PIXEL ENGINE SIM v0.6 build 112

1. ENGINE — design an engine (bottom end, air path, fuel & spark, cooling…), then press BUILD & DYNO. 2. DYNO reads the full power/torque curve and a graded scorecard. 3. LIVE is a running test bench: you set THROTTLE and LOAD and watch the engine respond in real time. You never set RPM directly — you drive it.
Then put it in a car on VEHICLE and take it to a circuit on TRACK. Every number in the app comes out of the same physics, so a change you make on the ENGINE tab shows up in the 0–100, the top speed and the lap time.

ENGINE — the design form: every part of the engine, a preset list to start from, an objective to aim at, and BUILD & DYNO.
VEHICLE — drop the engine into a chassis, choose tyres, gearbox, aero and brakes, and read the performance figures. The garage lives at the bottom.
TRACK — a solved lap of a real or imported circuit, with the racing line and what limits the car at every point.
LIVE — the engine running: a cutaway of one cylinder, the whole bank firing in order, and the controls on the right to drive it.
DYNO — the power and torque curves, the scorecard, and printable reports.
GUIDE — this manual. OPTIONS — sound, wear and language.

Start from a preset. On the ENGINE tab, open the build list at the top and pick something close to what you want — a 1.6 sport four, a turbo hot hatch, a V8. Press BUILD & DYNO.
Read the DYNO tab. Note the peak power, where it arrives, and the grade. The scorecard tells you what is costing you: a low band score means the power is peaky, a low safety score means it is close to knocking.
Change one thing. Swap the cam, the runner length or the boost, build again and compare. One change at a time is the only way to learn what each part does — the LOG PULL button on the LIVE tab keeps the previous curve on the dyno as a faint overlay.
Drive it. On LIVE, raise the throttle and use LOAD to hold an rpm. Watch the knock bar, the charge temperature and the EGT; with WEAR on, abuse breaks things.
Put it in a car. On VEHICLE, pick a chassis. The final drive is matched to your engine automatically, so the 0–100 and top speed are fair from the start. Then try TRACK.

THROTTLE — opens the intake; raises manifold pressure → torque → revs.
LOAD (dyno brake) — resistance that sets the hold RPM. 0% = free-rev to the limiter; 100% = holds near idle. It only resists above its target, so any engine can always rev up to it.
ELEC LOAD — accessory draw (lights/AC). Heavy load + small alternator at idle drains the battery → low voltage → weak spark.
AFR — air-fuel ratio. Petrol: peak power near λ0.86 (rich), best economy near λ1.1 (lean). Diesel: full fuel at the smoke limit, λ1.2; leaner is less torque but less fuel per kWh.
IAT / COOLANT / OIL — charge & engine temps (auto in AUTO THERMAL mode).
NITROUS — if fitted, sprays at wide-open throttle from a bottle that depletes.

PEAK TORQUE · MAX POWER · ECO — three buttons that put the engine you built on a real operating point and let it settle there. MAX POWER opens the throttle fully and has the dyno brake hold the rpm where this engine makes its peak power, at the best-power mixture (λ0.86 petrol, λ1.2 diesel) and the scorecard's conditions — so the LIVE OUTPUT should land on the same figure the DYNO tab prints. PEAK TORQUE does the same at the torque peak. ECO searches this engine's speed, load and mixture for the point that burns the least fuel per kWh — the lowest BSFC — and holds it; for a petrol engine that is usually a little lean of stoich at high load and modest rpm, for a diesel well lean. Each button flashes what it set and what to expect. With the mixture on AUTO, the ECU's fuel map chooses the mixture instead.

The LIVE tab is a cutaway of the engine you just designed. Everything on it moves because the simulation moved it. It is deliberately not drawn in real time: a crank at 5000 rpm turns 83 times a second, which at 60 fps is a strobe rather than a picture — you could not watch a valve open. So the cycle runs at a pace you can read, between about 3.3 s and 1.1 s per full four-stroke cycle, still ranked by rpm so idle is visibly lazier than the limiter. The true speed is on the tachometer above it and in the note; the animation's job is to make the cycle legible. The piston follows proper crank-slider geometry (not a sine wave — a real rod of finite length makes the piston linger at the bottom and snap through the top), the bore is drawn to your actual bore/stroke ratio so an oversquare engine really is short and wide, and the valves open on true four-stroke timing with the scavenging overlap visible around TDC. The chamber is colour-coded by what is in it right now — fresh charge, compressed charge, burning, burnt — and the stroke is named above the head, so the cycle is something you can watch rather than something you have to remember. A two-stroke shows ports uncovering in the liner instead of valves; a rotary shows a real epitrochoid housing with the rotor turning at a third of the eccentric shaft. The strip beside it draws every cylinder at its own phase, so you can see the firing order run through a V12.

