Hondata KPro and FlashPro fuel tables on a K-series

This procedure covers the port-injected engines below. The year gets you close. The ECU, the calibration and the installed hardware settle the rest.

The 2005–2006 RSX conversion uses a 2002–2004 RSX or 2002–2005 Civic Si primary oxygen sensor. The original ECU cannot accept KPro. An adapter alone does not convert an unmodified earlier ECU.

The 2013–2015 Accord K24W is a direct-injection engine and does not use the port-injector model on this page. No flow, dead-time or VE recipe here applies to it. K20C, L15, B/D and s300 work sit outside this wave, and KTuner is not covered.

FuelTech on a K has its own fuel model and its own page.

KPro calculates fuel from engine speed and manifold pressure. The fuel table is split by VTEC state and then by intake cam angle. The table window shows low-speed and high-speed fuel tables, each with six cam-angle tables at 0, 10, 20, 30, 40 and 50 degrees.

The ECU interpolates between breakpoints, so the active value is a blend unless the engine sits exactly on one.

• The number in a KPro fuel cell is the calibration's fuel value. Do not read it as a VE percentage that can be compared with another file. Apply a percentage change to the selected cells, then use the log to see whether that percentage moved lambda toward the target. The same number in a different calibration does not describe the same mixture.
• FlashPro can run an AFM calibration, a MAP calibration or a hybrid of the two. Fuel Table Type in the calibration says which. AFM systems calculate injector duration from the mass of air per cylinder and do not use the MAP fuel tables. MAP systems use volumetric efficiency tables indexed by engine speed and manifold pressure. On a hybrid, the ECU uses AFM below the Minimum pressure for speed/density and MAP at or above that pressure.
• That threshold is a load boundary, not a tuning preference. Above the switch, an AFM Flow edit does not control fuel. Below the switch, a MAP fuel cell does not control fuel. Check the type and the threshold before you decide a table had no effect.

Swipe or use the arrow keys to read every label.

• At 24 degrees actual cam angle, KPro blends the 20 and 30 degree tables. Editing only one changes part of the delivered fuel. Match both VTEC state and actual cam angle to the sample.
• FlashPro's part-throttle procedure gives two five-angle examples: 0, 10, 20, 30, 40 degrees or 0, 15, 30, 40, 50 degrees. Read the breakpoints from the calibration instead of assuming KPro's six-table layout.

The ECU estimates air and selects or blends a fuel table. It calculates a base injector duration, applies the injector model and then applies the target or feedback correction. A mistake at any stage looks like a fuel-table error at the others.

A fuel sample is only useful when the operating state is stable. Record rpm, manifold pressure, throttle, actual cam angle, cam target and VTEC state. Add lambda, short-term trim, long-term trim, injector pulse width and battery voltage. Include intake air temperature and coolant temperature. Add fuel pressure if the car has a sensor for it. The Honda logging guide covers the export and channel definitions.

• A useful trace holds rpm and load steady through the cell and holds the cam angle steady. Battery voltage stays stable, and the closed-loop state is known. Lambda should sit near the target without the trim swinging. A bad trace includes throttle tip-in or tip-out, injector cut, start-up, sensor warm-up, misfire, a pressure drop, or a moving cam angle. Those samples describe a transition, not a steady cell.
• During confirmed overrun fuel cut, lambda near 2.0 describes oxygen in the exhaust, not a mixture error. Confirm zero injector pulse before excluding the sample. Do not add fuel to correct a sample taken during fuel cut.
• Lambda on target with +15% short-term trim means feedback is adding fuel. Establish the injector, pressure and sensor model before correcting the base table. Apply a trim-aware overlay once; adding STFT again would count the same correction twice.

The injector model has to match the hardware before a fuel correction means anything. A table tuned around wrong dead time will hold a mixture at one voltage and drift at another. A table tuned around the wrong test pressure will be wrong as soon as the rail pressure changes.

