experiment: IMU + log rate to 416 Hz for SD writer stress test#643
Closed
sritchie wants to merge 2 commits into
Closed
experiment: IMU + log rate to 416 Hz for SD writer stress test#643sritchie wants to merge 2 commits into
sritchie wants to merge 2 commits into
Conversation
Closes #637. Wire format changes from a 127-byte / 50 Hz / 115200 frame to a 138-byte frame with four active groups (Common + Time + GNSS1 + AHRS). New per-sample timestamps land in vnTimeStartupNs (ns since boot, always advancing) and vnTimeGpsNs (ns since GPS epoch, valid once dateOk). The old 5 Hz GNSS1.UTC string column is dropped. OnSpeed's EFIS serial port now opens at 921600 baud for VN-300; the new frame at 400 Hz is 55.2 kB/s on the wire, which 115200 cannot carry. Other EFIS types (Dynon, Garmin, MGL) keep their 115200 baud and their native broadcast rates (10-50 Hz). The VN-300 hardware must be reconfigured to match: write reg 75 ($VNWRG,75,1,1,1B,01E3,0200,0090,0142), reg 5 ($VNWRG,5,921600,1), then $VNWNV to persist. Coupling is intentional and fail-closed — the parser memcmps the 10-byte header so an unconfigured VN-300 produces zero vn* data rather than corrupt output. Group masks verified against the canonical vnproglib enum: - Common 0x01E3 = TimeStartup + TimeGps + AngularRate + Position + Velocity + Accel - Time 0x0200 = TimeStatus - GNSS1 0x0090 = Fix + VelNed - AHRS 0x0142 = YawPitchRoll + LinearAccelBody + YprU Changes - software/Libraries/onspeed_core/src/efis/Vn300.{h,cpp}: 138-byte frame layout; new kHeader[10]; sync detection on 0xFA 0x1B; TimeStartup/TimeGps/TimeStatus decode; szTimeUTC + formatTimeHmsMs removed - software/Libraries/onspeed_core/src/types/LogRow.h: vnTimeUtc string + kLogRowUtcTimeLen replaced by three numeric fields - software/Libraries/onspeed_core/src/proto/LogCsv.{h,cpp}: column header line and row emission use the three new numeric columns; the comma-in-vnTimeUtc assertion is removed (u64/u8 can't produce commas) - software/Libraries/onspeed_core/src/proto/LogCsvHeaderIndex.{h,cpp}: three new idx fields; require-all list grows from 26 to 28; unused TakeString helper removed - software/sketch_common/src/io/EfisSerialPort.{h,cpp}: SuVN300Data carries the three new fields; Init() picks 921600 for VN-300, 115200 for everything else - software/sketch_common/src/tasks/LogSensor.cpp: plumb new fields into LogRow; drop VN-300 utcOrNull source for the sidecar meta builder (reconstruct offline from vnTimeGpsNs if needed) - software/Libraries/onspeed_core/src/test_frames/SynthFrames.cpp: 138-byte synth frame - test/test_efis_vn300: new fixture builder, new tests for per-sample timestamps and TimeStatus; old-format rejection test - test/test_efis_vn300_crc_parity: 137-byte fuzz buffer - test/test_log_csv, test/test_log_csv_header_index: new column names, new round-trip assertions; obsolete vnTimeUtc tests dropped - test/test_efis_dispatcher, test/test_synth_frames: 138-byte builder - docs/site: vectornav.md rewritten for 400 Hz / 921600 / new VNCC config recipe; log-columns.md describes the three new columns Verification pio test -e native 1169 passed, 1 skipped pio run -e esp32s3-v4p SUCCESS (15.7% flash, 22.3% RAM) tools/regression/run_snapshot all 5 goldens match Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
THROWAWAY BRANCH — not for merge. Built to flash on the bench and exercise the LogSensor / xWriteMutex / ring under 2x the production sample-rate load. Changes - HardwareMap.h: kImuSampleRateHz 208 → 416 - IMU330.cpp: CTRL1_XL and CTRL2_G ODR bits 0101 → 0110 (next LSM6 datasheet step) - OnSpeed-Gen3-ESP32.ino: hardcode g_Config.iLogRate=416 after LoadConfig so the writer fires every IMU sample, not every 4th - SensorIO.cpp: loosen iLogRate == 208 comparisons to >= 208 so the IMU-rate-write path stays active at 416 Hz What this does NOT do (and would need to before any real merge): - Retune Madgwick beta for the new dt - Update SavGol window sizes - Refit EKFQ tuning (gyro-bias process noise scales with dt) - Update regression goldens (test_ahrs, test_ekfq_octave, the snapshot harness all bake in 208 Hz assumptions) - Update the web UI radio to expose 416 as a selectable option Verification pio run -e esp32s3-v4p SUCCESS Bench protocol: .claude/skills/bench-stress/SKILL.md
sritchie
force-pushed
the
feat/vn300-400hz-baud-bump
branch
from
b296a8a to
85b227b
Compare
Collaborator
Author
|
spoiler alert, it now can! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
THROWAWAY BRANCH — not for merge. Stacked on top of #642 (VN-300 400 Hz / 921600 baud).
Purpose
Find out whether OnSpeed's LogSensor / xWriteMutex / ring can keep up at 2x the production sample-rate load. If yes, we plan a real 416 Hz PR with proper AHRS/EKFQ retuning. If no, we know the SD-writer pipeline is the bottleneck and we plan how to fix it.
What this does
kImuSampleRateHz01010110g_Config.iLogRate== 208>= 208The VN-300 (from the base PR) already runs at 400 Hz, so this is a true 2x sustained stress on the writer.
What this does NOT do
Things that would be required for a real merge but are deliberately omitted:
test_ahrs,test_ekfq_octave, regression snapshot harness all bake in 208 Hz)So native tests will drift on this branch; that's expected and ignored.
Bench protocol
.claude/skills/bench-stress/SKILL.mdis the canonical procedure. The stress script istools/bench/stress_web_handlers.py.🤖 Generated with Claude Code