LOG-007 — Exhaust Fabricator Brief: DC Rio Dual-Path Tee-Before-Valve System

Version: v1 — 2026-04-23 Vehicle: 2004 Kia Rio Sedan, DC chassis Engine: A5D 1493cc DOHC 16V inline-4 NA Scope: Post-cat exhaust section ONLY. Stock header + catalytic converter RETAINED. Do NOT touch anything forward of the cat outlet.

⚠️ v3 — OBJECTIVE CHANGED + CROSS-MODEL AUDIT APPLIED (2026-09-06). v1/v2 stated "the goal is sound, not power." That is RETIRED. The objective is now peak power (high-RPM horsepower), with the dual-path sound system kept as a secondary feature. Scope widened to: manifold-outlet-flange rearward, flanged removable cat, and porting of the stock cast manifold (custom header fabrication remains OUT).

Several v1/v2 acoustic claims were refuted by a two-round cross-model audit (GPT + Gemini + Groq, independent). Corrections are marked [AUDIT 2026-09-06] inline. Deleted rules are struck through rather than removed, so the reasoning trail survives. Full record + revised spec: _meta/audits/exhaust-audit-2026-09-06/ (see 04-REVISED-SPEC.md, 06-SYNTHESIS-R2.md).

Main pipe is now 2.0" (50mm) ID, not 2.25" (57mm) — sized on gas velocity at peak power. The 50→57mm reducer in §0 is deleted along with weld points W1/W2. Sources: STUB-047 (base design), STUB-048 (Branch A sound), STUB-049 (tee topology), S1 (PH sourcing), LOG-006 (brief format reference)

System Overview

Two exhaust branches after the cat:

• Branch B (always open) — stock-equivalent. Quiet. Routes to stock tip location. For LTO inspection and daily driving.
• Branch A (electric valve, on-demand) — open-pipe / racing character. Loud. For spirited driving.

Electric valve on Branch A. Default = Branch A closed. Switch ON = both branches flow simultaneously. Switch OFF = Branch A blocked, all flow routes through Branch B. Quiet again.

Diagram:

[CAT OUTLET] ──50mm/2.0"──► [REDUCER TO 57mm/2.25"] ──2.25"──► [TEE JUNCTION]
                                                                       │
                                                ┌──────────────────────┤
                                                │ Branch B (side)      │ Branch A (straight)
                                                │ always open          │ valve-controlled
                                                ▼                      ▼
                               [RESONATOR]                   [VALVE ≤5cm from tee]
                               [MUFFLER]                          │
                               [STOCK TIP]                        ▼
                                                         [30-40cm PIPE]
                                                         [SLIP-ON CAN]
                                                         [DUMP]

Section Details

0. Cat Outlet → Reducer (mandatory first step)

Stock Kia Rio DC post-cat pipe OD ≈ 50mm (2.0 inch). Measure the actual pipe first — confirm 50mm before cutting. (Reference: Kia-forums.com "04 Rio Exhaust Diameter" thread — "2 inches" confirmed by owner measurement.)

[AUDIT 2026-09-06] SECTION 0 IS DELETED. With the main run respecified to 2.0" (50mm) ID, there is nothing to expand into. Delete the reducer, delete weld points W1 and W2. If the stock post-cat pipe is sound and actually measures ~50mm ID, it may not need replacing at all.

⚠️ OD vs ID: the line below says "OD ≈ 50mm". All power sizing is on ID. A 50mm OD pipe with 1.5mm wall is ~47mm ID. Record which one you measure. If stock is 50mm OD, order 50mm ID (~53mm OD) for the new run.

Reducer piece: 50mm inlet → 57mm (2.25") outlet. Welded cone reducer or slip-on expander. Available at Banawe or custom-fabricated by the shop. This is where all 2.25" sections begin.

Minimum straight run from reducer to tee: 40–60cm. Two reasons:

1. Cat outlet temperature: >600°C — distance needed for gas to cool before the next weld point.
2. Flow settling: 10 pipe diameters required for flow to smooth out after cat substrate turbulence (10 × 57mm ≈ 57cm).

1. TEE Junction (most critical weld point)

Tee type: 30-degree mandrel wye (NOT 90-degree tee, NOT a saddle hole).

[MERGE 2026-09-22] ANGLE CHANGED 45° → 30°, AND THE FABRICATION METHOD IS REPLACED. Owner-approved. Two separate changes, both from the two-track reconciliation — see PLAN.md C1/C9 and _meta/audits/ph-dualpath-exhaust-audit-2026-09-10/02-SYNTHESIS.md A1.

Why 30° over 45°: when the valve is SHUT — 100% of legal driving and every inspection — all gas must turn into this branch. STUB-049 RQ1 records a 30° wye cutting branch loss ~70% against 45°'s ~50%. The shallower angle serves the dominant duty cycle.

⚠️ [ROUND 3, 2026-09-22] DOWNGRADED TO CONDITIONAL — 45° IS EQUALLY VALID.

