Ignition angle is the crank position at which the spark is commanded, almost always a number of degrees before top dead centre. The plug, the coil and the dwell decide whether a spark happens. The angle decides when it happens. This page is the calibration side of that. The hardware, including spark plugs, laser and pressure-sensing plugs, magnetos, distributors, wasted spark and coil-on-plug, is on Ignition systems and spark plugs.
What the angle is doing
The mixture does not burn the instant the spark jumps. A kernel has to grow, the flame has to cross the chamber, and cylinder pressure has to rise while the piston can still use it. Brake torque is the result of that pressure acting through the crank. Teaching material on MBT puts the useful peak of cylinder pressure roughly 14 to 18 degrees after top dead centre: late enough that the crank has leverage, early enough that the gases are not expanding into a piston that has already run away down the bore. HP Academy uses about 16 to 18 degrees in that same band. Those figures are a pressure-phasing target. They are not a spark-advance value to type into a cell.
MBT means both minimum advance for best torque and maximum brake torque. It is the smallest spark advance that produces the peak torque at that site. More advance than that does not add torque. It moves the pressure peak earlier, spends work compressing the burning gas, and usually raises knock and NOx. Less advance than MBT leaves the burn finishing late. Torque falls, exhaust temperature rises, and hydrocarbons eventually rise because the burn is still going when the exhaust valve opens.
Indicating work often watches the crank angle of 50 percent mass fraction burned, CA50, instead of the spark angle. A CA50 near 8 degrees after top dead centre is a common research target for combustion phasing. A Michigan State paper on closed-loop MBT control points out that peak-pressure location and CA50 are observed correlations. They can shift with operating point, so they still have to be checked against torque on that engine. A pressure trace that lands in the target window tells you the spark angle was right. It does not give you the spark angle without the burn duration of that cylinder.
Why the angle has to change
Combustion takes time. The crank measures angle. Anything that changes the burn time, or how many degrees pass during that time, moves the spark if the pressure peak is to stay in the same place. Knock then takes away the angles you would otherwise use.
| What changed | Effect on the burn | Effect on knock | Usual direction of the spark angle |
|---|---|---|---|
| Higher rpm | Similar burn time covers more crank degrees | Depends on the engine | More advance, to hold the pressure peak |
| Higher load | Denser, hotter charge burns faster | End-gas is more likely to auto-ignite | Less advance. At high load, knock often arrives before MBT |
| Lean of stoichiometric | Slower burn, then misfire | Usually less knock while the burn is stable | More advance until the kernel will not grow |
| Richer than stoichiometric | Slower burn, cooler combustion | Less knock | Often more advance is available. Enrichment is also used as a protection, not as a timing method |
| EGR or more residual | Diluted, slower burn | Less knock | More advance until the burn becomes unstable |
| Hotter intake air or coolant | Slightly faster burn | More knock | Less advance |
| Higher octane | Little change in flame speed | Less knock | High-load sites can move closer to MBT |
| More compression or boost | Faster burn | Much more knock | Often a long way retarded from MBT. This is knock-limited spark advance |
| Different cam timing or intake length | Changes residual, turbulence and effective compression | Can go either way | The old timing map does not transfer. Sweep it again |
| Cold start and catalyst heating | A late burn puts heat in the exhaust | Low at idle | Deliberate retard. The M112 warm-up description uses about 5 to 10 degrees of retard for this |
| Gearshift, traction or a torque limit | None. The burn is being spoiled on purpose | Not the reason | The controller retards. That retard is not the base map |
The surface that falls out of the first two rows is the familiar one: more advance as rpm rises, less advance as load rises. Real maps bend away from that wherever a cam switch, a variable intake, a lean cruise region or a knock limit cuts across it. A smooth table is the aim. A cell that sits several degrees away from both neighbours is a hole or a spike unless the engine really does change there.
MBT and the knock limit
At light and medium load the end-gas usually survives long enough for the flame to finish the burn. Those sites can sit on MBT. Torque stops rising, knock is not the reason you stopped, and further advance only adds NOx and heat in the piston.
At high load, and on almost any boosted site, the end-gas auto-ignites before the flame arrives. That is knock. The pressure spikes are fast enough to damage ring lands, bearings and the plug insulator, and they are not the same thing as a plug that is merely too hot. The most advanced angle that does not knock is knock-limited spark advance. EFI teaching material states the practical result directly: at wide-open throttle, KLSA rather than MBT is the ceiling on most engines.
A knock sensor, or a pressure plug, is what tells you that you have arrived. The ear is late. By the time knock is obvious in the cabin, the calibration is already past a sensible margin. The sensor is also not the timing table. A map that runs in continuous knock retard is using the protection as the calibration, and it will be wrong the moment the sensor, the fuel or the intake temperature changes.
Settle the charge before you move the angle
Timing is the phasing of a charge that already has the right density and the right strength. Fuel, airflow and boost come first.
- A wrong crank offset shifts every cell by the same amount. Check top dead centre and the controller's offset with a timing light at a known fixed advance, or against a locked-out reference, before treating the map as real.
- Log engine speed, load, delivered advance, lambda, knock retard, intake temperature, coolant temperature, and any per-cylinder trim.
- Hold lambda steady at the site. A lean misfire feels like a site that wants more advance. Adding advance to a lean, boosted cylinder makes the knock worse. A large retard used to hide a lean or over-boosted condition moves the heat into the exhaust and can damage the engine thermally instead.
- Confirm the knock path works. On one controlled pull, a degree or two of extra advance should produce retard or a pressure oscillation. A system that never counts is not a quiet engine. It is a deaf one.
