Tuning Wiki

This is a historical revision. Read the current page or return to its history.

Page organisation in this revision
Page type
Technical article
Subject
Switchable maps
Parent
Section level
Applies to
No ECU-specific applicability
Tool groups
None
Manufacturers
None

Changes from revision 2 to revision 4

Removed text is struck through; added text is highlighted.

Switchable maps let an ECU choose between calibrated behaviours without flashing a different tune each time. The word “map” is ambiguous: a mode may change a few tables, a larger calibration region, or an entire firmware image. These mechanisms have very different memory, hardware and transition requirements. A switch, cruise button or app is only the request interface; it does not reveal what the ECU actually changes.

Scope: This is a general engineering guide, not a claim that any particular EDC15, ME7, EDC17, MED17 or later ECU supports a given technique. ECU hardware, processor, software version, memory layout and the actual patch or product determine what is possible. Verify the exact part number and software. Modifying a vehicle can affect safety, emissions compliance, insurance and road legality.

ECU-specific guidesBosch EDC15 — code-block patches and dual-flash hardware.Bosch ME7 — ME7.1 L/M-box routine, ME7.5 five-mode patch and flash-board examples.Bosch EDC17 — diagnostic-app and pedal-based implementations, file coverage and diesel calibration.Bosch MED17 — software-specific Main Patch slots, limits and selected-map approaches.Bosch MG1 — B48/B58 custom-code slots and per-map enables.Simos 8.5 Main Patch, Simos 18 and Simos 19 RaceROM — detailed product-specific implementations.Detailed method guidesExternal memory bank switching — EPROM and flash chips, with EDC15P and ME7 board examples.Selected map and table switching — individual maps and functions, with ME7, EDC15, MG1/MED17 and VR30 examples.Calibration block and address switching — base-pointer selection, memory requirements and documented MEMS3 example.RAM calibration pages and overlays — XCP, AURIX and ETAS development systems.Reflashing and A/B firmware bank switching — programming and update banks compared with live calibration modes.

Methods at a glance

MethodWhat changesTypical benefit and limitation Physical chip or bankThe external memory image visible to the processor; potentially code and dataCan support independent images on suitable older hardware; a change while running can disrupt execution. Individual table selectionSpecified maps, curves, scalars or functionsEfficient use of space; everything outside the selected list remains shared. Calibration block or base-pointer selectionA larger defined set of calibration data, if all relevant accesses use the selected addressCan keep a coherent group together; copying bytes alone does not redirect accesses. RAM page or hardware overlaySelected calibration data accesses are redirected to another page or RAMUseful for online development; availability and persistence are implementation-dependent. Reflash or firmware bankA stored software or calibration image is programmed or selected for bootCan replace a large image, but programming and boot-bank swap are not necessarily live map switching.

1. Switching external chips or complete memory banks

Older ECUs often use an external EPROM or flash device. A dual-device board can make one of two memory devices visible to the ECU. Alternatively, where electrical and address compatibility allow, a larger device can hold two suitably aligned images: an extra high-order address input selects the image while the ECU continues to access the same apparent address range. The EcuVonix Dual EPROM instructions describe both arrangements and recommend switching with the ignition off. An EDC15P board is sold for switching two independent firmware images using AM29F400BT flash; this is a specific example, not evidence that every EDC15 uses that part or board.

If the device contains executable code as well as calibrations, bank selection may change both. Verify device capacity, voltage, pinout, access timing, bus width, chip-select behaviour, bank alignment, reset vectors and the complete image's compatibility with the ECU and vehicle. Selector inputs require defined electrical states, and the hardware must not allow two devices to drive the bus at once. Device identification, immobiliser data and coding can be stored elsewhere or within an image; the EcuVonix instructions report a key-code incompatibility in some installations. A socket or board does not establish compatibility by itself.

Live switching is not implicit. Changing the source during instruction fetch or while values and internal state belong to the old image can cause a crash, stall or unpredictable outputs. Use the device manufacturer's approved sequence; ignition-off switching and a clean restart are the conservative choice unless that exact implementation has been designed and validated for live changes. Do not change banks while programming or assume the selected position was correctly sampled at startup.

2. Switching selected maps, curves and functions

One common software approach retains a shared program and most shared calibration data. A mode variable selects between duplicates of particular tables or between mode-specific output values. For example, an ECU might select separate driver-demand, torque-limit, boost-target, fuel-target and ignition tables but share sensor scaling, protection limits and many other settings. The implementation may select table pointers, branch to different lookup calls, add an indexed dimension, or apply a correction; the interface alone does not identify which.

The space requirement is approximately the sum of the duplicated data for each additional mode plus alignment, axes, selector and code overhead. Shared axes can save space only when the lookup implementation supports that arrangement; otherwise each copy needs its own axis metadata. A correction or additional mode axis may be compact, but it changes only the parts of the calculation to which it is connected.