• The saving is ~0.04 hp (my figure; Gemini independently 0.03). Negligible. Both reviewers said so; neither thought it justified packaging cost.
• GPT: the argument only holds if the stock path is the turned leg when the valve is shut. It is on this car — but that is a condition, and I wrote the rule as universal.
• Gemini: choosing 30° lengthened the crotch and broke the stub rule (see §2a). I optimised a 0.04 hp coefficient and paid in buildability.

Use 30° where it packages naturally. Otherwise use 45° — it is not a fallback, it is an equally valid choice, and on a tight chassis the better one. Do not reroute to preserve 30°.

Why not 90°? A shallower wye has lower pressure drop on the Branch B side.

[AUDIT 2026-09-06] JUSTIFICATION CORRECTED. ASHRAE flow data — 45° wye branch loss coeff = 0.50 vs 90° = 1.20. Those coefficients are steady, incompressible, roughly isothermal duct data. Exhaust is pulsating, compressible and 300-600°C. The 45-vs-90 ranking survives for mean flow; the magnitudes do not transfer and must not be quoted as if they predict the flow split.

No passive angle optimises both valve states. Minimum loss into Branch B when the valve is SHUT, and minimum bleed into Branch B when it is OPEN, are opposing requirements. The valve-shut state is 100% of legal driving and every inspection — the dominant duty cycle. When the valve is shut the straight path is a dead end, so all gas must turn into the 45° branch, which is angled for flow continuing forward. The geometry is therefore optimised for the minority state.

Kept regardless: modelled penalty ~0.56 kPa at 6,500 RPM ≈ 0.06 hp. Negligible. ⚠️ That 0.56 kPa is a model output from a generic branch coefficient, not a measurement, and generic coefficients may not represent a tee whose straight continuation is closed.

Fabrication — read this before the shop touches the pipe.

⛔ [MERGE 2026-09-22] THE SADDLE-HOLE METHOD IS WITHDRAWN

Or shop-fabricate: cut 2.25" pipe at a 45-degree angle, weld as a side-branch. Straightforward for any experienced exhaust shop.

Three independent models rejected this unanimously (GPT, Gemini, Groq — 3/3), reviewing this project's own brief. Record: _meta/audits/ph-dualpath-exhaust-audit-2026-09-10/02-SYNTHESIS.md §1 A1. It was additionally endorsed in conversation on 2026-09-18 as "better than the LOG-007 cut list" — that advice is withdrawn. The reviewers had already refuted it; the endorsement was given without reading their synthesis.

Why it fails. A tube welded onto a hole in a live pipe is an oblique lateral aperture, not a formed wye with a guiding crotch:

• Two cylinders intersecting at an angle open a saddle-shaped window ≈1.41 × OD long. A hole-saw circle masks part of the stub mouth, so the effective throat is smaller than the branch bore and is not specified by any dimension on this sheet — two shops build "the same" system with materially different throats. This is the reproducibility failure.
• The acute crotch cannot be root-fused by ordinary MIG/stick without burn-through or lack of fusion. Fatigue initiates at that notch.
• The large oval window relieves drawing stress and ovalises thin-wall tube, which then distorts the valve seat a few cm downstream.
• A sharp internal lip is an edge-tone generator — whistle/chuff under load. On a build whose whole purpose is sound, that is a primary defect, not a flow nitpick.

Use one of these two instead:

Option 1 — formed 30° mandrel wye (preferred). A purchased or properly formed wye has a guiding crotch and a throat equal to its branch bore. Order 2.0"/Ø51 single-inlet to dual Ø51 outlet, 30° included where available. Measure every opening before ordering.

Option 2 — electric Y-pipe cutout kit. If the owned valve is (or is replaced by) an integrated Y-cutout, the aperture problem is solved by purchase — the Y body is a formed wye. The reviewers said exactly this. Caliper its throat, branch bore and flange bores before committing; a cast body can carry its own restriction.

⚠️ If the shop fabricates rather than buys, the branch must be fish-mouthed to the actual measured main-pipe OD and the aperture traced from the fitted stub — never hole-sawed to a generic circle. Cut undersize, finish to the saddle, dress the internal edge.

Acceptance criteria — inspect the inside, or do not use this method:

• No hanging penetration, no detached slag
• No un-fused internal lip
• No inward projection forming a scoop
• Finished throat not visibly smaller than the branch bore
• Edge profile continuous all the way around the saddle
• Verified by borescope or phone camera through a temporarily open flange

Gate: if the shop cannot inspect the inside, the welded-branch method is unsuitable for this build — buy a formed wye or the Y-cutout.

⚠️ Wall-thickness gate. Philippine replacement mid-pipe is often 1.0–1.2 mm aluminised, not 1.5 mm. Welding a branch into thin, pre-rusted steel burns through. Caliper the wall at the chosen station first. If it is thinned, dented or heavily scaled, replace that short section rather than welding into it.

