---

🧾 Overview

The factory ECU calibration for the 2005 Subaru WRX (EJ205) represents a conservative, reliability-focused baseline engineered by Denso and Subaru.

This ROM, identified by:

• ROM ID: A4TF800F
• Manufacturer: DENSO
• Year: 2003 calibration base

…is designed for:

• Emissions compliance
• Long-term engine durability
• Broad environmental tolerance
• Low-quality fuel survivability

---

⚙️ Factory Calibration Philosophy

Subaru did not optimize for power.

They optimized for:

• Worst-case fuel quality
• Heat soak conditions
• Manufacturing variance
• Driver abuse tolerance

What this means in practice:

Area	Factory Behavior
Boost	Conservative, slow ramp
Timing	Heavily retarded under load
AFR	Rich under boost (safe)
Knock Control	Extremely sensitive
WGDC	Mild and stable

---

🔍 Key Characteristics

1. Ignition Timing Strategy

• Low advance under load
• High knock sensitivity
• Rapid timing pull on feedback knock

👉 Result: Safe, but leaves power on the table

---

2. Fueling (AFR)

• Rich targets under boost (~10.5–11.0 AFR typical)
• Designed to protect engine under poor conditions

👉 Result: Cooling > Power

---

3. Boost Control

• Conservative WGDC tables
• Minimal overshoot
• Slow spool behavior

👉 Result: Predictable but underwhelming

---

4. Torque Delivery

• Soft midrange
• No aggressive torque spike

👉 Result: Daily-driver smoothness, not performance

---

⚠️ Limitations of Factory Map

• Not optimized for performance mods
• Leaves ~30–60whp unrealized
• Poor throttle response compared to tuned setups
• Inefficient boost utilization

---

🧠 Strategic Value (Project Black Pearl)

This map is your:

✔ Baseline reference
✔ Safety benchmark
✔ Diff comparison anchor

If you don’t understand this map…

👉 You are tuning blind.

---

📌 Final Verdict

The factory ROM is:

• Extremely safe
• Extremely conservative
• Intentionally underpowered

And that’s exactly why it’s the perfect starting point.

---

http://blackpearl.nexusinstruments.store/wp-content/uploads/2026/03/read_image.bin

📓 Project Black Pearl — Current Logbook Snapshot

Vehicle Platform

• Car: 2005 Subaru Impreza WRX
• Engine: EJ205
• Turbo: Mitsubishi TD04L-13T (PN 49177-04100)
• Exhaust: Blitz NUR-SPEC RX cat-back (SUS304 straight-through)

---

⚙️ Current Calibration State

(based on the latest documented targets)

Boost Target

• ~ 12 PSI

Wide Open Throttle AFR

• 11.3 – 11.5

Ignition Strategy

• Approx +1.5° midrange advance vs stock baseline
• Conservative timing ramp maintained for knock stability

Idle

• ~1000 RPM (stability-biased idle strategy)

Turbo Behavior

• Stock TD04L expected full spool:
• ~2700–3000 RPM

---

📊 System Philosophy (Active)

Black Pearl tuning loop remains:

Log → Analyze → Score → Adjust → Flash → Validate → Repeat

Safety rules currently enforced:

• No timing increase if IAM unstable
• No boost increase if AFR trends lean
• WGDC changes ≤ 2% per iteration
• Timing changes ≤ 0.5° per iteration

---

🎯 Project Performance Target

Safe Output Goal

Metric	Target
Wheel HP	~305 whp
Powerband	Strong midrange torque
Use Case	Rally-capable drivability
Reliability	Long-term engine health prioritized

---

🧠 Known System Constraints

• TD04 turbo will limit peak power potential
• Injector duty and airflow scaling must be monitored
• MAF scaling accuracy will strongly affect fueling stability
• Heat soak behavior needs validation under repeated pulls

---

📁 Current ROM Files Logged

From the uploaded ECU binaries:

• impreza_basemap_stage_1.bin
• impreza_stage_1p_vf22_13_5psi_idle1000_FINAL.bin
• impreza_city_knocksafe_eco.bin
• impreza_basemap_stage_ecuflash_model_boostpatched.bin
• impreza_basemap_tune.bin

These appear to represent different experimental calibration stages.

---

🔧 Recommended Next Logbook Entry

To move Black Pearl forward intelligently:

Run two controlled WOT pulls and log:

• RPM
• Load
• Boost target vs actual
• WGDC
• IAM
• FBKC / FLKC
• AFR
• Injector duty
• MAF voltage
• IAT

That will allow the AI analyzer to produce:

• Stability Score
• Knock Risk Score
• Boost Control Score
• Fueling Accuracy Score

---

💡 Important observation

If the car is still on the TD04, reaching 305 whp safely is extremely unlikely without a turbo upgrade.
Typical TD04 ceiling:

220–240 whp

Which means the uploaded VF22-labeled map suggests the project may already be preparing for a turbo swap stage.

---

⚡ One Question That Will Clarify the Entire Logbook

Are you currently running:

A) Stock TD04 turbo
B) VF22 turbo (as suggested by one ROM file)

That single detail changes the entire tuning strategy for Black Pearl.

A. Disregard all vf22 references. And add all known maintenance schedules with corresponding date and time to log book

📓 Project Black Pearl — Master Logbook (Updated)

Vehicle configuration baseline locked per instruction:

• Platform: 2005 Subaru Impreza WRX
• Engine: EJ205
• Turbo: TD04L-13T (stock)
• Exhaust: Blitz NUR-SPEC RX cat-back
• Tuning Framework: Project Black Pearl AI-assisted calibration protocol
• VF22 references: REMOVED / INVALIDATED

---

🧾 Calibration Baseline Entry

Date: March 10, 2026
Time: 14:00 EST

ROM State

• Stage 1 baseline calibration
• Idle target: 1000 RPM
• Boost target: 12 PSI
• WOT AFR target: 11.3–11.5
• Conservative ignition ramp

Engineering Note
TD04 turbo retained for stability phase of Black Pearl development.
Power development prioritized after full system reliability validation.

---

🔧 Master Maintenance Schedule

These entries now become part of the Black Pearl Logbook and should be updated each time work is performed.

---

🛢 Engine Oil & Filter

Interval: every 3,000 miles OR 3 months

Last Scheduled Entry

• Date: March 10, 2026
• Time: 14:05 EST
• Oil Spec: 5W-40 full synthetic (turbo safe)
• Filter: Subaru OEM / equivalent high-flow

Next Service Due

• June 10, 2026
• or +3,000 miles

Reason: EJ205 turbo lubrication reliability.

---

🔥 Spark Plugs

Interval: every 15,000 miles

Recommended spec:

• NGK BKR6EIX (Iridium)
• Gap: 0.028–0.030

Inspection Entry

• Date: March 10, 2026
• Time: 14:07 EST

Next inspection:

• September 10, 2026

---

🌬 Air Filter

Inspection Interval: every 5,000 miles

Entry

• Date: March 10, 2026
• Time: 14:08 EST

Next inspection:

• July 10, 2026

---

⛽ Fuel System

Fuel Filter Replacement

Interval:

• 30,000 miles

Entry:

• Date: March 10, 2026
• Time: 14:10 EST

Next service:

• March 10, 2028 (or mileage equivalent)

---

❄ Cooling System

Coolant Flush

Interval:

• Every 24 months

Entry:

• Date: March 10, 2026
• Time: 14:12 EST

Next service:

• March 10, 2028

Inspection items:

• Radiator hoses
• Thermostat function
• Overflow tank integrity

---

🌀 Turbo System Inspection

Interval: every 6 months

Inspection checklist:

• Compressor shaft play
• Oil feed line
• Oil return line
• Intercooler hoses
• Boost leaks

Entry:

• Date: March 10, 2026
• Time: 14:15 EST

Next inspection:

• September 10, 2026

---

🧠 ECU & Tuning System Validation

Interval: every new map revision

Entry:

• Date: March 10, 2026
• Time: 14:18 EST

Validation protocol:

1. Two clean WOT pulls
2. IAM stable
3. No sustained FBKC
4. Boost within tolerance
5. AFR within ±0.2 target

Status:

• BASELINE VALIDATION PHASE

---

📊 Sensor Health Audit

Interval:
Quarterly

Check:

• MAF voltage consistency
• Front O2 response
• Knock sensor noise
• Intake air temp stability
• Coolant temp accuracy

Entry:

• Date: March 10, 2026
• Time: 14:20 EST

Next audit:

• June 10, 2026

---

🛠 Relay Investigation Record

(From earlier troubleshooting memory)

Component under investigation:

• Mitsuba RC-5001 relay
• Function: starter/ignition circuit

Observed issue:

• No click on key-on

Entry recorded for diagnostics tracking.

---

📈 Black Pearl Development Status

Current phase:

Phase 1 — Mechanical Stability

Priority order:

1️⃣ Sensor integrity
2️⃣ Fueling accuracy
3️⃣ Boost control stability
4️⃣ Timing refinement
5️⃣ Power optimization

---

⚠ Important Engineering Reality

With the TD04 turbo, the safe power ceiling is roughly:

220–240 WHP

To reach the project target of 305 WHP, the roadmap will eventually require:

• VF34 / 16G / equivalent turbo
• Injector upgrade
• Fuel pump upgrade

But that discussion comes later once Black Pearl reaches stable Stage 1 calibration.

---

📊 Next Logbook Event (Recommended)

Run the first official Black Pearl diagnostic pull session:

Record:

• 3rd gear WOT pull
• 2500 → 6500 RPM

Log:

• Boost
• WGDC
• AFR
• IAM
• FBKC
• FLKC
• Injector duty
• MAF voltage
• IAT

---

💡 One thing that will massively improve the Black Pearl logbook system:

If we implement a “maintenance + tuning timeline dashboard”, the car becomes a data-driven engineering record, not just a build.

It also becomes incredibly compelling content for the website.

If you want, I can also build the Black Pearl Rally Engineering Logbook System that automatically tracks:

• maintenance
• map revisions
• dyno results
• sensor health
• performance metrics

It would make the project look like a professional motorsports development

---

The Problem Nobody Talks About

If you’ve ever tried using RomRaider or ECUFlash seriously… you already know:

• Half the builds don’t work
• Java dependencies break constantly
• Definitions are outdated or incomplete
• And worst of all — you’re tuning blind

Most people either:

• Give up
• Copy someone else’s map
• Or worse… blow something up chasing power

That wasn’t an option for Project Black Pearl.

🧠 Why WRX/STI Tuning Is More Complex Than People Think

Tuning a Subaru WRX/STI is not just “adding boost.”

It’s a balancing act between multiple tightly coupled systems:

• Airflow (MAF scaling, volumetric efficiency)
• Fueling (AFR targets, injector duty cycle)
• Ignition timing (knock threshold vs power)
• Boost control (WGDC, turbo dynamics)
• Thermals (IAT, coolant, heat soak)
• ECU logic (IAM, knock correction strategies)

Change one variable—and you affect all of them.

That’s where most people fail.

---

⚠️ The Reality Most Tuners Ignore

The factory ECU is not dumb—it’s adaptive, defensive, and constantly correcting:

• IAM drops = the engine has lost trust in your tune
• FBKC/FLKC = active knock correction
• Boost overshoot = unstable control, not “more power”
• Lean AFR under load = engine damage waiting to happen

You are not just tuning for power.

You are negotiating with the ECU’s safety systems.

---

🔬 The Actual Tuning Process (Done Properly)

A real tuning workflow looks like this:

1. Baseline Logging
   • No changes
   • Understand current behavior
2. Fueling First
   • MAF scaling
   • AFR accuracy
   • Injector headroom
3. Boost Control
   • Target vs actual
   • WGDC refinement
   • Eliminate oscillation
4. Timing Optimization
   • Only after stability
   • Incremental adjustments
   • Knock-free validation
5. Repeatability Testing
   • Multiple clean pulls
   • Heat consistency
   • No degradation

If you skip steps—you’re guessing.

---

🚫 Common Mistakes (That Kill Engines)

Let’s be blunt:

• ❌ Copy-pasting someone else’s map
• ❌ Increasing timing with active knock
• ❌ Chasing boost before fixing fueling
• ❌ Ignoring injector duty cycle limits
• ❌ Tuning based on “feel” instead of logs
• ❌ Doing single pulls and calling it “done”

These aren’t beginner mistakes.

People blow motors doing this every day.

---

✅ Non-Negotiable Rules

• If it’s not logged → it didn’t happen
• If IAM isn’t stable → stop tuning
• If AFR isn’t correct → don’t add boost
• If knock exists → don’t add timing
• If results aren’t repeatable → it’s not a tune

Power comes last.

Control comes first.

---

⚙️ The Problem With Existing Tools

If you’ve used RomRaider or ECUFlash, you already know:

• Broken builds
• Java issues
• Inconsistent definitions
• No structured decision-making

Worst of all:

They give you access—but not guidance.

That’s where people get hurt.

---

🧠 The Black Pearl Approach

Instead of relying on broken tools and guesswork…

I built a system.

---

🔧 What I Actually Did

Step 1 — Stabilize the Software

• Fixed RomRaider runtime issues
• Cleaned Java dependencies
• Aligned definitions with actual ECU behavior

Now it runs consistently.

---

Step 2 — Add AI (Without Risking the ECU)

The AI does NOT control the car.

It analyzes logs only and suggests controlled changes.

Think of it as:

A second set of eyes that never gets emotional.

---

📊 The AI Log Intelligence Engine

Every WOT pull is scored:

• Stability Score
• Knock Risk Score
• Boost Control Score
• Fueling Accuracy Score

Output:

SAFE / CAUTION / STOP

---

🔍 What It Detects Automatically

• Knock clustering by RPM/load
• Boost overshoot & oscillation
• AFR deviation vs target
• Heat soak sensitivity
• Injector duty risk
• MAF scaling errors

This is where real tuning happens—not in guesswork.

---

🎯 Controlled Change System

Every adjustment is limited:

• Timing: ≤ +0.5°
• WGDC: ≤ +2%
• Fueling: incremental only

Hard rules:

• No timing increase with knock
• No boost increase with unstable AFR
• No changes without validation

---

🚗 Current Black Pearl Configuration

• Platform: 2005 WRX EJ205
• Turbo: TD04L-13T
• Exhaust: Blitz NUR-SPEC RX
• Strategy: Stability → Fuel → Boost → Timing

We are not chasing peak power yet.

We are building a repeatable system.

---

📈 Where This Is Going

This project is becoming:

• AI log analysis tools
• WRX tuning frameworks
• Safe basemap strategies
• Step-by-step tuning systems
• Educational breakdowns

This is not just a build.

This is a methodology you can use—and eventually buy.

---

🧩 Final Thought

Anyone can increase boost.

Very few can:

• Build stable power
• Maintain engine health
• Repeat results consistently

That’s the difference between:

A tune… and a system.

---

🔧 What “Black Pearl Edition” Actually Changes

This is not a cosmetic rebuild.

This is a workflow-driven engineering upgrade to how RomRaider is used.

---

1. Environment Stabilization (Foundation Layer)

Before anything else:

• Repaired Java runtime inconsistencies
• Standardized execution environment
• Cleaned and aligned ECU definitions

Result:

• Consistent behavior
• No random crashes
• Reliable table interpretation

If the tool isn’t stable, the tune isn’t trustworthy.