The Wankel view. A rotary is drawn with the same instruments as a piston engine — throttle, turbo, exhaust and EGT, breathing, output, fuel and spark, and a rotor bank with each rotor numbered and lit as it fires. The housing is a true two-lobe epitrochoid. A Wankel has two "top dead centres", at the two waists of the housing: the ports sit at one and the spark plugs at the other. Each of the three chambers is coloured by what is in it, and unlike a piston engine every phase happens at a fixed place — intake in one lobe, power in the other — which is labelled on the drawing.

The bank, and the firing order. The panel on the right of the cutaway draws every cylinder, numbered, at its own point in the cycle. The number over a cylinder lights up while that cylinder fires, so you can read the order straight off the drawing, and the caption underneath spells it out.
Cylinders do not fire in the order they are numbered. An inline four fires 1-3-4-2, which is also why pistons 1 and 4 are always at the top together and 2 and 3 at the bottom: the firing order and the crank geometry are the same fact. An inline six fires 1-5-3-6-2-4, a V8 1-5-4-8-6-3-7-2, a V12 1-7-5-11-3-9-6-12-2-8-4-10.
V and boxer engines are numbered bank by bank — bank A is 1 to n/2 on the top row, bank B the rest below — which is how BMW, Audi, Ferrari and Porsche number theirs, and the orders above are in that numbering. (GM numbers a V8 odd-left, even-right, so its familiar 1-8-4-3-6-5-7-2 is the same engine written differently.) Counts with no production layout, such as a V5, fire in sequence, which is still evenly spaced.

The numbers that used to be printed here in 7 px type now live in the LIVE OUTPUT panel beside the canvas, where they are actual text; what stays on the cutaway is what benefits from being next to the part it describes — throttle plate angle, boost and compressor flow at the turbo, EGT at the exhaust, and the knock bar.

HP / kW — power = torque × rpm. TORQUE (Nm) — twisting force; peaks lower than power.
BMEP (bar) — brake mean effective pressure: torque normalised to displacement, so you can compare engines of any size. VE (%) — volumetric efficiency: how well the engine breathes vs its swept volume (boost can exceed 100%).
THERMAL EFF (%) — share of fuel energy turned into work. BSFC (g/kWh) — fuel burned per unit power; lower is better (best around stoich, part-load).
FUEL FLOW (L/h) — live consumption. MAP / BOOST (bar) — manifold pressure; boost is MAP above 1.0.
KNOCK (LOW/MED/HIGH) — detonation risk. AUTO timing retards to control it; sustained HIGH does damage. HEALTH (%) — condition; falls under abuse, and 0 = a blown engine (rebuild).
EMISSIONS (CLEAN/OK/DIRTY) — from the exhaust analyser: CO & HC climb rich, NOx climbs lean/hot; a catalytic converter only cleans all three near stoich.
λ (lambda) — AFR ÷ the fuel's stoichiometric AFR: λ<1 rich, 1 = stoich, >1 lean.
COST ($) — total build price of the parts you chose. REL (reliability) — a design estimate of how much abuse the build invites (boost/knock/heat/high-rpm stress); higher is tougher. hp/$1k — power per thousand dollars: the value metric.

DYNO RESULT — under the chart on the DYNO tab, the pull as a static sheet: peak power and torque with their rpm, in hp, kW, Nm and lb-ft, and a run-data table. Two pulls are shown, because they answer two different questions. The best-power pull is the engine at its best — full throttle, best-power mixture, standard conditions — and is what the grade and the objectives use. Your pull is the curve drawn on the chart: the engine as you have it tuned right now. The PDF report tabulates the same pull.