Weld quality note: No sharp burr inside the junction. A sharp edge at the branch opening creates vortex noise (singing/whistling). Keep the inside of the tee point smooth. (Source: Fauske.com acoustic resonance note cited in STUB-049.)

2. Branch A: Straight Pipe + Valve + Slip-On Can

2a. Dead-end stub (tee to valve) — CRITICAL MEASUREMENT

Distance from wye crotch to valve face: as short as packaging allows. Record the dead VOLUME, not just the length. Never exceed ~300mm.

[MERGE 2026-09-22] REWORDED — two rules reconciled, neither discarded.

• This brief said "≤5cm, shorter is better." The cross-model review amended that to a packaging limit, and asked for the internal dead volume (wye plane → butterfly face, including aperture neck, saddle thickness, housing counterbore and flange cavity) to be recorded instead — volume is comparable across valve designs, centreline distance is not.

• The parallel track independently derived a never-exceed ceiling: the λ/4 notch must stay above the 433 Hz 2nd harmonic. At the corrected mid-pipe c≈520 m/s that is ≈300 mm (their published 340–355 mm assumes c=615 m/s — see the temperature note in §Audit provenance).

• ⛔ [ROUND 3, 2026-09-22] THE "50–110 mm" BAND IS WITHDRAWN. I ruled these a floor and a ceiling. Both reviewers rejected that independently — they are different kinds of constraint, and combining them destroyed the provenance of each. GPT: "the clearest tidy-merge error."

  Worse, the 50 mm low end is probably unbuildable. A 30° wye on Ø53.8 OD tube has a saddle intersection spanning ~108 mm along the branch axis (45° spans ~76 mm). I invented a number that the geometry I had just specified does not permit.

CORRECTED RULE:

Stub = the shortest length the hardware physically permits. Normally ≤110 mm. The floor is mechanical — set by wye saddle, flange and valve body — and it must be MEASURED on the actual parts, not assumed. Do not deliberately extend toward ~300 mm (λ/4 at the 2nd harmonic, c≈520) unless you are intentionally notching that harmonic.

⛔ Gate: mock up the wye + valve and caliper crotch-to-butterfly-face before ordering. This is now a blocking measurement — §2 of the round-3 measurement list.

Why? This dead-end pipe section (when the valve is closed) acts as an acoustic resonator. Keep it as short as packaging allows. The rule stands.

[AUDIT 2026-09-06] RULE RETAINED, FIGURE DELETED. Unlike Branch B, this stub is genuinely capped when the valve is shut, so the closed-pipe model legitimately applies here. But 5cm resonates at 2,600 Hz is not a real number: physical length is not acoustic length. End correction, valve cavity volume, butterfly position and junction geometry make the effective length materially longer. Justify the rule as minimising dead cavity volume, not by a specific resonant frequency.

Practical instruction: Butt the valve body directly against the tee outlet — keep the pipe between them visibly short. If flanges are needed (for future valve removal), use a V-band flange set placed close to the tee.

2b. Electric valve specs

• Preferred: SPELAB 2.25" single remote electric exhaust cutout. Pipe: T-304 stainless. Valve body: CNC aluminum (acceptable for NA 1.5L at 200–400°C post-cat temperature, not for turbo). ~$129 USD / ~₱7,100 at current rate.
• Budget option: CN import equivalent via Lazada.ph — search "electric exhaust cutout 2.25" or "57mm exhaust valve". Target ₱2,500–₱4,500 for a stainless pipe body.
• NC wiring (Normally Closed): Default position = Branch A CLOSED. Switch/remote opens it. If power is lost (dead battery, loose wire), only Branch B flows — always defaults to the legal/quiet state.

Wiring:

[12V Power (fuse-protected)] ──► [Relay] ──► [Valve motor]
[Ground]                         │
[Switch/remote in cabin] ─────────┘

Switch ON  = valve open   = Branch A active
Switch OFF = valve closed = Branch B only

Use a 10A fuse. If the valve ships with a remote (most do), the relay is built into the controller box — mount it in a dry area near the cabin.

2c. Branch A pipe (valve exit to slip-on can)

Length: 30–40cm from valve exit to slip-on can inlet.

• Shorter = sharper sound (higher frequency character). Target 30cm.
• Longer (>40cm) = delays drone onset but starts to blur pulse separation.
• Pipe OD: 2.25" (57mm) throughout. For a shriller tone (superbike-like), downsize Branch A pipe to 1.75" (45mm) (higher gas velocity = higher dominant frequency). If power is the priority, keep 2.25" throughout.

[AUDIT 2026-09-06] BRANCH A IS NOW THE POWER PATH — SPEC BY MINIMUM FLOW AREA. Against a 50mm main run, a 38mm can inlet is 58% of pipe area; 45mm is 81%. Every slip-on in the §v2 sourcing table (38–51mm) throttles this branch. For a sound-only build that was a tone choice; for a power build it is the binding restriction.

Requirement: valve throat ≥50mm and can core ≥50mm ID. Specify by minimum effective flow area through valve throat, can inlet, core and outlet — not by nominal connection diameter.