EcuTek RaceROM documents up to four modes and multiple fuel and ignition maps on supported vehicle and ROM combinations. Haltech's rotary trim example uses switch position as an additional map axis for boost and other functions. These are examples of supported features, not proof of identical internal code or a complete replacement of every calibration. Maintain a mode matrix listing exactly which maps and scalars change and which protections remain shared.

Torque-based ECUs usually combine driver demand, limiters, air charge, boost, fuel, ignition or injection timing, transmission intervention and component-protection models. Changing only the most obvious target may have little effect if another limiter dominates, or may create a mismatch between requested and delivered torque. Conversely, raising only a limiter does not create a complete, validated tune. Inspect the control path and calibrate the interacting quantities as a set.

3. Switching a whole calibration address or region

A broader approach places multiple versions of a defined calibration segment in available memory and changes the base address used to resolve lookups. Conceptually, address = selected_base + table_offset: the same offset can refer to a corresponding table in either bank. Another design uses a table of pointers, selects a group of pointers, or redirects a region in hardware. Vector lists single-pointer, RAM-pointer and flash-pointer calibration concepts; these are architectural patterns, not proof of their use in a specific production ECU.

For example, if the exact software has one 64 KiB relocatable calibration region and one verified unused, accessible, suitably aligned 64 KiB region, two copies might be addressed through a chosen base. This is an illustrative calculation, not an ECU memory map. Two 64 KiB banks consume 128 KiB before extra code, metadata and alignment. Free-looking bytes are not necessarily unused: check linker or image layout, other variants, bootloader, diagnostic data, checksums, flash sector boundaries, erase behaviour and addressability. Flash or file size alone does not prove space is safe to use.

Every relevant reader must actually use the selected base or translation. Absolute references, literal constants, independently located axes, code-embedded values, tables outside the segment, cached values and other cores can remain on the original data. Changing one pointer is insufficient if other paths bypass it. Confirm pointer width, alignment, bank offsets, endianness, signedness, table shape and bounds; verify all affected consumers and any integrity checks. A whole calibration region is still not automatically the whole executable firmware or all EEPROM adaptations.

Documented example: Rover MEMS3

In his MEMS3 dual-map investigation, Andrew Revill describes a petrol MEMS3 implementation in which the main calculation loop loads a calibration base address from a RAM pointer; tables and scalars are accessed relative to that base. His patch stores an additional complete calibration in available EEPROM space and changes the selected base. Both calibrations must match the firmware, and several listed firmware variants lack sufficient space. His physical-switch implementation selects at the calculation-loop boundary, while his throttle-input variant samples at ignition-on and needs an ECU shutdown and restart. He also reports that learned idle and fuel adaptations carry between maps; his observations are based on limited testing and are not a general guarantee of safe switching. This is a documented example of one engineered patch, not a universal MEMS3 or other-ECU capability.

EDC15 and ME7: historical and software approaches

Physical dual-chip or bank hardware was a useful approach on some older external-memory ECUs, including the EDC15P example above. It should not be treated as the only method for the EDC15 or ME7 families. A published EDC15 MapSwitch product describes a software-based three-mode solution for specified 1.9 TDI files; the vendor says the user must correct checksums. A separate ME7.x MapSwitch product describes adding switching logic to compatible software and using existing controls. The publicly described interface does not establish its exact internal pointer or lookup mechanism. Product support, selected maps, feedback, persistence and switch conditions must be checked for the exact file.

4. RAM calibration pages and overlays

Development ECUs can expose a reference page in flash and a working page in RAM. Where implemented, a calibration tool can prepare a complete working page and then make the ECU use it. ASAM XCP page switching describes separate tool and ECU access to calibration pages and explains that an atomic switch helps avoid the ECU seeing a partly written map. XCP is a measurement and calibration protocol; page switching is optional and depends on the ECU implementation. The presence of a diagnostic connector or CAN bus does not establish XCP access.

On supported Infineon AURIX devices, data access overlay can redirect specified flash data accesses to suitable overlay memory. The TC3xx family manual describes block-based redirection and coordinated block activation. It also distinguishes data accesses from instruction fetches and other bus masters. Check the exact device manual, available memory, block alignment, cache handling, core ownership and synchronisation; a processor feature alone does not mean a locked production ECU exposes it or that it is configured as a driver-selectable mode.

RAM is normally volatile. On reset, a working page must be initialised again from an approved source; any persistence mechanism must be separately designed and verified. Overlay and XCP are valuable calibration-development methods, not automatic substitutes for a durable production calibration or a safe in-car mode switch.