1. Record the exact injector. Part number, flow, test fluid, test pressure, impedance and the latency points by voltage. Stop if the only number you have is a seller's nominal flow.
2. Establish the pressure reference. With an adjustable manifold-referenced regulator, set base pressure using its specified pump-running procedure, with no vacuum or boost on the reference port. A stock returnless system has no regulator to adjust; measure what it does rather than treating a swap-guide pressure as its specification.
3. Enter flow and latency in the right fields. In FlashPro, the current pressure and the injector test pressure are separate inputs. Entering both and then adding a global trim for the same pressure difference applies the correction twice.
4. Check the sensors that judge the result. Confirm MAP at key-on and the oxygen sensor scaling before you calculate a fuel error. A wrong pressure or oxygen reading makes a correct table look wrong.
5. Warm the engine and log idle and steady cruise. Include battery voltage and both trims. An error at every load suggests model or global scaling. An idle-only error that tracks voltage suggests opening time, an air leak or injector nonlinearity.
6. Load the electrical system at a steady rpm and load. Compare lambda and trim with the alternator charging and with accessories on. Only then move to the spatial fuel table.

Keep injector rating pressure separate from rail pressure. FlashPro has stock/current pressure and injector test-pressure inputs. KManager's size adjustment does not characterize pressure-dependent latency or short-pulse behavior; retain the pressure assumptions supplied with the calibration.

• Aftermarket data has to travel with its rating pressure. The ID1050-XDS is rated 1065 cc/min at 3.0 bar with gasoline at 52 degrees Celsius. The DeatschWerks 17U-08-1000-4 is a 1000 cc/min injector. Its Hondata-format latency data is published at 51 psi, or 3.5 bar, so do not treat the whole sheet as a 3 bar characterization.
• Pressure changes flow by the square root of the differential-pressure ratio: Q2 equals Q1 times the square root of dP2 divided by dP1. A 1000 cc/min injector at 3.0 bar becomes about 1080 cc/min at 3.5 bar. That assumes incompressible flow and an otherwise unchanged model. It is arithmetic, not a recommendation to raise pressure.
• Differential pressure is rail pressure minus manifold pressure, with both on the same reference. For example, a fixed 50 psi gauge rail at 10 psi boost gives 40 psi across the injector. Flow relative to 50 psi falls to the square root of 40/50, or 89.4%. A working 1:1 boost-referenced regulator instead raises rail pressure with boost to preserve the differential.
• Displacement is a sanity check, not a correction. A 2354 cc K24 is 1.178 times a 1998 cc K20, so the same breathing would need 17.8% more fuel. The head, cams, manifold and rpm change the shape as well as the scale. Do not multiply a K20 table by 1.178 and call it a K24 calibration.

Fix the cam at one safe angle, tune that table, then repeat for the angles and VTEC states the engine uses.

1. Save the file and record the composite cam targets. You will restore them after the fuel work. Keep the as-found file untouched and work from a copy.
2. Establish that the valvetrain has clearance for the angles you will command. A software angle is not a clearance measurement. The VTC and VTEC guide covers the mechanical check and the cam target tables.
3. Start on the low cam at one angle. Lock the low-speed cam table to that angle. Log actual cam, target, rpm, MAP, lambda, trims, injector pulse, battery voltage, IAT, ECT and VTEC state.
4. Collect steady samples across the region you intend to change. Exclude tip-in, tip-out, fuel cut, start-up, warm-up, misfire and any sample where the cam is still moving.
5. Choose one correction method. A zero-trim open-loop test calculates measured lambda divided by target lambda. A trim-aware overlay uses the trims and lambda together. Do not mix the two methods inside one correction and then attribute the result to the table.
6. Edit the table at the locked angle. Unselect Edit All Tables for an angle-specific correction. If actual cam is between breakpoints, both neighboring tables contribute; collect a settled fixed-angle sample before isolating either one.
7. Re-log the same region. Check the cell you changed and the neighbors. Repeat at the safe angles used on each VTEC lobe. Restore the composite targets, save and upload the tuned file, then check the transitions.

KManager's Fuel Adjustment overlay is calculated from the ratio of actual lambda to target lambda. The software notes that the estimate depends on many factors and is intended to show the regions that need work. Treat the overlay as a direction, then verify the result in the next log.

