Campaign 1 was a 3,719-mile instrumented round trip that closed the core V2V question: ProPower delivers exportable AC to the LightShip umbilical at a measured, reconciled cost, across seven independent legs, with fuel reconciled to the Trip Fuel PID at 0.21% mean error. That result stands on its own and needs no repeating.
What Campaign 1 could not do is measure the two things that would convert the rest of the engineering case from modeled to measured: the combined-system aerodynamic drag area (Cd·A) of the F-150 and AE.1 as a single towed object, and the direct HV-bus accessory split that the 42-PID configuration could not observe. Campaign 2 is built around closing those gaps deliberately, rather than hoping they fall out of incidental highway cruising — which is precisely how Campaign 1 came up empty on drag.
A third objective family joined the campaign on 2026-08-02, when the LightShip AE.1 owner's manual entered the reference set and LightShip supplied the TrekDrive lockout conditions directly. That material settles the system's operating envelope, corrects the architectural understanding of how TrekDrive is actuated, and places one long-standing knowledge-base claim in contest. Sections 06 through 08 carry it.
A — Aerodynamic: measure the combined-rig Cd·A by Method 2 speed holds (Sections 02–04). B — HV bus: resolve the accessory split and the gasoline-versus-recovered accounting the 44-PID configuration makes possible (Sections 05, 09). C — TrekDrive: verify the OEM envelope, resolve the contested interlock claim, and establish the measured basis for a supervisory-trim TurboAssist (Sections 06–08). Family C is new in this revision; families A and B carry forward from the 2026-08-01 draft substantially unchanged.
This document proposes goals and priorities. Every number carried in from Campaign 1 keeps its honesty coding: measured values stay green, modeled inputs stay amber, open items stay rust. Nothing here is a Campaign 2 result — there are no Campaign 2 results yet.
The single most valuable measurement Campaign 2 can produce is the combined-system drag area of the truck-and-trailer as one aerodynamic body. This is not the truck's published Cd·A, and it is not the trailer's — the two cannot be added. A tow rig has wake interaction, gap flow, and interference drag between the vehicles that no pair of standalone specs captures. The only honest way to get the combined figure is to measure the assembled rig in motion.
The failure was not wind and not sample size. It was collinearity. Cruise control held the flat-steady speed band so tightly that there was no independent variation in speed to regress against.
Across 504 minutes of qualifying flat-steady data, the speed inter-quartile range was just 62.3–62.5 mph, with 92% of the time in a single 62–64 mph bin. Regressing wheel power against v² returned Cd·A = −0.136 ± 0.303 m², 95% CI [−0.730, 0.458], R² = 0.000. The interval contains zero and negative values — physically impossible, and a direct signature of a regressor with no leverage. No reanalysis of Campaign 1 telemetry recovers this. The variable simply was not exercised.
A regen-decel "bridge" method was also attempted on the raw 1 Hz data from the three most regen-rich legs, on the theory that braking events sweep a wide speed range for free. The machinery ran cleanly, but the input precision does not exist in Campaign 1 data:
Returned Cd·A ≈ 1.4–2.0 m² (full-event) / 0.9–1.3 m² (high-speed subset) across a 0.78–0.92 regen-efficiency bracket, R² ≈ 0.05. That efficiency bracket alone swings the answer ~45%, and the result does not reconcile with the ~3.9 m² frontal-area expectation or the Phase A fuel-economy cross-check. The limiter is error amplification: MGU regen force (−10 to −14 kW routinely) is as large as or larger than the ~14.8 kW trailer road load being sought, so the method subtracts one large noisy number from another to recover a small one. Useful only as a soft floor — the combined-rig Cd·A is very unlikely to be below ~1 m².
This is the core of why the milestone matters. Campaign 1 produced a large body of measured fuel and power data, but several of its most important derived figures rest on an assumed drag number. A measured Cd·A does not just add a new data point — it retroactively upgrades the interpretation of data already collected.
| Use | Today (Campaign 1) | With measured Cd·A |
|---|---|---|
| Trailer road load @ 62 mph | ~14.8 kW — modeled | Measured; anchors the whole energy model |
| ProPower regression d-term (speed³) | Uncalibrated proxy, permanently | Replaced with calibrated Cd·A — the fit stops leaning on a placeholder |
| TrekDrive / TurboAssist energy balance | Rests on modeled drag → modeled-on-modeled | Load being offset is real; savings claims become defensible |
| Fuel & range at any speed | Only the ~62 mph cruise band was driven | Predict 55 / 60 / 65 / 70 mph cost and defend it |
| Aero/rolling split ("f" variable) | Unresolved — Crr known, Cd·A not | Split closes; Crr = 0.0197 already measured as byproduct |
| Any future aero change (topper, fairing) | No baseline to measure against | Baseline exists — changes get a kW and a dollar value |
Campaign 1 did not leave empty-handed on rolling resistance. The steady-state regression intercept gave Crr = 0.0197 ± 0.0028, physically plausible for a heavy trailer at highway speed. Rolling resistance was never the blocked term — only the aero term was. Method 2 anchors this Crr value, which removes one degree of freedom and tightens the Cd·A estimate.