5. Mode inputs, blending and feedback

Inputs include a dedicated digital switch, analogue rotary selector, steering-wheel or cruise controls, an authorised CAN message, an app, or measured conditions such as fuel composition. The ECU needs a defined input range, fault detection, debounce or hysteresis where appropriate, and a valid fallback on an open circuit, short circuit, invalid message or loss of communication. If existing vehicle controls are reused, their original functions and interaction with other modules must remain understood. Provide unambiguous mode indication and log both the requested and the actually active mode.

Discrete modes and continuous blending solve different problems. A mode chooses a defined set; a blend interpolates between calibrations or applies a continuously varying correction. Haltech's flex-fuel guide explains why a tank may contain an intermediate ethanol mixture and describes using a composition sensor and suitable corrections. A two-position petrol/E85 switch does not measure the fuel in the tank. Sensor faults, transport delay between sensor and engine, blending rules and calibration across intermediate mixtures still need validation.

Making the transition coherent

Latch the selection: capture one valid mode for a defined control cycle so one calculation does not combine old and new selection states. Coordinate multiple tasks or cores and any cached calibration values. Change a complete set: arrange related map and limit selections as one coherent change where the design permits. A pointer change or hardware block activation is only atomic for the scope guaranteed by the implementation; other tasks, cores and state must still be considered. Control the operating point: restrict or defer the transition where appropriate, for example until torque demand and boost are low. A switch that is safe at idle is not automatically safe at full load. Define behaviour when a request is denied or delayed. Manage transients: rate-limit or blend suitable targets where validated, consider controller integrators and adaptations, and retain independent protective functions. Do not blindly interpolate incompatible strategies, units, table axes or discrete operating states. Define startup: specify whether the default or last mode is restored, how stored selection is validated, and what happens after brownout, reset, a failed write or a missing selector. Avoid unnecessary nonvolatile writes and account for endurance.

Integrity, programming and security

A checksum can detect certain changes or errors, but is not equivalent to a cryptographic signature or message authentication. Modern systems may enforce secure boot, authenticated programming, protected regions, anti-rollback or runtime integrity; exact coverage varies. Extra data or code may be outside a basic checksum yet still violate another integrity mechanism. Use authorised tools and a supported workflow; do not infer that an accessible spare area is executable, writable, or accepted by the boot chain.

Programming a different tune each time is useful where a live switch is unavailable, but has programming time and recovery implications. Likewise, dual firmware banks used for an update or rollback are not evidence that a vehicle supports instantaneous switching between performance calibrations. NXP's automotive update discussion illustrates the separate firmware-update and security context. Observe the ECU and tool vendor's power supply and recovery requirements.

Practical verification checklist

Record ECU hardware and software identifiers, tool and patch versions, and the original, known-good file and recovery route. Confirm that a usable existing recovery copy is available. Identify the exact mechanism, input, data switched, shared data, mode count, default, persistence and switching conditions. Document sources and assumptions; do not infer capability from a family name. Map all relevant references and memory regions for the exact build. Verify space, alignment, axes, calibration units, integrity coverage, selector validity and any boot or security requirements. Check every mode independently and verify actual values or outputs, not just the dashboard indicator. Test invalid inputs, power loss, restart, repeated switching and behaviour during a rejected request. Validate transitions and steady-state operation in a controlled setting: begin with low load, log requested and actual mode, torque, boost, fuelling, knock or combustion indicators as applicable, temperatures, faults and interventions; progress only when results are acceptable. Keep the engine and drivetrain protections appropriate to every mode. Check limits, transmission coordination, fuel supply and sensor plausibility. Validate the complete operating range and legal requirements for the intended vehicle and use.

Related wiki pages

Simos 18 switchable mapsSimos 8.5 main patch: map switchingSimos 19 EA888 Evo4: RaceROM map switching and calibration

Evidence note: Product documents demonstrate the features their authors describe for supported applications. General implementation details and the verification checklist above are engineering explanations; they do not establish an undocumented ECU-specific implementation.

Rendered revisions

Open each version to inspect formatting, tables and photographs.

Revision 2: Switchable maps

Switchable maps let an ECU choose between calibrated behaviours without flashing a different tune each time. The word “map” is ambiguous: a mode may change a few tables, a larger calibration region, or an entire firmware image. These mechanisms have very different memory, hardware and transition requirements. A switch, cruise button or app is only the request interface; it does not reveal what the ECU actually changes.

Scope: This is a general engineering guide, not a claim that any particular EDC15, ME7, EDC17, MED17 or later ECU supports a given technique. ECU hardware, processor, software version, memory layout and the actual patch or product determine what is possible. Verify the exact part number and software. Modifying a vehicle can affect safety, emissions compliance, insurance and road legality.

Methods at a glance

MethodWhat changesTypical benefit and limitation
Physical chip or bankThe external memory image visible to the processor; potentially code and dataCan support independent images on suitable older hardware; a change while running can disrupt execution.
Individual table selectionSpecified maps, curves, scalars or functionsEfficient use of space; everything outside the selected list remains shared.
Calibration block or base-pointer selectionA larger defined set of calibration data, if all relevant accesses use the selected addressCan keep a coherent group together; copying bytes alone does not redirect accesses.
RAM page or hardware overlaySelected calibration data accesses are redirected to another page or RAMUseful for online development; availability and persistence are implementation-dependent.
Reflash or firmware bankA stored software or calibration image is programmed or selected for bootCan replace a large image, but programming and boot-bank swap are not necessarily live map switching.

1. Switching external chips or complete memory banks

Older ECUs often use an external EPROM or flash device. A dual-device board can make one of two memory devices visible to the ECU. Alternatively, where electrical and address compatibility allow, a larger device can hold two suitably aligned images: an extra high-order address input selects the image while the ECU continues to access the same apparent address range. The EcuVonix Dual EPROM instructions describe both arrangements and recommend switching with the ignition off. An EDC15P board is sold for switching two independent firmware images using AM29F400BT flash; this is a specific example, not evidence that every EDC15 uses that part or board.

If the device contains executable code as well as calibrations, bank selection may change both. Verify device capacity, voltage, pinout, access timing, bus width, chip-select behaviour, bank alignment, reset vectors and the complete image's compatibility with the ECU and vehicle. Selector inputs require defined electrical states, and the hardware must not allow two devices to drive the bus at once. Device identification, immobiliser data and coding can be stored elsewhere or within an image; the EcuVonix instructions report a key-code incompatibility in some installations. A socket or board does not establish compatibility by itself.

Live switching is not implicit. Changing the source during instruction fetch or while values and internal state belong to the old image can cause a crash, stall or unpredictable outputs. Use the device manufacturer's approved sequence; ignition-off switching and a clean restart are the conservative choice unless that exact implementation has been designed and validated for live changes. Do not change banks while programming or assume the selected position was correctly sampled at startup.

2. Switching selected maps, curves and functions

One common software approach retains a shared program and most shared calibration data. A mode variable selects between duplicates of particular tables or between mode-specific output values. For example, an ECU might select separate driver-demand, torque-limit, boost-target, fuel-target and ignition tables but share sensor scaling, protection limits and many other settings. The implementation may select table pointers, branch to different lookup calls, add an indexed dimension, or apply a correction; the interface alone does not identify which.

The space requirement is approximately the sum of the duplicated data for each additional mode plus alignment, axes, selector and code overhead. Shared axes can save space only when the lookup implementation supports that arrangement; otherwise each copy needs its own axis metadata. A correction or additional mode axis may be compact, but it changes only the parts of the calculation to which it is connected.

EcuTek RaceROM documents up to four modes and multiple fuel and ignition maps on supported vehicle and ROM combinations. Haltech's rotary trim example uses switch position as an additional map axis for boost and other functions. These are examples of supported features, not proof of identical internal code or a complete replacement of every calibration. Maintain a mode matrix listing exactly which maps and scalars change and which protections remain shared.

Torque-based ECUs usually combine driver demand, limiters, air charge, boost, fuel, ignition or injection timing, transmission intervention and component-protection models. Changing only the most obvious target may have little effect if another limiter dominates, or may create a mismatch between requested and delivered torque. Conversely, raising only a limiter does not create a complete, validated tune. Inspect the control path and calibrate the interacting quantities as a set.

3. Switching a whole calibration address or region

A broader approach places multiple versions of a defined calibration segment in available memory and changes the base address used to resolve lookups. Conceptually, address = selected_base + table_offset: the same offset can refer to a corresponding table in either bank. Another design uses a table of pointers, selects a group of pointers, or redirects a region in hardware. Vector lists single-pointer, RAM-pointer and flash-pointer calibration concepts; these are architectural patterns, not proof of their use in a specific production ECU.

For example, if the exact software has one 64 KiB relocatable calibration region and one verified unused, accessible, suitably aligned 64 KiB region, two copies might be addressed through a chosen base. This is an illustrative calculation, not an ECU memory map. Two 64 KiB banks consume 128 KiB before extra code, metadata and alignment. Free-looking bytes are not necessarily unused: check linker or image layout, other variants, bootloader, diagnostic data, checksums, flash sector boundaries, erase behaviour and addressability. Flash or file size alone does not prove space is safe to use.

Every relevant reader must actually use the selected base or translation. Absolute references, literal constants, independently located axes, code-embedded values, tables outside the segment, cached values and other cores can remain on the original data. Changing one pointer is insufficient if other paths bypass it. Confirm pointer width, alignment, bank offsets, endianness, signedness, table shape and bounds; verify all affected consumers and any integrity checks. A whole calibration region is still not automatically the whole executable firmware or all EEPROM adaptations.

