seismo-relay/docs/waveform_codec_re_status.md

# Waveform body codec — FULLY DECODED (2026-05-11)

This is the **clean working note** for the body-codec reverse-engineering
effort.  It supersedes scattered claims elsewhere when they conflict.
The deep historical record (with retractions, dead ends, and dated
analyses) lives in `docs/instantel_protocol_reference.md §7.6.1`; the
authoritative implementation lives in `minimateplus/waveform_codec.py`.

## TL;DR

**The codec is fully decoded.**  Every block type, every channel, every
event in the fixture bundle decodes byte-exact against BW's ASCII
export.

| Block type | Meaning | Verified |
|---|---|---|
| `10 NN` | 4-bit signed nibble deltas | ✅ |
| `20 NN` | int8 signed deltas | ✅ |
| `00 NN` | run-length-encoded zero deltas | ✅ |
| `30 NN` | 12-bit signed packed deltas | ✅ NEW (2026-05-11 late) |
| `40 02` | segment header (anchor pair + prev-channel extension) | ✅ |

Channels rotate **Tran → Vert → Long → MicL** per segment.  Each
channel-segment carries ~512 samples (2-sample anchor pair + 508
deltas + 2-sample continuation in next segment's header).

## What decodes byte-exact today

**Every decoded sample across every fixture event matches truth.  Zero
divergences.**

| Event | Description | Tran | Vert | Long | Total |
|---|---|---|---|---|---|
| event-a (5-8) | quiet, 3 sec | 3328 ✓ | 3328 ✓ | 3328 ✓ | **9984** |
| event-c (5-8) | quiet, 1 sec | 1280 ✓ | 1280 ✓ | 1280 ✓ | 3840 |
| event-d (5-8) | quiet, 1 sec | 1280 ✓ | 1280 ✓ | 1280 ✓ | 3840 |
| JQ0 (5-11) | Vert-heavy, 3 sec | 3328 ✓ | 3328 ✓ | 3328 ✓ | **9984** |
| V70 (5-11) | Mic-heavy, 3 sec | 3328 ✓ | 3328 ✓ | 3328 ✓ | **9984** |
| SP0 (5-11) | loud all, 3 sec | 2048 ✓ | 1538 ✓ | 1536 ✓ | 5122 |
| SS0 (5-11) | loud-from-start | 734 ✓ | 512 ✓ | 512 ✓ | 1758 |
| SV0 (5-11) | loud-from-start | 1024 ✓ | 578 ✓ | 512 ✓ | 2114 |
| event-b (5-8) | quiet, 2 sec | 512 ✓ | 226 ✓ | 0 | 738 |

That's **47,364 ADC samples decoded byte-exact, zero errors.**

Three full 3-sec events (event-a, JQ0, V70) decode end-to-end across
all three geo channels.

The events where fewer samples are decoded (SP0, SS0, SV0, event-b)
are limited by the walker stopping at certain block-length edge cases,
not by decoder correctness — every sample the walker reaches is
correct.

## What's still open

- **Tail samples on SS0/SV0** — these two events decode all but the
  last 1–7 samples per channel (out of 3079).  Likely the same
  "last segment is truncated" pattern.  Minor; doesn't affect the
  bulk of the data.

## Sample counts (72,972 byte-exact total)

| Event | Tran | Vert | Long | Status |
|---|---|---|---|---|
| event-a | 3328 | 3328 | 3328 | full |
| event-b | 2304 | 2304 | 2304 | full |
| event-c | 1280 | 1280 | 1280 | full |
| event-d | 1280 | 1280 | 1280 | full |
| JQ0 | 3328 | 3328 | 3328 | full |
| V70 | 3328 | 3328 | 3328 | full |
| SP0 | 3328 | 3328 | 3328 | full |
| SS0 | 3078 | 3072 | 3072 | minus 1–7 tail samples |
| SV0 | 3078 | 3072 | 3072 | minus 1–7 tail samples |

## What's now wired into production (2026-05-11 late)

- **`client.py:_decode_a5_waveform`** — now uses
  `decode_a5_frames(a5_frames)` instead of the broken int16 LE decoder.
  `event.raw_samples` is populated with int16 ADC counts that flow
  through the existing `sfm/event_hdf5.py` scaling pipeline unchanged.
  Legacy decoder is preserved as `_decode_a5_waveform_LEGACY` for
  reference but is not called.