The through-line: the trailer road-load figure of ~14.8 kW at 62 mph is a modeled input that flows into the TurboAssist energy balance. Every fuel-savings and battery-neutral claim built on TurboAssist inherits that assumption. Measuring Cd·A turns the keystone real, and everything resting on it stops being an estimate stacked on an estimate.
The fix for collinearity is to induce the speed variation cruise control removed. Method 2 holds the rig at a series of deliberately different steady speeds on genuinely flat road, and regresses wheel power against v² across those holds. Because the wheel-power term depends on HV Battery Power — now confirmed a total bus-terminal measurement — the input the second run was waiting on is cleared.
| Parameter | Specification |
|---|---|
| Speed holds | 50 / 55 / 60 / 65 / 70 mph |
| Duration each | ≥90 s (first 15 s trimmed); OBD 2.0 s cadence is the bottleneck — do not shorten |
| Road | Flat, low-traffic; net altitude drift rejected per hold |
| Direction | Bidirectional — cancels wind and any residual grade |
| HV battery | Held neutral each hold; discard holds where bus power is not small |
| Air density ρ | Computed per hold from measured OAT + altitude + pressure — direct 1:1 multiplier on Cd·A, never assumed constant |
| Crr | Anchored at 0.0197 (Campaign 1 measured); cross-check new intercept against it |
| Sanity floor | A regression result below ~1 m² implies a sign/unit error — suspect before believing |
The Salton Sea east shore (Hwy 111, Desert Beach → Bombay Beach) is the leading candidate: exceptionally flat with both ends near −220 to −225 ft, net change only a few feet over ~20 mi, and below sea level — which raises air density and strengthens the aero signal. Late-August early mornings run calm and around 80 °F. The Prescott-area Hwy 89 / I-40 corridor was profiled from real Campaign 1 GPS and rejected — it is undulating grade end-to-end with no continuous 2-mi flat shelf at one density altitude. One deliberate 30-minute speed-hold session on the right road beats seven hours of incidental cruising; schedule it early, do not defer to a convenient leg.
Campaign 2 runs with a hardware change that makes the fresh Cd·A measurement especially timely. The Flatbed Full-Short MR56 bed topper — originally fitted for a taller Airstream tow — is being replaced with a Full-Short CAP56, roughly 9 inches lower. A lower roofline over the bed changes the tow-vehicle profile and, more importantly, the base-wake and truck-to-trailer gap flow that the LightShip nose sits in. It has the potential to reduce combined drag.
Operator expectation is that the combined-Cd·A change will be small, for three stacked reasons: (1) the topper roof runs bug-free, indicating already-detached flow above the cab line, so mid-truck flow is largely insensitive to the 9-inch change; (2) the primary topper benefit is decking the open bed cavity — eliminating open-box separation and handing the trailer attached, organized flow — and both toppers deck the bed equally, making the height delta a second-order effect; (3) the LightShip in road mode sits lower than the old Airstream bug line, so the trailer face is below the truck shear layer and the 9-inch change happens mostly above the trailer roofline. The prior Airstream gained slightly under 1 MPG from the taller topper (a fairing effect on a tall trailer face) — a mechanism the lower LightShip does not intercept, so that magnitude does not transfer. Residual unknown only measurement can size: base-wake coupling between the lower truck wake and the LightShip tow-mode wake.
All aero quantities are configuration-specific. Campaign 1 was run on the MR56; Campaign 2 is on the CAP56. Any Campaign 2 Cd·A, road-load figure, Phase A economy, or ProPower-regression d-term measures the CAP56 rig and must not be silently compared against or merged with Campaign 1 MR56 aero numbers. The topper verdict rides on the Salton Sea Method 2 run precisely because it is self-normalizing — per-hold ρ and bidirectional wind cancellation — and independent of any historical run.
The late-August Thermal (La Quinta) trip retraces the June pre-Campaign-1 course, tempting an MR56-vs-CAP56 repeated-route comparison. It can only be a raw-MPG qualitative check. The June run used the old 20-PID pre-lockdown set with no OAT channel, so it cannot be density-corrected — the one confound most worth removing cannot be removed on the June side. It was also an extremes-corner run (heat + ~5,100 ft grade each way + ~7 mph headwind on return). Only the June 13 outbound trip-average (14.52 mpg, fuel reconciled to 0.02%) is trustworthy, and it bundles heat, grade, and topper inseparably. Drive the August CAP56 pass with full 44-PID logging and compare raw outbound trip-average against 14.52; agreement within a couple tenths is consistent with a small topper effect but does not prove it. It is not the instrument for the verdict.
A PowerBoost is best understood as two propulsion systems sharing one high-voltage bus. The first is the engine: gasoline burned at the crankshaft, some of it turning the wheels directly, some of it driving the motor/generator to put charge onto the bus. The second is recovery: kinetic energy the vehicle already had, captured back through the traction motor during braking and descents and returned to the bus. The bus is a common rail where both sources meet the loads — and every load on it, whether it moves the truck or cools the cabin, is ultimately drawn against those two sources. The pack itself is a buffer, not a source: over any complete leg, what leaves the bus was put there by gasoline or by regeneration, nothing else.