Petrol 4-stroke — the all-rounder: spark ignition, one power stroke every two revs.
Diesel (CI) — compression ignition (no spark, no spark-knock), very high compression, runs lean and unthrottled. Big low-end torque, low redline, and the best efficiency — but heavy, and its emissions signature is NOx + soot (PM), cleaned by a DPF/SCR system.
2-stroke — a power stroke every rev (ported, no valvetrain): high specific output for its size but thirsty and HC-dirty (it burns oil and short-circuits fresh charge).
Rotary (Wankel) — a spinning rotor firing every rev: turbine-smooth, revs sky-high, high power per litre, but thirsty, HC-heavy and hard on apex seals (lower reliability).
Switching type retargets the compression, redline and AFR ranges and disables what doesn't apply (spark for diesel; cam/VVT for ported engines).

Bottom end — engine type, bore × stroke × cylinders = displacement; oversquare & higher redline shift the powerband up.

Bore and stroke. Displacement is bore area × stroke × cylinders, but the shape matters as much as the size. A short stroke (oversquare) means the piston travels less per revolution, so the engine can rev higher before the piston speed becomes dangerous, and a wide bore leaves room for bigger valves. A long stroke (undersquare) gives more leverage and fills well at low rpm.
Mean piston speed is the limit that actually stops an engine revving: 2 × stroke × rpm / 60. Road engines sit at 15–25 m/s at the limiter and race engines reach about 30. Set a redline far past that and the reciprocating parts are no longer a road engine.
Compression ratio sets thermal efficiency — a higher ratio extracts more work from the same fuel — and it sets knock: the more the charge is squeezed, the hotter the end gas and the closer it is to detonating. An NA petrol engine runs 10.5–12.5:1, a turbo engine 8.5–10:1, a diesel 15–20:1.
Layout — inline, V or boxer — decides how the cylinders are arranged and, with the cylinder count, the firing order you see on the LIVE tab. A V or a boxer runs one exhaust system per bank.

Valvetrain — wilder cams move power up top & idle lumpy; VVT restores the bottom end. Valves per cylinder, valve material and springs are now modelled from real geometry rather than assumed. Four valves flow about 22% more than two for the same bore (and five flow no more than four — which is why nobody builds them any more). The app derives the valve diameter your bore can physically fit, works out what that valve weighs, and tells you the valve float rpm — the speed at which the spring can no longer pull the valve back down and it stops following the cam. If that number is below your redline, the engine cannot use the rev range it claims. Fit stiffer springs, lighter titanium valves, less lift, or lower the redline. Big-bore two-valve engines also choke on inlet airflow near the limiter, which is the real reason a big pushrod V8 makes its power low and a four-valve screamer can rev.

Induction — carbs → EFI → ITBs set breathing, metering precision and idle quality.

Fuel & spark — fuel type, octane, injector, ignition type, FIXED vs AUTO (ECU) timing, and the fuel mixture: MANUAL (you set one AFR) or AUTO, a 3D ECU map that runs near stoich at light load and enriches under load/boost for power — just like a real fuel map. In AUTO you can tune the map: tap any cell to select it, then − LEAN / + RICH to shift that load×rpm target (RESET MAP restores the factory grid).

Cooling & electrical — radiator/fan/oil-cooler/thermostat set temps; alternator size and ignition type set charging and the misfire RPM limit.

Air path — turbo (spools with rpm; size/config trade lag vs top-end) or supercharger (boost from idle); intercooler size cools the charge (denser, less knock) with a little lag; exhaust and catalytic converter trade flow vs back-pressure/emissions; nitrous is bottled power.

An engine is an air pump, and most of what separates a good one from a bad one happens in the pipes rather than the cylinder. Three things here are solved rather than assumed, and all three are yours to tune.

The intake runner is a tuned organ pipe. Every time a valve slams shut the column of air behind it bounces, and a well-chosen runner length puts that returning pressure wave at the valve just as it opens again — the engine is briefly supercharged by its own plumbing. Long runners tune low (torque down low), short runners tune high (power up top), and the VARIABLE setting is simply two lengths and a flap, which is all a DISA or a VarioRam ever was. Worth up to about 20–25% of airflow at the tuned speed, and the reason a torque curve has humps rather than being a smooth arc.

The exhaust header does the opposite job. The blowdown pulse races down the primary, hits the collector and reflects back as a rarefaction — a suction wave. Time it to arrive during valve overlap and it pulls the last of the burnt gas out and fresh charge in behind it. Primary length sets the arrival time, which is why headers are cut to a target rpm. The app sizes them from the engine's own flow rather than from a table.

Back-pressure is a tax on everything. Whatever the piston cannot push out, it pays for twice: once as pumping work on the exhaust stroke, and again as residual gas — hot burnt charge left in the clearance volume that takes the place of fresh air. The pumping loop is simply p_exhaust − p_intake, so a free-flowing exhaust is worth real power and a restrictive one is worth less than the back-pressure number alone suggests.

A turbo is a restriction you chose on purpose. The turbine sits in the exhaust and is a nozzle: gas can only leave through it, so manifold pressure rises until the flow fits. That pressure is what spins the compressor — and it is also back-pressure the engine has to push against. Boost is never free; it is bought with exhaust energy and paid for in pumping work. The wastegate is what holds a target: once the turbine can make more power than the compressor needs, the gate opens and dumps the surplus around it. Without one, routing every gram of exhaust through a fixed throat gives 3.5–4.5 bar, not the 0.5–1.5 a real engine runs.

Boost is not free, and the frame has a size. A compressor makes pressure by spinning, and that shaft speed has to be bought with exhaust energy — so the more boost you ask a given turbo for, the later it arrives and the longer it takes to build. The same small frame that is fully lit by 1200 rpm at 0.5 bar does not reach half of 2.5 bar until nearly 3000 rpm. The spool figure quoted in the design summary is for the boost you actually asked for, not a fixed property of the frame.

Each frame also has a rated flow — small 32, medium 55, large 82 lb/min, and multi-turbo configs flow more (twin ×1.85, sequential ×1.55, compound ×2.3). That is the axis a real compressor map is drawn against. Ask an engine to move more air than its turbo is rated for and you are off the map: compressor efficiency falls away, the top end chokes off, and — the one that really bites — the charge arrives hot. Squeezing air heats it by T·(PR^0.283 − 1) / efficiency, which grows faster than the boost does and grows again as efficiency drops. A 2.0 L asking 2.5 bar of a small frame wants 1.7× its rated flow, and the charge reaches the valves at 74 °C where the right frame delivers 58 °C — so it knocks, makes 120 hp less than the medium frame, and is slower round a lap than the same engine at 0.8 bar. Nothing stops you building it; it simply isn't the way to make power.

Lag is real in the 0–100 now. The acceleration run spins the turbo shaft alongside the car and reads torque off a two-dimensional rpm × boost map, because boost is a state with a history: at 4000 rpm a turbo might be at 0.1 bar because you have just floored it, or at full boost because you have been flat for two seconds. A single curve can only describe the second. The car also leaves the line with the turbo already part-lit, which is what brake-boosting is.

One limit left: the lap solver still works from the steady boost available at each rpm, so it feels the spool point move but not the transient. A very laggy build is therefore still flattered slightly by its lap time, though no longer by its 0–100.

Known limit: the turbine is fed by mean manifold pressure. A real one is fed by each cylinder's blowdown pulse, whose peak at low engine speed is far above the mean, so a real turbo gets 20–45% more turbine work at the threshold than this model gives it. That is measured rather than guessed, and it is worth most on a twin-scroll manifold, which this app does not yet reward. A second one: the turbine throat is taken as A/R × the wheel radius, where it should be the radius to the volute throat centroid — further out — so every turbine here is smaller than the frame it names, and the wastegate has always covered for it.

TURBINE A/R is the housing itself, and it is your call. TIGHT spools earlier and costs top end; LOOSE breathes at the top and spools later; STOCK is the frame's usual housing. Which side to take is a design decision, not physics — VW fits the GTI a small, tight turbine for response and accepts the power it costs. The model will not choose for you.

VARIABLE GEOMETRY is the throat as a control, and it is why diesels spool. A fixed housing has one throat area, and that one number is a compromise: tight enough to light at 1700 rpm is far too tight to breathe at 4500. A VGT puts a ring of pivoting vanes in the housing, so the ECU picks the throat at every operating point — shut for a quick spool, open to get out of the way once the target is made. Set TURBINE to VARIABLE GEOMETRY and watch the torque plateau walk down the rev range: a 2.0 diesel goes from full boost at 2575 rpm to full boost at 1800, and its torque peak moves from 2700 rpm to 1900. That is not a bonus applied to diesels, it is geometry — the ring's sweep (about 2.9:1, from 21° to 60° of vane angle) and its top-end loss (the vanes occupy 8–13% of the circumference, so a VGT at full open flows less than the plain housing of the same frame) both come straight out of vane count, thickness and the two end stops.