Documented example: Rover MEMS3

In his MEMS3 dual-map investigation, Andrew Revill describes a petrol MEMS3 implementation in which the main calculation loop loads a calibration base address from a RAM pointer; tables and scalars are accessed relative to that base. His patch stores an additional complete calibration in available EEPROM space and changes the selected base. Both calibrations must match the firmware, and several listed firmware variants lack sufficient space. His physical-switch implementation selects at the calculation-loop boundary, while his throttle-input variant samples at ignition-on and needs an ECU shutdown and restart. He also reports that learned idle and fuel adaptations carry between maps; his observations are based on limited testing and are not a general guarantee of safe switching. This is a documented example of one engineered patch, not a universal MEMS3 or other-ECU capability.

EDC15 and ME7: historical and software approaches

Physical dual-chip or bank hardware was a useful approach on some older external-memory ECUs, including the EDC15P example above. It should not be treated as the only method for the EDC15 or ME7 families. A published EDC15 MapSwitch product describes a software-based three-mode solution for specified 1.9 TDI files; the vendor says the user must correct checksums. A separate ME7.x MapSwitch product describes adding switching logic to compatible software and using existing controls. The publicly described interface does not establish its exact internal pointer or lookup mechanism. Product support, selected maps, feedback, persistence and switch conditions must be checked for the exact file.

4. RAM calibration pages and overlays

Development ECUs can expose a reference page in flash and a working page in RAM. Where implemented, a calibration tool can prepare a complete working page and then make the ECU use it. ASAM XCP page switching describes separate tool and ECU access to calibration pages and explains that an atomic switch helps avoid the ECU seeing a partly written map. XCP is a measurement and calibration protocol; page switching is optional and depends on the ECU implementation. The presence of a diagnostic connector or CAN bus does not establish XCP access.

On supported Infineon AURIX devices, data access overlay can redirect specified flash data accesses to suitable overlay memory. The TC3xx family manual describes block-based redirection and coordinated block activation. It also distinguishes data accesses from instruction fetches and other bus masters. Check the exact device manual, available memory, block alignment, cache handling, core ownership and synchronisation; a processor feature alone does not mean a locked production ECU exposes it or that it is configured as a driver-selectable mode.

RAM is normally volatile. On reset, a working page must be initialised again from an approved source; any persistence mechanism must be separately designed and verified. Overlay and XCP are valuable calibration-development methods, not automatic substitutes for a durable production calibration or a safe in-car mode switch.

5. Mode inputs, blending and feedback

Inputs include a dedicated digital switch, analogue rotary selector, steering-wheel or cruise controls, an authorised CAN message, an app, or measured conditions such as fuel composition. The ECU needs a defined input range, fault detection, debounce or hysteresis where appropriate, and a valid fallback on an open circuit, short circuit, invalid message or loss of communication. If existing vehicle controls are reused, their original functions and interaction with other modules must remain understood. Provide unambiguous mode indication and log both the requested and the actually active mode.

Discrete modes and continuous blending solve different problems. A mode chooses a defined set; a blend interpolates between calibrations or applies a continuously varying correction. Haltech's flex-fuel guide explains why a tank may contain an intermediate ethanol mixture and describes using a composition sensor and suitable corrections. A two-position petrol/E85 switch does not measure the fuel in the tank. Sensor faults, transport delay between sensor and engine, blending rules and calibration across intermediate mixtures still need validation.

Making the transition coherent

  • Latch the selection: capture one valid mode for a defined control cycle so one calculation does not combine old and new selection states. Coordinate multiple tasks or cores and any cached calibration values.
  • Change a complete set: arrange related map and limit selections as one coherent change where the design permits. A pointer change or hardware block activation is only atomic for the scope guaranteed by the implementation; other tasks, cores and state must still be considered.
  • Control the operating point: restrict or defer the transition where appropriate, for example until torque demand and boost are low. A switch that is safe at idle is not automatically safe at full load. Define behaviour when a request is denied or delayed.
  • Manage transients: rate-limit or blend suitable targets where validated, consider controller integrators and adaptations, and retain independent protective functions. Do not blindly interpolate incompatible strategies, units, table axes or discrete operating states.
  • Define startup: specify whether the default or last mode is restored, how stored selection is validated, and what happens after brownout, reset, a failed write or a missing selector. Avoid unnecessary nonvolatile writes and account for endurance.

Integrity, programming and security

A checksum can detect certain changes or errors, but is not equivalent to a cryptographic signature or message authentication. Modern systems may enforce secure boot, authenticated programming, protected regions, anti-rollback or runtime integrity; exact coverage varies. Extra data or code may be outside a basic checksum yet still violate another integrity mechanism. Use authorised tools and a supported workflow; do not infer that an accessible spare area is executable, writable, or accepted by the boot chain.

