Select a technical image to view it in detail.
I had wanted to upgrade the software on my N53B30 U0 for a long time. What held me back was that the engine already has chip-tuning. For this version the gain is considerable, around +50 hp and +40 Nm, which brings it almost level with the most powerful version of the engine. If the engine control unit is reprogrammed, the chip-tuning has to be redone, and I have no idea what the result would be.
Power supply during programming
During programming the vehicle must keep a stable supply voltage, and it must be quite high, above 13.0 V. This requires a dedicated, stabilized, high-current source of at least 50 A. An attempt with a 15 A source failed: the voltage dropped below 12 V even with every possible consumer (lights, climate fan, etc.) switched off.
Why I updated
The car showed all the signs of an old software release:
- Lambda 2.0 at idle (not 2.5) with Stratified charge active.
- False uneven-run error messages, and incorrect Rough run algorithms in both Homogeneous and Stratified charge modes.
- Quite incorrect switching between Homogeneous and Stratified charge.
- Incorrect creation of adaptations, among other things.
During the Rough run test in ISTA D, a software update notice appeared. The update improves evenness of engine run. That is direct acknowledgement that at least some of the symptoms above are known and have been addressed. Whether the fix works remains to be checked. The notice read:
The service data status ”Button designation: improve operational smoothness conversion” must be programmed via the ”conversation” path.
The original software release (MSD p/n) was 7611396.
Stage 1: ISTA P and ICOM
The first stage was done in the most correct way, with ISTA P and an ICOM. (There turned out to be several stages, although nothing suggested it at the outset.)
ISTA P reported an update for the MSD80 and programmed it in a few minutes. The software identification went from *ICB1S to *ICB4S. Afterwards several quite inadequate error messages appeared, for example 30DC, related to serial interfaces, and one for the outdoor temperature sensor. In the latter case the MSD80 did not command the sensor heating to start, yet a moment later recognized that the heating was not working, which is absolutely illogical. Putting the MSD80 to sleep and waking it again cleared these faults.
A short test series showed no significant improvement. Lambda at idle was still 2.0, and the Rough run algorithms still performed incorrectly in both Homogeneous and Stratified charge modes.
Stage 2: WinKFP and SP DATEN 2.53
Next I programmed with WinKFP. The downloaded SP DATEN v2.53 files contained a newer version. After a little while version 8603120 was flashed, and after that an upgrade to 8603156 was programmed.
A short test of this version showed Lambda still at 2.0, but the Rough run algorithm had changed.
In Homogeneous mode, with cylinder 1 switched off, the bar for cylinder 1 used to sit around -5 units but would unexpectedly jump to a maximum of +6 to +8. The bars for the other cylinders were also quite unstable. That is not good news.
The image shows the typical picture for the current software version when cylinder 1 is switched off.
Around 5 to 10% of test results look like this, and the readings fluctuate very strongly. For the initial software version, however, this result (+6 to +8 for cylinder 1) was typical.
The Stratified charge result is much more interesting:
The bar for cylinder 1 shows around 3.9, and the bars for all cylinders are stable despite significant unevenness of engine run. Evidently:
- The sensitivity to uneven cylinder performance has been lowered.
- The algorithm failures in this mode have been fixed.
On that basis I decided to test this version for a longer period.
Changes noticed in version 8603156
- The engine’s sensitivity to road unevenness has been reduced, but is still quite high.
- The false misfire of cylinder 1 when the diagnostic equipment is attached is still present, but reduced.
- Adaptations: the total bank corrections follow temperature more exactly, but the individual cylinder corrections are quite incorrect. In Stratified charge, the injector opening timing parameters are finally displayed correctly.
- Switching from Stratified charge to Homogeneous and back is correct.
- Rough run was still detected several times when accelerating evenly from 50 to 80 km/h, as the engine switches from Stratified to Homogeneous lean. It is rarer and much less pronounced than in the initial version.
Rough run management on uneven roads
A more interesting picture appears on really uneven roads. Driving on a very rough road, the MSD80 detected Rough run quite quickly and switched from Stratified charge to Homogeneous lean. But the switching behaved very differently from the initial version.
Initial version
- If Rough run was detected in the low load/RPM area of Stratified charge, Stratified charge was turned off completely and replaced by Homogeneous lean across its whole operating area.
- If Rough run was detected in the high load/RPM area of Stratified charge, for example at 100 km/h, the Stratified charge range was reduced, as seen in the image above (5).
- In both cases Stratified charge was switched off immediately at idle (6), and the engine ran in Homogeneous mode there.
New release
In the new version, if Rough run is detected under certain conditions (for example 1600 to 1800 RPM, 90 to 110 Nm), Stratified charge is switched off in that area only and replaced by Homogeneous lean (5). With a very even accelerator pedal, the engine then passes through these modes in turn: Stratified charge, Homogeneous lean, