Run any Skill in Manus with one click

flight-data-analysis

Analyze Dynon SkyView or Garmin GDU 460 flight data CSVs. Fuel pressure stability, MAP correlation, sensor diagnostics, pump performance comparison, engine trending.

Run Skill in Manus

Overview

Analyze Dynon SkyView or Garmin GDU 460 flight data CSVs. Fuel pressure stability, MAP correlation, sensor diagnostics, pump performance comparison, engine trending.

Install command

npx skills add https://github.com/sritchie/N720AK --skill flight-data-analysis

Copy and paste this command into Claude Code to install the skill

Source

sritchie/N720AK

Stars1

Forks0

UpdatedMarch 18, 2026 at 18:58

SKILL.md

readonly

name	flight-data-analysis
description	Analyze Dynon SkyView or Garmin GDU 460 flight data CSVs. Fuel pressure stability, MAP correlation, sensor diagnostics, pump performance comparison, engine trending.
argument-hint	["path to CSV file(s) and what to analyze"]
user-invocable	true

Flight Data Analysis

Analyze flight data from Dynon SkyView or Garmin GDU 460 CSV logs.

Input

$ARGUMENTS

Context

Read CLAUDE.md sections on:

"Dynon Fuel Pressure: DIFF vs Gauge Modes"
"Dynon MAP Sensor Details"
"Flight Data CSV Formats"
"Key Fuel System Findings"

These contain critical details about how N720AK's sensors work, the DIFF computation, known thresholds, and current pump/regulator configuration.

Step 1: Identify the Data

Determine the file format:
- Dynon: First row is column headers, Session Time is the first column, Manifold Pressure (inHg) present
- Garmin: First row starts with #airframe_info, column headers on row 2, data starts row 4, uses Manifold Press (inch Hg)
- EFII System32: First row starts with #System32 config info, column headers on row 2, data starts row 3, 1 Hz sample rate. Uses MAP1/MAP2 (inHg), AFR (air-fuel ratio from wideband O2), RPM1/RPM2, FUEL FLOW (GPH), FUEL TRIM (%), INJ DUTY (%), IGNITION (timing degrees), TPS (%), IAT1/IAT2 (°F), ENGTEMP (°F), BATT1/BATT2 (V), PUMP1/PUMP2 (ON/OFF), INJ EN1/INJ EN2 (RUN/STOP), HP, TORQUE. Date format: MM/DD/YYYY, Time: HH:MM:SS. No altitude column — cross-reference with Dynon/Garmin data if altitude needed. Config metadata in row 1 includes cylinder count, MAP sensor type, HP rating, injector size, and mag positions.
Find flight segments by RPM transitions (RPM > 500 = engine on)
Report: number of flights, duration of each, date/time range, altitude range

Step 2: Load and Parse

Use uv run --with numpy --with matplotlib python3 for all analysis. Key helper:

def safe_float(row, col):
    val = row.get(col, '') or ''
    try: return float(val.strip())
    except: return None

def p_atm_inhg(alt_ft):
    """Standard atmosphere pressure in inHg from pressure altitude."""
    return 29.92 * (1 - alt_ft / 145442.0) ** 5.25588

Computing Injector Differential

The quantity we always care about is injector differential = P_fuel_absolute - MAP_absolute (in PSI). How to get it depends on the data source:

Dynon DIFF mode (current N720AK setup, post 2026-02):

Fuel Pressure (PSI) column IS the injector differential — use directly
No conversion needed

Dynon gauge mode (old N720AK data, pre 2026-02):

Fuel Pressure (PSI) = gauge = P_fuel - P_atm
inj_diff = gauge_FP + (P_atm_inHg - MAP_inHg) * 0.49115
P_atm from pressure altitude via standard atmosphere

Garmin (always gauge):

Fuel Press (PSI) = gauge = P_fuel - P_atm
Same formula: inj_diff = gauge_FP + (P_atm_inHg - MAP_inHg) * 0.49115

How to tell DIFF vs gauge mode: On the ground before engine start, if FP = ~35 PSI and MAP = ambient (~24-25 inHg at KBDU), it's DIFF mode (because diff = absolute_FP - MAP ≈ 35). If FP = ~35 PSI and that equals what you'd expect for gauge (pumps on, no flow), it could be either — check once the engine starts. At ground idle (MAP ~11), DIFF mode reads ~35 PSI while gauge mode reads ~29 PSI.

Step 3: Core Analysis — Fuel Pressure vs MAP

This is the PRIMARY diagnostic. Always produce this analysis:

Scatter plot: Injector differential vs MAP (inHg) for all samples with RPM > 1500
Linear regression: slope (PSI/inHg) and r²
Binned averages: Mean + std dev of injector diff for each integer MAP value
Cruise stats: Average, std dev, and range of injector diff for MAP 18-24 inHg

Interpretation:

A perfect regulator/pump combo produces a flat horizontal line at 35 PSI
Negative slope = pump can't maintain pressure as fuel demand increases with power
slope > -0.05: excellent (Walbro 392 territory)
slope -0.05 to -0.15: acceptable (Walbro 391 territory)
slope < -0.20: pump struggling (Walbro 393 territory)

Step 4: Time Series

Plot MAP and injector differential vs flight time on a dual-axis chart.