Programming a different tune each time is useful where a live switch is unavailable, but has programming time and recovery implications. Likewise, dual firmware banks used for an update or rollback are not evidence that a vehicle supports instantaneous switching between performance calibrations. NXP's automotive update discussion illustrates the separate firmware-update and security context. Observe the ECU and tool vendor's power supply and recovery requirements.

Practical verification checklist

  1. Record ECU hardware and software identifiers, tool and patch versions, and the original, known-good file and recovery route. Confirm that a usable existing recovery copy is available.
  2. Identify the exact mechanism, input, data switched, shared data, mode count, default, persistence and switching conditions. Document sources and assumptions; do not infer capability from a family name.
  3. Map all relevant references and memory regions for the exact build. Verify space, alignment, axes, calibration units, integrity coverage, selector validity and any boot or security requirements.
  4. Check every mode independently and verify actual values or outputs, not just the dashboard indicator. Test invalid inputs, power loss, restart, repeated switching and behaviour during a rejected request.
  5. Validate transitions and steady-state operation in a controlled setting: begin with low load, log requested and actual mode, torque, boost, fuelling, knock or combustion indicators as applicable, temperatures, faults and interventions; progress only when results are acceptable.
  6. Keep the engine and drivetrain protections appropriate to every mode. Check limits, transmission coordination, fuel supply and sensor plausibility. Validate the complete operating range and legal requirements for the intended vehicle and use.

Related wiki pages

Evidence note: Product documents demonstrate the features their authors describe for supported applications. General implementation details and the verification checklist above are engineering explanations; they do not establish an undocumented ECU-specific implementation.

Revision 4: Switchable maps

Switchable maps let an ECU choose between calibrated behaviours without flashing a different tune each time. The word “map” is ambiguous: a mode may change a few tables, a larger calibration region, or an entire firmware image. These mechanisms have very different memory, hardware and transition requirements. A switch, cruise button or app is only the request interface; it does not reveal what the ECU actually changes.

Scope: This is a general engineering guide, not a claim that any particular EDC15, ME7, EDC17, MED17 or later ECU supports a given technique. ECU hardware, processor, software version, memory layout and the actual patch or product determine what is possible. Verify the exact part number and software. Modifying a vehicle can affect safety, emissions compliance, insurance and road legality.

ECU-specific guides

  • Bosch EDC15 — code-block patches and dual-flash hardware.
  • Bosch ME7 — ME7.1 L/M-box routine, ME7.5 five-mode patch and flash-board examples.
  • Bosch EDC17 — diagnostic-app and pedal-based implementations, file coverage and diesel calibration.
  • Bosch MED17 — software-specific Main Patch slots, limits and selected-map approaches.
  • Bosch MG1 — B48/B58 custom-code slots and per-map enables.
  • Simos 8.5 Main Patch, Simos 18 and Simos 19 RaceROM — detailed product-specific implementations.

Detailed method guides

Methods at a glance

MethodWhat changesTypical benefit and limitation
Physical chip or bankThe external memory image visible to the processor; potentially code and dataCan support independent images on suitable older hardware; a change while running can disrupt execution.
Individual table selectionSpecified maps, curves, scalars or functionsEfficient use of space; everything outside the selected list remains shared.
Calibration block or base-pointer selectionA larger defined set of calibration data, if all relevant accesses use the selected addressCan keep a coherent group together; copying bytes alone does not redirect accesses.
RAM page or hardware overlaySelected calibration data accesses are redirected to another page or RAMUseful for online development; availability and persistence are implementation-dependent.
Reflash or firmware bankA stored software or calibration image is programmed or selected for bootCan replace a large image, but programming and boot-bank swap are not necessarily live map switching.

1. Switching external chips or complete memory banks

Older ECUs often use an external EPROM or flash device. A dual-device board can make one of two memory devices visible to the ECU. Alternatively, where electrical and address compatibility allow, a larger device can hold two suitably aligned images: an extra high-order address input selects the image while the ECU continues to access the same apparent address range. The EcuVonix Dual EPROM instructions describe both arrangements and recommend switching with the ignition off. An EDC15P board is sold for switching two independent firmware images using AM29F400BT flash; this is a specific example, not evidence that every EDC15 uses that part or board.

If the device contains executable code as well as calibrations, bank selection may change both. Verify device capacity, voltage, pinout, access timing, bus width, chip-select behaviour, bank alignment, reset vectors and the complete image's compatibility with the ECU and vehicle. Selector inputs require defined electrical states, and the hardware must not allow two devices to drive the bus at once. Device identification, immobiliser data and coding can be stored elsewhere or within an image; the EcuVonix instructions report a key-code incompatibility in some installations. A socket or board does not establish compatibility by itself.