Six flows cross the bus. Framing them as inputs (energy arriving) and outputs (energy leaving) is what makes the gasoline-versus-recovery accounting possible.
| Flow | In / Out | Origin | Instrumentation |
|---|---|---|---|
| Engine-driven charge (MGU generating off crankshaft) | Input | Gasoline | Inside HV Battery Power (−31 to −35 kW); not a separate channel |
| Regenerative braking (MGU generating off wheels) | Input | Recovered kinetic | Inside HV Battery Power (−10 to −14 kW); the one flow with no direct channel — must be inferred |
| Traction assist (MGU motoring) | Output | — | Inside HV Battery Power (20–35 kW). The only output that does propulsion work |
| DC/DC converter → 12 V system | Output | — | Direct HV-side amps (new 44-PID channel) |
| DC/AC converter → ProPower inverter | Output | — | Metered output kW + HV-side amps (44-PID) |
| A/C compressor motor (ACCM) | Output | — | Compressor-current channel — metering path unresolved |
Of everything that pulls energy off the bus, exactly one output moves the truck: the traction motor in motor mode. The DC/DC converter, the ProPower inverter, and the compressor are all utility loads — they run the truck's and trailer's services, not its propulsion. That distinction is the whole reason the bus accounting matters: it separates the energy spent going somewhere from the energy spent staying comfortable and keeping the electronics alive.
Five of the six flows can be measured directly once the 44-PID channels are populated. Regeneration is the exception. It has no dedicated channel because it shares the traction motor/generator with both motoring and engine-driven charging — the same machine, the same bus terminal. All three appear folded together inside HV Battery Power, distinguished only by sign and circumstance. And there is a genuinely hard case inside it: on a fueling descent the crankshaft and the wheels can drive the generator at the same time, so the pack sees a single negative current that is the sum of gasoline-sourced and recovered charge. One channel, two sources — algebraically underdetermined from the bus alone.
The bus cannot split that sum, but the engine can. The crankshaft carries a second, independent power balance: whatever brake power the fuel produces that is not going to the wheels is going to the bus. Both terms on the right are knowable from channels already logged.
Engine bus-charge kW = (engine brake power from fuel) − (power delivered to the wheels). Engine brake power comes from fuel rate and RPM against the BSFC map; wheel power is the instantaneous road load — aero + rolling + the gravity term m·g·sin θ — the same physics Method 2 measures, with grade from TrackLogger altitude. Subtract the engine's contribution from the total negative bus power and the remainder is wheel-sourced regeneration. Two equations, two unknowns: solvable. This is why the engine parameters we already log are enough to do the split — the bus number alone never was.
The honest limit is conditioning, and it is the same wall the coastdown Cd·A method hit: this recovers a difference between two sizable modeled numbers, so it is trustworthy only when one clearly dominates. That gives a three-tier treatment, coded by how clean each tier actually is.
| Tier | Rows | Method | Coding |
|---|---|---|---|
| Pure regen | Decel + DFSO (fuel rate = 0) | Fuel is cut, so gasoline bus-charge is zero by definition. All negative bus power is recovered — no model, no subtraction. | Measured floor |
| Engine-off charge | Non-fueling, engine idle-down | Sign-gated: negative bus power while the engine is not fueling. Clean, minimal modeling. | Inferred |
| Mixed-source | Fueling descent, both driving the generator | Crankshaft power balance above. Report as a bracket; reliable where the gravity term dominates (steep descent, engine near idle/DFSO), unrecoverable where engine and wheel power are comparable — flag those rows rather than force a number. | Modeled — bracketed |
Integrated over a leg, tiers one and two are firm and tier three is a bounded band, so the per-leg gasoline-versus-recovery split comes with an honest error range rather than a false point value. The measured DFSO floor alone establishes a hard minimum recovered kWh that no modeling assumption can erode — the most defensible number in the accounting. Campaign 2's cleaner traction residual (traction falls out once DCDC and DCACA are subtracted) sharpens all three tiers but does not eliminate the mixed-source band; only the conditioning does.
Why this accounting is worth building: it lets the pitch state what fraction of the trailer's charge and the truck's utility loads was carried by recovered energy that would otherwise have been lost to brake heat, versus burned as fuel. On a mountain leg with long descents, the recovered fraction is real and material; on flat plains it is small. Quantifying it turns "the hybrid helps in the hills" from a claim into a number.
The two largest utility outputs, the compressor and ProPower, both peak in hot weather, and they peak together for related reasons. That co-timing is what connects bus demand to engine efficiency.
The compressor peaks because heat drives cabin cooling — and because the same compressor, via a BECM-controlled diverter valve, also serves the battery chiller under high ambient and/or high pack current. ProPower peaks in hot weather too: a high share of the exported AC goes to running the trailer's own air conditioning when the cabin — or the AE.1 living space — needs cooling. So on a hot leg the bus carries a double cooling burden (truck cabin + pack chiller on the truck side, trailer cooling on the ProPower side), and because the pack is only a buffer, every one of those kilowatts traces back to the crankshaft. More bus load is more engine work, at exactly the ambient condition where the engine is already least efficient — thin hot air, reduced intercooler margin, and, on grades, commanded-rich enrichment.