Finding the angle on a load-holding dyno
MBT is found by a spark sweep: hold speed and load, move only the advance, and record torque. HP Academy's practical account is that this needs a load-bearing dyno. An inertia dyno, or a road pull, does not hold the site still, so the torque change is mixed up with the speed change. The Michigan State paper describes the same traditional method, a sweep, and treats it as the measurement that online pressure criteria still have to be checked against.
- Pick a warmed-up site in the area you actually use. Start where the engine is unlikely to knock, so the shape of MBT is visible before the knock limit hides it.
- Step the advance in about one to two degrees. Wait until torque, lambda and temperatures settle. Record the delivered advance, not only the cell you edited. Temperature trims, knock retard and torque intervention may already have moved it.
- Stop at the first of two events: torque stops rising, or knock becomes consistent. If torque is still climbing and the cylinder is quiet, you have not reached MBT. If knock starts while torque is still climbing, the site is knock-limited. Write down both the knock threshold and the torque, so a later fuel or compression change has something to be compared with.
- Leave a margin under the knock threshold. The margin has to cover hotter intake air, the lowest octane the car will be given, and the worst cylinder, which is often not the one the knock sensor hears first. There is no single number of degrees that suits every engine. A margin that disappears on a hot day was not a margin.
- Repeat across enough speed and load points to see the surface, including both sides of any cam, intake or boost threshold. Then look at the table. Fill holes. Knock down spikes that have no mechanical reason.
- Run the same high-load site again at the intake temperature the car will see in traffic, not only at the temperature of a cool dyno cell.
Part-throttle cruise can usually be left on MBT. That is also where a little extra advance shows up as response, and where too much advance shows up as surge and NOx. Wide-open throttle on a boosted engine is where the sweep stops for knock, not for torque.
On the road
The road is where you find out whether the margin survives a hill, a hot day and a tank of ordinary fuel. Log knock retard, delivered advance, lambda and intake temperature. A site that spends the pull in retard needs to come back in the base map, or the cause is boost, mixture or temperature rather than the cell itself. Drivability, a flat spot, or a hesitation when the throttle opens is often a hole in the transient between two sites, not a reason to add five degrees everywhere.
A single acceleration run does not measure MBT. Torque is changing because the engine is accelerating. Use the road to confirm the dyno map, and to catch continuous knock retard. Do not use it as the sweep.
Cylinders do not all want the same angle
A wasted-spark pair shares one coil and one firing event, so those two cylinders get the same angle whether or not they knock together. Sequential coil-on-plug can trim a cylinder. The trim belongs to the cylinder that knocks or that makes less torque, not as a global addition that then has to be removed from the quiet cylinders.
A measuring spark plug, or a dedicated cylinder-pressure sensor, shows one cylinder's pressure against crank angle. From that trace you can see indicated mean effective pressure, where the peak sits, how fast the burn was, and the high-frequency spikes of knock. Kistler's measuring plugs are made for this when the head cannot be drilled. The method is on the hardware page. One cylinder's indicated torque is not the brake torque of the engine, but it will show a knock event and a late burn that the crank damper averages away. The Michigan State work is one published route from that trace toward closed-loop MBT. It still expects the criterion to be checked, not copied from another engine.
Angles that are not the combustion map
Cold start and catalyst light-off want a late burn and hot exhaust. That retard is deliberate. Taking it out to make the idle sound sharper slows the catalyst and raises the emissions the five-gas section describes. On a five-gas analyzer, more advance raises NOx, a small retard cuts NOx, and a large retard drops NOx further, raises exhaust temperature and then raises hydrocarbons. Sample before the catalyst if you are judging the burn. The patterns are in the ignition-systems article.
Torque, gearbox and traction intervention retard the spark to throw torque away quickly. If the log shows retard only during a shift or a traction event, the base map is not the fault. Correcting that retard in the main table puts the torque back into the shift.
A controller that already has a knock-learning map will walk the timing back in the regions where you were too close. Read the learned correction before deciding the base cell is what the engine ran.
A Bosch example, not a universal map
The names of the tables change with the controller. On Bosch ME7 the delivered angle is base timing, plus temperature and operating-state corrections, plus knock correction, plus a torque intervention, and the optimal-timing tables used by the torque model are not the same thing as the base map. That split is written up in the ME7 ignition and knock reference. Use it as an example of how one family separates those jobs. Do not copy its cell values, its axes or its cam-switching angles onto another ECU.
Sources
The phasing targets and the sweep method follow the cited teaching and research descriptions. A number from one engine is not a specification for another. This wiki has not repeated the sweeps on a vehicle.
- HP Academy, What is MBT? MBT varies with speed and load, and knock can arrive before MBT. Accessed 2026-11-11.
- HP Academy, MBT tuning technique. Peak pressure about 16 to 18 degrees after top dead centre, more advance with rpm, less with load, and a load-bearing dyno for the sweep. Accessed 2026-11-11.
- EFI Mastery, MBT timing. Peak-pressure band about 14 to 18 degrees after top dead centre, and knock-limited spark advance as the high-load ceiling. Accessed 2026-11-11.
- Zhu, Haskara and Winkelman, MBT timing detection and closed-loop control using cylinder pressure. Spark sweep as the traditional measurement, and the limits of peak-pressure and CA50 criteria. Accessed 2026-11-11.
- Szybist and Splitter, knock at high compression. An example of CA50 held near 8 degrees after top dead centre as a phasing target, with spark then delayed to the knock limit as load rises. Accessed 2026-11-11.
- Tuning Wiki, Ignition systems and spark plugs. Plugs, coils, dwell, the M112 warm-up retard, and the five-gas patterns.
- Tuning Wiki, ME7 ignition and knock reference. One controller's split between base timing, corrections, knock and torque intervention.