So why does nothing petrol-powered have one? Two reasons, both of which you can see in the app. The vane ring is the hottest moving part in the exhaust, and nickel alloys creep above about 1100 K. A diesel is quality-governed — always lean — so its gas arrives at the vanes near 880 K; a petrol engine runs rich for power and sends 1235 K. The design panel prints that temperature and tells you when you are over the limit, and the reliability score falls about 19 points when you are. The other reason is simpler: a vane ring, its bearings and an actuator cost about 1.7× a plain housing. Fit one to a petrol engine and it works beautifully and does not last — which is exactly its reputation.

Why boosted engines run low compression and lose timing. Higher manifold pressure means more residual trapped, and residual is hot — it drags the end gas toward detonation before the flame gets there. That is why a turbo engine is built at 8.5–9.5:1 where an NA engine of the same era runs 11:1 or more, and why the ECU is pulling timing long before you see the knock bar move. Push boost up on a high-compression engine and you can watch it happen: the AUTO ignition retards, and the power stops following the boost.

On the VEHICLE tab you drop your engine into a car. Pick a chassis and it arrives on its factory spec — the tyres, wheels, brakes, suspension, aero and gearbox that car would really wear, so a kei car turns up on skinny eco tyres and small steel discs while a hypercar gets wide semi-slicks, carbon-ceramics and a wing/diffuser package. Everything stays editable from there; REFIT FACTORY SPEC puts it all back.

The pictures are drawn from your build. The car at the top of the VEHICLE tab gets its length from the wheelbase, its wheels from the rim and tyre sizes, and its ride height from the suspension. It wears the splitter, wing and diffuser you fit. The driven wheels have green hubs, and a green block marks where the engine sits: in front of the cabin, behind it, or behind the rear axle. The engine at the top of the ENGINE tab follows the form as you edit, before you build: an inline is drawn side on so you can count its cylinders, and a V or a boxer is drawn from the front so you can see the angle of its banks. Its turbos, blower, carbs or throttle bodies, headers, cat, intercooler, radiator and nitrous bottle are all drawn. Saved cars and engines show the same pictures on their GARAGE cards, and on the TRACK map the car runs the lap in its own colours.

The GARAGE tab keeps your finished cars and engines. Name a car at the bottom of this tab and press SAVE CAR (SAVE TO GARAGE) and the whole thing is stored together — the body, every part bolted to it, and the engine inside, down to a hand-tuned fuel map. Name an engine at the bottom of the ENGINE tab and press SAVE ENGINE to keep it on its own: loading it later drops it into whatever car you have then. VIEW on a card opens the spec it will load. Each card shows the headline numbers it had when you saved it (power, weight, 0–100, top speed, lap time and total cost), so you can line three builds up and compare them without rebuilding any of them. LOAD puts a car back exactly as it was; typing a name that already exists highlights that card and saving overwrites it. This is separate from SAVE/LOAD, which is a single quicksave slot that each save overwrites. A garage entry is the same record SAVE FILE writes and a share link carries, so a build is portable between all three. Everything lives in this browser only — clearing site data clears the garage, and SAVE FILE is what makes a build survive that.

From that baseline you tune the drivetrain (FWD/RWD/AWD — AWD launches hardest), the gearbox (type, speeds and the gearing slider), tyres (compound + width/profile/ diameter set grip and rolling radius), aero (front/rear/floor add downforce and drag), brakes (type + rotor size set stopping power, how hot the rotors run and when the pads fade) and suspension (how far the body rolls and the camber it runs). Grip is not a multiplier on any of these: a tyre grips harder the less load it carries for its size, so a bigger tyre, a lighter car and less weight thrown onto the outside tyres in a corner (a low, wide car) all corner faster, and roll that tips the loaded tyre off its best camber angle costs grip.

Gearbox type changes four things: how long the drive is cut on an upshift, how much of the engine's output survives the driveline, how the ratios are spaced, and how the car leaves the line.
· Manual (H-pattern) — the most efficient box, but you lose ~350 ms per shift lifting off, moving the lever and re-engaging. Cheap. Best where there are few shifts to make.
· Sequential — a race box that bangs the next dog in: ~60 ms shifts and close ratios that keep a peaky engine in its band. Expensive.
· Dual-clutch (DCT) — the next gear is already engaged on the second clutch, so shifts are near-seamless (~30 ms). The quickest thing here, and priced like it.
· Torque-converter auto — slowest shifts of the modern boxes (~200 ms) and the worst driveline losses, because it drives through fluid. What it buys you is the launch: the converter stalls up against the brakes and multiplies engine torque (up to ~1.9× at a 2400 rpm stall), fading to 1:1 as the turbine catches up. That's why an automatic can out-launch a manual despite giving away power everywhere else — and why it suits a heavy, torquey car far better than a light, peaky one. It also spreads its ratios wider: a lower 1st to get the load moving, a taller top for cruising.