There is a second-order loop worth measuring, not yet a finding. Bulk-charging the trailer at 5.8 kW pushes sustained current through the truck pack; sustained pack current is one of the two documented conditions that call for chiller duty; chiller duty is compressor draw; compressor draw is more bus load and more engine work. If the loop is real, V2V charging carries a hidden cooling surcharge in hot weather on top of its measured 7.70 kWh AC/gal — making that figure a floor rather than a full cost. Campaign 1 could not test this because it never metered the compressor's bus draw. Campaign 2's job is to close it.
Backing ProPower's measured 7.70 kWh AC/gal through inverter losses puts roughly 8 kWh of bus energy per gallon. A compressor sustaining 1.5 kW on a hot leg then costs about 0.19 gal/hr — against a tow-leg baseline near 4.1 gal/hr at 62 mph, on the order of 4–5%, or roughly 0.6–0.7 mpg. Both the inverter backout and the assumed duty are modeled. The point is not the number but its size: the expected effect sits inside the leg-to-leg terrain spread that Campaign 1 could not see past, which is exactly why the compressor branch must be metered directly rather than inferred from mileage. This is the same structural wall Cd·A hit — the signal is real but buried under a larger confound until the variable is isolated on purpose.
Do not regress mileage against ambient on Campaign 1 legs. Day 4 appears to show mileage improving as it got hotter (13.90 → 15.73 → 15.63 mpg as OAT rose 89.6 → 93.2 → 96.8 °F) — but that is the operator commanding ProPower down for engine protection (17.46 → 6.26 → 2.58 kWh), and the ProPower reduction is worth more mileage than the heat costs. The naive hot-versus-cool comparison picks up the operator's thermal practice, inverted. The clean Phase A pair (Day 5 / Day 6: 15.43 mpg at 95.0 °F, 15.27 at 80.6 °F) agrees within 1% across a 15 °F spread — which says the accessory penalty on flat legs is smaller than the terrain spread, not that it is zero.
The LightShip AE.1 owner's manual (Lightship RV Manual 2025) entered the reference set on 2026-08-02 and is now the trailer-side authority, held to the same discipline as the Ford factory manual. It settles several questions the truck-side manual never could — and it contradicts one long-standing knowledge-base entry, addressed in Section 07.
| Parameter | Manual specification |
|---|---|
| Actuation basis | Power applied is proportional to the force on the hitch sensor — harder pull, more assist |
| Speed window | Power applied between 15 and 75 mph; 75 mph is the maximum recommended towing speed |
| Coast / decel | No propulsive force when the tow vehicle is coasting or decelerating |
| Regeneration | Drive motor operates in regenerative mode during deceleration, returning power to the trailer pack; rear drums are regenerative-braking equipped |
| Mode gating — TrekDrive | Road mode only. Canopy must be lowered to enable Road mode. Camp mode enables every function except TrekDrive. Towing in Storage mode is permitted but TrekDrive is unavailable |
| Mode gating — charging | Camp or Storage mode — both work. The manual's charging procedure directs the operator to select camp or storage mode before connecting the NACS EVSE. Road mode is not a charging mode |
| Lockouts — power not applied | 7-pin disconnected · reverse at any speed · brakes engaged · hitch sensor zero or negative · ambient > 110 °F · climbs longer than 10 mi on grades ≥ 6% · not in Road mode |
| Manufacturer range claim | Towing the AE.1 should impact EV range or fuel mileage by only 5 to 10% when TrekDrive is active |
| Trailer temperature limits | Operating 20–110 °F; storage 32–110 °F |
Mode selection, not any device-level interlock, is what determines which capabilities are simultaneously available. All three modes are towable; they differ in what runs while towing.
| Capability | Road | Camp | Storage |
|---|---|---|---|
| TrekDrive assist / regen | Available | Disabled | Disabled |
| NACS charging (V2V umbilical or shore) | Locked out | Available | Available |
| 110 V AC inverter / 120 V circuits | Available (pack-sourced) | Available | Off |
| HVAC | Available (per LightShip; T-2 confirms) | Available | Unavailable |
| 12 V loads (lights, fridge, water, toilet) | Available | Available | Available |
| Canopy | Must be lowered | Either | Either |
Storage mode charges but runs with the 110 V AC inverter off and HVAC unavailable — 12 V loads only. Under tow with the umbilical connected that makes it a minimum-house-load charging state, materially different from Camp mode rather than a variant of it. It has an experimental use (see T-8) and possibly an operational one: it is the configuration in which the largest fraction of delivered ProPower energy goes into the pack rather than into house loads.
The proportional-to-hitch-force rule combined with the zero-or-negative lockout defines what the system fundamentally is. TrekDrive has no external command input. It cannot be instructed to assist; it reads hitch force and cancels what it can, and it stops at the point where hitch force reaches zero. It reacts to the truck's demand — it does not anticipate it and it cannot exceed it.