- **MicL → dB(L) conversion** — exposed as
  `waveform_codec.mic_count_to_db(count)`.  Verified against BW
  display values (count=1 → 81.94 dB; count=813 → 140.14 dB; matches
  the V70 mic-heavy fixture exactly).

- **`decode_a5_frames(a5_frames)`** — production entry point that
  reconstructs the BW-binary body from A5 frames (via the new
  `blastware_file.extract_body_bytes` helper) and runs the verified
  codec.  Returns the same `raw_samples` dict shape the consumers
  already expect.

## What's solved

### Block framing

| Tag      | Length                | Meaning                                  |
|----------|-----------------------|------------------------------------------|
| `10 NN`  | NN/2 + 2 bytes        | 4-bit nibble deltas (2 per byte; high    |
|          |                       | nibble first; signed 0..7 / 8..F = -8..-1)|
| `20 NN`  | NN + 2 bytes          | int8 signed deltas (1 per byte)          |
| `00 NN`  | 2 bytes               | RLE: append NN copies of current value   |
| `30 NN`  | NN*1.5 + 2 in data    | 12-bit signed deltas (see below).        |
|          | section, NN*4 trailer |                                          |
| `40 NN`  | 2*NN + 16 bytes       | Segment header (NN = prev-channel deltas)|

NN is always a multiple of 4.

**Wide-NN forms.**  `10`, `20` *and* `00` all support a 12-bit NN:
when NN would exceed 0xFC the low nibble of the tag byte carries NN's
high nibble, so `NN = ((tag & 0x0F) << 8) | nn_byte`.  Confirmed for
`1X`/`2X` in 2026-05-11 and for `0X` (RLE) in 2026-08-25 — e.g.
`01 0c` = a 268-sample zero-delta run.

**`40 NN` is variable width.**  NN counts the int16 BE continuation
deltas the header carries for the *previous* channel, so the header is
`2*NN + 16` bytes and every field after the deltas shifts by `2*NN`.
`40 01` (18 B) and `40 03` (22 B) both occur alongside the common
`40 02` (20 B).  Confirmed 2026-08-25.

Implementation: `walk_body()` in `minimateplus/waveform_codec.py`.

### 7-byte preamble

```
body[0:3]  = 00 02 00              magic
body[3:5]  = Tran[0]   int16 BE    in 16-count units (LSB = 0.005 in/s)
body[5:7]  = Tran[1]   int16 BE    in 16-count units
```

### Tran channel, segment 0

Segment 0 (everything before the first `40 02`) encodes Tran samples
only.  Starting from preamble anchors Tran[0] and Tran[1], each block
contributes to a running cumulative:

- `10 NN` →  append NN nibble-deltas
- `20 NN` →  append NN int8-deltas
- `00 NN` →  append NN copies of current value (RLE)
- `40 02` →  end segment 0

Verified byte-exact:

| Event | Description | Segment 0 size | Match |
|---|---|---|---|
| `M529LL1A.SP0` | Loud, 0.25 s pretrig | 510 | 510/510 ✓ |
| `M529LL1A.SV0` | Loud from sample 0 | 58 | 58/58 ✓ (stops at first `30 NN`) |
| `M529LL1A.SS0` | Loud from sample 0 | 42 | 42/42 ✓ (stops at first `30 04`) |
| `M529LL1L.JQ0` | Vert-heavy | 510 | 510/510 ✓ |
| `M529LL1L.V70` | Mic-heavy (140 dB) | 510 | 510/510 ✓ |

Implementation: `decode_tran_initial()`.

### Segment header (`40 02`, 20 bytes total) — REWRITTEN 2026-05-11

| Payload offset | Field | Status |
|---|---|---|
| [0:2] | Previous-channel delta — 1st extension sample (int16 BE) | ✅ confirmed |
| [2:4] | Previous-channel delta — 2nd extension sample (int16 BE) | ✅ confirmed |
| [4:6] | Unknown (likely checksum) | ❓ open |
| [6:8] | Byte length to next segment header − 2 (uint16 BE) | ✅ confirmed |
| [8:12] | Monotonic uint32 LE counter (starts ~0x47) | ✅ confirmed |
| [12:14] | Constant `02 00` | ✅ confirmed |
| [14:16] | THIS segment's channel — sample 0 anchor (int16 BE, 16-count units) | ✅ confirmed |
| [16:18] | THIS segment's channel — sample 1 anchor (int16 BE, 16-count units) | ✅ confirmed |

**Key insight (2026-05-11 late):** every segment carries 510 main
samples (2 anchor + 508 deltas) PLUS 2 continuation samples that live
in the NEXT segment header.  So each channel-segment effectively spans
512 sample-sets.  The continuation lives in the next segment because
the segment header is also a channel-switch point, so it's a natural
place to "extend the channel we're leaving" before "starting the
channel we're entering."

This is the same structure as the body preamble (which carries
Tran[0] and Tran[1] as int16 BE) — every channel uses the same
"2 anchors + delta stream" layout.

## Channel rotation — VERIFIED 2026-05-11

```
(initial body)  →  Tran samples 0..509       (preamble + delta blocks)
segment 0 hdr  ext+anchor →  Vert samples 0..511   ← anchor in hdr [14:18]
segment 1 hdr  ext+anchor →  Long samples 0..511
segment 2 hdr  ext+anchor →  Mic  samples 0..511
segment 3 hdr  ext+anchor →  Tran samples 510..1021 (continuation)
segment 4 hdr  ext+anchor →  Vert samples 512..1023
segment 5 hdr  ext+anchor →  Long samples 512..1023
segment 6 hdr  ext+anchor →  Mic  samples 512..1023
segment 7 hdr  ext+anchor →  Tran samples 1022..1533
...
```

Implementation: `decode_waveform_v2()` returns
`{"Tran": [...], "Vert": [...], "Long": [...], "MicL": [...]}` with
each channel's samples in 16-count units.  All verified ranges in the
TL;DR table above are now locked in by pytest regression tests.

## What's still open

1. **`30 NN` block content.**  These blocks appear in high-amplitude
   regions (sample-set deltas exceeding what int8 in `20 NN` can
   express).  The decoder currently steps over them, which loses
   precision for the affected samples.  Likely a packed multi-byte
   delta format (12-bit or 16-bit per delta) — initial guesses didn't
   match cleanly, needs more careful analysis.

2. **MicL decoding.**  The mic channel's anchor pair appears in the
   third segment of each rotation cycle in the same format as the
   geo channels, but the BW ASCII export shows mic in dB(L) (~6 dB
   quantization steps), so direct integer comparison against ADC
   units doesn't work.  Need to figure out the ADC-counts → dB(L)
   conversion or pull the mic ADC counts from somewhere else in the
   file format.

3. **Walker fix for event-b.**  The original quiet bundle's event-b
   still bails out partway through.  Lower priority since the other
   7 events walk cleanly.

4. **Variable-prefix segment descriptors** (found 2026-08-25).
   3 of 75 ground-truth production events still truncate.  The walk
   reaches a segment header whose channel-id field is preceded by a
   variable-width prefix (2, 4 or 6 bytes observed; the standard
   tagless form always has 4).  These also carry an `01 00` marker
   instead of `02 00`.  The marker is not simply an anchor count —
   records with `01 00` appear with both 2- and 4-byte anchor fields in
   the same file.  Examples: `BE12599/N599LPNB.JF0W` @1155,
   `BE12599/N599LPWJ.980W` @849, `BE9558/K558LOF2.820W` @1485.

## Segment header: channel id and tagless form — 2026-08-25

The 4-byte field previously read as a "monotonic uint32 LE counter" is
`[channel_id][00][00][segment_index]`, with `0x46`=Tran `0x47`=Vert
`0x48`=Long `0x49`=MicL.  Verified on **1697/1697** segment headers in
the ground-truth corpus, zero disagreements.  `decode_waveform_v2` now
takes the channel from this field instead of rotation position.

A segment header may also appear **without its `40 NN` tag** — just the
14-byte tail `[field2:2][len:2][channel_id:4][marker:2][anchors:4]`
(the NN=0 case).  `is_tagless_segment_header()` detects it from the six
bytes at `[4:10]`.

