Drive-by-wire engine ECUs contain a second, independent calculation whose only job is to catch the main control software delivering more torque than the driver asked for. German manufacturers standardised it as the three-level E-Gas monitoring concept: level 1 runs the engine, level 2 recalculates a permissible torque and an actual torque from its own inputs and compares them, and level 3 checks that the processor doing both is working. If level 2 finds the actual torque above the permissible torque for longer than a short tolerance time, the ECU forces a safe state: a reset, a throttle or power-limited limp-home mode, or an injection cut-off.
This is why a remap can produce "internal control module" torque faults on an engine that is mechanically healthy. A calibration change that raises what level 1 delivers, or alters how torque is modelled, without keeping the level 2 calculation consistent makes the modified engine look to the monitor like a failed ECU. How level 1 arbitrates between competing torque requests is a separate subject, covered in Torque requests and the limit that wins.
Why the monitoring exists
With an electronic throttle, or a diesel whose pedal is only a sensor, a software or hardware fault inside the ECU could command torque the driver did not request. The standardised concept was written by a working group of Audi, BMW, Daimler, Porsche and Volkswagen together with ECU suppliers, on the stated basis that monitoring offers no scope for brand differentiation and should be supplier-independent. Version 5.5 is dated 5 July 2013.
Its hazard analysis defines four safety goals and rates only one above ordinary quality management: prevention of unintended acceleration, classed ASIL B. ASIL is the risk classification of ISO 26262, the road-vehicle functional safety standard first published in 2011, which grades hazards from QM through A to D by severity, exposure and controllability. Missing acceleration, unintended deceleration and missing deceleration are treated as controllable by the driver. The whole structure is therefore one-sided: it limits excess torque and takes no interest in a shortfall.
The three levels
| Level | Name | Content |
|---|---|---|
| 1 | Functional level | All engine control functions, including conversion of the requested torque into air, fuel and ignition, plus the diagnosis of the sensors and actuators that matter for monitoring: both pedal tracks, both throttle position tracks, brake and cruise inputs, and torque requests arriving from other control units. |
| 2 | Function monitoring | A separately developed calculation in the same processor that detects a faulty level 1 by comparing permissible and actual torque, or permissible and actual vehicle acceleration. Its memory areas are checked cyclically. |
| 3 | Controller monitoring | A physically separate monitoring module (an ASIC or second controller with its own clock) that tests the main processor through a question-and-answer exchange and can switch off the output stages independently of it. |
Independence is the point. The concept requires level 2 to be developed separately so that a systematic error does not affect both levels in the same way, and requires that faults in level 1 or in the hardware cannot influence level 2 undetected.
Level 2: the torque comparison
On petrol engines the monitor forms a permissible torque from the pedal request, the cruise control request and any torque-increasing requests from other control units such as the gearbox, read as raw signals and not as the values level 1 has already processed. It forms an actual torque by calculating backwards from what the engine is doing: the load signal (air mass or manifold pressure) and the ignition angle on port-injected engines, with injection time and lambda added on direct-injection engines. The actual torque must stay below the permissible torque.
Level 2 is deliberately simpler than level 1 and cannot reproduce every correction, so some values are taken across from level 1. The concept allows this for loss torque, adaptation values, correction factors and injection efficiency factors, but only if level 2 validates them against permissible limits first. Those limits, and the tolerance between the two torques, are left to each manufacturer.
The monitoring maps are part of the calibration. Community documentation of one Volkswagen Simos 18.1 software lists separate monitoring copies of the torque reference, reference ignition and ignition efficiency tables, with threshold maps for the instantaneous and integrated torque difference; see Simos 18 torque monitoring and safety plausibility. A tuning-vendor guide to ME7.5 describes a permissible-torque map addressed by pedal position and engine speed, with a filter and an offset map, feeding the maximum permissible indicated torque. Neither description has been checked against manufacturer documentation.
Diesel equivalents
A diesel has no throttle to close and its torque follows injected quantity, so the concept specifies different structures. Diesel monitoring consists of overrun monitoring combined with either a torque comparison or an acceleration comparison.
- Overrun monitoring. When the pedal is released and no cruise, idle-speed or external request is asking for torque, no torque-producing injection should be taking place. The monitor checks the measured injector actuation in that state.
- Continuous torque comparison. The electrical trigger signals actually sent to the injectors are measured and, with rail pressure, converted back into a torque that is compared with the permissible torque. Rail pressure has its own plausibility check.
- Acceleration comparison. As an alternative, the driver request is interpreted as a target vehicle acceleration; if actual acceleration stays above it for a calibrated time, level 2 limits the drive torque to zero and overrun monitoring is used to find the fault.
Ford's 2023 model-year diesel OBD summary shows how this appears in a production ECU. P061A sets when commanded inner torque exceeds permissible inner torque for the operating condition, with a detection time of 0.1 s. P061B covers the injection path, including a difference between programmed and actual energising time above 127.2 µs. P061C sets when engine speed calculated by the control software and by the monitor differ by more than 400 rpm. P062B includes a requested energising time above 200 µs in overrun fuel shut-off.
Level 3: controller monitoring
The monitoring module cyclically sends one of at least ten questions to the main processor at an interval of no more than 80 ms. The answer is a number built from test paths in the main processor: a program flow check that the level 2 modules ran in the right order and time slots, and an instruction-set test that runs copies of the level 2 calculations on test data representing fictitious faults. A wrong or mistimed answer increments an error counter that rises faster than it recovers. The main processor in turn tests the module by deliberately sending wrong answers and checking that the counter reacts. Bosch's diesel handbook describes the same arrangement: processor and monitoring module supervise each other, and either can trigger back-up functions independently of the other.