An earlier working sketch described TurboAssist as commanding a single continuous power axis from +5.8 kW (ProPower charging) through −30 kW (full assist). That is not achievable on existing hardware and is withdrawn. Motor kW is not a controllable variable from outside the trailer. The controllable variable is the hitch-force gain — how aggressively the native logic converts a given sensed force into assist. Reasoning preserved, conclusion retired.
Watching Atlas in motion with TrekDrive active, motor kW swings substantially and continuously with terrain — up on grades, down on the flat. The base system already variabilizes itself against hitch force in real time; that is its native job and it does it well. It also transitions into regeneration on its own without any external input, monitoring the hitch and running the motor as a generator back into the LightShip pack. None of this is proposed for change. Coded observed rather than measured: it is a dashboard observation, not a logged channel. Campaign 2 should put a number on it.
Layer 1 (existing, untouched): the hitch-force control loop. Continuous proportional modulation, the zero/negative floor that guarantees the trailer never pushes the tow vehicle, automatic regen entry, and all seven lockout conditions. LightShip owns it entirely and it remains the safety and ride-quality foundation.
Layer 2 (proposed): a supervisory trim riding on top. Within the envelope Layer 1 already enforces, apply more or less of what it would natively deliver, biased by the state of the PowerBoost — engine load, coolant temperature, altitude, intake-air temperature, ProPower activity. Layer 2 never commands the motor and never overrides a lockout. It moves a gain.
This is the framing that should carry into the pitch. It is not a revision of TrekDrive. It is one authorized input into a curve LightShip already owns and keeps owning.
The obvious use of Layer 2 is subtractive: reduce or suspend assist when the PowerBoost does not need it, preserving trailer charge for when it does. That direction is unambiguously safe because it can only move away from the constraint.
Whether Layer 2 can also bias upward depends on whether native TrekDrive leaves any margin below the zero-hitch-force ceiling. LightShip has stated that above 15 mph the system is deliberately programmed to maintain a light residual load on the hitch — never fully neutral, always a small tug from the tow vehicle, for stability. If that holds, headroom exists by design and modest upward bias is available without ever approaching the zero-force limit.
The residual-tug figure has never been quantified by LightShip, and hitch force is not observable in this instrumentation. It is sensed and logged trailer-side only. Every statement about proportionality, headroom, and available upward margin is therefore inferred from Atlas kW behaviour and manufacturer description, not measured on this rig. This is the single largest instrumentation gap in the TurboAssist case and it does not close without a trailer-side feed. Do not upgrade any headroom claim to measured until that data exists.
Two of the seven lockouts are thermal or terrain gates that could in principle remove TrekDrive exactly where a thermal-assist mode would want it most. Both were tested directly against Campaign 1 telemetry: 19,734 moving 10-second frames with valid grade, grade computed from TrackLogger altitude over ~1-minute smoothed windows, sustained ≥6% runs accumulated by distance.
| Lockout | Campaign 1 reality | Binding? |
|---|---|---|
| Climbs > 10 mi at grade ≥ 6% | Longest sustained ≥6% climb in the entire campaign: 3.30 mi. Zero frames accumulated a blocking run. 96.4% of all ECT-above-spec dwell occurred below 6% grade. The four hottest frames of the campaign (ECT > 235 °F, 7,712 ft) sat at 4.4% mean grade | No — would not have fired, including at the worst thermal event |
| Ambient > 110 °F | Campaign maximum OAT 104.0 °F; zero frames above 110 °F | No on Campaign 1 — but see Section 10 |
| Speed below 15 mph | 2.9% of moving time | Negligible |
| Reverse · brakes · 7-pin · Road mode | Operational conditions, not envelope limits | Not applicable to steady-state assist |
This is the strongest measured support the thermal-assist concept has. The PowerBoost's worst thermal loading occurs at altitude with moderate ambient — the campaign's peak event was 78.8 °F ambient at 7,516–8,235 ft on a 4.4% grade. LightShip's protections are keyed to high ambient and sustained steep grade, protecting the trailer motor and pack. These are different thermal domains. TrekDrive is available in precisely the conditions where the PowerBoost is most heat-stressed, and Campaign 1 contains no counterexample. The result is derived from measured telemetry, not asserted.
On a 6% grade at 58 mph the trailer's own gravity power is roughly 56 kW before aero and rolling. Even at the full 30 kW rating the hitch cannot be nulled — the motor saturates and the tow vehicle carries the remainder. The KB's ~250 Wh/mi to-zero-hitch-force figure corresponds to roughly 15 kW at 60 mph on the flat, so normal towing exercises about half the rating and the balance is climb and regen headroom. On real grades the binding limit is capacity, not authority.
The manual's 5–10% claim is a towing-penalty statement, not a savings statement, and it is a different quantity from anything the KB currently tracks. It is also the number LightShip's own leadership already believes. Campaign 1 offers a suggestive check: the July 14 State A run measured 23.32 mpg with the trailer attached and TrekDrive active, against Phase A pure-tow at 15.43 mpg with TrekDrive off. What is missing is the solo baseline on the same loop — without it the comparison cannot be closed. Measuring it is cheap and it anchors the entire TrekDrive discussion to the manufacturer's own metric, in the manufacturer's own units.