---

2. Structured Logging Integration

Default tuning workflows are chaotic.

Black Pearl enforces a required logging schema:

• RPM
• Load
• Boost (target vs actual)
• WGDC
• IAM
• FBKC / FLKC
• AFR
• Injector Duty
• IAT / Coolant Temp
• Throttle

Why this matters:

If you’re not logging correctly, you’re not tuning—you’re guessing.

---

3. AI-Assisted Log Intelligence Layer

This is the core upgrade.

Instead of manually interpreting logs with bias or inconsistency…

The system applies structured analysis:

Every pull is evaluated for:

• Stability
• Knock risk
• Boost control accuracy
• Fueling accuracy

Output:

SAFE / CAUTION / STOP

---

4. Pattern Detection Engine (Where Most Tuners Fail)

The system identifies patterns automatically:

• Knock clustering by RPM/load
• Boost oscillation or overshoot
• AFR drift vs target
• Heat soak sensitivity
• Injector duty cycle risk zones
• MAF scaling inconsistencies

This removes:

• Emotional tuning
• One-off misinterpretation
• Inconsistent decision-making

---

5. Controlled Change Generator (Safety Layer)

This is where the system becomes definitive.

All tuning changes are constrained:

• Timing adjustments: ≤ +0.5° per iteration
• WGDC adjustments: ≤ 2% per iteration
• Fueling: incremental only

Hard rules:

• No timing increase with knock present
• No boost increase with unstable AFR
• No changes without validation logs

---

6. Validation Protocol (Non-Negotiable)

Before any power increase:

• Two repeatable clean pulls
• Stable IAM
• No sustained knock
• Boost tracks target
• AFR within margin

If validation fails:

The change is rejected.

---

🧩 The Result: A Definitive Tuning Workflow

This transforms tuning from:

❌ Trial-and-error
❌ “Feels faster”
❌ Copy-paste maps

Into:

✅ Structured decision-making
✅ Repeatable outcomes
✅ Log-validated changes

---

🔁 The Black Pearl Workflow Loop

Every iteration follows:

Log → Analyze → Score → Adjust → Flash → Validate → Repeat

Each revision includes:

• Map version name
• Change summary
• Reason for change
• Before/after comparison
• Risk evaluation

This creates:

A professional-grade, repeatable tuning system.

🧩 Hardware Constraint Profiles (HCP) — Enforcing Safe Tuning Boundaries

A core addition in the Black Pearl Edition patch is the introduction of Hardware Constraint Profiles (HCPs)—a system-level safeguard that aligns all AI-assisted recommendations with the physical limitations of the vehicle’s installed components.

Each HCP defines a validated configuration based on specific bolt-on upgrades (e.g., stock TD04, VF22, injectors, MAF housing, fuel system, intake/exhaust combinations). These profiles establish hard boundaries for critical parameters such as airflow range, injector capacity, boost ceiling, and safe AFR targets.

The purpose is simple:

The AI is not allowed to “improvise” beyond what the hardware can safely support.

When analyzing logs and generating recommendations, the AI is constrained to operate only within the limits of the active HCP. This prevents dangerous scenarios such as:

• Suggesting leaner AFR targets without sufficient fuel system capacity
• Recommending boost increases beyond turbo efficiency or injector headroom
• Misinterpreting airflow due to unsupported MAF scaling configurations

If the system detects conditions or goals outside the defined hardware profile, it does not compensate—it flags the condition and halts escalation.

This transforms the AI from a generic analytical tool into a hardware-aware decision engine, ensuring that every recommendation is grounded in real-world mechanical capability—not theoretical performance.

In practice, HCPs eliminate one of the biggest risks in ECU tuning:

Applying “correct” tuning logic to the wrong hardware setup.

By binding software intelligence to verified hardware configurations, Black Pearl enforces a closed-loop, safety-first tuning environment—where performance gains are achieved only within proven, supportable limits.

---

🧠 From System to Product — Standardizing Safe WRX Tuning

The Black Pearl framework is not just a tuning methodology—it is designed to be packaged, repeatable, and usable by others without guesswork. To achieve this, the system is being formalized into two practical tools: a WRX Mod Compatibility Matrix and a guided Setup Wizard. The Compatibility Matrix translates Hardware Constraint Profiles (HCPs) into a clear, decision-ready reference, mapping real-world bolt-on configurations (turbo, injectors, fuel system, MAF scaling) to their corresponding safe operating limits—including boost ceilings, AFR targets, and injector duty boundaries. This ensures that both the user and the AI operate within verified mechanical constraints, eliminating unsafe assumptions. Building on this, the Setup Wizard evolves the same logic into an interactive workflow—where users input their hardware and log data, and receive structured, constraint-aware guidance on what to adjust, what to avoid, and when to stop. Together, these tools convert Black Pearl from a personal tuning system into a standardized, scalable framework—where safe results are not dependent on experience alone, but on enforced logic, validated limits, and repeatable process.

🧩 How I Built and Trained a Local LLM From ECU ROM Data

One of the most important developments behind Black Pearl was not just fixing the tuning workflow—but teaching a system to understand the ECU itself.

This started by working directly with raw Denso ECU ROM (.bin) files, which are fundamentally 16-bit hex-based representations of calibration data and logic. On their own, these files are not human-readable—they are dense, encoded, and require definition files to interpret correctly. Instead of relying entirely on existing definitions, I approached the problem differently:

Break the ROM down, translate it, and make it understandable to an AI system.

---

🔧 Reverse Engineering the ROM Data

Using a combination of:

• Decompilation techniques
• Hex inspection
• RomRaider/ECUFlash definitions
• VSCode with AI-assisted tooling

I extracted segments of the ROM and began converting raw hex values into structured representations of:

• Tables (fuel, timing, boost)
• Axes (RPM, load)
• Scalars and limits
• Behavioral relationships between maps

---

🔄 Converting Hex → Structured Data

The key step was transforming binary data into formats that AI models can actually reason about:

• .bin → parsed hex blocks
• Hex → mapped values using known definitions
• Output → structured into JSON / XML-like formats

Each table was reconstructed into something like:

• Human-readable labels
• Defined axes
• Interpreted values (not just raw hex)

This effectively turns the ECU from a “black box” into a machine-readable dataset.

---

🧠 Training a Local AI to Understand ECU Logic

Once the data was structured, it was used to guide a local LLM to:

• Recognize relationships between tables
• Understand how changes in one area affect another
• Interpret logs in the context of actual ECU behavior
• Stay aligned with hardware constraints (via HCPs)

This is critical:

The AI is not guessing—it is referencing structured representations of how the ECU actually operates.

---

⚠️ Why This Matters for Safety

Most tuning tools expose tables without context.

This system does the opposite:

• It understands what the table does
• It understands how it interacts with other tables
• And it is constrained by real hardware limits

That combination is what allows Black Pearl to move toward:

AI-assisted tuning that is controlled, explainable, and grounded in real ECU logic—not abstraction.

---

🧩 The End Result

By converting raw ROM data into structured, human-readable formats and aligning it with a constraint-based system, Black Pearl achieves something most tools do not:

• Transparency
• Repeatability
• Safety through enforced logic

This is not about replacing the tuner.

It’s about building a system where:

Every recommendation can be traced, explained, and validated—before it ever touches the engine.

---

🔬 Technical Deep Dive: Reverse Engineering Denso ECU ROMs to Build an AI-Assisted WRX Tuning System

By admin | March 21, 2026

---

🧠 Introduction — Why Understanding the ECU Matters

Most people tuning a WRX/STI interact with the ECU through tables:

• Fuel maps
• Ignition timing
• Boost targets

But very few actually understand what’s underneath:

A tightly coupled, 16-bit encoded system controlling airflow, combustion, and engine safety in real time.

Tools like RomRaider and ECUFlash expose this data—but they rely heavily on definition files and assumed interpretations.

That creates a problem:

• You can modify values
• But you don’t fully understand the system behavior

And that’s where tuning becomes dangerous.

---

⚙️ The Foundation — Denso ECU ROM Architecture

The 2002–2005 Subaru WRX (EJ205) uses a Denso ECU with ROM data stored in a binary .bin format.

At its core, the ROM consists of:

• 16-bit encoded data blocks
• Lookup tables (2D / 3D maps)
• Scalars and limits
• Axis definitions (RPM, load, airflow)
• Control logic relationships

None of this is inherently human-readable.

Without interpretation, a ROM file is just:

Hexadecimal data with no meaning.

---

🔍 The Limitation of Traditional Tools

RomRaider and ECUFlash rely on:

• XML-based definition files
• Predefined memory addresses
• Assumed scaling formulas

While effective, this approach has limitations:

• Definitions can be incomplete or outdated
• Relationships between tables are not explicit
• No contextual understanding of system interaction

You are editing isolated values, not a connected system.

---

🔧 The Black Pearl Approach — Decode, Structure, Understand

Instead of relying entirely on definitions, the Black Pearl workflow introduces a different method:

Reverse engineer → structure → interpret → constrain

---

Step 1 — Binary Extraction and Inspection

Using:

• Raw .bin ROM files
• Hex-level inspection
• Existing definition references

Segments of the ROM are identified and isolated:

• Fuel tables
• Timing tables
• Boost control tables
• Compensation tables

---

Step 2 — Mapping Raw Hex to Real Values

Each segment is:

• Parsed from raw hex
• Converted using known scaling logic
• Validated against expected behavior

This transforms:

0x3A 0x7F 0x42...

Into:

• AFR targets
• Timing degrees
• Load values
• Boost targets

---

Step 3 — Structuring the Data (Critical Step)

Once decoded, the data is reorganized into structured formats:

• JSON
• XML-like schemas

Each table becomes:

• Named
• Indexed by axis (RPM/load)
• Represented with real-world values

This converts the ROM into:

A machine-readable model of ECU behavior

---

🧠 Training a Local LLM to Understand the ECU

With structured data in place, a local language model can now interact with the ECU logic meaningfully.

Instead of feeding it raw hex, the model sees:

• “Primary Open Loop Fueling Table”
• “Base Timing Advance”
• “Target Boost vs RPM”

With:

• Defined axes
• Interpreted values
• Contextual relationships

---

What the AI Learns

The model is guided to understand:

• How fueling affects knock
• How boost affects airflow and injector duty
• How timing interacts with load and combustion stability
• How ECU safety systems respond (IAM, knock correction)

This is not guesswork.

It is pattern recognition grounded in real ECU data structures.

---

🔒 Hardware Constraint Profiles (HCP) — Preventing Unsafe AI Behavior

A major risk with AI-assisted tuning is overgeneralization.

To prevent this, Black Pearl introduces:

Hardware Constraint Profiles (HCPs)

Each HCP defines:

• Turbo capability
• Injector limits
• Fuel system capacity
• MAF scaling range

The AI is restricted to operate within these boundaries only.

---

What This Prevents

• Lean AFR recommendations on undersized injectors
• Boost targets beyond turbo efficiency
• Misinterpreted airflow from incompatible MAF setups

If a condition exceeds hardware limits:

The system flags it—and stops escalation.

---

🔁 From Data to Decision — Closing the Loop

The final system connects three layers:

1. ECU Structure (ROM Data)

• Tables and logic decoded and structured

2. Real-World Data (Logs)

• RPM, AFR, boost, knock, etc.

3. AI Interpretation (Constrained)

• Pattern detection
• Risk scoring
• Controlled recommendations

---

The Result

A closed-loop system:

Log → Analyze → Constrain → Recommend → Validate

Not:

❌ Guess → Change → Hope

---

⚠️ Why This Matters

Most tuning failures are not due to lack of tools.

They are due to:

• Lack of structure
• Lack of constraint
• Lack of validation

By converting raw ECU data into structured formats and binding it to hardware-aware logic, Black Pearl enables:

• Explainable tuning decisions
• Repeatable results
• Safer performance optimization

---

🧩 Final Perspective

The ECU is not a collection of tables.

It is a system.

And once you:

• Decode it
• Structure it
• Constrain it
• And validate it

You move from:

Editing values…

To:

Engineering outcomes.

---

🔥 Strategic Insight (For Builders & Tuners)

The future of tuning isn’t:

• More aggressive maps
• Bigger numbers

It’s:

Systems that enforce safe, repeatable performance through structured intelligence.

That’s what Black Pearl is building.

🔬 Research & Development — Building the Black Pearl Toolchain

Black Pearl did not emerge from theory—it was built through continuous research, iteration, and custom tool development to solve real limitations in existing ECU tuning workflows. Throughout the process, several internal utilities and cross-platform tools were developed to support data extraction, visualization, and system-level understanding of the WRX ECU ecosystem.

Projects such as WrxDash and ImprezaDash were created to provide simplified, real-time visualization of critical engine parameters—allowing faster recognition of patterns in boost behavior, knock activity, and fueling trends without relying on cluttered or inconsistent logging interfaces. In parallel, tools like TorqueTap were designed to capture and interpret performance data in a more accessible format, bridging the gap between raw ECU output and actionable insight.

On the data engineering side, utilities such as ecuextractor were developed to directly interface with .bin ROM files—enabling structured extraction of tables, scalars, and calibration data for further analysis. These tools form the backbone of the Black Pearl workflow, allowing raw ECU data to be converted into formats that can be validated, visualized, and ultimately understood by both humans and AI systems.

This R&D layer is critical:

It transforms fragmented, tool-dependent tuning into a unified, cross-platform ecosystem.

Instead of relying on disconnected software with limited context, Black Pearl integrates custom-built tools into a cohesive pipeline—where data flows from ROM → structure → analysis → decision. Each tool serves a specific purpose, but together they reinforce the same objective:

• Reduce ambiguity
• Increase visibility
• Enforce consistency
• Enable safe, repeatable tuning outcomes

This is what separates Black Pearl from traditional tuning approaches.

It’s not just a better tune—it’s a purpose-built engineering stack designed around control, clarity, and reliability.

---

“How I hacked a non-working java-compiled open-source ecu tuning software that NEVER works, and added my own custom AI without compromising my ecu!”

AI English-based Summarization of hexdecimal map for: “Max Wastegate Duty Cycle” map

“Here is a concise explanation of the Max Wastegate Duty table:nnWhat it controls: This table sets the maximum wastegate duty cycle, which determines how much air can bypass the turbocharger to control boost pressure.nnTypical usage:nn* This table is used in conjunction with the Boost Control – Wastegate Compensation table to fine-tune the wastegate’s response to changing throttle and engine speed conditions.n* The goal is to ensure proper boost pressure and airflow at various engine speeds and throttle positions while minimizing turbocharger lag and maximizing power delivery.nnRisks:nn* Improperly set max wastegate duty values can lead to inadequate boost pressure, reduced power output, or even damage to the turbocharger or other engine components.n* Under-tuning may result in too much boost pressure, causing excessive EGTs (exhaust gas temperatures) and potentially damaging the engine or catalytic converter.n* Over-tuning can cause premature wear on moving parts due to excessive turbocharger cycling.nnOverall, this table is a critical component of boost control, requiring careful consideration and tuning to optimize performance and minimize risks.”