⚠️ A nominal 2.0" valve housing is not a 2.0" effective throat. Butterfly disc, shaft and seat reduce it by an unknown amount. Measure the owned valve's actual throat before ordering anything else in Branch A.

⚠️ The v1 claim that Branch A carries 75–85% of flow is withdrawn as invented. Parallel paths divide until both have equal total pressure drop tee-to-outlet — a function of valve throat, bends, can construction and Branch B's muffler, not of area ratio.

2d. Branch A tip — ⛔ SUPERSEDED by the v4 tail spec (see end of file)

[MERGE 2026-09-22] THE SLIP-ON CAN IS WITHDRAWN AS THE SPEC. The owner's target is an LFA-adjacent bright/hard note. A 38–45mm can inlet is 58–81% of the Ø51 pipe area — it throttles the loud branch and its packing absorbs the very harmonics the target depends on. Branch A now terminates in the megaphone + reverse-cone tail ("Build A") specified in §v4 at the end of this file. The text below is kept for its sourcing and reasoning trail.

2d (v2). Motorcycle Slip-On Can (PREFERRED over megaphone)

Target spec: Motorcycle-style straight-through can, 38–45mm inlet, short body (100–200mm), stainless steel.

Why slip-on > megaphone (update from STUB-048): The small chamber in a slip-on adds a slight "snap" that defines pulse character — each cylinder hit is more distinct. A blunt-cut straight pipe sounds more diffuse at lower RPM.

Reducer needed: 57mm → 45mm (or whichever size matches the slip-on inlet). Available at Banawe or any exhaust shop. ~₱100–₱300.

Shopee.ph confirmed listing: Universal 38–45mm stainless motorcycle muffler. ~₱800–₱1,500.

Megaphone option (alternative): If the owner wants a megaphone tip on Branch A instead:

• Inlet: 2.25" (57mm)
• Outlet: 4.0–4.25" (101–108mm) — ratio ~1.8× inlet
• Body length: 20–25cm
• Taper angle: 7–9 degrees
• Fabrication: roll and weld mild/stainless sheet. Any skilled exhaust shop can do this.
• (Source: STUB-047 RQ1)

Sound expectation (Branch A OPEN):

• 800–2,000 RPM idle: low hum, aftermarket-muffler-like
• 2,000–3,500 RPM: dry mechanical character, audible pulse separation
• 3,500–6,500 RPM: aggressive shriek/rasp, motorcycle-like character (CBR600RR-like at lower RPM — approx 60–70% of superbike character per frequency analysis)
• Engine braking / decel: crackles and pops (existing A5D behavior amplified)

3. Branch B: Resonator + Muffler + Stock Tip

[2026-10-07] Branch B is now specified elsewhere. The quiet branch has its own spec — _meta/audits/silent-muffler-audit-2026-10-07/01-BUILD-SPEC-R1.md (round-1 audited, readings pending) — and its universal method is LOG-035. This section stays as the 2026-04 to 2026-09 history and is not maintained. Two items that audit raised for the main run, not Branch B: (1) no flex joint is recorded anywhere in this log — if the 130 cm run is rigid from the manifold flange, a Ø51 stainless flex immediately after the flange is the fix for wye-crotch and hanger fatigue; (2) LOG-035 §7 carries the measurement protocol every later exhaust change on this car should reuse.

3a. Branch B pipe run (tee to resonator inlet)

[AUDIT 2026-09-06] THIS ENTIRE LENGTH CONSTRAINT IS WITHDRAWN. REFUTED.

Route Branch B by packaging. Keep bends smooth. There is no forbidden-length rule.

[MERGE 2026-09-22] AMENDED — this was over-corrected. GPT's review of the follow-up brief: killing the c/4L forbidden-length rule was right; concluding Branch B has no acoustic length constraint at all went a step too far. Branch B is an open, lossy, flow-carrying line terminated through a muffler and tailpipe. Its reflection coefficient need not approach 1 for its length to matter — length still sets its input impedance, and therefore how it shunts the loud branch in the open state.

Action: route by packaging, but RECORD the length as a design variable. It is not acoustically irrelevant; it is merely not governed by the rule that was deleted. (ph-dualpath-exhaust-audit-2026-09-10/02-SYNTHESIS.md D6.)

Two errors, found in sequence:

1. The avoid-figures below are the endpoints of a continuous interval, not two isolated notches. At c=520 m/s every length from 60cm to 130cm resonates at some firing frequency in the 3,000–6,500 RPM band. The "70–80cm safe" half is in-band at 4,875–5,571 RPM — arguably the worst place to put it.
2. More fundamentally, the model does not apply at all. The closed-pipe relation f = c/(4L) needs a reflection coefficient near 1. Branch B is a flow-through transmission line — tee → resonator → muffler → stock tip → atmosphere — and is never capped. GPT, Gemini and Groq independently rejected it; verified against this section. A tightened correction of "≤55cm or ≥150cm" produced during the audit is also withdrawn.

