_x_simple_cfg3_walker_populates_five_channel_data. Total:
    551 -> 560.

• Round 38 — LFE body decoder + cfg2_back_mono end-to-end decode +
  ACPL_3 centre channel via IMDCT (TS 103 190-1 §4.2.7.2 / §4.2.6.6
  Cfg2 / §5.7.7.6.2):

  • parse_mono_data(b_lfe=true) now walks the trailing sf_data(ASF)
    body via the new decode_asf_long_lfe_body_with_max_sfb_lfe
    helper. sf_info_lfe() (Table 35) forces long-frame / single
    window group, so the LFE body shape is the regular ASF long-frame
    quartet (`asf_section_data + asf_spectral_data + asf_scalefac_data
    • asf_snf_data) — only the max_sfb[0] bit-width changes (n_msfbl_bitsfrom Table 106 column 4 instead ofn_msfb_bits). The dequantised + scaled spectrum lands on MonoLfeData::scaled_spec`,
      matching the round-37 non-LFE behaviour.
  • Ac4Decoder::dispatch_5x_cfg2_simple_aspx — new helper that runs
    end-to-end IMDCT + overlap-add for the 5_X SIMPLE/ASPX
    coding_config == 2 channel layout. Walks the parsed
    four_channel_data.scaled_spec_per_channel[0..4] into output
    slots 0/1/3/4 (L/R/Ls/Rs per Table 180) and the trailing
    cfg2_back_mono.scaled_spec into slot 2 (C). Fires when
    five_x_mode in {Simple, Aspx} and five_x_coding_config == Cfg2FourMono. Cfg0 / Cfg1 / Cfg3 dispatch helpers remain deferred.
  • Ac4Decoder::receive_frame ACPL_3 path: replaces the round-37
    silence-placeholder for the centre carrier with an IMDCT of
    cfg0_centre_mono.scaled_spec via imdct_mono_lfe_data_f32. The
    Pseudocode 118 multichannel synthesis now emits a non-silent
    centre when the trailing mono_data(0) body decoded successfully;
    falls back to silence when the centre body is absent (LFE / SSF /
    Huffman miss / non-Cfg0 frame).
  • New unit + integration tests:
    • parse_mono_data_lfe_walks_sf_data_body — replaces the round-37
      "LFE body deferred" test; verifies the LFE body now decodes into
      scaled_spec for an all-zero stream.
    • dispatch_5x_cfg2_populates_l_r_c_ls_rs — cfg2 dispatch IMDCTs
      every channel into the right output slot with non-zero energy
      from per-channel ramp spectra.
    • dispatch_5x_cfg2_noop_on_length_mismatch — cfg2 dispatch
      leaves output slots untouched when the carrier transform length
      differs from the requested sample count.
    • five_x_simple_cfg2_walker_populates_four_plus_back_mono —
      integration test threading parse_5x_audio_data_outer → cfg2
      tools layout, asserting the four per-channel scaled spectra +
      back-mono body all populate cleanly.

• Round 37 — 7_X ASPX_ACPL_1 / ASPX_ACPL_2 dispatch + cfg0 centre
  end-to-end decode (TS 103 190-1 §5.7.7.6.3 Pseudocode 120 +
  §5.7.7.6.1 Pseudocode 117 cfg0 centre-channel wiring):