Based on the provided TABLE_3D, I’ve analyzed the Wastegate Duty Cycle (%) and identified potential areas for optimization while ensuring a conservative approach.nnAdjustment Idea 1:n”Smooth Out” the Mid-Range SectionnSuggested adjustment: Gradually increase the duty cycle in the mid-range section (approximately 30-70%) by 2-3% to improve efficiency. This adjustment aims to reduce the “kinks” and create a more linear power delivery.nnSafety Warning: Be cautious when making changes in this region, as it may affect the overall engine performance and responsiveness.nnAdjustment Idea 2:n”Taper Off” at High-Load RegionsnSuggested adjustment: Slightly decrease the duty cycle in high-load regions (above 80%) by 1-2% to reduce heat generation and potential overheating. This adjustment aims to maintain optimal performance while minimizing thermal stress.nnSafety Warning: Be aware that excessive reduction of duty cycle in high-load regions may impact engine performance, leading to reduced power output or even engine damage.nnAdjustment Idea 3:n”Fill the Gap” at Low-Load RegionsnSuggested adjustment: Increase the duty cycle in low-load regions (below 20%) by 1-2% to improve efficiency and reduce fuel consumption. This adjustment aims to optimize engine performance during light-load conditions.nnData Gaps/Risks: The low-load region (u003c20%) has relatively few data points, which may indicate a lack of calibration or testing in these conditions. It is essential to verify the accuracy of the base map table and consider additional testing or calibration to ensure optimal performance.nnAdjustment Idea 4 (optional):n”Fine-Tune” at Critical Load PointsnSuggested adjustment: Make minor adjustments (+/-1-2%) at specific critical load points (e.g., 50%, 60%, or 70%) to fine-tune the duty cycle and optimize engine performance. This adjustment aims to improve overall efficiency and responsiveness.nnSafety Warning: When making changes at critical load points, be cautious not to over-correct, as this may lead to instability or performance degradations.nnPlease note that these suggestions are based on a general analysis of the provided TABLE_3D and should be verified through additional testing or calibration to ensure optimal performance and safety.

AI Based Summarization Overview of: “Target Boost” map

“Here’s a concise explanation of the Target Boost table:nnWhat it controls: The Target Boost table sets the desired boost pressure (psi relative to sea level) based on throttle position and engine speed.nnTypical usage:nn* This table is used in conjunction with other tables (e.g., “Target Boost Compensation…”) to fine-tune the boost control system.n* It helps the engine management system determine the optimal amount of boost pressure needed for a given driving scenario.n* The target boost pressure is typically higher at higher throttle positions and engine speeds.nnRisks:nn* Incorrectly set targets can lead to engine knock, pinging, or detonation (which can cause damage).n* Insufficient boost pressure may result in reduced power and poor performance, while excessive boost pressure can lead to over-boosting and potential engine damage.n* The table’s settings should be carefully tuned to ensure proper air-fuel mixture and cooling system operation.nnOverall, the Target Boost table is a critical component of the boost control system, requiring careful calibration to achieve optimal engine performance and reliability.”

AI-Based “Target Boost” Tuning Suggestions (Example)

Based on the provided table and statistics, here are 3-5 conservative adjustment ideas to help you achieve your goal of a safe 15PSI max boost target:nn1. Tentative Adjustment: Introduce a small, gradual change by increasing the average boost target by 2-3% (around 0.18-0.28 PSI). This could be achieved by adjusting the lower and upper bounds of the table simultaneously, without exceeding the current maximum value. Safety Warning: Be cautious when making changes to the entire table scope, as this may affect other dependent systems or calculations.nnRisk: Data gaps or inconsistencies if not carefully implemented.nData gap: No selected cells or stats available for analysis.nn2. Segmented Adjustment: Focus on specific segments of the table (e.g., top 20% or bottom 30%) and make targeted adjustments to boost targets within those ranges. This approach allows you to maintain a conservative overall adjustment while still addressing potential bottlenecks or areas of improvement. Safety Warning: Segmental adjustments may introduce data inconsistencies if not properly synchronized.nnRisk: Discrepancies between adjusted and unadjusted segments.nData gap: No information provided about the table’s structure or relationships between cells.nn3. Threshold-Based Adjustment: Establish a threshold value (e.g., 5 PSI) and adjust boost targets above this threshold to achieve a more gradual increase towards the desired maximum of 15 PSI. This approach helps distribute the adjustment evenly across the table, reducing the risk of abrupt changes. Safety Warning: Be cautious when introducing thresholds, as this may create discontinuities in the data.nnRisk: Potential for inconsistencies if threshold is not carefully set.nData gap: No information provided about potential correlations or relationships between cells and boost targets.nn4. Linear Interpolation: Use linear interpolation to connect nearby data points (e.g., neighboring cells) and create a more gradual increase towards the desired maximum. This approach helps maintain the overall structure of the table while introducing a controlled change. Safety Warning: Be cautious when using interpolation, as this may introduce artifacts or inconsistencies if not properly implemented.nnRisk: Potential for inaccuracies or data distortions.nData gap: No information provided about the table’s structure or relationships between cells.nn5. Hybrid Approach: Combine two or more of the above adjustment ideas to create a conservative and controlled change. This approach allows you to leverage the strengths of each individual method while minimizing potential risks. Safety Warning: Be cautious when introducing hybrid approaches, as this may create complexities or data inconsistencies if not properly implemented.nnRisk: Potential for overcompensation or underestimation.nData gap: No information provided about potential correlations or relationships between cells and boost targets.nnBefore making any adjustments, consider the following:nn* Review the table’s structure and relationships between cells to identify potential dependencies or constraints.n* Verify the statistics and ensure that they accurately reflect the data.n* Perform sensitivity analyses to assess the impact of different adjustment approaches on the overall system.n* Implement changes incrementally and monitor the results to ensure the desired outcome is achieved.

---

---

---

NOTE/WARNING/DISCLAIMER: “The following is an English-to-English translated staged tuning plan for ‘exaggerated tuning’ stage-3 tune, based on user inputted desired power goals. And absolutely NOT based on or currently matching my exact/current hardware profiles. {this was merely generated to constrast the Ai’s overall understanding in comparison to ai summarizations above – and nothing else)

AI Staged Plan

Stage 1
Here is the 3-stage plan for the EJ20 Turbo engine:nnStage 1: Foundationnn* Stage name: “Stock-Boost”n* Required mods:nt+ VF39 turbocharger (replace existing stock turbo)nt+ Updated intake manifold (Cobb or similar) to accommodate new turbont+ High-flow air filter (to ensure consistent airflow)n* Tuning focus:nt+ Boost pressure: 12-14 PSI (targeted boost level for this stage)nt+ Timing: no significant advance, focus on maintaining stock timing settingsnt+ Fueling: maintain stock fuel injection system and adjust as needed to prevent knockn* Safety checks and logging targets:nt+ Monitor intake temperature (u003c120°C) and coolant temperature (u003c105°C)nt+ Log boost pressure, engine RPM, and A/F ratio to ensure operating within safety limitsnt+ Conduct regular compression tests to detect any potential issuesn* Conservative expected gains: +10-15 horsepower (compared to stock)nnStage 2: Power Enhancementnn* Stage name: “Power Surge”n* Required mods:nt+ High-performance spark plugs (for improved combustion efficiency)nt+ Upgraded fuel injectors (Cobb or similar) for increased flow ratent+ Engine calibration software upgrade (to optimize tuning parameters)n* Tuning focus:nt+ Boost pressure: 14-16 PSI (targeted boost level for this stage)nt+ Timing: slightly advanced timing (up to 2-3 degrees) to improve combustion efficiencynt+ Fueling: adjust fuel injection system to maintain optimal A/F ratio and prevent knockn* Safety checks and logging targets:nt+ Monitor intake temperature (u003c120°C) and coolant temperature (u003c105°C)nt+ Log boost pressure, engine RPM, and A/F ratio to ensure operating within safety limitsnt+ Conduct regular compression tests to detect any potential issuesn* Conservative expected gains: +20-25 horsepower (compared to previous stage)nnStage 3: High-Power Optimizationnn* Stage name: “Turbocharged Performance”n* Required mods:nt+ Advanced engine calibration software upgrade (to optimize tuning parameters)nt+ Upgraded intake manifold (Cobb or similar) for improved airflownt+ Engine braking system upgrade (for reduced stress on components)n* Tuning focus:nt+ Boost pressure: 16 PSI (targeted boost level for this stage)nt+ Timing: advanced timing (up to 4-5 degrees) to optimize combustion efficiencynt+ Fueling: adjust fuel injection system to maintain optimal A/F ratio and prevent knockn* Safety checks and logging targets:nt+ Monitor intake temperature (u003c120°C) and coolant temperature (u003c105°C)nt+ Log boost pressure, engine RPM, and A/F ratio to ensure operating within safety limitsnt+ Conduct regular compression tests to detect any potential issuesn* Conservative expected gains: +25-30 horsepower (compared to previous stage)nnRemember to always prioritize safety over performance when modifying your Subaru WRX. It’s essential to monitor your engine’s behavior, log data, and adjust your modifications accordingly to ensure a healthy and reliable powerplant.

Summary:
Here is the 3-stage plan for the EJ20 Turbo engine:nnStage 1: Foundationnn* Stage name: “Stock-Boost”n* Required mods:nt+ VF39 turbocharger (replace existing stock turbo)nt+ Updated intake manifold (Cobb or similar) to accommodate new turbont+ High-flow air filter (to ensure consistent airflow)n* Tuning focus:nt+ Boost pressure: 12-14 PSI (targeted boost level for this stage)nt+ Timing: no significant advance, focus on maintaining stock timing settingsnt+ Fueling: maintain stock fuel injection system and adjust as needed to prevent knockn* Safety checks and logging targets:nt+ Monitor intake temperature (u003c120°C) and coolant temperature (u003c105°C)nt+ Log boost pressure, engine RPM, and A/F ratio to ensure operating within safety limitsnt+ Conduct regular compression tests to detect any potential issuesn* Conservative expected gains: +10-15 horsepower (compared to stock)nnStage 2: Power Enhancementnn* Stage name: “Power Surge”n* Required mods:nt+ High-performance spark plugs (for improved combustion efficiency)nt+ Upgraded fuel injectors (Cobb or similar) for increased flow ratent+ Engine calibration software upgrade (to optimize tuning parameters)n* Tuning focus:nt+ Boost pressure: 14-16 PSI (targeted boost level for this stage)nt+ Timing: slightly advanced timing (up to 2-3 degrees) to improve combustion efficiencynt+ Fueling: adjust fuel injection system to maintain optimal A/F ratio and prevent knockn* Safety checks and logging targets:nt+ Monitor intake temperature (u003c120°C) and coolant temperature (u003c105°C)nt+ Log boost pressure, engine RPM, and A/F ratio to ensure operating within safety limitsnt+ Conduct regular compression tests to detect any potential issuesn* Conservative expected gains: +20-25 horsepower (compared to previous stage)nnStage 3: High-Power Optimizationnn* Stage name: “Turbocharged Performance”n* Required mods:nt+ Advanced engine calibration software upgrade (to optimize tuning parameters)nt+ Upgraded intake manifold (Cobb or similar) for improved airflownt+ Engine braking system upgrade (for reduced stress on components)n* Tuning focus:nt+ Boost pressure: 16 PSI (targeted boost level for this stage)nt+ Timing: advanced timing (up to 4-5 degrees) to optimize combustion efficiencynt+ Fueling: adjust fuel injection system to maintain optimal A/F ratio and prevent knockn* Safety checks and logging targets:nt+ Monitor intake temperature (u003c120°C) and coolant temperature (u003c105°C)nt+ Log boost pressure, engine RPM, and A/F ratio to ensure operating within safety limitsnt+ Conduct regular compression tests to detect any potential issuesn* Conservative expected gains: +25-30 horsepower (compared to previous stage)nnRemember to always prioritize safety over performance when modifying your Subaru WRX. It’s essential to monitor your engine’s behavior, log data, and adjust your modifications accordingly to ensure a healthy and reliable powerplant.

🔥 Project Black Pearl – Refined Directive

Project Black Pearl is a structured, AI-assisted performance build of a 2005 Subaru Impreza WRX into a reliable ~305whp street-focused performance car with entry-level rally capability.

This is not just a build.
It is:

1. A performance engineering project
2. A documented learning laboratory
3. A monetizable content engine

---

🎯 Primary Objectives

1. Performance Target

• ~305 wheel horsepower
• Safe, conservative, reliable tune
• Balanced street drivability
• Rally-capable suspension, braking, and cooling
• Data-driven ECU calibration

No dyno-queen numbers. No ego tuning.
Reliable power > peak power.

---

2. Engineering Philosophy

• Reliability first
• Data > guesswork
• Incremental upgrades
• Log everything
• Tune conservatively
• Always build with future headroom

Every change must answer:

• Does this improve reliability?
• Does this improve response?
• Does this improve consistency?
• Does this move us closer to 305whp safely?

---

3. ECU & Tuning Strategy

• Controlled boost targets
• Conservative timing
• Knock-safe fueling
• Gradual power scaling
• Log → Analyze → Adjust → Repeat

No random map changes.
Every revision is documented, named, and justified.

---

4. Rally-Oriented Capability

Entry-level rally capability means:

• Suspension travel and control
• Brake endurance
• Cooling resilience
• Strong midrange torque
• Predictable throttle modulation
• Chassis stability under imperfect surfaces

This is a usable car, not a parking lot build.

---

5. Content Production Strategy (YouTube)

Black Pearl is also a content asset.

Each phase of the build must generate:

• Install walkthroughs
• Tuning breakdowns
• Mistake analysis
• Data log reviews
• Cost transparency
• Before/after comparisons
• Real-world driving impressions

Content pillars:

• “AI Tuned WRX”
• Budget vs Performance Reality
• What Actually Works
• What I’d Do Differently
• Data-Driven Builds

The goal:
Document the journey in a way that builds trust and authority.

---

💰 Monetization Intent

This project should evolve into:

• Affiliate links (parts, tools, tuning gear)
• Digital downloads (base maps, logging templates, tuning checklists)
• Consulting/tuning assistance
• Sponsorship positioning
• Branded merch
• Educational breakdown content

The build must fund itself over time.

---

📊 Operational Discipline

For every modification:

Document:

• Cost
• Purpose
• Expected gain
• Risk
• Measured result
• Long-term effect

If it doesn’t produce measurable value, it doesn’t move the project forward.

---

🧠 Strategic Mindset

Black Pearl is:

• A system
• A case study
• A brand asset
• A revenue engine
• A proof of capability

It is not:

• Random parts thrown at a car
• Blind tuning
• Ego horsepower chasing

---

🚀 Next-Level Refinement

If we want to elevate this:

Answer these:

1. Are we staying VF22 long-term or planning turbo evolution?
2. What is the final boost target at 305whp?
3. Are we building bottom-end headroom or staying stock internals?
4. What’s the suspension philosophy — street bias or rally bias?
5. What’s the realistic total budget cap?

Clarity here determines the entire roadmap.

---

If you want, I can now:

• Turn this into a pinned project charter
• Convert it into a YouTube channel “About” positioning statement
• Or build a 12-month roadmap to hit 305whp + monetization simultaneously

Where are we pushing next — power refinement, reliability reinforcement, or revenue extraction?

AI-DRIVEN INTELLIGENT ECU Tuning ResearchThought for 8s

Here’s a research-grade, practical roadmap for AI-driven intelligent ECU tuning for Project Black Pearl (05 WRX), split into what’s real, what’s hype, and what we can actually build into your workflow.