Retained concern (nobody had considered it): the relevant acoustic length runs through the resonator, muffler and tailpipe to the radiating outlet, and the cruise band (2,000–3,500 RPM = 67–117 Hz) is where cabin boom actually bites. This brief only ever considered 3,000–6,500 RPM. Not a fabrication constraint — a thing to listen for.

⚠️ Contrast with §2a: the Branch A tee-to-valve stub is genuinely capped when the valve is shut, so quarter-wave logic legitimately applies there. That is exactly why it does not apply here. See _meta/CONTRADICTIONS.log 2026-09-06.

Length: 40–80cm.

Why the min/max?

• Minimum 40cm: keeps the resonator far enough from the tee junction (heat + turbulence).
• Range guidance from STUB-049 acoustic formula: avoid Branch B lengths that create quarter-wave resonance in the A5D operating band (100–216 Hz):
  • λ/4 at 100 Hz (3,000 RPM) = ~130cm — AVOID ~130cm Branch B.
  • λ/4 at 216 Hz (6,500 RPM) = ~60cm — AVOID exactly 60cm.
  • Safe range: 40–55cm or 70–80cm

Pipe OD: 2.25" (57mm) for consistency throughout the system. → [AUDIT] 2.0" (50mm) ID throughout. See header.

45-degree tee branch for Branch B: Setting Branch B at a 45° angle (opposite of Branch A's straight-through path) gives lower resistance at Branch B entry from the main pipe flow. This is correct — the 45° branch faces more head-on to the incoming flow, matching the lower-resistance path for stealth-mode routing.

3b. Resonator (Branch B — drone suppression)

Target spec: Straight-through resonator, 2.25" ID/OD, perforated core with mineral wool or fiberglass packing, multi-layer body.

Import option:

• 1320Performance Ultra Quiet Mini Muffler 2.25" — $75 USD / ~₱4,125. Best performing option for this build. (Source: 1320performance.com — confirmed price)
• Vibrant 1140 equivalent — any Vibrant distributor, ~$45–60 USD.

Local budget option:

• Any local exhaust shop can fabricate a straight-through resonator body. Use 2.25" perforated pipe as the core, wrap with mineral wool/glass wool, weld the outer can. Cost: ~₱500–₱800 in parts and labor. Cheapest PH path.
• Do NOT use a chambered muffler as the resonator — it MUST be straight-through (perforated inner pipe with packing material, no baffles or diversion plates).

3c. Muffler (Branch B — volume suppression, stealth goal)

Target: Compact S-bend / multi-chamber muffler. 2.25" inlet. Stock-tip location exit.

Cheapest (recommended): Surplus muffler from Banawe junkyard.

• Donor cars: Honda City ~2000–2004, Toyota Vios Gen 1, Mitsubishi Mirage/Lancer, Suzuki Swift — any compact car with 50–57mm exhaust pipe.
• Estimated price: ₱500–₱1,500 surplus at Banawe (used-parts grade).
• Any muffler with a matching inlet and compact body works — the goal is maximum noise suppression, not performance.

Fit note: Donor muffler inlet is likely 50mm (2.0"). A 50→57mm expander/reducer coupling (~₱100–₱200) will be needed at the muffler inlet. Or find a donor with a 57mm inlet already.

Sound expectation (Branch A CLOSED, Branch B only):

• Should be QUIET. Stock-equivalent or quieter than stock.
• LTO checkpoint safe: Branch B tip exits at stock location, stock-looking.
• If questioned by a traffic enforcer: "Modified exhaust" is not obvious because the external appearance matches stock.

Complete Cut List

⚠️ MEASURE EVERYTHING ON THE ACTUAL VEHICLE BEFORE CUTTING ANY PIPE. The values here are estimates — actual routing under the DC Rio chassis will dictate final measurements.

Weld Point List

Hanger/bracket welding:

• Minimum 2 hangers on the Branch B run (one near the tee, one near the muffler).
• 1 hanger on the Branch A pipe section.
• Stock Rio DC exhaust hangers are bolt-on rubber loops — reuse stock mounting points if possible, or weld new brackets to a nearby subframe/chassis rail.

Alignment Issues (Field Notes)

Exhaust routing under the DC Rio is tight — some bends will be required depending on ground clearance and chassis rails. Avoid:

• Sharp 90-degree bends — use a mandrel-bent piece or butt-weld two 45-degree pieces for curves.
• Bends within 20cm of the tee — flow needs to be clean through the tee junction.
• Pipe touching the underside of the floor pan — maintain clearance for heat and vibration.

If Branch B needs additional bends to reach the stock tip location: that's fine. Branch B is not flow-critical — the muffler dominates the flow resistance, not the pipe length or bends.

