E85 and flex fuel on a K20 or K24 with Hondata and FuelTech

This page covers the port-injected engines below. The year gets you close. The ECU, the calibration and the fuel system settle the rest.

The US 2013–2017 Accord 2.4 K24W is direct injection and uses a different fuel system. No port-injector flow, dead time or flex recipe on this page applies to it. K20C, L15, B/D and s300 work sit outside this wave.

KTuner is not covered here, and FuelTech on a K has its own setup page.

The current US pump range is 51% to 83% ethanol. The fuel specification uses the same window. That is the range the pump can deliver, not a measurement of the tank you just filled.

• The blend moves stoichiometry, energy per gallon, vaporization and knock margin. It does not carry one fixed octane number. No exact content-to-AKI conversion is established here, and no fixed AKI rating is printed.
• Lambda is actual air-fuel ratio divided by the fuel's stoichiometric ratio. A wideband calibrated to gasoline displays lambda times 14.7. At lambda 1 it can show 14.7 on E85 even though the physical ratio is lower. The display is a scale, not a fuel identity.
• Take a display of 11.8 on the gasoline scale. Divide by 14.7 and the sample is lambda 0.803. If the blend's stoichiometric ratio is near 9.8:1, the physical ratio is near 7.9:1. The same 11.8 display on gasoline is 11.8:1 physical. That is why “11.8 AFR on E85” needs the scale named.
• Work in lambda, or write down both the display scale and the assumed stoichiometric ratio. The fuel table pages use lambda for the same reason. A gasoline-scale target can represent the same lambda on either fuel. Changing the display unit does not change fuel delivery or calibrate an ethanol blend.
• A blend estimate is arithmetic. Ten gallons of E70 plus five gallons of E10 gives (10 × 70 + 5 × 10) / 15 = E50 by volume, assuming no other contents. That is a planning number for the tank. The rail sees the blend only after the fuel has circulated. Compare the sensor with a fuel sample measured independently. A matching pump label is not that check.

Three routes look similar on a parts list and behave differently at the ECU. KPro4 reads a frequency. FlashPro ECT2 reads a conditioned 0 to 5 V signal. FlashPro CANFlex reads content over CAN. FuelTech uses its own flex input and FTManager setup.

A sensor that suits one route may not suit another.

On KPro4, the sensor is a digital input. The documented sensors produce a pulse-width signal whose frequency runs from 50 Hz at 0% to 150 Hz at 100%. The input has a built-in 2.4 k ohm pull-up, so no extra resistor is wired between output and supply.

The sensor belongs in the return line from the rail to the tank, where fuel flows past it whenever the pump runs. Flow direction does not matter.

• For this KPro4 installation, confirm content with fuel flowing. With the pump stopped, the displayed value is not a fresh tank measurement. KPro4 also holds the last content value for cranking. At restart, that value can be the old tank until the pump has circulated the new fuel. KManager describes an update after 10 to 15 seconds of running in normal conditions.
• The frequency convention is linear in the documented range. Subtract 50 from the frequency to get ethanol percent: 120 Hz is 70%, not 120%. An out-of-range value needs diagnosis and a check of the ECU's actual response. Do not assume every ECU recognizes every sensor fault.

Swipe or use the arrow keys to read every label.

• The documented FlashPro examples show the limit. A Zeitronix ECA is listed at 0.34 V for 0% and 4.97 V for 100%, and it will not read below 15% ethanol. An Innovate ECF-1, MTX-D or ECB-1 is listed at 0.00 V and 5.00 V, and it will not read below 1.5%. Those are examples for those units, not a universal transfer table. Calibrate the transfer you actually have.
• CANFlex also exists only on some vehicles. There is no blanket support for every older K. Open the calibration and look for the ethanol parameters before assuming the route exists. The 2006–2011 Civic Si feature list includes an ethanol input. If the menu is missing, check the application, software and calibration before assuming it needs a fixed-blend tune.

Start with the calibration identity. KPro4 native flex is documented for PRB or PRA. A PNF EP3 calibration does not meet that requirement. ECU case identity alone does not establish the calibration family. KPro3 and earlier do not have this input.

• The reference blend is a tuning decision. If the car already runs on high ethanol, use that as the reference deliberately. Enabling a gasoline-based enrichment table on an E85-referenced file can add a second correction on top of the first.
• On KPro4, Ethanol content input enabled enables both the input and the compensation tables. A monitoring-only setup still has to define those tables so they do not add fuel or spark. Decide what the reference is before the first start.