---

What “AI ECU tuning” actually means (3 levels)

Level 1 — AI as a tuning assistant (real + useful now)

• Learns patterns from your logs and suggests targeted map edits (MAF scaling, fueling trims, timing deltas, boost control tweaks).
• Flags risk: knock trends, IAM drops, heat-soak sensitivity, injector duty issues, boost oscillation.
  This is the best ROI path for open-source Subaru workflows.

Level 2 — Automated calibration with constraints (real in research + pro industry)

This is the big one: automatic tuning under safety bounds (EGT/knock/torque/pressure constraints), often using optimization + risk limits. There’s active research explicitly on risk-aware automated engine control calibration.

Level 3 — “Fully autonomous tuning” (mostly marketing)

A lot of “AI tuning file” services are basically template retrieval + pattern matching, not true closed-loop learning. Useful sometimes, but don’t confuse it with intelligent control.

---

The best research directions to steal for your WRX

1) Risk-aware optimization (your safest path to power)

Instead of “maximize power,” the objective becomes:

• hit requested load/boost/AFR targets
• minimize knock probability
• obey constraints (timing limits, boost limits, temps, fuel system headroom)

This is exactly the framing used in recent automated calibration research.

Translation to WRX reality:
We can create an optimizer that proposes small deltas to:

• Timing (base + compensations where relevant)
• Boost WGDC (and dynamics)
• OL fueling targets
  …and only accepts changes if logs show safer margins.

---

2) Model Predictive Control (MPC) + learned models (industry-grade approach)

MPC is widely used because it can enforce constraints while optimizing behavior, and it pairs well with learned/ML models.

Translation to WRX:
You likely won’t replace Subaru’s control loops, but you can use MPC ideas offline:

• predict boost response to WGDC changes
• predict knock sensitivity vs IAT/load/RPM
• choose edits that reduce oscillation and knock risk

---

3) Knock-control literacy is mandatory (or AI will hurt you)

Subaru knock strategy is nuanced; AI can’t “save you” if you don’t model the rules of the ECU (IAM, FBKC, FLKC, learned knock, etc.). Use RomRaider’s knock-control docs and deep community breakdowns as your baseline truth.

---

What to build for Black Pearl (high ROI, low risk)

A. “Log Intelligence Layer” (first milestone)

Create an AI analyzer that:

• scores each pull: safe / caution / stop
• detects patterns:
  • knock clustered by RPM/load?
  • IAM decay after heat soak?
  • boost overshoot then correction?
  • AFR error vs MAF scaling?
• outputs a tight action list (“Change X table by Y in Z cells”)

This is where you get results without gambling the engine.

B. “Change Generator” with guardrails (second milestone)

The AI proposes map deltas, but your rules must be hard:

• max timing change per iteration (ex: +0.5° / cell)
• max WGDC change per iteration (ex: +1–2%)
• no changes if IAM not stable
• no changes if FBKC/FLKC appear in relevant regions
• require repeatability across ≥2 pulls

This aligns with the “robust performance + monitoring + RL/optimization direction” highlighted in ML engine control work.

C. Content engine (this prints money if you do it clean)

Each iteration becomes a video:

• “AI found knock pattern at 4400rpm”
• “AI reduced boost oscillation by adjusting WGDC dynamics”
• “AI fixed MAF scaling error causing lean tip-in”

You’re not just building a car — you’re building a repeatable “AI tuning methodology” brand.

---

Tools + ecosystem reality (Subaru open-source)

Your ecosystem is still:

• RomRaider (logging + definitions + visualization)
• ECUFlash (read/flash)
• Tactrix OpenPort

Be strict: never log + flash simultaneously (common conflict).

---

The 5 questions that decide your whole research path

Answer these straight — they determine what “intelligent tuning” can safely do:

1. Fueling hardware: injectors + pump are what right now? (stock? upgraded?)
2. Sensor strategy: do you have a wideband integrated into logs or only external?
3. Turbo + boost target: VF22 still the plan and what peak boost do you want on pump gas?
4. Fuel type: 93 only, or are you going flex fuel later?
5. Primary constraint: are we optimizing for reliability, response, or peak power first?

If you reply with those 5, I’ll turn this into a concrete Black Pearl research plan with:

• required log channels
• safety thresholds
• scoring model
• and the exact “AI recommendation format” that feeds your current tuning workflow.

Give me the new refined directive

Here is the new refined directive for:

🧠 Project Black Pearl

AI-Driven Intelligent ECU Tuning Program

---

Mission

Develop a data-driven, AI-assisted ECU calibration framework for a 2005 Subaru WRX that produces:

• ~305whp safely
• Predictable, repeatable power delivery
• Rally-capable drivability
• Engine longevity
• Documented, monetizable methodology

This is not “auto-tune.”
This is intelligent, constraint-based performance engineering.

---

Core Philosophy

1. Reliability is non-negotiable.
2. Every change must be measured.
3. AI assists — it does not replace judgment.
4. Power is earned through stability, not forced through aggression.
5. If it cannot be logged, it cannot be trusted.

---

System Architecture

Layer 1 — Data Acquisition

All tuning decisions are based on structured logs including:

• RPM
• Load
• Boost (target vs actual)
• WGDC
• IAM
• FBKC / FLKC
• AFR (wideband integrated preferred)
• MAF voltage
• Injector duty cycle
• IAT / Coolant temp
• Throttle angle

No blind pulls. No street guessing.

---

Layer 2 — AI Log Intelligence Engine

The AI performs:

1. Pull Scoring

Each WOT pull receives:

• Stability Score
• Knock Risk Score
• Boost Control Score
• Fueling Accuracy Score

Output:
SAFE / CAUTION / STOP

---

2. Pattern Detection

The system identifies:

• Knock clustering by RPM/load
• Heat soak sensitivity
• Boost overshoot/oscillation
• AFR deviation vs target
• MAF scaling error zones
• Injector duty risk regions

No emotional tuning. Only patterns.

---

Layer 3 — Controlled Change Generator

The AI may propose changes under strict constraints:

• Timing delta ≤ +0.5° per iteration
• WGDC delta ≤ 2% per iteration
• Fueling changes incremental only
• No timing increases if:
  • IAM unstable
  • Active knock present
  • Boost overshooting target
• No boost increases if:
  • Injector duty > safe margin
  • AFR trending lean
  • IAT excessive

Changes are small, deliberate, reversible.

---

Layer 4 — Validation Protocol

Before any power escalation:

1. Two repeatable clean pulls
2. Stable IAM
3. No sustained knock events
4. Boost tracking within tolerance
5. AFR within target margin

If validation fails → revert.

---

Performance Objective Framework

We optimize in this order:

1. Mechanical stability
2. Fueling accuracy
3. Boost control precision
4. Timing refinement
5. Torque smoothing
6. Peak output

Power is the last variable adjusted — not the first.

---

Safety Doctrine

Hard limits must exist:

• Conservative boost ceiling
• Conservative AFR targets on pump gas
• Injector duty margin preserved
• Thermal management respected

An extra 10whp is never worth engine damage.

---

Rally-Oriented Calibration Bias

Tuning emphasis:

• Strong midrange torque
• Predictable throttle modulation
• Smooth boost ramp
• No violent torque spikes
• Heat consistency under repeated pulls

This is a usable car, not a dyno screenshot.

---

Iterative Loop Model

Log → Analyze → Score → Adjust → Flash → Validate → Repeat

Each iteration documented:

• Map version name
• Change summary
• Reason for change
• Before/after log comparison
• Risk evaluation

Every revision becomes content and data.

---

Monetization Integration

This directive must evolve into:

• AI log analysis templates
• WRX tuning checklists
• Safe-iteration playbooks
• Map revision documentation frameworks
• Educational breakdown videos
• Digital download assets

You are building:

1. A performance car
2. A documented methodology
3. A monetizable system

---

Non-Negotiables

• No ego tuning
• No blind timing jumps
• No “it feels faster” conclusions
• No skipping validation

---

Strategic Outcome

At completion, Project Black Pearl should demonstrate:

• A stable ~305whp calibration
• A repeatable AI-assisted tuning framework
• A documented roadmap others can follow
• A performance car built on engineering discipline

---

If you want the next level, I can now:

• Convert this into a printable “Black Pearl ECU Control Doctrine”
• Build a scoring model formula for your log intelligence layer
• Or design the exact log-analysis pipeline structure for automation

Where do you want to tighten the system next — scoring logic, boost control modeling, or timing strategy?

Black Pearl ECU Control Doctrine

🧠 BLACK PEARL ECU CONTROL DOCTRINE

Project Black Pearl – AI-Governed Calibration Framework
2005 Subaru WRX | Target: ~305whp | Rally-Capable | Reliability-First

---

I. Foundational Principle

The ECU is not a power lever.
It is a thermal, combustion, and load management system.

The objective is not maximum output.
The objective is controlled, repeatable combustion under defined constraints.

Power is the byproduct of stability.

---

II. Chain of Authority

1. Mechanical Reality — highest authority
2. Sensor Data — second authority
3. Structured Logs — third authority
4. AI Analysis — advisory only
5. Human Decision — final control

AI may recommend.
It may never override physical evidence.

---

III. Operational Hierarchy (Order of Optimization)

Tuning must follow this strict sequence:

1. Mechanical integrity
2. Fuel delivery accuracy
3. MAF scaling correctness
4. Boost control stability
5. Knock control stability
6. Timing refinement
7. Torque shaping
8. Peak output

Skipping levels is forbidden.

---

IV. Data Governance Standard

No map revision may occur without complete logs including:

• RPM
• Load
• Boost (target vs actual)
• WGDC
• IAM
• FBKC
• FLKC
• AFR (wideband preferred)
• MAF voltage
• Injector duty cycle
• IAT
• Coolant temp
• Throttle angle

If it is not logged, it is not changed.

---

V. Pull Authorization Protocol

A WOT pull is authorized only if:

• Coolant temp stable
• IAT within safe margin
• IAM at full or stable baseline
• No active knock history in recent driving

If any condition fails → no pull.

---

VI. Pull Scoring Doctrine

Each pull must receive 4 scores:

1. Stability Score
   • Boost tracking consistency
   • WGDC behavior
   • No oscillation
2. Knock Risk Score
   • FBKC presence
   • FLKC learning
   • IAM behavior
3. Fueling Accuracy Score
   • AFR vs target deviation
   • Injector duty margin
4. Thermal Margin Score
   • IAT sensitivity
   • Coolant stability

Output classification:

• SAFE
• CAUTION
• STOP

STOP mandates reversion or mechanical inspection.

---

VII. Change Governance Rules

No change may exceed:

• +0.5° timing per iteration
• ±2% WGDC per iteration
• Incremental fueling correction only

Timing increases are forbidden if:

• IAM unstable
• Any knock present in that cell
• Boost not fully controlled

Boost increases are forbidden if:

• Injector duty approaching unsafe threshold
• AFR leaner than target
• Heat soak detected

Changes must be:

• Small
• Reversible
• Documented

---

VIII. Validation Protocol

Before advancing power:

• Two consecutive clean pulls
• No sustained knock
• IAM stable
• Boost within tolerance
• AFR within acceptable deviation

If validation fails → revert to last stable revision.

No emotional overrides.

---

IX. Rally Calibration Bias

The calibration must prioritize:

• Strong midrange torque (3k–5k RPM emphasis)
• Smooth boost ramp
• No torque spikes
• Predictable throttle modulation
• Heat resilience across repeated pulls

Street drivability > dyno bragging rights.

---

X. Hard Safety Limits

Non-negotiable ceilings must exist for:

• Boost
• Timing advance
• Injector duty
• AFR under load
• IAT compensation

An additional 10whp is never worth a ringland.

---

XI. Iterative Loop Doctrine

Every revision must follow:

Log → Analyze → Score → Adjust → Flash → Validate → Document

Documentation must include:

• Map version name
• Exact table changes
• Reason for change
• Before/after log comparison
• Risk evaluation

Every map is a data artifact.

---

XII. Monetizable Methodology Mandate

This doctrine must evolve into:

• AI log analysis templates
• ECU tuning checklists
• Safe iteration frameworks
• Map revision documentation templates
• Educational breakdown content
• Digital downloads

The system must outlive the car.

---

XIII. Prohibited Actions

• No blind timing jumps
• No boost spikes for “testing”
• No tuning by feel
• No ignoring knock
• No escalating power during instability
• No undocumented changes

---

XIV. Strategic Outcome

Upon completion, Project Black Pearl will demonstrate:

1. A stable ~305whp calibration
2. Repeatable AI-assisted tuning workflow
3. Rally-capable street drivability
4. Engine longevity under stress
5. A documented, teachable framework

---

If you want to tighten this further, the next upgrade would be:

• A formalized numeric scoring model (weighted formula)
• Defined safe thresholds for EJ205 pump gas operation
• Or a version-control naming architecture tied to your actual .bin files

Which layer do you want to formalize next?

A formalized numeric scoring model (weighted formula)

Below is a formal, numeric, weighted scoring model you can run on every WOT pull. It outputs:

• 0–100 Overall Score
• SAFE / CAUTION / STOP
• Component subscores you can graph over time

It’s built to be knock- and safety-dominant (because that’s what keeps EJ205s alive).

---

Black Pearl Pull Scoring Model v1.0

1) Overall Score (0–100)

$$S_{overall} = 0.35S_{knock} + 0.25S_{fuel} + 0.20S_{boost} + 0.15S_{thermal} + 0.05S_{repeat}$$Soverall​=0.35Sknock​+0.25Sfuel​+0.20Sboost​+0.15Sthermal​+0.05Srepeat​

Why these weights: knock + fueling are the biggest “engine-ending” failure modes on pump gas; boost control and thermal stability drive repeatability; repeatability is important but not more important than safety.

---

2) Subscores (each 0–100)

A) Knock Score $S_{knock}$Sknock​ (0–100)

Use 3 penalties: IAM, FBKC, FLKC.

IAM penalty

• Let $IAM_{start}, IAM_{end}$IAMstart​,IAMend​
• $\Delta IAM = IAM_{start} – IAM_{end}$ΔIAM=IAMstart​−IAMend​

$$P_{IAM} = \begin{cases} 0 & \Delta IAM = 0 \\ 60 & \Delta IAM > 0 \end{cases}$$PIAM​={060​ΔIAM=0ΔIAM>0​

(Any IAM drop during/after the pull is a major red flag → heavy penalty.)

FBKC penalty

• Let $FBKC_{min}$FBKCmin​ be most negative FBKC during pull (e.g., -1.41, -2.81)

$$P_{FBKC} = \min(60,\ 20 \cdot |FBKC_{min}|)$$PFBKC​=min(60, 20⋅∣FBKCmin​∣)

So: -1.41 → 28.2 penalty, -2.81 → 56.2 penalty, -4.22 caps at 60.

FLKC penalty

• Let $FLKC_{min}$FLKCmin​ be most negative learned knock in the pull region

$$P_{FLKC} = \min(40,\ 15 \cdot |FLKC_{min}|)$$PFLKC​=min(40, 15⋅∣FLKCmin​∣)

Knock Score$$S_{knock} = \max(0,\ 100 – (P_{IAM}+P_{FBKC}+P_{FLKC}))$$Sknock​=max(0, 100−(PIAM​+PFBKC​+PFLKC​))

---

B) Fuel Score $S_{fuel}$Sfuel​ (0–100)

This score cares about AFR error + injector duty headroom.

AFR tracking penalty

You need: measured wideband AFR $AFR_{meas}$AFRmeas​ and commanded/target AFR $AFR_{tgt}$AFRtgt​ (or OL target if available).

Compute error over the pull window (e.g., 3500–6500 rpm):

• $E_{AFR} = \text{median}(|AFR_{meas} – AFR_{tgt}|)$EAFR​=median(∣AFRmeas​−AFRtgt​∣)
• $E_{AFR,max} = \text{max}(|AFR_{meas} – AFR_{tgt}|)$EAFR,max​=max(∣AFRmeas​−AFRtgt​∣)

$$P_{AFR} = 30\cdot E_{AFR} + 10\cdot E_{AFR,max}$$PAFR​=30⋅EAFR​+10⋅EAFR,max​

(Example: median error 0.2 and max 0.5 → $P_{AFR}=30(0.2)+10(0.5)=11$PAFR​=30(0.2)+10(0.5)=11.)

IDC penalty

Let $IDC_{max}$IDCmax​ be peak injector duty cycle (%).$$P_{IDC} = \begin{cases} 0 & IDC_{max} \le 85 \\ 2\cdot (IDC_{max}-85) & 85 < IDC_{max} \le 95 \\ 40 & IDC_{max} > 95 \end{cases}$$PIDC​=⎩⎨⎧​02⋅(IDCmax​−85)40​IDCmax​≤8585<IDCmax​≤95IDCmax​>95​

Fuel Score$$S_{fuel} = \max(0,\ 100 – (P_{AFR} + P_{IDC}))$$Sfuel​=max(0, 100−(PAFR​+PIDC​))

---

C) Boost Control Score $S_{boost}$Sboost​ (0–100)

Reward accurate tracking; punish overshoot and oscillation.

You need boost target $B_{tgt}$Btgt​ and actual $B_{act}$Bact​ (same units).

Compute:

• Peak overshoot: $O = \max(B_{act}-B_{tgt})$O=max(Bact​−Btgt​)
• Mean absolute tracking error: $E_B = \text{mean}(|B_{act}-B_{tgt}|)$EB​=mean(∣Bact​−Btgt​∣)
• Oscillation proxy: $J = \text{std}(\Delta B_{act})$J=std(ΔBact​) (std of boost “rate of change”)

$$P_{overshoot} = 25\cdot O$$Povershoot​=25⋅O$$P_{track} = 20\cdot E_B$$Ptrack​=20⋅EB​$$P_{osc} = 10\cdot J$$Posc​=10⋅J

Boost Score$$S_{boost} = \max(0,\ 100 – (P_{overshoot}+P_{track}+P_{osc}))$$Sboost​=max(0, 100−(Povershoot​+Ptrack​+Posc​))

(If you don’t have a clean $J$J, skip it and redistribute weight inside boost.)

---

D) Thermal Score $S_{thermal}$Sthermal​ (0–100)

We’re scoring IAT and heat soak sensitivity.

Let:

• $IAT_{start}, IAT_{end}$IATstart​,IATend​
• $\Delta IAT = IAT_{end}-IAT_{start}$ΔIAT=IATend​−IATstart​

$$P_{IATlevel} = \begin{cases} 0 & IAT_{start} \le 90F \\ 0.5\cdot (IAT_{start}-90) & 90F < IAT_{start} \le 130F \\ 30 & IAT_{start} > 130F \end{cases}$$PIATlevel​=⎩⎨⎧​00.5⋅(IATstart​−90)30​IATstart​≤90F90F<IATstart​≤130FIATstart​>130F​$$P_{IATrise} = 2\cdot \Delta IAT$$PIATrise​=2⋅ΔIAT

Thermal Score$$S_{thermal} = \max(0,\ 100 – (P_{IATlevel}+P_{IATrise}))$$Sthermal​=max(0, 100−(PIATlevel​+PIATrise​))

(If you log °C, adjust thresholds accordingly.)

---

E) Repeatability Score $S_{repeat}$Srepeat​ (0–100)

This only makes sense once you have two pulls under similar conditions.

Let:

• $\Delta B_{peak}$ΔBpeak​ = |peak boost pull1 – pull2|
• $\Delta AFR$ΔAFR = median |AFR error pull1 – pull2|
• $\Delta FBKC$ΔFBKC = |FBKCmin pull1 – pull2|

$$P_{rep} = 20\cdot \Delta B_{peak} + 50\cdot \Delta AFR + 10\cdot \Delta FBKC$$Prep​=20⋅ΔBpeak​+50⋅ΔAFR+10⋅ΔFBKC$$S_{repeat} = \max(0,\ 100 – P_{rep})$$Srepeat​=max(0, 100−Prep​)

---

3) Classification Rules (hard gates + score bands)

Even if the weighted score looks okay, these gates override:

STOP (immediate)

• $\Delta IAM > 0$ΔIAM>0 (IAM drop)
• $FBKC_{min} \le -2.8$FBKCmin​≤−2.8 in the WOT load cells
• AFR goes lean beyond a hard limit you set (example: > 0.8 AFR lean vs target at high load)
• $IDC_{max} > 95\%$IDCmax​>95%

CAUTION

• $FBKC_{min}$FBKCmin​ between -1.4 and -2.8
• Any FLKC in the WOT region
• Boost overshoot > 1.0 psi (or equivalent)
• Thermal score < 70

SAFE

• No IAM drop, minimal/no knock, fueling and boost stable

Score bands

• SAFE: $S_{overall} \ge 85$Soverall​≥85
• CAUTION: 70–84
• STOP: <70 or any STOP gate triggered

---

4) How you use it (the money part)

Every revision gets a line item like:

• v1.07: Overall 78 (CAUTION) → “Reduce timing 0.5° @ 4.2g/rev 4200–5200, fix WGDC overshoot”
• v1.08: Overall 89 (SAFE) → “Proceed to small WGDC +1% in midrange”

This turns your tuning into a measurable product and your videos into evidence-based episodes.

---

Quick probing questions (so we set the right thresholds)

1. Are you logging wideband AFR into the ECU/log, or is it external only?
2. What units are you using for boost (psi or kPa), and do you have boost target logged?
3. What’s your typical fuel (91/93, ethanol blend, etc.)?

Answer those and I’ll tailor the hard limits + coefficients so the scoring matches your exact setup.

91/93, ethanol blend,

Got it. With 91/93 + ethanol blend, you should run the scoring model with fuel-profile switchable thresholds, because ethanol dramatically changes knock margin and “acceptable” timing/boost behavior.

Below is Black Pearl Pull Scoring Model v1.1 (Fuel-Aware): same structure, but with profiles for 91, 93, and E-blend (default E30).

---

Fuel Profiles (choose one per pull/session)

Profile A — 91 Pump (most conservative)

• Knock penalties harsher
• Lean events less tolerated
• IDC headroom more protected

Profile B — 93 Pump

• Moderately conservative

Profile C — Ethanol Blend (assume E30 unless you specify)

• Knock tolerance can be slightly wider
• Still zero tolerance for IAM drops and major knock
• Fuel system limits matter more (ethanol increases fuel demand)

---

1) Weights (slightly revised for ethanol)

91 / 93

$$S_{overall} = 0.38S_{knock} + 0.25S_{fuel} + 0.20S_{boost} + 0.12S_{thermal} + 0.05S_{repeat}$$Soverall​=0.38Sknock​+0.25Sfuel​+0.20Sboost​+0.12Sthermal​+0.05Srepeat​

Ethanol blend (E30 default)

$$S_{overall} = 0.34S_{knock} + 0.30S_{fuel} + 0.18S_{boost} + 0.13S_{thermal} + 0.05S_{repeat}$$Soverall​=0.34Sknock​+0.30Sfuel​+0.18Sboost​+0.13Sthermal​+0.05Srepeat​

Why: ethanol usually improves knock margin, but it stresses fueling capacity, so fueling gets more weight.

---

2) Hard STOP Gates (these do not negotiate)

All fuels (91/93/E-blend)

STOP immediately if any:

• IAM drops during/after pull
• FBKC ≤ -2.8 in WOT/load region
• IDC > 95%
• Boost overshoot > 2.0 psi (or equivalent) with oscillation

Fueling STOP (AFR error vs your target)

This uses error vs commanded, not absolute AFR.

• 91: STOP if max error ≥ +0.6 AFR lean
• 93: STOP if max error ≥ +0.7 AFR lean
• E30: STOP if max error ≥ +0.8 AFR lean
  (Still not “safe,” just slightly more tolerant before hard stop.)

---

3) Knock Score $S_{knock}$Sknock​ (fuel-adjusted penalties)

IAM penalty (unchanged)

$$P_{IAM} = \begin{cases} 0 & \Delta IAM = 0 \\ 70 & \Delta IAM > 0 \end{cases}$$PIAM​={070​ΔIAM=0ΔIAM>0​

FBKC penalty coefficient by fuel

Let $FBKC_{min}$FBKCmin​ be most negative value.

• 91: $\;P_{FBKC} = \min(70,\ 25\cdot |FBKC_{min}|)$PFBKC​=min(70, 25⋅∣FBKCmin​∣)
• 93: $\;P_{FBKC} = \min(65,\ 22\cdot |FBKC_{min}|)$PFBKC​=min(65, 22⋅∣FBKCmin​∣)
• E30: $\;P_{FBKC} = \min(60,\ 18\cdot |FBKC_{min}|)$PFBKC​=min(60, 18⋅∣FBKCmin​∣)

FLKC penalty coefficient by fuel

• 91: $\;P_{FLKC} = \min(50,\ 18\cdot |FLKC_{min}|)$PFLKC​=min(50, 18⋅∣FLKCmin​∣)
• 93: $\;P_{FLKC} = \min(45,\ 16\cdot |FLKC_{min}|)$PFLKC​=min(45, 16⋅∣FLKCmin​∣)
• E30: $\;P_{FLKC} = \min(40,\ 12\cdot |FLKC_{min}|)$PFLKC​=min(40, 12⋅∣FLKCmin​∣)

$$S_{knock} = \max(0,\ 100 – (P_{IAM}+P_{FBKC}+P_{FLKC}))$$Sknock​=max(0, 100−(PIAM​+PFBKC​+PFLKC​))

---

4) Fuel Score $S_{fuel}$Sfuel​ (ethanol tightens IDC rules)

AFR penalty (same math, different sensitivity)

Let:

• $E_{AFR} = median(|AFR_{meas}-AFR_{tgt}|)$EAFR​=median(∣AFRmeas​−AFRtgt​∣)
• $E_{AFR,max} = max(|AFR_{meas}-AFR_{tgt}|)$EAFR,max​=max(∣AFRmeas​−AFRtgt​∣)
• 91: $P_{AFR} = 35E_{AFR} + 12E_{AFR,max}$PAFR​=35EAFR​+12EAFR,max​
• 93: $P_{AFR} = 32E_{AFR} + 11E_{AFR,max}$PAFR​=32EAFR​+11EAFR,max​
• E30: $P_{AFR} = 30E_{AFR} + 10E_{AFR,max}$PAFR​=30EAFR​+10EAFR,max​

IDC penalty (this is where ethanol gets strict)

Let $IDC_{max}$IDCmax​ be peak IDC.

91 / 93

• 0 penalty ≤ 85%
• linear penalty 85–95%
• hard cap >95%

$$P_{IDC} = \begin{cases} 0 & IDC_{max} \le 85 \\ 2.2\cdot (IDC_{max}-85) & 85 < IDC_{max} \le 95 \\ 45 & IDC_{max} > 95 \end{cases}$$PIDC​=⎩⎨⎧​02.2⋅(IDCmax​−85)45​IDCmax​≤8585<IDCmax​≤95IDCmax​>95​

E30 (more fuel required → protect headroom)

• 0 penalty ≤ 82%
• linear penalty 82–92%
• hard cap >92% (because you’ll run out of injector fast with E content swings)

$$P_{IDC,E30} = \begin{cases} 0 & IDC_{max} \le 82 \\ 2.8\cdot (IDC_{max}-82) & 82 < IDC_{max} \le 92 \\ 55 & IDC_{max} > 92 \end{cases}$$PIDC,E30​=⎩⎨⎧​02.8⋅(IDCmax​−82)55​IDCmax​≤8282<IDCmax​≤92IDCmax​>92​$$S_{fuel} = \max(0,\ 100 – (P_{AFR} + P_{IDC}))$$Sfuel​=max(0, 100−(PAFR​+PIDC​))

---

5) Boost Score $S_{boost}$Sboost​ (keep it tight on all fuels)

Same structure, but overshoot matters slightly less on ethanol only if knock is clean (still not a free pass).

• All fuels:
  $P_{overshoot} = 25O,\quad P_{track} = 20E_B$Povershoot​=25O,Ptrack​=20EB​
• Oscillation penalty stays:
  $P_{osc} = 10J$Posc​=10J

$$S_{boost} = \max(0,\ 100 – (P_{overshoot}+P_{track}+P_{osc}))$$Sboost​=max(0, 100−(Povershoot​+Ptrack​+Posc​))

---

6) Thermal Score $S_{thermal}$Sthermal​ (ethanol still heat-soaks)

Keep thermal strict. Ethanol helps knock, not physics.

Use your existing IAT-based penalties, but trigger CAUTION earlier if:

• Thermal score < 75 (all fuels)
• Or IAT rise is steep pull-to-pull

---

7) Classification Bands (same, but smarter)

• SAFE: $S_{overall} \ge 85$Soverall​≥85 and no gates triggered
• CAUTION: 70–84
• STOP: <70 or any STOP gate

---

The one detail I need to lock this down

When you say ethanol blend, are you usually:

• E20–E30, or
• E40–E60, or
• “whatever the pump gives” (variable)?

Because IDC limits and AFR tolerance need different ceilings once you push past ~E30. If you tell me your typical blend range, I’ll finalize the E-profile into E30 vs E50 with correct IDC headroom rules.

Pump Options

Good. Let’s define clear fuel operating profiles for Black Pearl so the scoring model, boost ceilings, and timing strategy automatically adapt.

We’ll assume three real-world pump scenarios:

1. 91 Octane
2. 93 Octane
3. E30 Blend (controlled mix of pump 93 + E85)

Anything beyond E30 on stock-style fuel systems becomes a hardware conversation, not a tuning conversation.

---

🔵 Fuel Profile 1 — 91 Octane (Most Conservative)

Intended Use

• Daily driving
• Hot weather
• Long pulls
• Worst-case fuel quality

Target Philosophy

• Stability > output
• Slightly richer under load
• Lower peak boost
• Conservative timing

Recommended Calibration Envelope (EJ205 Safe Bias)

• Peak Boost: ~15–16 psi (VF22 dependent)
• WOT AFR target: 10.8–11.0
• Timing ramp conservative midrange
• Zero tolerance for repeat knock
• Injector duty ceiling: 90%

Scoring Sensitivity

• Highest knock penalty
• Tightest AFR lean tolerance
• CAUTION if any FLKC in WOT

Use this profile when ambient temps are high or fuel quality is questionable.

---

🔵 Fuel Profile 2 — 93 Octane (Primary Power Target)

Intended Use

• Performance street
• Controlled pulls
• Main 305whp attempt baseline

Target Philosophy

• Controlled midrange torque
• Smooth boost ramp
• Clean combustion

Recommended Calibration Envelope

• Peak Boost: ~17–18 psi (depending on turbo efficiency)
• WOT AFR target: 10.8–11.2
• Timing modest but progressive
• Injector duty ceiling: 92–93%

Scoring Sensitivity

• Moderate knock penalty
• AFR tracking must stay tight
• Boost overshoot tolerance slightly higher than 91

This should be your primary tuning fuel for developing the 305whp calibration.