Pre-Test Checklist (After Fabrication)

• No exhaust leak at any weld point (soapy-water test at idle)
• Valve operation verified: switch ON → Branch A opens (exhaust dump audible). Switch OFF → Branch A closed (quiet, Branch B only)
• Branch A dump clearance: no contact with undercarriage, spare tire, or fuel tank
• Exhaust hangers: all mounted, no tension on weld points (rubber hangers absorb shock)
• Branch B tip: located at stock position and stock-looking
• Cold test: start the engine, listen to Branch B for drone. If drone is present at a specific RPM (humming/buzzing), report back and adjust Branch B pipe length by ±5cm.
• Hot test: after warmup, repeat leak check. Exhaust joints contract/expand with heat.

Budget Summary

Stub References

• STUB-047: Base system design, megaphone geometry, Branch B stealth spec, parts constraint
• STUB-048: Branch A sound design, frequency analysis, slip-on vs straight pipe, RPM ceiling
• STUB-049: Tee-before-valve topology, 45° vs 90° flow data, dead-end stub calc, Branch B length guidance
  • ⚠️ [AUDIT 2026-09-06] RQ4 (Branch B length) marked refuted. RQ1 scope-limited to mean flow only — it does not settle the acoustic question. RQ3 (stub) retained at high but caveated on effective-vs-physical length. RQ5 (tee placement) rationale weakened.
• STUB-048: ⚠️ [AUDIT 2026-09-06] RQ5's "packing only absorbs above ~5 kHz" claim corrected as unsupported — dissipative attenuation has no universal threshold.
• S1 (research-reports/s1-ph-exhaust-parts-sourcing...md): PH sourcing, confirmed Shopee.ph slip-on listing, AutoBane price, 1320Performance resonator pricing
• LOG-006: Carpenter brief format reference (keyboard build)

v1 complete. Update to v2 after actual cat outlet OD is measured and chassis routing is confirmed.

v2 Sourcing Audit — 2026-05-09 (Shopee.ph re-check via ph-scraper MCP)

Audit scope: Confirm Shopee.ph availability + current pricing for every line item EXCEPT the electric valve (already owned). Lazada/Carousell scrapers offline at audit time — Shopee-only data.

Owner status note: Electric valve (2"–2.5") already in hand. Skip valve sourcing. Adapt Branch A flange sizing to the owned valve's actual inlet/outlet OD (verify before ordering V-band sets).

Confirmed picks (best price-quality matches)

Cross-check vs LOG-007 v1 budget

Audit findings (action items)

1. Y-wye listing ambiguity. Listing 29837991608 markets as "Y Shape Connector" but does not state branch angle. Action: message seller to confirm 45° (not 90°) before ordering. Fallback: shop-fabricate from straight 2.25" pipe.
2. Resonator decision. v1 had two options (US import ₱4,125 vs local fab ₱500–₱800). v2 adds a third: Shopee CN ₱2,380 (ITSOK, 5.0★). Recommendation: local fab still cheapest and serviceable; ITSOK is the import middle-ground if local quality is poor.
3. Pipe stock is the one item Shopee fails on. No 2.25" exhaust tubing sold by-the-meter on Shopee. Buy raw pipe at Banawe. Lazada was offline during audit — re-check Lazada later for pipe stock listings.
4. V-band flange sizing depends on owned valve. v1 spec'd 2.25". The owned valve is "2"–2.5"" range. Action: measure valve flange OD before ordering V-band kits. ₱499 SKU 28130781737 covers 2/2.5/3" — pick the matching size at checkout.
5. Slip-on can choice = sound character. All three confirmed picks are 38–51mm range. Cheaper (₱529 GP Universal) = simpler tone; mid (₱999 MJMOTO retro) = tapered cafe-racer character; (₱1,080 Yoshimura w/ DB killer) = tunable. Recommend MJMOTO ₱999 for stated "motorcycle-like character" goal in v1.
6. All confirmed picks are stocked or shippable to PH. No deal-breaker sourcing gaps remain post-valve-owned.

Skipped (already owned or unchanged)

• Electric valve — owned. Re-confirm Branch A pipe ID matches valve OD before ordering reducers.
• Banawe surplus muffler — local-only sourcing channel, not on Shopee.
• Banawe pipe stock — same.
• Cat-outlet OD measurement — still pending field measurement (v1 noted ≈50mm to be confirmed).

v2 status

Sourcing-audit ready. Field-measurement still pending (cat outlet OD, valve flange OD, chassis routing clearances). Update to v3 once shop fits up the actual vehicle.

v3 — Power Objective + Audit Corrections (2026-09-06)

The objective changed. v1/v2 were written and reviewed under "sound, not power"; that is retired. Full analysis: _meta/audits/exhaust-audit-2026-09-06/04-REVISED-SPEC.md.

Sizing: 2.0" (50mm) ID main run

Sized on exhaust gas velocity at peak power. 1.493L, ~400°C, VE 0.80–0.88 (realistic for a stock-cam 1990s-design 1.5L 16V at high RPM — the earlier 0.94–1.05 sweep was too optimistic).