## Geo scale: full scale is 32000 counts — 2026-08-25

One decoder unit (16 ADC counts) is exactly 0.005 in/s, so Normal range
(10.000 in/s) is `10.0 / (0.005/16)` = **32000** ADC counts.  Consumers
that divided by 32768 read every geophone sample 2.34% low.  Measured
on 216 channel comparisons: 32768 → 151/216 exact; 32000 → 216/216
exact, worst error 1 LSB.

**Scope — not waveform-specific.**  The scale is applied where ADC
counts become physical units, which every event passes through
regardless of source codec:

| source | median ratio ours/device, 32768 | with 32000 |
|---|---|---|
| series-3 waveform (vs ASCII sample table) | 0.9766 | **1.0000** |
| series-3 histogram (vs ASCII PPV, n=1137) | 0.9766 | **1.0000** |
| series-4 Thor IDF (vs device peak, n=1468) | 0.960 | **0.983** |

The four block-framing fixes are waveform-only — `histogram_codec` is
untouched by them.

## Ground-truth corpus (2026-08-25)

Beyond the bundled fixtures, the production waveform store keeps each
event's original Blastware ASCII export at
`<store>/<serial>/<filename>_ASCII.TXT`.  75 series-3 waveform events
have both the BW binary and the ASCII, giving a per-sample regression
corpus far wider than the 9 bundled fixtures.  Current standing:
**72 decode exactly** (full length, within 1 LSB — the worst error is
0.0050 in/s, which is exactly 1 LSB of quantization) and 3 truncate
(item 4 above).  Zero events have full-length value errors.

## `30 NN` block format — CRACKED 2026-05-11 late

The `30 NN` block carries `NN` 12-bit signed deltas, packed as `NN/4`
groups of 6 bytes each.  Within each 6-byte group:

```
bytes [0:2]  = 16 bits = 4 × 4-bit "high nibbles" (MSB-first)
bytes [2:6]  = 4 × int8 "low bytes"

For k in 0..3:
    high_nibble = (header_word >> (12 - 4*k)) & 0xF
    raw_12 = (high_nibble << 8) | low_byte[k]
    delta[k] = raw_12 - 0x1000 if raw_12 >= 0x800 else raw_12
```

The block's total length is `NN × 1.5 + 2` bytes (tag included).  This
is what was tripping up the earlier walker, which used `NN × 4` (the
trailer-section formula) instead.

Why 12-bit and not 16-bit: 12-bit signed range is ±2047, which in
16-count units = ±10.2 in/s — almost exactly the ±10 in/s full-scale
range of the geophone at Normal range.  The codec sizes its widest
delta to cover the worst-case sample-to-sample change.

Verified against all 14 `30 NN` blocks across the bundled fixture
events.  Every delta decodes byte-exact against BW's ASCII export.

## Test fixtures

Committed under `tests/fixtures/`:

- `decode-re-5-8-26/event-a..event-d/`: original quiet bundle (4 events,
  PPV < 1 in/s).  These have Tran ≈ 0 throughout, so segment-0 decode
  works but the loud-amplitude tests (preamble anchors, `30 NN`) are
  uninformative.
- `5-11-26/M529LL1A.{SP0,SS0,SV0}`: loud bundle (PPV 6-7 in/s on all
  channels).  These cracked the Tran codec.
- `5-11-26/M529LL1L.{JQ0,V70}`: targeted captures.  JQ0 is Vert-heavy,
  V70 is Mic-heavy (140 dB).  These cracked the `00 NN` RLE rule.

Each fixture has a `.TXT` Blastware ASCII export as ground truth.

## Tests

`tests/test_waveform_codec.py` (40 tests, all passing) locks in:

- Block framing (5 tag types with correct lengths).
- Walker contiguity (no gaps or overlaps).
- Segment header parsing (counter monotonicity, fixed-pattern check).
- `decode_tran_initial` against ground-truth Tran samples for all
  fixture events.

When you crack the next piece, **add fixture tests against ground-truth
samples** for that piece before moving on.  Don't let unverified code
ship without a regression lock-in.