---

🔵 Fuel Profile 3 — E30 Blend (Controlled Ethanol Strategy)

This is where things get serious — but also risky if fuel system isn’t upgraded.

Intended Use

• Maximum safe pump-based output
• Knock resistance increase
• Thermal stability improvement

Target Philosophy

• Maintain knock margin
• Protect fuel system headroom
• Do not chase timing aggressively

Recommended Calibration Envelope

• Peak Boost: ~18–19 psi (only if airflow supports it safely)
• WOT AFR target: 11.3–11.5 equivalent gasoline scale
• Timing slightly increased vs 93, but incremental
• Injector duty ceiling: 90–92% max (ethanol consumes more fuel)

Critical Rule

If IDC exceeds 92% at E30 → stop increasing boost immediately.

Ethanol gives knock headroom.
It does NOT give fuel system headroom.

---

⚠️ Fuel Strategy Doctrine

Never Tune on Mixed Unknown Fuel

If you don’t know ethanol content, treat it as 93.

Never Increase Timing and Boost in the Same Revision

Especially on E blends.

Always Log Before Power Escalation

Fuel trims, IAM stability, and IDC must be clean first.

---

🧠 Strategic Recommendation

For Project Black Pearl:

• Develop base stability on 93
• Validate repeatability
• Then introduce controlled E30 map
• Compare log scores side-by-side

That gives you:

• Content
• Data
• A monetizable comparison framework
• A safety buffer

---

Now the real question:

Are your injectors and pump stock or upgraded?

Because if they’re stock, your ethanol ceiling determines your entire 305whp strategy.

Stock OEM 440cc injectors (20 years old) / Fuel Pump = AEM 440 High Performance Fuel Pump

Alright — with stock OEM 440cc injectors (20 years old) + AEM 340 lph pump, your hard limit is the injectors, not the pump.

Tell-it-like-it-is:

• E30 is not a “go make more power” button on 440cc injectors.
• Ethanol increases fuel demand ~25–35% depending on blend.
• That pushes IDC into the danger zone fast, and old injectors add unknown flow variance.
• Your pump is fine. Your injectors are the choke point.

So your “Pump Options” doctrine needs to be injector-aware.

---

Black Pearl Fuel Doctrine (Injector-Limited Edition)

✅ Allowed Operating Profiles (with 440cc injectors)

1) 91 Octane Map (Daily / Hot Weather / Worst Fuel)

Purpose: survive anything.

• Boost: conservative
• Timing: conservative
• Goal: reliability + knock-free

IDC guardrail: keep ≤ 85–88% at WOT.

---

2) 93 Octane Map (Primary Performance Map)

This is your best path to ~305whp if the hardware/turbo supports it.

• Boost: moderate-high but controlled
• Timing: incremental
• Fueling: stable and rich enough to protect

IDC guardrail: keep ≤ 90–92% (absolute).

---

3) “Light Ethanol” Blend Only (E10–E20 max)

This is the only ethanol strategy that makes sense on 440s unless you accept high risk.

Target range: E10–E20

• Use it for knock margin and consistency
• Not for turning the boost knob to the sky

IDC guardrail: keep ≤ 88–90%.

E30? On 440s: not recommended for WOT tuning unless you’re okay with living on the edge.

---

🚫 Prohibited Profile (for your current setup)

E30 “Power Map” on 440cc injectors

Hard no as a repeatable strategy.
If you do it anyway, the scoring model should treat:

• IDC > 92% as STOP
• Any lean deviation as STOP
• Any IAM drop as STOP

Because you have no headroom, and ethanol content variation will bite you.

---

Update to the Scoring Model for Your Setup

Injector Duty STOP Gates (tightened)

Because injectors are old and small:

• STOP if IDC > 92% (all fuels)
• CAUTION if IDC 88–92%
• SAFE zone ≤ 88%

(Yes, this is stricter than before — that’s what “20-year-old 440s” demands.)

Fuel Profile Weights (injector-limited)

Even on pump gas, fueling deserves more weight because you’re near the ceiling.$$S_{overall} = 0.36S_{knock} + 0.30S_{fuel} + 0.18S_{boost} + 0.11S_{thermal} + 0.05S_{repeat}$$Soverall​=0.36Sknock​+0.30Sfuel​+0.18Sboost​+0.11Sthermal​+0.05Srepeat​

---

What this means for your 305whp target (realistic)

With 440cc injectors, the build is usually capped by:

• IDC headroom
• AFR stability at high airflow
• consistency between cylinders (old injectors)

You can still build a fast WRX, but 305whp is “possible but not guaranteed” depending on turbo efficiency, intercooling, and your boost curve.

If you want 305whp reliably and safely, injectors are the next mandatory move.

---

The smartest money move (ROI)

Option A — Replace injectors with something modern

• STI “pinks” (565cc) if you want simple and proven
• ID1050x / similar if you want future flex / bigger turbo headroom

Option A is the move that prevents you from wasting time “tuning around” injector limits.

Option B — Keep 440s and cap the mission

• Focus on response and midrange
• Target a lower whp goal but very reliable
• Build the rally drivability angle

---

Two quick, high-value questions

1. What turbo are you actually on right now (still VF22?) and what boost are you commanding?
2. Do you have wideband AFR integrated into logs, or external only?

Answer those and I’ll give you a tight, injector-safe boost/AFR envelope for each fuel profile and lock in the STOP/CAUTION thresholds for your scoring model.

NOTE: Block Stamp Says = EJ20, Turbo = VF22 OR TD04 (unsure)

Good. This matters a lot.

You just identified two completely different airflow ceilings.

VF22 and TD04 are not remotely the same conversation.

Before we even talk about 305whp, we need to anchor reality.

---

Step 1 — Engine Reality

Block stamp “EJ20” tells us almost nothing useful.

What matters:

• Is it still an EJ205 (USDM 2005 WRX)?
• Has it been swapped?
• Compression?
• AVCS?
• Heads original?

For now, we assume EJ205 longblock unless proven otherwise.

---

Step 2 — Turbo Identification Is Critical

Scenario A — You Have a TD04 (Stock WRX Turbo)

Airflow Reality:

• ~14–16 psi efficient range
• ~240–260whp realistic
• Spools very fast
• Falls off up top

With 440cc injectors:

This setup is actually balanced.
Injector ceiling aligns with turbo airflow.

305whp?

Not happening.
Physics says no.

But you can build:

• Very responsive
• Rally-friendly midrange
• Extremely repeatable

If you’re on TD04, your mission changes from “305whp” to:
“Maximum usable torque within injector limits.”

---

Scenario B — You Have a VF22

Different animal entirely.

Airflow Reality:

• Much larger compressor
• Supports 300+ whp
• Slower spool
• More boost potential

With 440cc injectors:

You will hit injector ceiling before turbo ceiling.

That means:
You will not safely exploit the VF22’s potential on stock injectors.

---

Step 3 — How to Identify Turbo Immediately

Do this:

1. Look at compressor housing.
2. Look for stamping plate.
3. TD04 usually marked “TD04L”.
4. VF22 will have IHI stamping and specific compressor shape.
5. Measure inlet size — VF22 inlet is larger.

If you want, describe what you physically see and I’ll narrow it down.

---

Now Let’s Be Strategic

With 440cc injectors:

If TD04:

• Peak boost ceiling: ~16 psi
• IDC limit: 88–90%
• Target AFR: 10.8–11.2
• 305whp target: abandon
• Build torque curve, not peak power

If VF22:

• You must upgrade injectors before chasing 305whp
• Or cap boost early and accept wasted turbo potential

---

Here’s the Bigger Question

Are you trying to:
A) Build the cleanest, smartest TD04 rally street weapon?
B) Or build a legitimate 300+ whp VF22 car?
C) Or are you unsure what hardware you actually have?

Because the next 10 tuning decisions depend entirely on that answer.

Let’s lock the hardware truth first.

EJ205 (USDM 2005 WRX) = verified USDM motor, NOT JDM

Good. That removes a lot of unknowns.

You have:

• EJ205 USDM
• 8.0:1 compression
• AVCS intake
• Stock 440cc injectors (aging)
• AEM 340 pump
• Turbo = still unknown (VF22 or TD04)

Now we build the doctrine around what that engine actually tolerates.

---

EJ205 Reality Check (Non-Negotiable)

The EJ205 is:

• Ringland-sensitive
• Knock-intolerant under high cylinder pressure
• Not forgiving of lean spikes
• Very sensitive in the 3,500–4,800 RPM high-load range

That midrange area is where most 205s die.

So your scoring model must weight midrange knock more heavily than top-end knock.

---

Updated Black Pearl Knock Doctrine (EJ205-Specific)

High-Risk Zone Multiplier

If knock occurs between:

• 3,200–5,000 RPM
• Load > 1.8 g/rev (or equivalent high load region)

Then multiply FBKC/FLKC penalties by:$$1.5\times$$1.5×

This makes midrange knock significantly more expensive in your scoring system.

Because that’s where ringlands crack.

---

EJ205 Safe Operating Envelope (Pump Gas, 440cc Injectors)

On 93 Octane (realistic max scenario)

With stock injectors:

• Peak Boost: 16–17 psi
• Midrange Boost: must ramp smoothly
• AFR Target: 10.8–11.2
• IDC ceiling: 90% hard
• Timing: conservative in 3.5–5k

Trying to push 18–19 psi on 440s is where things get unstable.

---

If You Actually Have a TD04

Your car becomes:

• Extremely responsive
• Torque-biased
• Safer at moderate boost
• Limited in peak power

And honestly?
That’s a fantastic rally-street configuration.

---

If You Actually Have a VF22

Then:

• You’ll run out of injector before turbo
• You’ll see IDC climb fast above 15–16 psi
• 305whp requires injector upgrade
• Ethanol becomes fuel-system-limited immediately

---

Hard Truth About 305whp on 440cc Injectors

Possible?
Maybe.

Stable and repeatable?
Very unlikely.

Sustainable?
Not if you’re near 95% IDC.

If your mission is “documented, intelligent, repeatable 305whp,” injectors are the gating item.

---

Strategic Fork in the Road

You now have two intelligent paths:

Path A — Stay 440cc

• Target ~270–285whp (depending on turbo)
• Optimize torque curve
• Make it bulletproof
• Market it as “Smart Stock Injector Build”

Path B — Upgrade Injectors

• 565cc minimum
• Or 1000cc modern injectors
• Then 305whp becomes a controlled engineering target

---

Now I need one final piece of truth:

When you roll into boost, how does it feel?

• Full boost by ~2,800–3,000 RPM? → likely TD04
• Doesn’t wake up until ~3,800–4,200 RPM? → likely VF22

Describe spool behavior and I’ll narrow it down before we finalize your control doctrine.

3rd gear feels like the kick-in spot

That description alone doesn’t confirm the turbo — but it tells me something important.

If it only “feels alive” in 3rd gear, that usually means one of three things:

1. The turbo is modest (TD04-like) and load in 3rd allows it to hit full boost cleanly
2. The boost curve is ramping hard midrange (aggressive WGDC spike)
3. You’re not feeling real spool delay — you’re feeling load-dependent torque

So let’s cut through it logically.

---

Turbo Identification Through Behavior

Answer this precisely:

When you floor it in 2nd gear at ~2,500 rpm:

• Does it hit strong boost by ~3,000–3,200 rpm?
• Or does it feel soft until ~3,800–4,000+ rpm?

If boost comes in by ~3k → TD04 behavior

If boost doesn’t wake up until near 4k → VF22 behavior

The VF22 is noticeably lazier down low.

---

Why This Matters for Doctrine

Because:

If TD04:

• 440cc injectors are correctly matched
• 305whp goal is unrealistic
• Best strategy: torque curve shaping + repeatability
• Safer build, more rally-friendly

If VF22:

• Injector ceiling becomes immediate constraint
• You’re airflow-capable of 300+ whp
• 440cc injectors become the bottleneck
• Boost escalation becomes risky without injector upgrade

---

Now Let’s Think Like Engineers

You said “3rd gear feels like the kick-in spot.”

That’s normal for both turbos because:

• 3rd gear loads the engine more
• Boost builds more completely
• Torque feels stronger

That doesn’t distinguish turbo type.

---

Quick Physical Check (Fastest Answer)

Look at the compressor inlet:

• TD04: small inlet, compact housing, stock WRX look
• VF22: visibly larger compressor housing and inlet

Or:

Do you hear strong spool whistle early (~2.5–3k)?
TD04 tends to spool early and audibly.

---

Strategic Reality Right Now

Until we confirm turbo:

Your scoring model must assume injector-limited strategy.

So:

• IDC > 90% = CAUTION
• IDC > 92% = STOP
• No midrange timing escalation
• No boost increases until clean repeatable pulls

---

Now I’m going to ask the question that determines everything:

Have you logged a WOT pull yet with IDC and boost?

Because one single 3rd gear log will tell us turbo type indirectly.

### If boost doesn’t wake up until near 4k → VF22 behavior The VF22 is noticeably lazier down low.

I’m going to say this – because no matter the throttle pressure 1,2 gear NEVER wakes up, AVCS = non-existent (as far as I can tell)

Good. That statement changes things.

If:

• 1st and 2nd never really “wake up”
• Boost feels lazy down low
• Power doesn’t come alive until later
• AVCS feels nonexistent (no early torque swell)

Then this strongly leans toward VF22 behavior, not TD04.

A TD04 on an EJ205 absolutely wakes up in 2nd gear. It hits early and hard around 2.8–3.2k RPM. You would feel it.

A VF22, on the other hand:

• Spools later
• Feels soft below ~3.5k
• Comes alive closer to 3.8–4.2k
• Feels stronger in 3rd due to load

What you described fits that profile much better.

---

Now the Important Part

If you are on a VF22 with:

• Stock 440cc injectors
• AEM 340 pump
• EJ205

Then your system is airflow-capable but fuel-limited.

That means your current doctrine must shift from:

“Find more boost”

to

“Protect injector headroom at all costs.”

---

What This Likely Means in Practice

On VF22 + 440cc:

• Boost above ~15–16 psi will push IDC very quickly
• Midrange cylinder pressure becomes dangerous fast
• Ethanol blends will crush injector headroom
• 305whp is mechanically possible from airflow
• But not safely possible on 440s long-term

This is the classic “big turbo, small injector” mismatch.

---

Critical EJ205 + VF22 Risk Zone

The danger band becomes:

3,500–4,800 RPM
High load
High boost ramp
Old injectors near saturation

That is where ringlands crack.

So your scoring model must:

• Apply 1.5× knock penalty in 3.2–5k band
• Treat IDC > 88% as CAUTION
• Treat IDC > 92% as STOP
• Forbid boost increases unless IDC < 85%

---

Strategic Decision Point

If this is truly VF22:

You have two intelligent paths.

Path A — Cap the VF22

• Keep boost ~15–16 psi
• Focus on smooth ramp
• Build midrange torque
• Stay injector-safe
• Accept lower peak number

Path B — Upgrade Injectors

• 565cc minimum
• Or modern 1000cc
• Unlock real VF22 potential
• 305whp becomes realistic

---

Hard Truth

Chasing 305whp on VF22 with 440cc injectors is not intelligent engineering.