Restriction begins ~130 m/s. 50mm never exceeds 88 m/s across every combination of the disputed power-peak RPM (5,800–6,500) and realistic VE. Not restrictive; no reason to go larger.

⚠️ 50mm is adequate, not optimal. The v3-draft argument that "larger costs velocity" was wrong — it imports a scavenging benefit into a section where velocity is not independently valuable. Velocity earns its keep in the port, primaries and collector, not in an untuned tailpipe downstream of a cast manifold. Building from zero, 54mm would be marginally preferable; the difference is not reliably measurable.

Backpressure → power: use pumping work, not folklore

The "~1% power per 1.75 kPa" heuristic is retired — untraceable, and wrong by ~5×.

swept-volume flow @6,500 RPM = 0.0809 m³/s
1 kPa mean exhaust pressure   = 81 W = 0.108 hp

Diameter is worth ~0.1 hp. It is not where power lives on this engine.

Revised priority — find the loss before fixing it

The v3-draft ranking (cat delete > porting >> everything) is withdrawn. The 3–7 kPa cat figure was assigned from the vehicle's age and never measured.

Manifold porting: expect near zero

0 hp if the casting is already acceptable; 0.5–1.5 hp with correctable outlet lips, port misalignment or flash; 2–3 hp only if unusually defective. Highest-value region is the collector outlet and flange transition — the one place full engine flow passes. Deburr, do not polish. Never open the manifold wider than the head port (creates an expansion pocket). Stop before removing sound wall material — cast manifolds crack.

Compare losses per operating state

Four cases, not one: cat / test-pipe × valve-open / valve-closed. Comparing tee loss against the cat is invalid in the test-pipe state, where there is no cat.

Measure before cutting

1. Paired in-gear acceleration runs, cat vs test pipe — same flat road, both directions, same gear and start RPM. Answers priority #1. Free.
2. OBD2 MAF at WOT — replaces the largest invented input (VE) with a measured one.
3. Acceleration runs, Branch A open vs closed, in both cat states.
4. Every actual minimum ID — valve throat (including butterfly shaft/disc), flange bores, pipe ID after bends, tee throat, muffler inlet tubes, resonator core, can outlets.
5. Photograph manifold ports with a caliper in frame — decides whether porting has a target.
6. ⛔ Cat outlet OD and ID — BLOCKING GATE, outstanding since v1. Do not cut before this exists.

[MERGE 2026-09-22] This measurement has never been taken. The figure quoted in §0 ("OD ≈ 50mm") is a Kia-forums owner report, and the ~47mm ID is a derivation from an assumed 1.5mm wall — not a reading off this car.

⚠️ The parallel track's 31-A5D-STOCK-SIZE-CORRECTED.svg cites this line as "LOG-007's own 2026-09-06 caliper measurement … the one real caliper reading in this chain." It is not a caliper measurement. Both tracks independently land on ≈Ø51 for the main run — this one by pumping-work arithmetic, that one by the false anchor — so the number survives the error, but the measurement is still owed. Corrected in that sheet's index entry on 2026-09-22.

A cheap 0–15 psi pressure gauge with a bung upstream of the cat would settle priority #1 outright and is worth reconsidering.

Audit provenance

Round 1 (sound objective): GPT + Gemini + Groq, independent. Round 2 (power objective): GPT only — Gemini's relay did not complete, Groq rejected for demonstrable arithmetic errors. Round 2 findings are single-source except where independently verified by calculation (the pumping-work figures above were reproduced and confirmed exact). The revised priority ranking and the porting estimate rest on one reviewer.

Open empirical question: is the cat actually restrictive? Prediction P6 in 03-PREDICTIONS.md says no — falsifiable by paired acceleration runs.

v4 — Two-Track Merge, Tail Spec, and New Risks (2026-09-22)

Two Claude sessions designed this same car in parallel from 2026-09-10 without reading each other. This section records what the merge changed. Full reconciliation: PLAN.md (repo root). Parallel track: _meta/audits/ph-dualpath-exhaust-audit-2026-09-10/.

v4.1 Confirmed architecture (owner decision, 2026-09-22)

[MANIFOLD] ─── 130 cm straight, Ø50.8 (2.0") ID ───┐
                                                    │
                            ┌───────────────────────┤ 30° MANDREL WYE
                            │                       │ (formed or Y-cutout —
                            │                       │  NEVER a saddle hole)
                  Branch B  │                       │  Branch A
                  SILENT    │                       │  LOUD
                  always open                       │  valve-controlled
                            ▼                       ▼
                    [STOCK MUFFLER]         [VALVE — 50–110 mm from crotch]
                    [STOCK TIP]                     │
                                                    ▼
                                            [BUILD A TAIL — v4.2]
                                            [OPEN DUMP, mid-car]

Note the split of objectives. This car is performance-first on the main run with a bright/hard loud path. That is deliberately different from the universal deep-sound reference (research-reports/universal-deep-*), which optimises the opposite way. Do not reconcile the two — they are different targets on purpose.

v4.2 Branch A tail — "Build A"

Fabricator sheet with flat-pattern development: _meta/audits/exhaust-tail-audit-2026-09-18/drawings/tail-BUILD-A-fabsheet.svg

[ROUND 3] What the reverse cone actually is. My brief called it a "4:1 impedance step". Wrong — Ø50.8→Ø76.2 is 2.25:1. The 4:1 label is stale from the superseded Ø38.1 exit. The 38% reflection figure is correct for 2.25:1; I updated the number and not the label.