Look for:

Blank/missing data: In Dynon data, empty strings mean the sensor went invalid (NOT zero). Mark these visually (orange bands). If MAP blanks, check if it's near 3.05 inHg (the min_val=1.5 PSI threshold on sensor 100434-000).
Transients: Rapid FP drops during power changes — may explain pump cutover events
Steady-state sag: Does FP settle at a lower value during extended cruise?

Step 5: Additional Analysis (as requested)

CHT/EGT analysis: Spread, trends, leaning state
Fuel burn: Total fuel flow integration, GPH averages by phase
Pump switchover detection (Garmin only): Check Fl Pmp 1 Amps, Fl Pmp 2 Amps, FUEL PP 2 ON (discrete)
Multi-flight comparison: Overlay binned FP-vs-MAP curves from different flights/configurations
Before/after comparison: Compare old vs new regulator, old vs new pumps, etc.
EFII analysis: AFR trends (target ~14.7 stoich for cruise, richer for climb/takeoff), fuel trim % (how much the ECU is correcting), injector duty cycle, ignition timing advance, MAP1 vs MAP2 correlation, pump status transitions, HP/torque estimates. EFII data has no altitude — merge with Dynon/Garmin by timestamp if altitude-dependent analysis needed.

Step 6: Generate Graphs

Always use uv run --with matplotlib --with numpy python3 and save to /tmp/ then copy to ~/Downloads/ and open them.

Standard graph set:

Time series (dual panel: MAP top, FP bottom, shared x-axis)
Scatter (injector diff vs MAP with trend line and r²)
Binned comparison (error bars, multiple datasets if comparing)

Step 7: Report

Provide a concise summary with:

Flight date, duration, altitude range
Cruise injector differential: average, std dev
MAP slope and r²
Any anomalies (blanking events, transients, pump switchovers)
Comparison to baseline if applicable

Existing Script

There is also scripts/regulator_diagnostic.py which automates much of this. Run it with:

cd ~/code/rv10
uv run --with numpy --with matplotlib python3 scripts/regulator_diagnostic.py /path/to/log.csv

For Garmin data add --alt-correct. For multi-flight: pass multiple CSV paths. For altitude correction scan: --alt-scan.

The script auto-detects Dynon vs Garmin format.

Reference Data

N88810 (Garmin, Walbro 392): ~/Downloads/log_20260308_*.csv — the gold standard baseline
N720AK old regulator (Walbro 393): ~/Downloads/2011-01-01-N720AK-*.csv
N720AK new regulator (Walbro 391): ~/Downloads/2026-03-16-N720AK-*.csv

Flight Data Analysis

Analyze flight data from Dynon SkyView or Garmin GDU 460 CSV logs.

Input

$ARGUMENTS

Context

Read CLAUDE.md sections on:

"Dynon Fuel Pressure: DIFF vs Gauge Modes"
"Dynon MAP Sensor Details"
"Flight Data CSV Formats"
"Key Fuel System Findings"

These contain critical details about how N720AK's sensors work, the DIFF computation, known thresholds, and current pump/regulator configuration.

Step 1: Identify the Data

Determine the file format:
- Dynon: First row is column headers, Session Time is the first column, Manifold Pressure (inHg) present
- Garmin: First row starts with #airframe_info, column headers on row 2, data starts row 4, uses Manifold Press (inch Hg)
- EFII System32: First row starts with #System32 config info, column headers on row 2, data starts row 3, 1 Hz sample rate. Uses MAP1/MAP2 (inHg), AFR (air-fuel ratio from wideband O2), RPM1/RPM2, FUEL FLOW (GPH), FUEL TRIM (%), INJ DUTY (%), IGNITION (timing degrees), TPS (%), IAT1/IAT2 (°F), ENGTEMP (°F), BATT1/BATT2 (V), PUMP1/PUMP2 (ON/OFF), INJ EN1/INJ EN2 (RUN/STOP), HP, TORQUE. Date format: MM/DD/YYYY, Time: HH:MM:SS. No altitude column — cross-reference with Dynon/Garmin data if altitude needed. Config metadata in row 1 includes cylinder count, MAP sensor type, HP rating, injector size, and mag positions.
Find flight segments by RPM transitions (RPM > 500 = engine on)
Report: number of flights, duration of each, date/time range, altitude range

Step 2: Load and Parse

Use uv run --with numpy --with matplotlib python3 for all analysis. Key helper:

def safe_float(row, col):
    val = row.get(col, '') or ''
    try: return float(val.strip())
    except: return None

def p_atm_inhg(alt_ft):
    """Standard atmosphere pressure in inHg from pressure altitude."""
    return 29.92 * (1 - alt_ft / 145442.0) ** 5.25588