It’s gambling.

And Project Black Pearl is not a gambling platform.

---

Now I want one more data point:

When you roll into WOT from 3k in 3rd, does boost climb smoothly and then surge near 4k?

Or does it spike quickly and then settle?

That tells me whether boost control is also part of the “kick” you’re feeling.

Feels like a spike

Good. That’s the most useful answer so far.

If it feels like a spike, here’s what that usually means on an EJ205:

• Boost is ramping too aggressively in midrange
• WGDC is front-loaded
• Turbo is hitting target abruptly instead of progressively
• Torque rise is steep, not linear

That “kick” you feel is likely boost control behavior, not just turbo size.

Now let’s dissect this properly.

---

What a Spike Usually Means

In 3rd gear:

1. Boost undershoots slightly
2. WGDC overcompensates
3. Boost overshoots
4. Torque jumps suddenly
5. ECU trims it back

That creates a surge sensation.

This can happen with both TD04 and VF22.

So the spike does NOT confirm turbo size.

It confirms:

⚠️ Boost control is not smooth.

And on an EJ205, that’s dangerous.

---

Why This Matters (Very Important)

The EJ205 hates:

• Sudden midrange torque spikes
• Rapid cylinder pressure increases
• High load + aggressive timing

Ringlands fail when torque rises too sharply, not just when peak power is high.

Your doctrine must now include:

“Torque rate-of-change control.”

---

Immediate Doctrine Adjustment

Your scoring model needs an added metric:

Torque Ramp Penalty (Boost Ramp Proxy)

We use boost rate of change as a torque proxy.

Let:$$R = \max\left(\frac{dBoost}{dt}\right)$$R=max(dtdBoost​)

If boost rises more than a defined threshold per second:

• Add ramp penalty.

Example logic:

• Mild ramp → 0 penalty
• Moderate spike → -10 points
• Aggressive spike → -25 points

Because violent torque ramps are mechanically abusive.

---

What This Suggests About Your Current Tune

One of the following is likely true:

1. WGDC table is too aggressive in midrange
2. Turbo dynamics compensation is not smooth
3. Boost target ramp is steep
4. Timing is stacking with boost ramp

You don’t “feel horsepower.”
You feel torque change.

---

Now Let’s Think Strategically

If you are on:

TD04

A spike usually means overly aggressive WGDC.

VF22

A spike can mean:

• Turbo lag + hard hit when it finally lights
• WGDC ramp too aggressive

Either way, the fix is similar:

Smooth the boost ramp.
Protect the 3.5–5k zone.
Don’t chase peak psi.

---

Critical Question

Have you logged:

• Boost target
• Boost actual
• WGDC
• RPM
• IAM
• FBKC
• FLKC
• IDC

If not, that’s the next move.

Because we need one clean 3rd gear pull log to:

• Identify turbo behavior indirectly
• Confirm injector headroom
• Score midrange knock risk
• Quantify the spike

---

Right now, based on behavior alone:

You likely have:

• Larger-than-stock turbo
• Aggressive midrange WGDC
• Injector headroom risk
• No AVCS torque smoothing benefit

But we need data to stop guessing.

Are you ready to log a structured 3rd gear pull?

Fold the updated doctrine under the original

Understood. Below is the fully integrated Black Pearl ECU Control Doctrine, with all updated injector, EJ205, fuel, and boost-ramp protections folded cleanly under the original structure.

This is now the authoritative version.

---

🧠 BLACK PEARL ECU CONTROL DOCTRINE

EJ205 USDM | Injector-Limited | Boost-Ramp Controlled | AI-Governed

---

I. Foundational Principle

The ECU is a combustion stability controller.

Power is a byproduct of:

• Controlled cylinder pressure
• Stable fueling
• Smooth torque delivery

The EJ205 does not tolerate:

• Lean spikes
• Injector saturation
• Midrange torque surges

---

II. Chain of Authority

1. Mechanical integrity
2. Sensor data
3. Structured logs
4. AI scoring
5. Human decision

AI advises. It does not command.

---

III. Hardware Reality (Locked Assumptions)

Engine: EJ205 USDM (verified)
Injectors: OEM 440cc (20 years old)
Fuel Pump: AEM 340
Turbo: Likely VF22 (behavior-based assumption)
AVCS: Present (USDM), but torque behavior suggests minimal smoothing

Primary constraint: Injector headroom

---

IV. Optimization Hierarchy

1. Mechanical stability
2. Fuel delivery accuracy
3. Injector headroom preservation
4. Boost ramp smoothness
5. Knock stability
6. Timing refinement
7. Torque shaping
8. Peak output

Peak power is last.

---

V. Data Governance Standard

Required logging channels:

• RPM
• Load
• Boost target
• Boost actual
• WGDC
• IAM
• FBKC
• FLKC
• AFR (wideband)
• Injector Duty Cycle
• IAT
• Coolant Temp
• Throttle angle

No log → no revision.

---

VI. Fuel Profiles (Injector-Limited Edition)

91 Octane

• Conservative boost
• AFR 10.8–11.0
• IDC SAFE ≤ 88%

93 Octane (Primary Development Fuel)

• Boost moderate (likely 15–17 psi ceiling)
• AFR 10.8–11.2
• IDC SAFE ≤ 90%

Light Ethanol (E10–E20 Only)

• Not for power escalation
• IDC SAFE ≤ 88%
• No E30 power tuning on 440cc injectors

---

VII. Hard STOP Gates (Absolute)

STOP if any:

• IAM drops
• FBKC ≤ -2.8 in WOT region
• IDC > 92%
• AFR error exceeds fuel-profile lean limit
• Boost overshoot > 2 psi
• Severe boost oscillation

CAUTION if:

• IDC 88–92%
• Any FLKC in high-load zone
• Aggressive boost ramp
• Thermal score < 75

---

VIII. Midrange Risk Multiplier (EJ205-Specific)

If knock occurs between:

3,200–5,000 RPM
AND high load

Then:

Knock penalties × 1.5

This protects ringlands in the most failure-prone band.

---

IX. Boost Ramp Control Doctrine (New Layer)

Sudden torque rise is prohibited.

Define Boost Ramp Rate:

R = max(dBoost/dt)

If ramp exceeds threshold:

• Mild spike → -10 score
• Moderate spike → -20
• Aggressive spike → -30

Boost must build progressively.
No midrange “hit.”

Torque smoothness > sensation.

---

X. Injector Preservation Doctrine

Because injectors are:

• Small
• Aged
• Primary bottleneck

Rules:

• No boost increase unless IDC < 85%
• IDC 85–88% = caution zone
• IDC 88–92% = danger zone
• IDC > 92% = STOP

Injector saturation is silent engine damage.

---

XI. Weighted Scoring Model (Revised)

Injector-limited weighting:$$S_{overall} = 0.36S_{knock} + 0.30S_{fuel} + 0.18S_{boost} + 0.11S_{thermal} + 0.05S_{repeat}$$Soverall​=0.36Sknock​+0.30Sfuel​+0.18Sboost​+0.11Sthermal​+0.05Srepeat​

Knock remains dominant.
Fueling nearly equal importance due to injector ceiling.

---

XII. Validation Protocol

Before power escalation:

• Two clean repeatable pulls
• IAM stable
• IDC below safe threshold
• Boost tracking stable
• No ramp spike penalties
• AFR within tolerance

If validation fails → revert.

---

XIII. Rally Bias Reinforced

Calibration must deliver:

• Linear throttle response
• Smooth torque rise
• No violent midrange surge
• Repeatable pull-to-pull behavior
• Thermal resilience

Dyno shock factor is irrelevant.

---

XIV. Strategic Fork Doctrine

Given current hardware:

Path A — Injector-Limited Optimization

• Cap boost
• Shape torque
• Maximize drivability
• Likely 270–285whp range
• Market as intelligent stock-injector build

Path B — Injector Upgrade (Required for 305whp Goal)

• 565cc minimum
• Or 1000cc modern injectors
• Unlock VF22 potential
• Ethanol strategy becomes viable

Without injector upgrade:
305whp is unstable engineering.

---

XV. Prohibited Actions

• No boost increase while IDC > 85%
• No timing increase in 3.2–5k until knock-free for 2 sessions
• No simultaneous boost + timing escalation
• No tuning on unknown ethanol content
• No spike-based “feels fast” conclusions

---

XVI. Strategic Outcome

Black Pearl will demonstrate:

• Injector-aware engineering discipline
• Torque-controlled boost strategy
• Knock-weighted scoring system
• Data-backed calibration decisions
• A monetizable tuning methodology

---

This is now a fully unified doctrine:

Original framework

• Fuel profile logic
• Injector ceiling governance
• EJ205 midrange multiplier
• Boost ramp control layer

Next escalation options:

1. Formal numeric ramp-rate thresholds
2. Define exact boost ceiling per suspected turbo
3. Build version-control naming tied to your .bin files

Where do you want to tighten the system next?

Below is a formal, numeric, weighted scoring model you can run on every WOT pull. It outputs:

• 0–100 Overall Score
• SAFE / CAUTION / STOP
• Component subscores you can graph over time

It’s built to be knock- and safety-dominant (because that’s what keeps EJ205s alive).

---

Black Pearl Pull Scoring Model v1.0

1) Overall Score (0–100)

$$S_{overall} = 0.35S_{knock} + 0.25S_{fuel} + 0.20S_{boost} + 0.15S_{thermal} + 0.05S_{repeat}$$Soverall​=0.35Sknock​+0.25Sfuel​+0.20Sboost​+0.15Sthermal​+0.05Srepeat​

Why these weights: knock + fueling are the biggest “engine-ending” failure modes on pump gas; boost control and thermal stability drive repeatability; repeatability is important but not more important than safety.

---

2) Subscores (each 0–100)

A) Knock Score $S_{knock}$Sknock​ (0–100)

Use 3 penalties: IAM, FBKC, FLKC.

IAM penalty

• Let $IAM_{start}, IAM_{end}$IAMstart​,IAMend​
• $\Delta IAM = IAM_{start} – IAM_{end}$ΔIAM=IAMstart​−IAMend​

$$P_{IAM} = \begin{cases} 0 & \Delta IAM = 0 \\ 60 & \Delta IAM > 0 \end{cases}$$PIAM​={060​ΔIAM=0ΔIAM>0​

(Any IAM drop during/after the pull is a major red flag → heavy penalty.)

FBKC penalty

• Let $FBKC_{min}$FBKCmin​ be most negative FBKC during pull (e.g., -1.41, -2.81)

$$P_{FBKC} = \min(60,\ 20 \cdot |FBKC_{min}|)$$PFBKC​=min(60, 20⋅∣FBKCmin​∣)

So: -1.41 → 28.2 penalty, -2.81 → 56.2 penalty, -4.22 caps at 60.

FLKC penalty

• Let $FLKC_{min}$FLKCmin​ be most negative learned knock in the pull region

$$P_{FLKC} = \min(40,\ 15 \cdot |FLKC_{min}|)$$PFLKC​=min(40, 15⋅∣FLKCmin​∣)

Knock Score$$S_{knock} = \max(0,\ 100 – (P_{IAM}+P_{FBKC}+P_{FLKC}))$$Sknock​=max(0, 100−(PIAM​+PFBKC​+PFLKC​))

---

B) Fuel Score $S_{fuel}$Sfuel​ (0–100)

This score cares about AFR error + injector duty headroom.

AFR tracking penalty

You need: measured wideband AFR $AFR_{meas}$AFRmeas​ and commanded/target AFR $AFR_{tgt}$AFRtgt​ (or OL target if available).

Compute error over the pull window (e.g., 3500–6500 rpm):

• $E_{AFR} = \text{median}(|AFR_{meas} – AFR_{tgt}|)$EAFR​=median(∣AFRmeas​−AFRtgt​∣)
• $E_{AFR,max} = \text{max}(|AFR_{meas} – AFR_{tgt}|)$EAFR,max​=max(∣AFRmeas​−AFRtgt​∣)

$$P_{AFR} = 30\cdot E_{AFR} + 10\cdot E_{AFR,max}$$PAFR​=30⋅EAFR​+10⋅EAFR,max​

(Example: median error 0.2 and max 0.5 → $P_{AFR}=30(0.2)+10(0.5)=11$PAFR​=30(0.2)+10(0.5)=11.)

IDC penalty

Let $IDC_{max}$IDCmax​ be peak injector duty cycle (%).$$P_{IDC} = \begin{cases} 0 & IDC_{max} \le 85 \\ 2\cdot (IDC_{max}-85) & 85 < IDC_{max} \le 95 \\ 40 & IDC_{max} > 95 \end{cases}$$PIDC​=⎩⎨⎧​02⋅(IDCmax​−85)40​IDCmax​≤8585<IDCmax​≤95IDCmax​>95​

Fuel Score$$S_{fuel} = \max(0,\ 100 – (P_{AFR} + P_{IDC}))$$Sfuel​=max(0, 100−(PAFR​+PIDC​))

---

C) Boost Control Score $S_{boost}$Sboost​ (0–100)

Reward accurate tracking; punish overshoot and oscillation.

You need boost target $B_{tgt}$Btgt​ and actual $B_{act}$Bact​ (same units).

Compute:

• Peak overshoot: $O = \max(B_{act}-B_{tgt})$O=max(Bact​−Btgt​)
• Mean absolute tracking error: $E_B = \text{mean}(|B_{act}-B_{tgt}|)$EB​=mean(∣Bact​−Btgt​∣)
• Oscillation proxy: $J = \text{std}(\Delta B_{act})$J=std(ΔBact​) (std of boost “rate of change”)

$$P_{overshoot} = 25\cdot O$$Povershoot​=25⋅O $$P_{track} = 20\cdot E_B$$Ptrack​=20⋅EB​ $$P_{osc} = 10\cdot J$$Posc​=10⋅J

Boost Score$$S_{boost} = \max(0,\ 100 – (P_{overshoot}+P_{track}+P_{osc}))$$Sboost​=max(0, 100−(Povershoot​+Ptrack​+Posc​))

(If you don’t have a clean $J$J, skip it and redistribute weight inside boost.)

---

D) Thermal Score $S_{thermal}$Sthermal​ (0–100)

We’re scoring IAT and heat soak sensitivity.

Let:

• $IAT_{start}, IAT_{end}$IATstart​,IATend​
• $\Delta IAT = IAT_{end}-IAT_{start}$ΔIAT=IATend​−IATstart​

$$P_{IATlevel} = \begin{cases} 0 & IAT_{start} \le 90F \\ 0.5\cdot (IAT_{start}-90) & 90F < IAT_{start} \le 130F \\ 30 & IAT_{start} > 130F \end{cases}$$PIATlevel​=⎩⎨⎧​00.5⋅(IATstart​−90)30​IATstart​≤90F90F<IATstart​≤130FIATstart​>130F​ $$P_{IATrise} = 2\cdot \Delta IAT$$PIATrise​=2⋅ΔIAT

Thermal Score$$S_{thermal} = \max(0,\ 100 – (P_{IATlevel}+P_{IATrise}))$$Sthermal​=max(0, 100−(PIATlevel​+PIATrise​))

(If you log °C, adjust thresholds accordingly.)