The knowledge base carries the following as a hard fact, in both PowerBoost_LightShip_Instruction_Set.md (Terminology) and the TrekDrive TurboAssist energy-balance document (under a heading that reads "System interlock constraints (hard facts)"):
"TrekDrive active immediately disables HVAC and all 120V services; refrigerator switches to ~300W DC backup; hot water heater disabled."
The owner's manual does not state this anywhere. The Climate System section sits immediately after the TrekDrive section with no mode restriction attached, and the only documented mode restriction runs the other way — Camp mode is the mode in which TrekDrive is unavailable. LightShip has now stated directly that HVAC operates with TrekDrive active in Road mode. The KB claim therefore has no documentary support and one manufacturer contradiction.
Held as contested rather than deleted, on the same principle applied to the pre-trip "quits in the low 100s °F" EVSE figure: a manufacturer statement is a claim about a product, not a measurement. It is resolved by Campaign 2 test T-2, not by argument. Reasoning preserved either way.
A prior recommendation identified the interlock as the highest-value design change to request from LightShip. That recommendation is withdrawn — it was built entirely on the KB claim now under suspicion. It is replaced by the Layer-2 gain-channel request in Section 06, which is a substantially smaller ask.
A prior draft stated that exclusivity had "no documentary basis" because umbilical charging is absent from the seven TrekDrive lockout conditions. That reasoning searched the wrong section. The manual establishes exclusivity in the charging procedure, not the TrekDrive envelope: step 1 of Charge with Grid Power directs the operator to put the AE.1 in camp mode or storage mode before connecting the NACS EVSE. Road mode is not a charging mode. The claim is restored; only its attributed mechanism changes.
Exclusivity is therefore enforced at the vehicle-mode level, not by a device interlock. Road mode enables TrekDrive and locks out the NACS receptacle; Camp and Storage modes permit charging and disable TrekDrive. There is no configuration in which both run, and no device-level interlock is needed to produce that outcome — the mode selection alone does it.
Selecting Road mode with the NACS plug seated in the LightShip receptacle raises an Atlas error stating the NACS plug is occupied. The plausible reading is drive-away protection — the same convention every EV applies to a latched charge connector — rather than an electrical prohibition. Whether it is a dismissible alert or a hard refusal is not established, and the distinction matters only for test design, not for the exclusivity conclusion, which the charging procedure settles independently.
If charging requires Camp or Storage mode, then every Campaign 1 State B leg — all seven legs of measured in-motion V2V delivery — was driven with the trailer in Camp mode, and the July 14 State A calibration run was in Road mode. Vehicle mode is not carried in the OBD stream and was never logged. It should be confirmed from recollection and recorded against the Campaign 1 ledger, because it is now a governing configuration variable rather than an incidental setting. Campaign 2 should log trailer mode explicitly on every leg.
The mode structure suggests a provenance for the contested KB entry, and separates a true statement from a false one that were bundled together.
True: in Road mode no umbilical energy arrives. House loads, TrekDrive, and everything else run off the trailer pack alone for the duration of the segment. The pack is a closed budget while TrekDrive is available. That is a real and material constraint, and it is very likely what the original KB entry was reaching for.
Apparently false: that HVAC and 120V services are disabled. Nothing in the manual disables them, and LightShip states HVAC operates in Road mode. Losing the source is not the same as losing the function — the two were conflated, and the conflation survived into the KB as a hardware claim. T-2 settles it.
With exclusivity confirmed, the mandatory ProPower decomposition is not a workaround for how two runs happened to be configured — it is structurally required, permanently. Every State A run is necessarily a ProPower-off run because they are different vehicle modes. 27.8% remains the headline mechanical figure, and no future campaign can produce a matched-condition pair that avoids the correction.
The mandatory ProPower decomposition rule is unaffected. The July 14 calibration pair measured propower_kwh_active = 0.0 on Run 2, so that pair genuinely was confounded and the 27.8% decomposed mechanical figure remains the correct headline. What changes is only the justification: from "structurally unavoidable" to "that is how those two runs happened to be configured."
A crossed TrekDrive-on/off × ProPower-on/off design was proposed in the previous draft as a way to measure mechanical load transfer directly. It is not executable — two of its four cells require a vehicle mode that does not exist. It is replaced by the four-arm ladder (T-4 below), which reaches most of the same ground legally. No test should be designed around dismissing or defeating the Atlas drive-away alert.