Correct description, per GPT and Gemini: a frequency-dependent reactive/radiation interaction of uncertain magnitude. Not cosmetic, not a strong resonator. It shapes timbre at the harmonics; it does not lower pitch or "add depth". Both reviewers agree it serves the bright/hard target — Gemini: "perfectly serves the bright loud path."

Note the whole 216 mm tail is only compact at the fundamental (0.09 λ at H1, but 0.45 λ at H5). That is why it can shape harmonics — a better justification than the one I gave.

Sourcing (2026-09-22): _meta/audits/exhaust-tail-audit-2026-09-18/07-SOURCING-MEGAPHONE-2026-09-22.md. Shortest path found: Lmtless Megaphone Scav, 2.0" variant (PH-made 304 SS, Ø50.8→Ø76.2 straight megaphone, ₱2,480–2,880) is this section's diverging cone off the shelf; add a ₱326–400 Ø76→Ø51 reducer for the reverse cone. Lmtless (QC/Cebu) can also fabricate the fab sheet exactly — they already roll this cone.

Why the exit is NOT smaller. Three rounds established that shrinking the exit buys ~2 dB of reflection and costs 9–20 dB of broadband jet hiss (Lighthill U⁸). A 1.5" exit is +19.1 dB of hiss. Do not let a fabricator neck it down — the instinct that smaller equals sportier is exactly backwards here.

Critical, and more important than any dimension above:

• Outlet faced square, fully deburred inside and out. A ragged lip sheds vortices into a whistle. Both reviewers rated this above any angle error.
• No weld bead protruding into the bore anywhere. Dress every internal seam flush.

v4.3 Two corrections to this project's own earlier statements

C1 — The LFA frequency gap was understated, then corrected. An earlier figure of 300 Hz for the LFA applied the inline-4 relation RPM/30 to a V10. The LFA fires five pulses per revolution: 750 Hz at 9,000 rpm, against this engine's 217 Hz at 6,500. The gap is 3.5×, not 1.4× — worse than the superbike gap, not better. LOG-032:32 had this right and was contradicted. The parallel track derived the same 750 Hz independently.

What "LFA tone" therefore means here: the character — bright, hard-edged, rising, with a megaphone bloom — not the pitch. Pitch is set by cylinder count and rpm and no pipework moves it.

C2 — The A-vs-B cone comparison was argued wrongly. A steeper reverse cone was claimed to cause "more separation and hiss." Contractions accelerate flow — a favourable pressure gradient, inherently stabilising. Diffusers separate; contractions do not. Both cones are also acoustically compact (<0.11 λ across all engine content), so they act as lumped area steps and the taper angle is very nearly inaudible. Build A was selected for being shorter at equal acoustic effect, not for a hiss advantage it does not have.

v4.4 New risks — live because the cat is deleted

⚠️ Actuator thermal soak is now a real case, not a conditional. The parallel track's review flagged this as conditional on cat-delete platforms — this is now one. The owned electric valve was specified against a post-cat 200–400°C environment. With the cat gone, gas at the wye station runs hotter. Fit a standoff or heat shield, and IR-check the actuator body after a loaded pull.

Additional risks carried in from the parallel track's review (02-SYNTHESIS.md §4.10):

• Butterfly leak-by at valve-closed WOT. Heat-tolerant butterflies have no elastomeric seal; under a few psi they hiss along the spindle — directly under the front seats, in the state that is supposed to be quiet.
• Command two states only. Partial valve positions are the worst case for whistle, edge tones, shaft vibration and unstable flow division. Verify full-stop repeatability.
• Condensate. Orient so the valve is not the lowest pocket. Do not drill a drain — that is a permanent leak and a whistle source.
• Cabin drone from the mid-car dump. Floorpan cavity coupling at 2,500–3,500 rpm. Two outside models and this brief's own §Branch B note (67–117 Hz) agree.

v4.5 Temperature station — why some numbers here differ from the parallel track

The parallel track's own 2026-09-14 audit found unanimously that its c = 615 m/s is a manifold-station figure; mid-chassis is 620–720 K, giving c ≈ 500–540 m/s. That correction is still unapplied in its current sheets.

This brief uses c ≈ 520 m/s, inside the corrected band. Verified locally: the two tracks' apparent disagreement on Ø51 velocity (109 vs 82 m/s) is entirely this station difference on the same pipe at the same VE — not a design disagreement.

⛔ An IR reading at the planned wye station after a loaded pull is the single measurement that settles it. Until then, treat every tuned length in either track as ±10%.