---

E) Repeatability Score $S_{repeat}$Srepeat​ (0–100)

This only makes sense once you have two pulls under similar conditions.

Let:

• $\Delta B_{peak}$ΔBpeak​ = |peak boost pull1 – pull2|
• $\Delta AFR$ΔAFR = median |AFR error pull1 – pull2|
• $\Delta FBKC$ΔFBKC = |FBKCmin pull1 – pull2|

$$P_{rep} = 20\cdot \Delta B_{peak} + 50\cdot \Delta AFR + 10\cdot \Delta FBKC$$Prep​=20⋅ΔBpeak​+50⋅ΔAFR+10⋅ΔFBKC $$S_{repeat} = \max(0,\ 100 – P_{rep})$$Srepeat​=max(0, 100−Prep​)

---

3) Classification Rules (hard gates + score bands)

Even if the weighted score looks okay, these gates override:

STOP (immediate)

• $\Delta IAM > 0$ΔIAM>0 (IAM drop)
• $FBKC_{min} \le -2.8$FBKCmin​≤−2.8 in the WOT load cells
• AFR goes lean beyond a hard limit you set (example: > 0.8 AFR lean vs target at high load)
• $IDC_{max} > 95\%$IDCmax​>95%

CAUTION

• $FBKC_{min}$FBKCmin​ between -1.4 and -2.8
• Any FLKC in the WOT region
• Boost overshoot > 1.0 psi (or equivalent)
• Thermal score < 70

SAFE

• No IAM drop, minimal/no knock, fueling and boost stable

Score bands

• SAFE: $S_{overall} \ge 85$Soverall​≥85
• CAUTION: 70–84
• STOP: <70 or any STOP gate triggered

---

4) How you use it (the money part)

Every revision gets a line item like:

• v1.07: Overall 78 (CAUTION) → “Reduce timing 0.5° @ 4.2g/rev 4200–5200, fix WGDC overshoot”
• v1.08: Overall 89 (SAFE) → “Proceed to small WGDC +1% in midrange”

This turns your tuning into a measurable product and your videos into evidence-based episodes.

---

Quick probing questions (so we set the right thresholds)

1. Are you logging wideband AFR into the ECU/log, or is it external only?
2. What units are you using for boost (psi or kPa), and do you have boost target logged?
3. What’s your typical fuel (91/93, ethanol blend, etc.)?

Answer those and I’ll tailor the hard limits + coefficients so the scoring matches your exact setup.

---

Fuel Profiles (choose one per pull/session)

Profile A — 91 Pump (most conservative)

• Knock penalties harsher
• Lean events less tolerated
• IDC headroom more protected

Profile B — 93 Pump

• Moderately conservative

Profile C — Ethanol Blend (assume E30 unless you specify)

• Knock tolerance can be slightly wider
• Still zero tolerance for IAM drops and major knock
• Fuel system limits matter more (ethanol increases fuel demand)

---

1) Weights (slightly revised for ethanol)

91 / 93

$$S_{overall} = 0.38S_{knock} + 0.25S_{fuel} + 0.20S_{boost} + 0.12S_{thermal} + 0.05S_{repeat}$$Soverall​=0.38Sknock​+0.25Sfuel​+0.20Sboost​+0.12Sthermal​+0.05Srepeat​

Ethanol blend (E30 default)

$$S_{overall} = 0.34S_{knock} + 0.30S_{fuel} + 0.18S_{boost} + 0.13S_{thermal} + 0.05S_{repeat}$$Soverall​=0.34Sknock​+0.30Sfuel​+0.18Sboost​+0.13Sthermal​+0.05Srepeat​

Why: ethanol usually improves knock margin, but it stresses fueling capacity, so fueling gets more weight.

---

2) Hard STOP Gates (these do not negotiate)

All fuels (91/93/E-blend)

STOP immediately if any:

• IAM drops during/after pull
• FBKC ≤ -2.8 in WOT/load region
• IDC > 95%
• Boost overshoot > 2.0 psi (or equivalent) with oscillation

Fueling STOP (AFR error vs your target)

This uses error vs commanded, not absolute AFR.

• 91: STOP if max error ≥ +0.6 AFR lean
• 93: STOP if max error ≥ +0.7 AFR lean
• E30: STOP if max error ≥ +0.8 AFR lean
  (Still not “safe,” just slightly more tolerant before hard stop.)

---

3) Knock Score $S_{knock}$Sknock​ (fuel-adjusted penalties)

IAM penalty (unchanged)

$$P_{IAM} = \begin{cases} 0 & \Delta IAM = 0 \\ 70 & \Delta IAM > 0 \end{cases}$$PIAM​={070​ΔIAM=0ΔIAM>0​

FBKC penalty coefficient by fuel

Let $FBKC_{min}$FBKCmin​ be most negative value.

• 91: $\;P_{FBKC} = \min(70,\ 25\cdot |FBKC_{min}|)$PFBKC​=min(70, 25⋅∣FBKCmin​∣)
• 93: $\;P_{FBKC} = \min(65,\ 22\cdot |FBKC_{min}|)$PFBKC​=min(65, 22⋅∣FBKCmin​∣)
• E30: $\;P_{FBKC} = \min(60,\ 18\cdot |FBKC_{min}|)$PFBKC​=min(60, 18⋅∣FBKCmin​∣)

FLKC penalty coefficient by fuel

• 91: $\;P_{FLKC} = \min(50,\ 18\cdot |FLKC_{min}|)$PFLKC​=min(50, 18⋅∣FLKCmin​∣)
• 93: $\;P_{FLKC} = \min(45,\ 16\cdot |FLKC_{min}|)$PFLKC​=min(45, 16⋅∣FLKCmin​∣)
• E30: $\;P_{FLKC} = \min(40,\ 12\cdot |FLKC_{min}|)$PFLKC​=min(40, 12⋅∣FLKCmin​∣)

$$S_{knock} = \max(0,\ 100 – (P_{IAM}+P_{FBKC}+P_{FLKC}))$$Sknock​=max(0, 100−(PIAM​+PFBKC​+PFLKC​))

---

4) Fuel Score $S_{fuel}$Sfuel​ (ethanol tightens IDC rules)

AFR penalty (same math, different sensitivity)

Let:

• $E_{AFR} = median(|AFR_{meas}-AFR_{tgt}|)$EAFR​=median(∣AFRmeas​−AFRtgt​∣)
• $E_{AFR,max} = max(|AFR_{meas}-AFR_{tgt}|)$EAFR,max​=max(∣AFRmeas​−AFRtgt​∣)
• 91: $P_{AFR} = 35E_{AFR} + 12E_{AFR,max}$PAFR​=35EAFR​+12EAFR,max​
• 93: $P_{AFR} = 32E_{AFR} + 11E_{AFR,max}$PAFR​=32EAFR​+11EAFR,max​
• E30: $P_{AFR} = 30E_{AFR} + 10E_{AFR,max}$PAFR​=30EAFR​+10EAFR,max​

IDC penalty (this is where ethanol gets strict)

Let $IDC_{max}$IDCmax​ be peak IDC.

91 / 93

• 0 penalty ≤ 85%
• linear penalty 85–95%
• hard cap >95%

$$P_{IDC} = \begin{cases} 0 & IDC_{max} \le 85 \\ 2.2\cdot (IDC_{max}-85) & 85 < IDC_{max} \le 95 \\ 45 & IDC_{max} > 95 \end{cases}$$PIDC​=⎩⎨⎧​02.2⋅(IDCmax​−85)45​IDCmax​≤8585<IDCmax​≤95IDCmax​>95​

E30 (more fuel required → protect headroom)

• 0 penalty ≤ 82%
• linear penalty 82–92%
• hard cap >92% (because you’ll run out of injector fast with E content swings)

$$P_{IDC,E30} = \begin{cases} 0 & IDC_{max} \le 82 \\ 2.8\cdot (IDC_{max}-82) & 82 < IDC_{max} \le 92 \\ 55 & IDC_{max} > 92 \end{cases}$$PIDC,E30​=⎩⎨⎧​02.8⋅(IDCmax​−82)55​IDCmax​≤8282<IDCmax​≤92IDCmax​>92​ $$S_{fuel} = \max(0,\ 100 – (P_{AFR} + P_{IDC}))$$Sfuel​=max(0, 100−(PAFR​+PIDC​))

---

5) Boost Score $S_{boost}$Sboost​ (keep it tight on all fuels)

Same structure, but overshoot matters slightly less on ethanol only if knock is clean (still not a free pass).

• All fuels:
  $P_{overshoot} = 25O,\quad P_{track} = 20E_B$Povershoot​=25O,Ptrack​=20EB​
• Oscillation penalty stays:
  $P_{osc} = 10J$Posc​=10J

$$S_{boost} = \max(0,\ 100 – (P_{overshoot}+P_{track}+P_{osc}))$$Sboost​=max(0, 100−(Povershoot​+Ptrack​+Posc​))

---

6) Thermal Score $S_{thermal}$Sthermal​ (ethanol still heat-soaks)

Keep thermal strict. Ethanol helps knock, not physics.

Use your existing IAT-based penalties, but trigger CAUTION earlier if:

• Thermal score < 75 (all fuels)
• Or IAT rise is steep pull-to-pull

---

7) Classification Bands (same, but smarter)

• SAFE: $S_{overall} \ge 85$Soverall​≥85 and no gates triggered
• CAUTION: 70–84
• STOP: <70 or any STOP gate

---

The one detail I need to lock this down

When you say ethanol blend, are you usually:

• E20–E30, or
• E40–E60, or
• “whatever the pump gives” (variable)?

Because IDC limits and AFR tolerance need different ceilings once you push past ~E30. If you tell me your typical blend range, I’ll finalize the E-profile into E30 vs E50 with correct IDC headroom rules.

🔵 Fuel Profile 1 — 91 Octane (Most Conservative)

Intended Use

• Daily driving
• Hot weather
• Long pulls
• Worst-case fuel quality

Target Philosophy

• Stability > output
• Slightly richer under load
• Lower peak boost
• Conservative timing

Recommended Calibration Envelope (EJ205 Safe Bias)

• Peak Boost: ~15–16 psi (VF22 dependent)
• WOT AFR target: 10.8–11.0
• Timing ramp conservative midrange
• Zero tolerance for repeat knock
• Injector duty ceiling: 90%

Scoring Sensitivity

• Highest knock penalty
• Tightest AFR lean tolerance
• CAUTION if any FLKC in WOT

Use this profile when ambient temps are high or fuel quality is questionable.

---

🔵 Fuel Profile 2 — 93 Octane (Primary Power Target)

Intended Use

• Performance street
• Controlled pulls
• Main 305whp attempt baseline

Target Philosophy

• Controlled midrange torque
• Smooth boost ramp
• Clean combustion

Recommended Calibration Envelope

• Peak Boost: ~17–18 psi (depending on turbo efficiency)
• WOT AFR target: 10.8–11.2
• Timing modest but progressive
• Injector duty ceiling: 92–93%

Scoring Sensitivity

• Moderate knock penalty
• AFR tracking must stay tight
• Boost overshoot tolerance slightly higher than 91

This should be your primary tuning fuel for developing the 305whp calibration.

---

🔵 Fuel Profile 3 — E30 Blend (Controlled Ethanol Strategy)

This is where things get serious — but also risky if fuel system isn’t upgraded.

Intended Use

• Maximum safe pump-based output
• Knock resistance increase
• Thermal stability improvement

Target Philosophy

• Maintain knock margin
• Protect fuel system headroom
• Do not chase timing aggressively

Recommended Calibration Envelope

• Peak Boost: ~18–19 psi (only if airflow supports it safely)
• WOT AFR target: 11.3–11.5 equivalent gasoline scale
• Timing slightly increased vs 93, but incremental
• Injector duty ceiling: 90–92% max (ethanol consumes more fuel)

Critical Rule

If IDC exceeds 92% at E30 → stop increasing boost immediately.

Ethanol gives knock headroom.
It does NOT give fuel system headroom.

---

⚠️ Fuel Strategy Doctrine

Never Tune on Mixed Unknown Fuel

If you don’t know ethanol content, treat it as 93.

Never Increase Timing and Boost in the Same Revision

Especially on E blends.

Always Log Before Power Escalation

Fuel trims, IAM stability, and IDC must be clean first.

---

🧠 Strategic Recommendation

For Project Black Pearl:

• Develop base stability on 93
• Validate repeatability
• Then introduce controlled E30 map
• Compare log scores side-by-side

That gives you:

• Content
• Data
• A monetizable comparison framework
• A safety buffer

---

Black Pearl Fuel Doctrine (Injector-Limited Edition)

✅ Allowed Operating Profiles (with 440cc injectors)

1) 91 Octane Map (Daily / Hot Weather / Worst Fuel)

Purpose: survive anything.

• Boost: conservative
• Timing: conservative
• Goal: reliability + knock-free

IDC guardrail: keep ≤ 85–88% at WOT.

---

2) 93 Octane Map (Primary Performance Map)

This is your best path to ~305whp if the hardware/turbo supports it.

• Boost: moderate-high but controlled
• Timing: incremental
• Fueling: stable and rich enough to protect

IDC guardrail: keep ≤ 90–92% (absolute).

---

3) “Light Ethanol” Blend Only (E10–E20 max)

This is the only ethanol strategy that makes sense on 440s unless you accept high risk.

Target range: E10–E20

• Use it for knock margin and consistency
• Not for turning the boost knob to the sky

IDC guardrail: keep ≤ 88–90%.

E30? On 440s: not recommended for WOT tuning unless you’re okay with living on the edge.

---

🚫 Prohibited Profile (for your current setup)

E30 “Power Map” on 440cc injectors

Hard no as a repeatable strategy.
If you do it anyway, the scoring model should treat:

• IDC > 92% as STOP
• Any lean deviation as STOP
• Any IAM drop as STOP

Because you have no headroom, and ethanol content variation will bite you.

---

Update to the Scoring Model for Your Setup

Injector Duty STOP Gates (tightened)

Because injectors are old and small:

• STOP if IDC > 92% (all fuels)
• CAUTION if IDC 88–92%
• SAFE zone ≤ 88%

(Yes, this is stricter than before — that’s what “20-year-old 440s” demands.)

Fuel Profile Weights (injector-limited)

Even on pump gas, fueling deserves more weight because you’re near the ceiling.$$S_{overall} = 0.36S_{knock} + 0.30S_{fuel} + 0.18S_{boost} + 0.11S_{thermal} + 0.05S_{repeat}$$Soverall​=0.36Sknock​+0.30Sfuel​+0.18Sboost​+0.11Sthermal​+0.05Srepeat​

---

What this means for your 305whp target (realistic)

With 440cc injectors, the build is usually capped by:

• IDC headroom
• AFR stability at high airflow
• consistency between cylinders (old injectors)

You can still build a fast WRX, but 305whp is “possible but not guaranteed” depending on turbo efficiency, intercooling, and your boost curve.

If you want 305whp reliably and safely, injectors are the next mandatory move.

---

The smartest money move (ROI)

Option A — Replace injectors with something modern

• STI “pinks” (565cc) if you want simple and proven
• ID1050x / similar if you want future flex / bigger turbo headroom

Option A is the move that prevents you from wasting time “tuning around” injector limits.

Option B — Keep 440s and cap the mission

• Focus on response and midrange
• Target a lower whp goal but very reliable
• Build the rally drivability angle

---