  • Ac4Decoder::receive_frame now dispatches the §5.7.7.6.3
    Pseudocode 120 channel-pair multichannel synthesis when the 7_X
    walker resolved seven_x_mode to AspxAcpl1 or AspxAcpl2. The
    dispatch reuses dispatch_acpl_5x_pair (the Pseudocode 117
    channel-pair core that Pseudocode 120 wraps for 7.X) — for
    ASPX_ACPL_{1,2} the additional 2 channels (z6/z7 in Pseudocode
    120) live outside the A-CPL pair so the 5-channel core (L/R/C/Ls/Rs)
    is identical between 5_X and 7_X. Slots 5..7 stay at silence until
    the additional-channel decode path lands.
  • MonoLfeData (mch::parse_mono_data) now decodes the trailing
    sf_data(ASF) body for non-LFE, ASF-frontend, long-frame /
    single-window-group mono channels — replacing the round-36
    silence-placeholder for the centre channel in the 5_X / 7_X
    Pseudocode 117 / 120 dispatches with a real IMDCT / overlap-add
    centre carrier. The new MonoLfeData::scaled_spec field carries
    the dequantised + scaled MDCT spectrum; LFE / SSF-frontend / short
    / grouped / Huffman-error cases still leave this None and
    preserve the prior outer-shell-only behaviour.
  • Ac4Decoder::dispatch_acpl_5x_pair now accepts optional
    centre_pcm / ls_pcm / rs_pcm carrier overrides — None
    falls back to the round-36 silence-placeholder; Some(...) threads
    real per-channel PCM through Pseudocode 117's z4 = x2 centre
    passthrough (and the x3in = 2*x3 / x4in = 2*x4 Ls/Rs pre-
    multiplications for ACPL_1 mode). The Cfg0 centre is wired from
    the parsed cfg0_centre_mono.scaled_spec via a new
    imdct_mono_lfe_data_f32 helper.
  • New unit + integration tests:
    • parse_mono_data_non_lfe_walks_sf_data_body — non-LFE long-frame
      ASF-frontend mono walks the trailing body into scaled_spec.
    • parse_mono_data_lfe_skips_body_walk — LFE body decoder remains
      deferred (round-36 behaviour preserved).
    • parse_mono_data_non_lfe_ssf_frontend_skips_body_walk —
      SSF-frontend mono skips the ASF body walk.
    • dispatch_acpl_5x_pair_centre_pcm_passthrough_emits_centre_energy
      — supplying a real centre PCM produces non-silent ch2 output.
    • seven_x_pair_dispatch_resolves_same_mode_as_five_x — the 7_X
      dispatch maps SevenXCodecMode::AspxAcpl{1,2} to the same
      Acpl5xPairMode selectors as the 5_X path.
    • imdct_mono_lfe_data_f32_returns_none_when_no_scaled_spec /
      imdct_mono_lfe_data_f32_imdcts_when_scaled_spec_present —
      cover the new IMDCT helper.
    • seven_x_aspx_acpl_2_walker_to_synthesis_glue — integration
      test threading parse_7x_audio_data_outer →
      acpl_data_1ch_pair → run_acpl_5x_pair_pcm for a non-iframe
      7_X ACPL_2 substream.

• Round 36 — 5_X ASPX_ACPL_1 / ASPX_ACPL_2 decoder dispatch wiring
  (TS 103 190-1 §5.7.7.6.1, Pseudocode 117):

  • Ac4Decoder::receive_frame now dispatches the §5.7.7.6.1
    Pseudocode 117 channel-pair multichannel synthesis when the 5_X
    walker resolved five_x_mode to AspxAcpl1 or AspxAcpl2 and
    populated the matching acpl_config_1ch_* + acpl_data_1ch_pair
    tools slots. The dispatch:
    • Reads L/R carrier PCM from pcm_per_channel[0]/[1] (already
      filled by the stereo ASF / ASPX decode path), zero-filling on
      absence to keep QMF analysis history consistent.
    • Resolves the active acpl_config_1ch per mode:
      acpl_config_1ch_partial for AspxAcpl1,
      acpl_config_1ch_full for AspxAcpl2 (per Tables 25 + 59).
    • Builds zero-filled centre / Ls / Rs carrier placeholders
      (centre matches the existing ACPL_3 wiring; Ls/Rs only for
      ACPL_1 mode per Acpl5xPairMode mode-vs-surround consistency
      check). The carrier-decode paths gain real signal when
      cfg0_centre_mono and the surround mono carriers acquire
      end-to-end decoders in a future round.
    • Calls acpl_synth::run_acpl_5x_pair_pcm and writes the five
      output channels (L, R, C, Ls, Rs) into pcm_per_channel[0..5],
      growing the slot vector as needed.
    • Skipped when the substream is already an AspxAcpl3 5_X frame
      (the two pipelines are mutually exclusive per Table 97).
  • The dispatch logic is extracted into a private
    Ac4Decoder::dispatch_acpl_5x_pair helper so the path can be
    unit-tested without building a full 5_X TOC + body.
  • Five new unit tests in decoder::tests cover the synthesis
    arithmetic + dispatch behaviour:
    • dispatch_acpl_5x_pair_aspx_acpl_2_emits_five_channels —
      ACPL_2 dispatch with ±2000 carrier-PCM tones produces five
      channels with non-zero L / R energy.
    • dispatch_acpl_5x_pair_aspx_acpl_1_emits_five_channels —
      ACPL_1 dispatch with zero-filled Ls/Rs surround placeholders
      still emits five-channel output.
    • dispatch_acpl_5x_pair_rejects_unaligned_sample_count —
      dispatch is a no-op when sample count isn't a multiple of
      NUM_QMF_SUBBANDS (64).
    • dispatch_acpl_5x_pair_zero_fills_missing_carriers — when L/R
      slots are None, dispatch synthesises silence-grade output
      from zero-filled fallbacks.
    • dispatch_acpl_5x_pair_resolves_partial_for_aspx_acpl_1 —
      regression check that qmf_band differs between partial
      (1..8) and full (always 0) configs per Table 59.
  • Test count: 535 → 540 (+5).
  • The §5.7.7.6.2 ACPL_3 dispatch (round 34 — Pseudocode 118)
    remains unchanged; the new ACPL_1 / ACPL_2 path takes its place
    when those modes are signalled.

