Hemi spark and knock: final timing and retard with HP Tuners

A Gen III Hemi calibration in this window commonly exposes a part-throttle spark surface and a separate wide-open surface. The exact labels, axes and switch point come from the file. What matters is that a cell in one surface is not automatically the cell used at wide-open throttle.

• The selection condition belongs to the operating system, and it can be a pedal threshold, a torque condition, or both. A throttle switch point from a standalone conversion is an owner's setting. Read the selection condition in this calibration before changing it.
• Axes differ across files. Some use rpm by MAP in kPa. Others use rpm by aircharge in grams per cylinder. Write down which one you have, and whether the pressure is absolute. “Load” alone is not a unit, and a cell copied between two different axes is a different request.
• Spark values are crank degrees before top dead center when positive. Keep the sign convention from the file. A negative number can mean retard rather than a smaller advance, depending on the display.

A VVT file can also expose base and lock-pin spark surfaces, with thermal versions of each. Check the cam state and the file's selection rules when an edit appears to do nothing. A change to one base surface does not establish what the lock-pin or thermal surface will command.

Do not copy all of them together just to make the log move.

• Temperature correction is a separate stage. Some files expose a hot intake-air table, a hot coolant table or both. Read the reference temperatures, deadband and interpolation rules in that file before deciding when the correction applies. A later Hellcat file can show useful labels. Do not assume its reference temperatures or table IDs transfer to a 2005–2014 naturally aspirated calibration.
• If the file exposes an MBT or optimal table, do not assume it is a hard ceiling on final spark. That role is not established for these operating systems. Do not raise it four degrees to make more timing available. If a minimum-spark table is exposed, check its applicable mode and tooltip before treating it as an active floor.
• The working relationship is simple to state and easy to misread. Final spark is the selected base plus the active modifiers minus knock retard and any spark reduction requested by torque management. A final 17° with 5° of knock retard means the base and modifiers totaled 22° before that subtraction. That holds if knock is the only correction in play. It does not mean the base cell is 17°, and it does not mean 22° belongs in the cell.

1. Save the read and write down the controller. Record the OS, the editor build and the exact spark labels before an edit. A label from another file is not an identification.
2. Name the axis and the unit. Write rpm by kPa absolute, or rpm by grams per cylinder. Do not write “load” and move on.
3. Find the selection condition. Read the condition that selects the wide-open surface. It can be a pedal threshold, a torque condition, or both. Confirm it in the log, not from another calibration.
4. List every modifier that can touch the region. Temperature, cam angle, tip-in and torque intervention each move final spark. If a channel is unavailable, write that down.
5. Prove the surface with a small change. Make a small reduction in the surface you believe is active and check the same region. Account for interpolation before expecting the full change in final spark.

The 2008 Ram 5.7 diagnostic description and the 2012 Charger 5.7 sensor description both describe short-term and long-term knock spark retard. A sustained short-term event can initiate long-term compensation. Recovery reduces retard that was already applied. The 2012 description adds that long-term information is retained in battery-backed RAM.

It also says short-term memory is lost at key-off, and the PCM ignores knock-sensor input at idle.

• The Hemi does not simply repeat short-term events and forget them. It can store a response. That is not the same as a GM-style high-octane and low-octane table pair with a knock-learn blend. These files do not have a verified high-octane and low-octane blend. Do not import a GM knock-learn procedure into a Hemi calibration.
• The 2012 description applies retard to all cylinders, even though the sensors correspond to banks. Use that distinction when you look at a per-cylinder threshold table. The threshold changes the level at which that cylinder's signal is treated as knock. It does not prove that final retard is independent for each cylinder.
• A general sensor passage in the 2012 description says short-term authority excludes wide-open operation. The 2008 strategy describes event-driven correction without that exception. Do not turn either sentence into a claim that knock protection is absent at wide open. The log and the operating system settle it.

The sensors listen to the engine, not only to combustion. The 2008 description says the output includes mechanical background noise that varies with engine, speed, load and temperature. A fixed voltage test therefore has no valid basis.

A signal below 1.5 V is not automatically false, and a signal above 2.5 V is not automatically detonation. The calibrated threshold and the noise model decide what the controller does with the signal.

