fabi/Sylpheed
1
0
Fork 0
Code Issues 9 Pull Requests Actions Packages Projects Releases Wiki Activity

re: re-run the two findings resting on the truncating reader

Browse Source 
Pulse ratio HOLDS: 0.1000 (f6b, 16 clean cycles) and 0.0993 (f6). 60 leaf
units = 120 title units = ptbtn00f's declared loop. The independent leg
under leaf/title = 0.5 stands.

Unit 10's conclusion survives but MY NUMBERS WERE WRONG. I quoted
254.0-256.9 / 253.9-254.9 as the implied parent's range; those were the rows
I had printed, every 20th frame, not the series. Real first-cycle figures
are 250.9-260.5 (f6, n=1128) and 253.1-255.0 (f6b, n=560).

Post-wrap the spread widens to 237-283, but only in f6 -- the run with
dropped frames -- while f6b holds at median 254.3. That is my linear phase
model drifting, not the parent varying; a parent that varied would degrade
in both.

Same shape as an error already in my record: a printed subset doing the work
of the population.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Jc4pciRArGHfxGGhEbwp5t
This commit is contained in:
sylph-decoder
2026-09-02 21:10:31 +00:00
parent 568a2c004f
commit 1f149a1863
1 changed files with 51 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
docs/re/reverify-pulse-and-decomposition.md Normal file
View File

@@ -0,0 +1,51 @@
# Re-verifying the two findings that rested on the truncating reader
The port asked me to **re-run these rather than reason about them**, having
watched me retract a "these are unaffected" argument once already. Right call.
**Instrument:** ⟨capture⟩, same logs, re-read by
[`read_draws.py`](../../tools/re-capture/read_draws.py).
## ✅ The `ptbtn00f` pulse ratio — holds
| | median pulse | sweep loop | ratio | leaf units |
|---|---|---|---|---|
| `f6b` | 60 frames, 16 clean cycles | 600 | **0.1000** | 60.0 |
| `f6` | 116 frames | 1168 | **0.0993** | 59.6 |
Unchanged by the reader. **60 leaf units × 2 = 120 title units =
`ptbtn00f`'s declared loop.** The independent leg under `leaf/title = 0.5`
stands.
## 🟡 Unit 10's decomposition — the conclusion survives, **my numbers were wrong**
Unit 10 said the implied parent "pins at 255.0 and holds: **254.0–256.9** (`f6`)
and **253.9–254.9** (`f6b`)". Those came from the **rows I printed** — every 20th
frame — quoted as if they were the range of the whole series. The real figures:
| | first cycle, post-ramp | after the leaf wraps |
|---|---|---|
| `f6b` | 253.1–255.0, median **254.4**, n=560 | median 254.3, one 41.3 outlier |
| `f6` | 250.9–260.5, median **252.9**, n=1128 | 237.3–283.2, median 267.9, n=584 |
**The conclusion holds on the first cycle** — a flat ~254 across 560 and 1128
frames while the drawn alpha swings 8→255→132. The parent multiplies in.
**The post-wrap spread is my phase model, not the parent.** `t = (f−onset)×rate`
accumulates error, and it degrades **only in `f6`** — the run whose pulse periods
show dropped frames (65 and 93 against a median 116). `f6b`, which paces cleanly
(all pulse periods exactly 60), stays at median 254.3 after its wrap. A parent
that actually varied would degrade in both.
So: ✅ the finding, with the bounds above and **restricted to the first cycle**;
the post-wrap series needs a phase model fitted per cycle, which I have not done.
## The error worth naming
Quoting a range off the sampled rows I happened to print, as though off the
series. It flatters — sampling every 20th frame skips the excursions — and it is
**the same shape as an error already in my record**: presenting
`sorted(set(...))[:9]` as a trajectory when it was a 14-sample tail of 1754.
Both times the printed subset was doing the work of the population, and both
times nothing in the output looked wrong.