Live switching is not implicit. Changing the source during instruction fetch or while values and internal state belong to the old image can cause a crash, stall or unpredictable outputs. Use the device manufacturer's approved sequence; ignition-off switching and a clean restart are the conservative choice unless that exact implementation has been designed and validated for live changes. Do not change banks while programming or assume the selected position was correctly sampled at startup.

2. Switching selected maps, curves and functions

One common software approach retains a shared program and most shared calibration data. A mode variable selects between duplicates of particular tables or between mode-specific output values. For example, an ECU might select separate driver-demand, torque-limit, boost-target, fuel-target and ignition tables but share sensor scaling, protection limits and many other settings. The implementation may select table pointers, branch to different lookup calls, add an indexed dimension, or apply a correction; the interface alone does not identify which.

The space requirement is approximately the sum of the duplicated data for each additional mode plus alignment, axes, selector and code overhead. Shared axes can save space only when the lookup implementation supports that arrangement; otherwise each copy needs its own axis metadata. A correction or additional mode axis may be compact, but it changes only the parts of the calculation to which it is connected.

EcuTek RaceROM documents up to four modes and multiple fuel and ignition maps on supported vehicle and ROM combinations. Haltech's rotary trim example uses switch position as an additional map axis for boost and other functions. These are examples of supported features, not proof of identical internal code or a complete replacement of every calibration. Maintain a mode matrix listing exactly which maps and scalars change and which protections remain shared.

Torque-based ECUs usually combine driver demand, limiters, air charge, boost, fuel, ignition or injection timing, transmission intervention and component-protection models. Changing only the most obvious target may have little effect if another limiter dominates, or may create a mismatch between requested and delivered torque. Conversely, raising only a limiter does not create a complete, validated tune. Inspect the control path and calibrate the interacting quantities as a set.

3. Switching a whole calibration address or region

A broader approach places multiple versions of a defined calibration segment in available memory and changes the base address used to resolve lookups. Conceptually, address = selected_base + table_offset: the same offset can refer to a corresponding table in either bank. Another design uses a table of pointers, selects a group of pointers, or redirects a region in hardware. Vector lists single-pointer, RAM-pointer and flash-pointer calibration concepts; these are architectural patterns, not proof of their use in a specific production ECU.

For example, if the exact software has one 64 KiB relocatable calibration region and one verified unused, accessible, suitably aligned 64 KiB region, two copies might be addressed through a chosen base. This is an illustrative calculation, not an ECU memory map. Two 64 KiB banks consume 128 KiB before extra code, metadata and alignment. Free-looking bytes are not necessarily unused: check linker or image layout, other variants, bootloader, diagnostic data, checksums, flash sector boundaries, erase behaviour and addressability. Flash or file size alone does not prove space is safe to use.

Every relevant reader must actually use the selected base or translation. Absolute references, literal constants, independently located axes, code-embedded values, tables outside the segment, cached values and other cores can remain on the original data. Changing one pointer is insufficient if other paths bypass it. Confirm pointer width, alignment, bank offsets, endianness, signedness, table shape and bounds; verify all affected consumers and any integrity checks. A whole calibration region is still not automatically the whole executable firmware or all EEPROM adaptations.

Documented example: Rover MEMS3

In his MEMS3 dual-map investigation, Andrew Revill describes a petrol MEMS3 implementation in which the main calculation loop loads a calibration base address from a RAM pointer; tables and scalars are accessed relative to that base. His patch stores an additional complete calibration in available EEPROM space and changes the selected base. Both calibrations must match the firmware, and several listed firmware variants lack sufficient space. His physical-switch implementation selects at the calculation-loop boundary, while his throttle-input variant samples at ignition-on and needs an ECU shutdown and restart. He also reports that learned idle and fuel adaptations carry between maps; his observations are based on limited testing and are not a general guarantee of safe switching. This is a documented example of one engineered patch, not a universal MEMS3 or other-ECU capability.

EDC15 and ME7: historical and software approaches

Physical dual-chip or bank hardware was a useful approach on some older external-memory ECUs, including the EDC15P example above. It should not be treated as the only method for the EDC15 or ME7 families. A published EDC15 MapSwitch product describes a software-based three-mode solution for specified 1.9 TDI files; the vendor says the user must correct checksums. A separate ME7.x MapSwitch product describes adding switching logic to compatible software and using existing controls. The publicly described interface does not establish its exact internal pointer or lookup mechanism. Product support, selected maps, feedback, persistence and switch conditions must be checked for the exact file.

4. RAM calibration pages and overlays

Development ECUs can expose a reference page in flash and a working page in RAM. Where implemented, a calibration tool can prepare a complete working page and then make the ECU use it. ASAM XCP page switching describes separate tool and ECU access to calibration pages and explains that an atomic switch helps avoid the ECU seeing a partly written map. XCP is a measurement and calibration protocol; page switching is optional and depends on the ECU implementation. The presence of a diagnostic connector or CAN bus does not establish XCP access.