Computing Injector Differential

The quantity we always care about is injector differential = P_fuel_absolute - MAP_absolute (in PSI). How to get it depends on the data source:

Dynon DIFF mode (current N720AK setup, post 2026-02):

Fuel Pressure (PSI) column IS the injector differential — use directly
No conversion needed

Dynon gauge mode (old N720AK data, pre 2026-02):

Fuel Pressure (PSI) = gauge = P_fuel - P_atm
inj_diff = gauge_FP + (P_atm_inHg - MAP_inHg) * 0.49115
P_atm from pressure altitude via standard atmosphere

Garmin (always gauge):

Fuel Press (PSI) = gauge = P_fuel - P_atm
Same formula: inj_diff = gauge_FP + (P_atm_inHg - MAP_inHg) * 0.49115

Step 3: Core Analysis — Fuel Pressure vs MAP

This is the PRIMARY diagnostic. Always produce this analysis:

Scatter plot: Injector differential vs MAP (inHg) for all samples with RPM > 1500
Linear regression: slope (PSI/inHg) and r²
Binned averages: Mean + std dev of injector diff for each integer MAP value
Cruise stats: Average, std dev, and range of injector diff for MAP 18-24 inHg

Interpretation:

A perfect regulator/pump combo produces a flat horizontal line at 35 PSI
Negative slope = pump can't maintain pressure as fuel demand increases with power
slope > -0.05: excellent (Walbro 392 territory)
slope -0.05 to -0.15: acceptable (Walbro 391 territory)
slope < -0.20: pump struggling (Walbro 393 territory)

Step 4: Time Series

Plot MAP and injector differential vs flight time on a dual-axis chart.

Look for:

Blank/missing data: In Dynon data, empty strings mean the sensor went invalid (NOT zero). Mark these visually (orange bands). If MAP blanks, check if it's near 3.05 inHg (the min_val=1.5 PSI threshold on sensor 100434-000).
Transients: Rapid FP drops during power changes — may explain pump cutover events
Steady-state sag: Does FP settle at a lower value during extended cruise?

Step 5: Additional Analysis (as requested)

CHT/EGT analysis: Spread, trends, leaning state
Fuel burn: Total fuel flow integration, GPH averages by phase
Pump switchover detection (Garmin only): Check Fl Pmp 1 Amps, Fl Pmp 2 Amps, FUEL PP 2 ON (discrete)
Multi-flight comparison: Overlay binned FP-vs-MAP curves from different flights/configurations
Before/after comparison: Compare old vs new regulator, old vs new pumps, etc.
EFII analysis: AFR trends (target ~14.7 stoich for cruise, richer for climb/takeoff), fuel trim % (how much the ECU is correcting), injector duty cycle, ignition timing advance, MAP1 vs MAP2 correlation, pump status transitions, HP/torque estimates. EFII data has no altitude — merge with Dynon/Garmin by timestamp if altitude-dependent analysis needed.

Step 6: Generate Graphs

Always use uv run --with matplotlib --with numpy python3 and save to /tmp/ then copy to ~/Downloads/ and open them.

Standard graph set:

Time series (dual panel: MAP top, FP bottom, shared x-axis)
Scatter (injector diff vs MAP with trend line and r²)
Binned comparison (error bars, multiple datasets if comparing)

Step 7: Report

Provide a concise summary with:

Flight date, duration, altitude range
Cruise injector differential: average, std dev
MAP slope and r²
Any anomalies (blanking events, transients, pump switchovers)
Comparison to baseline if applicable

Existing Script

There is also scripts/regulator_diagnostic.py which automates much of this. Run it with:

cd ~/code/rv10
uv run --with numpy --with matplotlib python3 scripts/regulator_diagnostic.py /path/to/log.csv

For Garmin data add --alt-correct. For multi-flight: pass multiple CSV paths. For altitude correction scan: --alt-scan.

The script auto-detects Dynon vs Garmin format.

Reference Data

N88810 (Garmin, Walbro 392): ~/Downloads/log_20260308_*.csv — the gold standard baseline
N720AK old regulator (Walbro 393): ~/Downloads/2011-01-01-N720AK-*.csv
N720AK new regulator (Walbro 391): ~/Downloads/2026-03-16-N720AK-*.csv

flight-data-analysis

Flight Data Analysis

Input

Context

Step 1: Identify the Data

Step 2: Load and Parse

Computing Injector Differential

Step 3: Core Analysis — Fuel Pressure vs MAP

Step 4: Time Series

Step 5: Additional Analysis (as requested)

Step 6: Generate Graphs

Step 7: Report

Existing Script

Reference Data

More from this repository

More from this repository

Flight Data Analysis

Input

Context

Step 1: Identify the Data

Step 2: Load and Parse

Computing Injector Differential

Step 3: Core Analysis — Fuel Pressure vs MAP

Step 4: Time Series

Step 5: Additional Analysis (as requested)

Step 6: Generate Graphs

Step 7: Report

Existing Script

Reference Data