1. Log both terms if the file exposes them. A short-term channel and a long-term or learned channel tell different stories. If only one is available, say so in the note.
2. Watch the shape in time. A decaying trace can be short-term recovery; a lasting correction can include learned retard. Confirm the channel before naming the behavior. A single maximum value hides the history.
3. Do not reset memory to make a clean log. A reset erases evidence. If you reset for a controlled comparison, record it and verify the tune again after the memory relearns.
4. Check the circuit before the calibration. Sensor mounting torque, wiring, grounds and stored knock-circuit codes come first. The 2012 description specifically warns that mounting torque affects response.

A spark log without the selection and correction channels is a single number with no history. Add the channels the file supports. If a channel is unavailable, record it in the log notes.

Final spark is reported by the PCM; it is not an independent measurement at the crankshaft. Keep the channel definition and sign convention with the log. A useful trace repeats in the same region.

Final spark follows the requested change, short-term retard stays at zero or decays after tip-in, and long-term retard does not step up through the pull. A trace that shows final spark lower than the base value with no retard channel is ambiguous.

It does not tell you whether the difference came from temperature, cam, torque intervention or knock.

1. Start with a warm, stable baseline. Record the mode, gear, cam state and temperatures. A cold pull and a heat-soaked pull are different tests.
2. Log the same route or the same load steps. Changing the road, the tire pressure or the load changes the result.
3. Keep the event in view. Log a window before and after the retard, not only the moment it peaks. The decay is part of the evidence.
4. Write down unavailable channels. Include those limits in the review notes.
5. Separate fuel, air and spark before the first edit. A lean or misfiring pull is not a spark test.

Octane is a specification, not a timing number. The 2014 Challenger fuel specifications distinguish the 5.7 automatic, the 5.7 manual and the 6.4. Use the application, not the engine size alone. This table is a 2014 Challenger specification.

It does not include the 6.1, and this page does not print a 6.1 fuel row. The 5.7 automatic 87 AKI row is not a 6.1 specification.

These specs do not license aggressive timing. They tell you the fuel the combination was built to tolerate. A calibration written for 93 AKI assumes 93 AKI in the tank. A calibration that knocks on 91 is not repaired by a higher-octane label alone.

• Fuel math causes another common error. A 14.7-scale AFR of 12.5 is lambda 0.8503. A 14.7-scale AFR of 13.0 is lambda 0.8844. Neither number is a mixture recommendation. Equivalence ratio is the reciprocal of lambda: lambda 0.8503 is an equivalence ratio of 1.176. Those two AFR values cannot both be described as 0.82 to 0.84 lambda. Decide whether the number in the file is AFR on a 14.7 scale, lambda, or an equivalence ratio. Compare a target from another calibration only after you know which one you have.
• Pump E85 is a blend, not always 85% ethanol, and this page has no verified Hemi E85 spark map. Do not add four degrees across the table because the fuel has more knock resistance. Measure the ethanol content, keep a wideband in the log, and treat the spark side as unverified until the file and the log support it. The airflow and fuel guide owns the fuel-mass and injector side of that work.

1. Confirm the fuel in the tank. Record the octane, ethanol content and fuel age. A file written for 93 AKI is not a file for 91 AKI.
2. Confirm the mixture under load. Compare commanded and measured lambda, fuel pressure and injector data. A lean excursion is a stop condition.
3. Check the ignition system and misfire status. A misfire can show up as a rough trace or a false knock signal.
4. Only then compare spark. If fuel, pressure or misfire is unexplained, fix that first. Spark work on top of it measures the fault.

The test is a controlled change. A torque plateau is a stop condition even when the retard channel is still quiet. The useful result is the least advance that meets the output objective on that run. Whether it holds on hotter air or worse fuel is a separate validation.