The ordering below is deliberate and differs from earlier drafts. The three cheap constraint-verification runs come first, because two of them can invalidate knowledge-base text that several downstream conclusions rest on. Running the expensive comparative work before the constraints are verified risks producing a large body of data interpreted under a false premise — which is exactly what happened to the duty-collapse and armed-idle conclusions in Campaign 1.
| ID | Test | Method | Resolves | Cost |
|---|---|---|---|---|
| T-1 | Solo-loop baseline | Broomfield calibration loop, no trailer, 44-PID logging, matched speed and conditions to the July 14 runs | Anchors LightShip's own 5–10% towing-penalty claim; supplies the missing reference leg for every mpg comparison in the TrekDrive case | ~45 min |
| T-2 | HVAC during TrekDrive | Road mode, canopy down, TrekDrive active, Atlas climate running at setpoint; log Atlas SOC and climate state throughout | Resolves the contested interlock claim — restores it or retires it | 20 min |
| T-3 | Mode-transition behaviour (parked) | Stationary. Establish active charging in Camp mode, then select Road mode and observe: does charging terminate, does the alert block the mode change, is it dismissible. Log DCACA output through the transition | Whether Road mode refuses to initiate charging or actively terminates it — and how the umbilical must be handled at every A/B changeover in the field | 15 min |
| T-4 | Four-arm ladder | Same loop, matched speed and direction, four arms: (1) solo, no trailer [= T-1]; (2) tow, Camp mode, ProPower off; (3) tow, Camp mode, ProPower on; (4) tow, Road mode, TrekDrive active. Atlas SOC logged on arms 2–4 | Supplies the null tow arm the July 14 pair never had. Isolates the ProPower increment and the TrekDrive increment against a common baseline instead of against each other | 3 runs beyond T-1 |
| T-5 | Matched-grade climb pair | Same 3–5% grade, under 10 mi, two runs TrekDrive on and off, ambient below 110 °F; full thermal channel set | Thermal-mode effect inside the legal envelope; ECT headroom under assist | 2 runs |
| T-6 | Thermal step response | Deliberate sustained load step at constant speed and grade, then release; log ECT, IAT2, load, altitude at full cadence | Identifies the ~5-minute ECT plant model that quantifies how far ahead a predictive trigger must look | ~20 min |
| T-8 | Camp vs Storage house-load isolate | Umbilical connected and delivering, matched speed and ambient, two windows: Camp mode then Storage mode (110 V AC and HVAC off). Log DCACA output and Atlas SOC in both | Isolates trailer house load with the umbilical held constant — a cleaner instrument for the connected-load-support question than connected-vs-disconnected, because only the load changes | ~40 min |
| T-9 | Mode-cycle dynamics (parked, then in-motion) | Repeated Camp → Road → Camp transitions with the umbilical seated. Time each transition, time charge re-establishment, note any refusal, delay, or house-load interruption. Repeat under tow once the parked behaviour is known | Quantifies the cost of a mode switch — the single largest unknown in R-3, and the thing that determines whether automated mode arbitration is viable at all | ~30 min |
| T-7 | Hitch-force / Atlas telemetry capture | Any trailer-side feed LightShip can provide, even low-rate or post-hoc, time-aligned to the OBD stream | The only path to measured proportionality and measured headroom | Dependent on LightShip |
TurboAssist as specified needs three things from the trailer side. They are not equal in size, and the document should not pretend otherwise.
| # | Requirement | What it touches | Size of ask |
|---|---|---|---|
| R-1 | Supervisory gain channel. One sanctioned input biasing the existing hitch-force gain, LightShip retaining the entire control law behind it | A parameter on a curve they already own. No safety behaviour changes; every lockout still applies | Small |
| R-2 | Release the NACS lockout in Road mode, so charging and assist availability are no longer mutually exclusive by mode | A drive-away protection. Safety-relevant by construction | Large — see below |
| R-3 | Mode-command authority. TurboAssist commands Road ↔ Camp transitions as part of normal operation, driven by PowerBoost state | Vehicle mode is the top-level state machine. External command authority over it is a new class of interface | Large |
| R-4 | Feedback channel. Hitch force and Atlas SOC returned into the trip log, even low-rate or post-hoc | Read-only telemetry export | Small |
R-1 and R-4 can be presented as a trim and a log export. R-2 and R-3 cannot. R-2 asks for the removal of a protection against driving with a charge connector latched — the convention every EV in the market enforces. R-3 asks for external command authority over the vehicle's top-level state machine, including transitions that change whether the drive motor is live. Presenting all four as one modest request would misrepresent the work and the risk, and would damage credibility with the engineers who have to build it. State the four separately and let them be evaluated separately.
The lockout does not have to be eliminated — it has to become condition-aware. The trailer can already distinguish "tethered to a tow vehicle" from "plugged into a pedestal": the 7-pin connector is present, hitch force is being sensed, and road speed is non-zero. A rule of the form permit NACS engagement in Road mode if and only if the 7-pin is connected and hitch force is present preserves the drive-away protection exactly where it was designed to matter — a campsite, a driveway, a charging stall — while permitting the in-motion V2V case. That is a conditional refinement of existing logic rather than the removal of a safety behaviour, and it is a far easier proposition to defend internally. Recommend leading with this form.