• Round 35 — ETSI float-table validation suite (TS 103 190-1
  Annex C.1, C.3, C.11 + Annex D.2, D.3):

  • New parse_c_float_arrays() parser in tests/etsi_table_validation.rs
    extracts every const float / const float32 array (1-D and 2-D) from
    the ETSI accompaniment file docs/audio/ac4/ts_10319001v010401p0-tables.c.
    The existing integer parser (round 20) bailed on the float keyword, so
    the five float-typed reference tables had been silently un-validated.
  • New assert_float_table() / assert_complex_float_table() helpers
    compare Rust &[f32] (and &[(f32, f32)]) constants against the
    parsed reference under a 1 ppm absolute / 10 ppm relative epsilon —
    tight enough to catch a single-digit transcription typo in the visible
    decimal prefix, loose enough that f32 literal-rounding noise is
    invisible.
  • Four new tests:
    • etsi_source_parses_floats — sanity-check that the 5 expected float
      arrays are parsed with the right lengths (20, 32, 256, 640, 1024).
    • validate_ssf_float_tables — POST_GAIN_LUT[20],
      PRED_GAIN_QUANT_TAB[32], RANDOM_NOISE_TABLE[256] against the
      Annex C.1 / C.3 / C.11 reference data.
    • validate_qmf_window — the 640-entry Annex D.3 QWIN prototype
      window against the reference.
    • validate_aspx_noise_table — the Annex D.2 ASPX_NOISE[512][2]
      complex-noise table against the reference (flattened row-major to
      compare &[(f32, f32)] against the parsed flat Vec<f32>).
  • Outcome: all 1,860 published float reference values
    (20 + 32 + 256 + 640 + 1,024) cleared the epsilon test on first run,
    proving the existing transcriptions in ssf_tables.rs, qmf.rs, and
    aspx_noise.rs are byte-correct against the ETSI accompaniment data.
    Future regressions caught at compile-time of the test target rather
    than at decode-time of a fixture.

• Round 34 — FIXVAR / VARFIX / VARVAR atsg border derivation + SNF injection + 5_X ASPX_ACPL_3 synthesis (TS 103 190-1 §5.7.6.3.3.2 + §5.1.4 + §5.7.7.6.2):

  • New derive_fixvar_atsg() (§5.7.6.3.3.2 Pseudocode 77, FIXVAR arm): builds
    the signal-envelope border vector right-to-left from T - var_bord_right using
    rel_bord_right[] deltas, then reverses to ascending order.
  • New derive_varfix_atsg() (Pseudocode 77, VARFIX arm): builds left-to-right
    from var_bord_left using rel_bord_left[] deltas, appends T.
  • New derive_varvar_atsg() (Pseudocode 77, VARVAR arm): left-side anchors +
    right-side internal anchors + T, totalling num_env + 1 entries.
  • New derive_atsg_borders() dispatcher: routes FIXFIX / FIXVAR / VARFIX /
    VARVAR framing to the matching derivation and computes noise borders
    ([0, T] for 1 noise envelope, [0, mid, T] for 2).
  • Both the TNS path and envelope-adjustment path in decoder.rs now call
    derive_atsg_borders instead of the FIXFIX-only derive_fixfix_atsg, enabling
    A-SPX bandwidth extension for all four interval classes.
  • New inject_snf_noise() in asf_data.rs (§5.1.4 SNF): injects
    shaped noise (gain = 2^((idx * 1.5 - 84) / 4)) into zero-energy MDCT
    bins using a 16-bit LCG (multiplier 69069, addend 1). Previously the
    parse_asf_snf_data() result was discarded; now it is consumed by the
    long-mono ASF decode path.
  • 5_X ASPX_ACPL_3 synthesis wired into Ac4Decoder::receive_frame:
    Ac4Decoder gains two new persistent fields (acpl_5x_pair_state,
    acpl_5x_mch_state); when the walker populates acpl_config_2ch +
    acpl_data_2ch + stereo carrier spectra, run_acpl_5x_mch_pcm
    (Pseudocode 118) runs and fills pcm_per_channel[0..5] with L/R/C/Ls/Rs.
  • Unit tests: derive_fixvar_atsg (2 cases + reject), derive_varfix_atsg
    (2 cases + reject), derive_varvar_atsg (2 cases), derive_atsg_borders
    dispatch (FIXFIX + FIXVAR), inject_snf_noise (fill, gain formula, LCG).

Changed

• register entry point unified on RuntimeContext (task #502).
  The legacy pub fn register(reg: &mut CodecRegistry) is renamed to
  register_codecs and a new pub fn register(ctx: &mut oxideav_core::RuntimeContext) calls it internally. Breaking change
  for direct callers passing a CodecRegistry; switch to either the
  new RuntimeContext entry or the explicit register_codecs name.