The final drive is chosen for your engine, not by the slider. A car is geared for maximum speed when the engine sits at its power peak exactly as drag consumes that power — not at the limiter, where it is already past peak and cannot hold the speed. So the app solves the drag balance for the speed your car could theoretically hold, P_peak·η = ½ρ·C_dA·v³ + C_rr·m·g·v, and picks the ratio that puts peak power there. A 3800 rpm diesel and a 9000 rpm screamer therefore get completely different ratios, as they should — before this, both got the same ones, and the diesel ran out of gear at 167 km/h while the air could not have stopped it until 219.
The GEARING slider is now a bias about that — roughly 1.28× shorter at one end for acceleration, 0.74× taller at the other for cruising, and neutral in the middle. Two consequences worth knowing. Top speed has an optimum in the middle of the slider, not at the tall end: gear taller than ideal and the engine can no longer reach the power it needs at that road speed, so the car gets slower. And on a twisty circuit short gearing costs you almost nothing, because you trade a top speed you never see for acceleration you use in every corner.
Where the car changes gear. The 0–100 and quarter-mile runs upshift where the next gear pulls harder at the wheels than the current one — the torque crossover, which is what a good driver and a good automatic both do — or at the limiter if that comes first. For a peaky engine that is right up at the redline anyway. For a diesel it is not: its torque has faded long before the limiter, and holding it there used to cost a 2.0 TDI four seconds to 100 km/h. The lap uses the same rule. One step remains and it is real: if a taller final drive pushes 100 km/h into the next gear, the run makes one more shift and the time jumps by about one shift.

The GEARING slider — moving it shortens or lengthens the final drive about the ratio the app matched to your engine: about 1.28× shorter at 0 and 0.74× taller at 100, neutral near the middle. The resulting ratio is shown next to it. Short gearing multiplies torque at the wheels, so the car accelerates harder and reaches the limiter sooner in every gear; tall gearing does the opposite. Top gear at the limiter is a hard ceiling — the limiter cuts fuel there, so no amount of power pushes the car past the speed its own gearing allows. The final drive is held to 2.2–6.0:1, the mechanical range of a real one.

The differential splits drive between the two wheels on an axle — and decides how much of that axle's grip you can actually use. The two wheels are never loaded identically (road camber, engine torque reaction, axle wind-up, suspension geometry), and an open diff can only feed each wheel what the lighter-loaded one will hold: it spins that wheel and throws the rest away, leaving about 78% of the axle's grip usable at launch. That 78% is not a constant — most of the asymmetry it is fighting is torque reaction (driveshaft wind-up, anti-lift geometry, torque steer), which is proportional to the torque actually going through the axle. In first gear under full multiplication it is at its worst; in third at 100 km/h the axle sees a fraction of that torque and little more than road camber is left, so an open diff recovers toward ~94%. A viscous LSD ties the two together through fluid shear and recovers most of it (~92%); a clutch-plate LSD locks them against a preload for effectively all of it; a spool welds them solid and is the best of all in a straight line, which is why drag cars run one.
The key thing: a diff only matters while you're traction-limited. Fit a spool to a slow car and nothing changes — it never troubles its tyres. Fit one to a car that's spinning up off the line and it can be worth more than a power upgrade. Watch the LIMITED-BY behaviour: if your 0–100 barely moves when you change diff, the engine is the limit, not grip.

Why a front-driver runs out of grip so early. Three things multiply, and the first one surprises people: a FWD car only drives the front axle, which carries about 60% of the weight standing — and accelerating transfers weight off it, down to roughly 53%. Multiply that by an open diff (0.78) and by eco tyres (μ 0.81) and the whole car is capped at 0.33 g, whatever the engine makes. At 100 km/h that ceiling is about 2500 N, and even a 109 hp engine is already asking for 2700 N — so a 217 hp version of the same car reaches 100 km/h at exactly the same time. Both are just riding the ceiling. This is not a quirk of the simulation; it is why nobody builds a 200 hp front-drive kei car. The fix is grip, not power — on that same 217 hp car: a clutch LSD alone takes 9.9 s to 8.0 s, sport tyres 6.6 s, slicks 5.5 s, and switching to AWD 4.4 s. Same engine throughout.

And now the trade is measurable. On the TRACK tab a locked diff pays for its straight-line advantage: it scrubs in every corner (less lateral grip) even though it never wastes drive. On an 842 hp muscle car a spool is ~0.2 s quicker to 100 km/h than a clutch LSD and ~2.5 s slower per lap, while an open diff is worst at both — it spins the inside wheel hardest exactly when cornering has already unloaded it.

The engine's mass and torque curve are added in and the Performance panel computes 0–100 km/h, 400/800/1000 m (time + trap speed), top speed, 100–0 braking and the 0–200–0 test live — a traction-limited launch and a tyre/brake-limited stop with fade, weight transfer, downforce, drag and rolling resistance. Sticky tyres, big brakes and AWD help most.

The TRACK tab puts the car on a circuit and solves a lap. First it works out the racing line: nobody drives the centreline, because using the full width of the track straightens a corner and so raises the speed you can carry through it. The line is found by minimising its total curvature inside the track edges, and out-in-out through corners, straightened esses and linked complexes all fall out of that on their own. Two things bend it away from pure minimum curvature. First, the track is not one width: every circuit here narrows through its tight sections and opens out on the fast ones, and a 7 m corner simply cannot be straightened as much as a 13 m one — the figure quoted on the map is the range, and the circuit's headline width is its average. Second, a late apex: a metre of radius at a corner exit is worth far more than a metre at the entry, because exit speed is carried the whole length of whatever follows while at the entry you are braking anyway. The solver prices curvature by how open the track is ahead, so it gives up entry radius to straighten the exit — moving the apex about 11 m later on corners that lead onto a straight, and leaving corners that lead into more corners exactly where they were. Circuits are three-dimensional, so the lap also accounts for elevation: gravity pulling you back up a climb and pushing you down a descent, the slightly reduced weight-on-tyres of a slope, and — the one that really bites — vertical curvature. Over a crest the car is thrown light and loses grip exactly where it is going fastest; through a compression it is pressed into the road and can carry far more speed. On the test circuits a dip is worth about a third more corner speed than the same corner on a brow. Where the hills fall matters as much as how big they are: a climb in a braking zone lets you brake later and gains time, while the same climb on a straight simply costs you.

Zoom and pan the map. Scroll (or pinch on a phone) to zoom in on any corner, drag to move around, and double-click to fit the whole circuit again. The track is drawn to scale, so zoomed in you can see where the racing line uses the width.

It is worth 5–7% of lap time over driving the centreline, and a wider circuit is worth more. Then it finds the cornering speed at every point (v = √(μ·(mg + downforce)·r / m)), then works backwards from each corner to find the braking points and forwards out of each corner on the available drive, taking the lower speed at every point. A friction circle ties the two together: grip already spent turning isn't available for accelerating or braking, so a car at the cornering limit can't also put its power down. The coloured line on the map shows what is limiting the car at each point, and the legend totals how much of the lap each one accounts for — a fast circuit is mostly power and braking, a tight one mostly traction out of slow corners. Shift time is charged per gear change around the lap, so a slow-shifting box loses real time.

One limit worth knowing: the line is still only an approximation of the minimum-time line. The late-apex bias is priced from the geometry ahead, not from the car's own speed, so the same line is driven by a 90 hp hatchback and an 800 hp aero car — where in reality the fast car would brake later and apex later still. It is worth about half a percent, and it is the same half percent for everyone, so build-to-build comparisons stay honest.

IMPORT GPX TRACK loads a real circuit from a GPX trace — a recorded lap, an export from a mapping tool, anything with track points. The trace is projected to metres, resampled to an even spacing (GPX points are wildly uneven, and a starved spline invents corners) and lightly smoothed so GPS jitter doesn't read as curvature. Elevation comes straight from the file, so an imported circuit gets its real gradients. Imported tracks are marked * in the list and are kept in your browser only — nothing is uploaded, and nothing ships with the app.

On accuracy: an imported circuit's lap time will read a little slow. A GPX trace carries no width at all, so the width profile is modelled from the layout — roads are built wide where they are fast and narrow where they are tight — rather than surveyed. The line is still an approximation of the minimum-time line, and the driver never makes a mistake or takes a compromise line. Treat it as a consistent yardstick for comparing your own builds, not as a real lap record.