1. Fix the fuel and the load before the first pull. Use a repeatable load step or a dyno run, healthy lambda and stable temperatures. Record the conditions.
2. Identify the active surface and the visited cells. If aircharge exceeds the last breakpoint, check how this OS handles the table boundary. Match the axis units, interpolation and active modifiers before the write.
3. Set a conservative baseline for this engine. This page does not give a starting map because no same-OS stock extract was verified. Start from a conservative value and change only the cells you intend to test.
4. Change one limited region in degrees. A 1° step is only an example increment, not a required increase or an established safety margin. An engine already showing unexplained retard needs diagnosis before more advance. Check the adjacent cells and every modifier that can reach them.
5. Repeat comparable runs and read final spark. Compare torque, logged final spark, knock response and temperatures. If requested spark changed but logged final spark did not, stop and resolve table selection or modifiers.
6. Stop at the plateau or the first adverse response. A torque plateau, a sustained retard, a hotter charge or an unexplained lean event ends the step. Back up one step to the last point that did not show the adverse response on that run.
7. Recheck heat soak, low-speed high-load, tip-in and shifts. Those checks decide whether the previous step held. A result that only holds on a cool dyno pull is not a validated result.

Cam angle is a confounder. If the engine has variable cam timing, fix the cam position for the comparison or log the actual angle on every run. A cam move changes cylinder filling independent of spark, so a torque change can come from the cam schedule rather than the timing cell.

Hold the cam still for the comparison.

Retard that stays after a timing change is evidence about the cause, not proof that the table did not write. Work through the conditions in order. The table below is the short version.

If the blade closes while the pedal stays down and the retard channel is quiet, the airflow and fuel guide owns that diagnosis. Spark is not the first fix.

1. Save the faults and the baseline log. Record knock-circuit codes, misfire status, fuel data and the exact operating conditions before the first change.
2. Check the sensor installation and the circuit. Check mounting torque, wiring and grounds. The service description warns that torque affects response. A knock-circuit fault can make the retard channel unreliable.
3. Plot the event in time. Check whether it follows a shift request, a heat soak, a tip-in, a lean excursion or a repeatable load and rpm region. Correlation narrows the cause, but correlation does not prove false knock.
4. Inspect the external noise paths. Check for exhaust contact with the body or crossmember, loose mounts or brackets, and accessory or valvetrain changes. Repair the physical cause first.
5. Compare only with a matching baseline. A factory calibration is not a valid loaded-test baseline for different injectors, a cam or a blower. Hardware has to match before the comparison means anything.
6. Review sensitivity only after confirming a noise problem. Repair the cause first. Any remaining sensitivity change needs the correct OS definition and independent knock assessment. A zero retard channel after desensitizing is not validation.

Start at the first change in the time trace before using a retard histogram to choose cells. Recovery can carry one event into later rpm and load cells. Compare the event onset on repeated logs; do not subtract the maximum recorded retard from every cell visited afterward.

Keep the surrounding spark surface smooth enough that interpolation does not introduce an abrupt step.

• For example, suppose retard starts at 2,800 rpm and decays as the engine reaches 4,000 rpm. That is an illustrative trace, not a Hemi recovery-rate specification. It does not establish that every cell between those speeds caused a fresh knock event.
• A clean spark plug or a clean borescope does not prove there was no detonation. It is one piece of evidence. The retard channel, the measured mixture, the temperature and the mechanical inspection have to agree before you call an event false.

Use signed corrections consistently. A channel showing −3° can mean a three-degree timing reduction on these Dodge logs. A channel displaying retard as a positive magnitude needs subtraction instead. Read the definition before building a math channel.

The second example shows why a negative-retard sign does not explain the whole spark difference. Select one timestamp and record the interpolated base value, available corrections and final spark together. If the correction channels are incomplete, leave the difference unresolved rather than adding it to the table.

One clean pull does not validate a spark change. The conditions that produce retard are often the ones a single cool run does not include. Repeat the same route warm and again after a heat soak, and check the transitions as well as the steady state.

1. Start with a warm baseline. Record the mode, gear, cam state and temperatures for every event. Do not compare a cold log with a hot one.
2. Repeat the same route or load step. A different road grade, tire pressure or passenger load is a different test.
3. Check the transitions. Tip-in, a shift and a low-speed high-load pull can each produce a response that a steady pull does not.
4. Keep the fuel and hardware constant. If either changed, label the run as a new baseline instead of comparing it with the old one.
5. Send the same package for review. Send the read, the log, the channel list, the fuel, the hardware and the exact change. A final spark number without those is not enough to diagnose.

If the retard appears at a shift on a NAG1 car, read the NAG1 shift guide before you pull timing. The engine computer and the transmission computer both act on torque. A spark dip at the shift can be a request rather than a knock response. The log has to show which one moved first.