On supported Infineon AURIX devices, data access overlay can redirect specified flash data accesses to suitable overlay memory. The TC3xx family manual describes block-based redirection and coordinated block activation. It also distinguishes data accesses from instruction fetches and other bus masters. Check the exact device manual, available memory, block alignment, cache handling, core ownership and synchronisation; a processor feature alone does not mean a locked production ECU exposes it or that it is configured as a driver-selectable mode.

RAM is normally volatile. On reset, a working page must be initialised again from an approved source; any persistence mechanism must be separately designed and verified. Overlay and XCP are valuable calibration-development methods, not automatic substitutes for a durable production calibration or a safe in-car mode switch.

5. Mode inputs, blending and feedback

Inputs include a dedicated digital switch, analogue rotary selector, steering-wheel or cruise controls, an authorised CAN message, an app, or measured conditions such as fuel composition. The ECU needs a defined input range, fault detection, debounce or hysteresis where appropriate, and a valid fallback on an open circuit, short circuit, invalid message or loss of communication. If existing vehicle controls are reused, their original functions and interaction with other modules must remain understood. Provide unambiguous mode indication and log both the requested and the actually active mode.

Discrete modes and continuous blending solve different problems. A mode chooses a defined set; a blend interpolates between calibrations or applies a continuously varying correction. Haltech's flex-fuel guide explains why a tank may contain an intermediate ethanol mixture and describes using a composition sensor and suitable corrections. A two-position petrol/E85 switch does not measure the fuel in the tank. Sensor faults, transport delay between sensor and engine, blending rules and calibration across intermediate mixtures still need validation.

Making the transition coherent

  • Latch the selection: capture one valid mode for a defined control cycle so one calculation does not combine old and new selection states. Coordinate multiple tasks or cores and any cached calibration values.
  • Change a complete set: arrange related map and limit selections as one coherent change where the design permits. A pointer change or hardware block activation is only atomic for the scope guaranteed by the implementation; other tasks, cores and state must still be considered.
  • Control the operating point: restrict or defer the transition where appropriate, for example until torque demand and boost are low. A switch that is safe at idle is not automatically safe at full load. Define behaviour when a request is denied or delayed.
  • Manage transients: rate-limit or blend suitable targets where validated, consider controller integrators and adaptations, and retain independent protective functions. Do not blindly interpolate incompatible strategies, units, table axes or discrete operating states.
  • Define startup: specify whether the default or last mode is restored, how stored selection is validated, and what happens after brownout, reset, a failed write or a missing selector. Avoid unnecessary nonvolatile writes and account for endurance.

Integrity, programming and security

A checksum can detect certain changes or errors, but is not equivalent to a cryptographic signature or message authentication. Modern systems may enforce secure boot, authenticated programming, protected regions, anti-rollback or runtime integrity; exact coverage varies. Extra data or code may be outside a basic checksum yet still violate another integrity mechanism. Use authorised tools and a supported workflow; do not infer that an accessible spare area is executable, writable, or accepted by the boot chain.

Programming a different tune each time is useful where a live switch is unavailable, but has programming time and recovery implications. Likewise, dual firmware banks used for an update or rollback are not evidence that a vehicle supports instantaneous switching between performance calibrations. NXP's automotive update discussion illustrates the separate firmware-update and security context. Observe the ECU and tool vendor's power supply and recovery requirements.

Practical verification checklist

  1. Record ECU hardware and software identifiers, tool and patch versions, and the original, known-good file and recovery route. Confirm that a usable existing recovery copy is available.
  2. Identify the exact mechanism, input, data switched, shared data, mode count, default, persistence and switching conditions. Document sources and assumptions; do not infer capability from a family name.
  3. Map all relevant references and memory regions for the exact build. Verify space, alignment, axes, calibration units, integrity coverage, selector validity and any boot or security requirements.
  4. Check every mode independently and verify actual values or outputs, not just the dashboard indicator. Test invalid inputs, power loss, restart, repeated switching and behaviour during a rejected request.
  5. Validate transitions and steady-state operation in a controlled setting: begin with low load, log requested and actual mode, torque, boost, fuelling, knock or combustion indicators as applicable, temperatures, faults and interventions; progress only when results are acceptable.
  6. Keep the engine and drivetrain protections appropriate to every mode. Check limits, transmission coordination, fuel supply and sensor plausibility. Validate the complete operating range and legal requirements for the intended vehicle and use.

Related wiki pages

Evidence note: Product documents demonstrate the features their authors describe for supported applications. General implementation details and the verification checklist above are engineering explanations; they do not establish an undocumented ECU-specific implementation.