• Ctrl+J opens Adjust Selected Values for the selected cells. Fuel edits applied across cam tables are always percentage changes, even when the edit in the current table is an absolute value. That behavior matters when Edit All Tables is selected.
• Edit All Tables changes all related cam-angle tables together. It is useful when the tables are similar and the correction is a broad shape change. It is wrong when the error is local to one angle, because it moves five tables you did not log. Check the box state before every apply, not after.
• For a simplified blend example, let matching 20-degree and 30-degree cells both contain 100 calibration units. At 24 degrees, their weights are 60% and 40%. Adding 10% only to the 30-degree cell gives 0.6 × 100 + 0.4 × 110 = 104, a 4% blended increase. These are invented values to show interpolation, not cells to enter. Re-log with the composite cam targets restored.
• Stop the session if a change produces a lean condition you cannot explain, a new misfire, or a fuel pressure drop. A fuel table is not the place to diagnose a failing pump or a blocked filter.

Use the procedure for the active model: MAP edits volumetric efficiency; AFM edits the voltage-to-flow transfer. A hybrid needs each region checked separately.

MAP calibration

The part-throttle MAP procedure can run in closed loop. The software reads the fuel trims and the measured lambda to calculate the fuel change, so forcing open loop is not a prerequisite. That is a difference from the older KManager instruction to zero the short-term trim limits.

1. Confirm Fuel Table Type and the cam breakpoints. Write down the angles the calibration uses. Do not assume the KPro 0-to-50 set.
2. Select the low-speed cam fuel tables and confirm the cam angle. Unselect All so the percentage does not land on every cam table. Lock the cam to the first angle and save the test calibration under a different name. Keep the original cam-angle table for restoration.
3. Record a light-throttle pull in first and second. Hold a steady throttle so load stays consistent as rpm rises. Start near 1000 rpm and stop about 1000 rpm past the VTEC point. The first and last rows often carry tip-in and release error.
4. Show Fuel Change with Shift+F3. The overlay uses lambda, the target and fuel trims to suggest a percentage. Look for a pattern across a column or a block of rows.
5. Apply the pattern, not single cells. Keep the 2D load lines smooth and evenly spaced. If the suggested change would make a bump, apply less to a larger area.
6. Repeat for the same angle until the suggested change is close to zero. Then lock the next safe angle and repeat. Finally, copy only the original low-speed cam-angle table into the tuned file. Save and upload that tuned file; uploading the whole original would discard the fuel work.

The documented method can take three or four cycles per angle. Recheck the 2D shape after every correction.

AFM calibration

The AFM procedure starts from stock injectors, stock AFM flow values and stock WOT compensation values. If those are not the baseline, the trim pattern is describing the previous changes as well as the sensor transfer.

1. Warm the engine and record 5 to 10 minutes of varied driving. Keep the log in closed loop and include steady cruise, light acceleration and some deceleration.
2. Open the XY graph from Advanced Graphs. Put AFM.v on the X axis and S.TRIM on the Y axis. Show the mean, select closed loop and exclude open loop.
3. Read the trend by voltage band. A band that consistently pulls fuel is reporting more air than the engine is using. A band that consistently adds fuel is reporting less.
4. Change a small number of bands at a time. The software's own procedure limits this to two or three changes, then a re-test. Keep a copy of the original transfer.
5. Re-log and check the same bands. The aim in the procedure is trims within 1%. If a band gives inconsistent trims on repeat tests, check temperatures, learned trims, leaks and sensor signal before changing it again.

• If the airflow estimate is too high, the ECU delivers too much fuel and feedback goes negative; reduce the relevant AFM Flow values. An under-reading estimate needs the opposite change. Verify direction with one small edit and a repeat log before extending it across more bands.
• Do not calibrate changing turbulence into the transfer. Check intake leaks, sensor installation and repeatability when the same voltage produces inconsistent results.

Fuel both lobes before you judge the crossover. The high-cam table can be wrong. The phaser may still be moving toward its target. Sensor transport delay and a pressure change can also shift the reading. A one-cell enrichment spike at the switch treats the symptom and leaves the cause.

None is a target for your engine unless the row says so.

With zero feedback correction, measured lambda 0.98 and target 0.90 require 0.98/0.90 = 1.0889 times the existing fuel. A hypothetical cell of 100 becomes 108.89. For a rich example, 12.43/13.10 = 0.9489, so remove 5.11%.

Both AFR readings must use the same display scale. These examples assume stable air, pressure and injector behavior.

A gasoline AFR display is lambda times 14.7. An ethanol blend changes the physical ratio at stoichiometry, but a gasoline-calibrated gauge can still show 14.7 at lambda 1. Work in lambda when the fuel is not gasoline, and keep the display convention in the note.