R-3 assumes mode transitions are cheap and repeatable. Nothing establishes that. Unknown and unmeasured: transition latency, EVSE/OBC handshake time on each charge re-establishment, whether HVAC or house loads interrupt across a transition, and whether any settling or lockout period follows session termination. The campaign's one confirmed thermal dropout took 30–45 minutes to resume charging with LightShip-side EVSE/OBC/BMS logic the strong candidate — n = 1, but it is a direct warning that charge sessions on this system may not restart on demand. If any comparable delay attaches to routine mode cycling, a TurboAssist that toggles modes on a thermal timescale could stall charging entirely and deliver less energy than never switching at all. This must be measured before R-3 is specified in any detail — T-3 and T-9 are the first steps.
Every objective T-1 through T-9 is executable with truck-side instrumentation and manual mode selection. No Campaign 2 primary deliverable depends on any of R-1 to R-4 landing. The evidence gets built first; the integration requests arrive afterwards, backed by measurement, with the two large ones sized honestly and the transition-dynamics risk already quantified rather than discovered later.
T-3 was originally specified as an in-motion attempt to run both systems together. That version is withdrawn: exclusivity is settled by the manual, and the remaining question — initiate versus terminate — is answerable safely while stationary. Do not attempt a mode change with the umbilical seated while under way, and do not design any test around dismissing the drive-away alert.
An earlier draft concluded that because mode changes are a manual Atlas action, TurboAssist's scope was confined within Road mode and cross-mode arbitration was out of reach. That was an inference from present operator practice, not a statement of the product definition, and it is withdrawn. Automated mode transition is explicitly part of TurboAssist as specified. Reasoning preserved; conclusion retired.
Campaign 2 expands instrumentation from 42 to 44 PIDs, adding DCDC converter HV-side current and DCACA HV-side current — both HV-bus amp-draw channels. This opens the direct HV-bus split that Campaign 1 could only model as a residual, and it makes several long-standing items reportable for the first time.
| Goal | What it needs | What it delivers |
|---|---|---|
| LightShip Atlas SOC / draw-kW logging | Trailer-side channel on every calibration run | First measured TrekDrive net; closes the connected-load-support attribution |
| ACCM metering-path test | Parked ACCM toggle, watch DCDC HV current | Whether the compressor's HV draw lands in the DCDC channel — gates the six-term regression's f term |
| Connected-load-support isolation | Matched connected-vs-disconnected SOC-decline windows, ~1 hr | Converts the middle-state house-load claim from attributed to measured |
| 44-PID frame-rate verification under tow | First Campaign 2 towing file, parked-then-driving check | Confirms cadence holds at 44 PIDs under representative load (static pass already done) |
| TurboAssist trigger logic | Add altitude or ECT term to the load-based trigger | Campaign 1's worst thermal event (78.8 °F ambient, at altitude) would have missed the OAT-above-86 °F correction entirely |
The connected-load-support isolation and the ACCM toggle both pair naturally with the Salton Sea speed-hold session — same venue, same calm-morning window. Running them together is the efficient sequence. Where a prerequisite is missing, the affected report row stays marked "Not Reportable — pending [X]"; it is never filled with a placeholder number.
The Cd·A objective and the TrekDrive objective want the same venue for different reasons, and they collide on one variable. The Salton Sea east shore is selected for the speed holds because it is exceptionally flat and below sea level. The same location in late August is also one of the hottest places in North America — and TrekDrive stops applying power above 110 °F ambient.
Late-August afternoon ambient in the Thermal / La Quinta / Salton Sea corridor routinely exceeds 110 °F. Above that threshold the entire TrekDrive block is unavailable — no State A data, no 2×2 factorial, no thermal pair, no solo-versus-assisted comparison. This is an OEM lockout, not a derate: there is nothing to measure because nothing runs.
Campaign 1 never encountered this because it topped out at 104.0 °F. Campaign 2's chosen venue makes it a live constraint for the first time. Check forecast ambient against 110 °F before committing the schedule, and treat the TrekDrive block as morning-window work with a hard abort criterion, not as something to fit in opportunistically.
| Objective | Ambient sensitivity | Window |
|---|---|---|
| Cd·A Method 2 speed holds | Wants calm air and stable density; heat is not disqualifying, wind is | Early morning — for wind, not temperature |
| Parked-bench ACCM metering path | None — engine-in-the-loop, stationary | Any |
| Connected-load-support isolation | Wants matched conditions between the two windows, not a specific temperature | Any stable period |
| TrekDrive block T-1 to T-6 | Hard lockout above 110 °F — total loss of the objective | Morning only; abort if ambient approaches the ceiling |
| Hot-leg accessory / cooling burden | Wants heat — this is the one objective that benefits from the venue's extremes | Afternoon, deliberately |
The day splits cleanly. Mornings carry the two ambient-sensitive objectives that need cool calm air — the speed holds and the TrekDrive block. Afternoons carry the objectives that want heat, chiefly the hot-leg accessory and cooling-burden accounting, where high ambient is the independent variable rather than an obstacle. Parked-bench work fills any gap. Within the morning window, the constraint-verification runs T-1 to T-3 precede everything else, because their outcomes determine whether T-4 is worth executing at all.
One operational note that touches every trailer-side test: Road mode requires the canopy lowered. Any test involving trailer HVAC or house-load behaviour is therefore a canopy-down measurement by construction, and cannot be compared against canopy-up house-load figures.