Level 3 also includes the memory tests that underlie ECU checksums (RAM and ROM at least once per driving cycle, with start released only after a fault-free result), an A/D converter test, and a test of the shut-off paths to the output stages once per driving cycle.
Fault reactions
Reactions are staged to keep the vehicle usable where that is safe.
- Reset. A level 2 torque deviation first causes a processor reset. Only after a calibrated number of resets without recovery does level 2 demand an irreversible injection cut-off, with the throttle actuator de-energised on petrol engines.
- Injection cut-off. In the concept's terms this limits the maximum permitted engine speed by cutting the torque-relevant injections. It is a power-limited limp-home mode and not a complete shutdown: the engine runs up to a speed threshold and is cut above it.
- Throttle limp-home. For throttle sensor faults the engine continues on the remaining sensor with the opening limited as a function of engine speed. If both sensors are implausible the actuator is de-energised and injection cut-off is applied, and injection cut-off is also applied if the throttle does not reach its limp-home position when requested.
- Output stages off. When the level 3 error counter reaches its limit the monitoring module switches off the power output stages and allows a limited number of resets. After that they stay off until the vehicle is restarted.
The concept gives 500 ms as a benchmark for the time from detection to reaction in torque monitoring and leaves the exact figures to the manufacturer. A Mitsubishi service bulletin for the 2018 Eclipse Cross gives a production example of the outcome: P061A with a fail-safe mode limiting the vehicle to 40 km/h.
What a remap changes
A stock calibration is released with level 1 and level 2 agreeing across the operating range. Changing one side opens a gap. The mechanisms below follow from the structure of the concept; no manufacturer document describing tuned vehicles has been identified.
- Raising driver-request or torque-limit maps so that level 1 delivers more torque at a given pedal position than the level 2 permissible-torque calculation allows.
- Rescaling the level 1 torque model (torque to load, optimum ignition, torque to quantity) so that the same real output is reported differently, without the corresponding change to the monitoring copies. Model consistency within level 1 is discussed for ME7 and Simos 18.
- Increasing air mass, boost or injection duration beyond the range the monitor's reverse calculation was calibrated for, so that calculated actual torque exceeds what the pedal position permits.
- On diesels, altering injection in a way that leaves energising time present when the monitor expects none.
The manufacturer evidence that such a mismatch is enough on its own comes from the Mitsubishi bulletin: on an unmodified car the monitor did not recognise turbocharged operation, P061A set under boost, and the remedy was revised ECU software. Owner reports after remapping follow the same pattern. A BMW N47 with EDC17CP02 stored internal control unit faults after a remap while driving normally, with torque deviation suspected and no verified diagnosis; see BMW EDC17 tuning torque checks and adaptations.
The identifying sign is the fault text: internal control module torque, torque calculation, engine speed or monitoring processor performance (generic codes P061A, P061B, P061C, P060A and P0607 in the Ford document, or manufacturer-specific equivalents). The staged reactions above imply the driving symptoms: a momentary loss of drive as the processor resets, then an engine-speed or road-speed restriction that persists until the ignition is cycled or the fault is cleared. Because the comparison only fails where delivered torque exceeds the permitted value, the fault can be absent in gentle driving. A boost, smoke or temperature limit reached within level 1 reduces torque without setting these codes.
The monitoring thresholds are calibration data and can be altered, and commercial editors list torque monitoring among their maps. Widening them until the fault disappears removes the protection against unintended acceleration that the concept exists to provide, for every fault and not only the one the tune created. The engineering answer is to bring the torque model and the monitoring model back into agreement, so that the monitor still reflects what the engine does.
Not established
- Which manufacturers and ECU families implement the concept exactly as published. The document is a guideline from five German manufacturers; Ford and Mitsubishi material shows equivalent monitoring but does not cite it.
- Tolerances, debounce times and reset counts for any specific ECU. The concept leaves them to each project.
- A version 6.0 dated 2015 is listed by secondary sources; differences from version 5.5 are not described here.
- No manufacturer document naming the level 2 maps of any production ECU has been identified. Map names in circulation come from community and vendor definitions.
Sources
- Standardized E-Gas Monitoring Concept for Gasoline and Diesel Engine Control Units, version 5.5, EGAS Workgroup (Audi, BMW, Daimler, Porsche, VW), 5 July 2013, independent PDF mirror; the copy in NHTSA docket 2014-0108 returned an access error. Safety goals, three levels, level 2 structures for petrol and diesel, level 3 question-and-answer, fault reactions (sections 8 to 11). Accessed 2026-11-19.
- Electronic Diesel Control (EDC), Robert Bosch GmbH, 2001, independent PDF mirror, printed pages 41 and 79. Monitoring module with question-and-answer cycle; monitoring of internal ECU functions and switch-off paths. Accessed 2026-11-19.
- 2023 MY OBD System Operation Summary for Diesel Engines, Ford Motor Company, revision 25 August 2023, pages 103 to 105. ECU monitor fault codes and malfunction thresholds for torque, injection, engine speed and monitoring module. Accessed 2026-11-19.
- Mitsubishi TSB-18-13-002, Mitsubishi Motors, 7 June 2018, summary on CarComplaints (NHTSA TSB 10142734). P061A torque monitoring during turbocharged driving, 40 km/h fail-safe, software remedy. Accessed 2026-11-19.
- ISO 26262, Wikipedia, undated. Publication dates and the ASIL classification; the ISO catalogue page returned an access error. Accessed 2026-11-19.
- Audi TT ME7.5 tuning guide, ECUFILES technical blog, 29 December 2023. Vendor description of the ME7.5 permissible-torque maps; unsourced and unverified. Accessed 2026-11-19.
- Simos 18 torque monitoring and safety plausibility and BMW EDC17 tuning torque checks and adaptations on this wiki. Community map definitions, editor coverage of torque monitoring and the owner report cited above.