HOT LAP REPORT — on the TRACK tab, LAP REPORT PNG / PDF saves the lap as a telemetry sheet: the circuit map with full-throttle and braking sections picked out and every corner tagged, a speed trace against distance with the braking zones shaded, lap and sector times, the full car spec, and a speed trap at every corner — entry, apex and exit speed, gear and radius. Corners are numbered T1…Tn and named per circuit (Riverside's are Millrace, Weir, Ferryman…). "Full throttle" means the car is accelerating at its limit, engine or tyres; the footer also reports how much of the lap is specifically engine-limited.
Bring your own circuit. The TRACK tab imports a GPX trace — the file any phone, watch or lap timer records — and turns it into a drivable layout: the points are projected to metres, resampled to even spacing, the lap is closed if you stopped the clock a few metres from where you started, and the elevation is kept, so climbs cost you power and descents give it back. Corner detection then names the turns T1…Tn. Imported tracks appear in the circuit dropdown alongside the built-in four and survive a reload. A real 20 km trace imports and solves in about a quarter of a second.
Imported circuits have modelled width rather than surveyed width, derived from the layout, and corners cannot carry real names because a GPX file does not hold any.
DYNO REPORT — the REPORT PNG / REPORT PDF buttons save the current pull as a printable sheet laid out like a real dynamometer print-out: the power and torque traces with their peaks called out, a peak-result band, the full engine specification, the conditions the run was made under, and a tabulated run-data list. Any pulls you've logged are overlaid faintly for comparison. The graph is the engine as tuned right now, so the sheet also prints the design potential (the best-power figure the scorecard grades) — that's why the two numbers can differ. Everything is generated offline in the browser; nothing is uploaded.

Every build is graded S → F from seven things: specific power & torque (per litre), powerband width, thermal efficiency, knock safety, driveability (low-end torque) and emit (emissions). Chasing peak power usually costs drivability, safety and emissions — the score rewards a balanced build.

Track objectives judge the whole car, not just the engine: four of them set a lap target on a named circuit (Club Sprint at Ashdown, Grand Prix Pace at Riverside, Under Seven at the Ring on the Nordschleife, and Lap per Dollar for the cheapest car that can still lap quickly). Their budget covers engine plus vehicle, so a cheap engine in an expensive chassis counts against you — build the car on the VEHICLE tab, then watch the objective card.
The scorecard also reports your lap time for the circuit selected on the TRACK tab. It is reported, not folded into the letter grade: the grade rates the engine, and an engine's grade should not move because you fitted different tyres.
Pick an OBJECTIVE at the top of the ENGINE tab to turn free building into a puzzle. Each challenge sets targets — power, economy, emissions, reliability or power-per-dollar — and some cap your budget. The card ticks off each target live (✓/✗) and shows a PASSED verdict when the whole brief is met; the build's cost and reliability are always on show. Cheap displacement wins on value; boost and race parts win on power but cost money and reliability. Balance the brief.

ENGINE SOUND — a synthesised engine note pitched to the firing frequency; a wild cam makes the idle lope.
WEAR & DAMAGE — with wear on, detonation, overheating, running lean under boost and over-revving damage the engine and can blow it; with it off the engine is invincible, which is the way to experiment.
LANGUAGE — English or Greek. Pages added since the Greek translation was made appear in English.

Tip: turn WEAR off in OPTIONS to experiment freely, then back on for consequences.

This is a physics model, and like every model it leaves things out. These are the ones worth knowing, because they can move a result you care about:
· The VE curve is still partly fitted for some engines. Petrol engines with poppet valves get their breathing shape from the parts: valve timing, the tuned intake and exhaust, port flow and valve float. Diesels, rotaries and two-strokes still use a fitted curve.
· Turbines are fed by mean pressure. Real ones get 20–45% more work from each cylinder's exhaust pulse at low rpm, so turbos here spool a little later than real ones, and twin-scroll manifolds are not yet rewarded.
· The lap uses steady boost. The 0–100 integrates turbo lag; the lap solver does not, so a very laggy build is slightly flattered on track.
· The racing line is an approximation of the minimum-time line, and the driver never makes a mistake.
· Imported circuits have modelled width, because a GPX file carries none.
Each of these is measured and written up in the project, and each one is on the list.