1. Establish the low-blend baseline. Record the calibration, injector data, pressure, pump supply and a lambda log on the fuel the car runs now.
2. Check the hardware gate. KPro4, PRB or PRA, and a calibration that exposes the Ethanol menu. Save the as-found file before the edit.
3. Install the sensor in the return line. The documented GM sensors use a 9 to 18 V supply, ground and an output to the KPro4 ethanol input. Confirm connector orientation before wiring.
4. Enable the input and inspect every compensation table. Fuel, ECT, cranking, cranking ECT, ignition and boost. Write down the reference and the current values.
5. Compare the sensor with a known sample. Check it first at idle with fuel circulating. Resolve discrepancies before checking the reading under load.
6. Tune running fuel at the endpoints first. Low blend and high blend, then intermediate blends. Use lambda and fuel pressure, not the pump label.
7. Calibrate cranking and ECT separately. Repeated genuine cold soaks. Running fuel compensation does not cover cranking.
8. Set ignition from torque and knock. After fuel is correct. One region at a time. No fixed adder.
9. Add boost only if the fuel system has capacity. The ethanol number does not create pump or injector headroom.
10. Test the failure behavior and archive the result. Record unplugged-sensor behavior, retained content, refuel transition and the exact software build. Do not assume an automatic fallback exists.

The ECT table is applied on the basis of 100% ethanol. A table value of 20% at E100 contributes 15% at E75. That is an example of proportional application, not a cold-start value to copy. The same logic applies to the cranking ECT table.

• Smoothing is a real tuning variable. A factor of 1000 can take about 10 minutes for a full-scale change. A factor of 10 responds almost immediately. The content after a refuel is described as updating after 10 to 15 seconds of running. A stable but stale reading can be more dangerous than a fault because it looks valid.
• The maximum engine speed for update must stay above fast idle. If it is set below the speed the engine actually runs, the content never updates. That is a configuration error, not a sensor failure.

FlashPro ECT2 needs a sensor and an analyzer with a 0 to 5 V output. CANFlex supplies content over CAN instead. Both routes are application-specific; confirm the calibration and hardware support.

• On ECT2, the signal wire goes to the ECT2 pin and ECT2 is disabled in the miscellaneous parameters. Verify the ground reference and the transfer at both ends. The ECU pull-up can affect the low-content reading, which is why a 5 V full-scale match does not prove the low end is correct. The Zeitronix and Innovate examples above show how different the low-end limits can be.
• On CANFlex, content arrives over CAN without repinning the ECT2 signal. Install the CANFlex hardware and sensor using the instructions for that vehicle. The hardware and the software build still have to support that application. A CANFlex menu on one chassis does not prove it exists on another.
• The compensation structure is not two complete files blended by a straight line. Its generic help names Ethanol fuel compensation, Ethanol ignition multiplier, Ethanol ignition compensation, Ethanol WOT fuel compensation and Boost by ethanol limit. Table availability varies with the application. Where supported, advanced ethanol replaces the simple 2D fuel and boost tables with 3D fuel and boost tables. The ignition result is the compensation value multiplied by the multiplier, then added to commanded timing.
• Do not import K20C direct-injection WOT targets or starting tables into a port-injected K. The fuel system, the injection timing and the ethanol behavior are different. The port-injection fuel page covers the active AFM, MAP or hybrid model; that model still has to be right before the ethanol correction means anything.
• If the loaded calibration has no usable ethanol control, resolve compatibility before attempting flex operation. A dash sensor alone does not supply automatic compensation. Any interim fixed-blend calibration must match the measured fuel.

FuelTech's setup sequence is explicit. Enable flex under Map Options. Choose the failure content under Other Functions and Flex Fuel. Run the Flex Fuel Wizard to set the reference from the current fuel. Then calibrate the corrections in order.

• The corrections are main fuel injection, engine start and prime pulse, acceleration enrichment, O2 closed-loop target, main ignition and wastegate pressure. They are separate jobs. A running fuel correction does not fix a prime pulse, and an O2 target change is not an air-model correction.
• Record the base map's actual reference blend before running the wizard, and keep a copy. If the base map is an ethanol reference, the low-content correction can be negative. If it is a gasoline reference, the high-content correction is usually positive. The sign is relative to the base. It is not proof that the fueling is correct.
• The failure content is a deliberate choice. An E0 fallback can be conservative for spark and still underfuel a real high-ethanol tank. A high-ethanol fallback can protect fuel and be too much advance for gasoline. Decide which failure is safer for the build, test it at low load, and make fuel, spark and boost fallbacks agree.
• FT450 runs flex fuel and does not drive the throttle. FT550 and FT600 add drive-by-wire. The ECU choice is on the comparison page.

Ethanol needs more volume for the same mass of air. The exact increase depends on the blend, the fuel density, the stoichiometric ratio and the base fuel. A 30% figure is an example, not a calibrated requirement for every E85 tank.

• If measured demand rises 30%, a first estimate takes 75% duty to 97.5%, or 80% to 104%. Neither leaves usable reserve; an injector cannot deliver above 100% duty. This assumes unchanged differential pressure and ignores dead time, so use the actual pulse width and pressure to check capacity.
• Hold air mass and lambda constant. Changing assumed stoichiometry from 14.7 to 9.8 requires 14.7 / 9.8 = 1.50 times the fuel mass. Volume demand also depends on density: multiply that result by baseline density divided by new-fuel density. This is why the 30% duty example cannot set an ethanol compensation table.
• Read the injector size and the pressure it was rated at. Read the fuel pressure, the duty cycle, the pump voltage and the rail differential pressure. Fuel temperature, filter condition and line and seal compatibility matter with ethanol. An ethanol sensor does not supply more flow.
• The ID1050-XDS is rated at 1065 cc/min at 3.0 bar with gasoline at 52 °C. That is a flow rating at a test pressure. It is not an E85 horsepower rating, and it is not a recommendation for a particular build. The same injector can be enough on one engine and short on another.
• Do not change the stoichiometric ratio and then scale the injector flow by the same ethanol fraction. That double-counts the same correction. Choose one reference and one correction path, then measure lambda. If the base fuel model is wrong, fix the model before adding an ethanol multiplier.
• If a running-fuel compensation already adds 20% and testing calls for 35% total, change +20% to +35% in that path. Adding a separate 35% multiplier produces 1.20 × 1.35 = 1.62, or 62% total. The intended increase from the present quantity is 1.35 / 1.20 = 1.125, or 12.5%. These are arithmetic examples, not fuel settings.

Ethanol vaporizes poorly compared with gasoline. It has a higher latent heat of vaporization and a higher flash point. A cold engine can need more start fuel than the running correction supplies. KPro4 has separate cranking and cranking ECT tables. FlashPro has its own starting-fuel controls.

FuelTech separates Engine Start and Prime Pulse Compensation from running fuel.

• KManager already rescales cranking fuel when Injector Size changes. Do not apply another injector-size ratio to the cranking table after entering the new size. Use the correct voltage compensation, then adjust the ethanol cranking and temperature corrections for the measured blend.
• Test on genuine cold soaks. Record ethanol content, ECT and IAT, battery voltage, time to fire and sustained idle. Repeat the same soak before and after the change. A warm restart after a short key-off is a different test from an overnight soak, and a heat-soaked restart is different again.
• The content retained at restart can be the old tank. The fresh fuel needs circulation before the rail sees it. Do not chase a start problem with a cranking multiplier until the content reading and the fuel pressure are known. A weak pump or the wrong injector dead time can look like a cold-start calibration error.
• The idle and injector page owns dead time, injector size and idle air. This page owns the blend and the ethanol-specific start corrections.

E85 is not an automatic timing adder. Pump E85 does not carry one fixed octane number, and no exact content-to-AKI conversion is established here. No fixed AKI rating and no fixed degree adder are printed. A higher ethanol content can improve knock margin without requiring more advance.

One tank is not a validation. The car has to start, idle, cruise, take load, refuel and recover from a sensor fault at more than one content. The table below is the minimum sheet for a flex calibration.

A matching lambda with a steady +15% short-term trim means feedback is adding fuel. Diagnose pressure, injector data and sensor scaling before changing the base model. In a stable cell with no other trim, the delivered quantity is base × 1.15.

Moving that correction into the base requires another log as feedback settles. Include long-term trim when it is active; do not count the same correction twice.

• A stable content display is not proof the sensor is valid. Check the signal against a known sample, check the update after a refuel, and check the software version. A frozen plausible value can hide a real blend change.
• The Honda logging page covers the export and the channel definitions. Archive the calibration, the native log, the CSV export, the software build, the measured content and the fuel pressure. Include content and software version so another tuner can interpret the log.

None of these is a target unless the row says so. Everything else comes from the calibration, the sensor or the log.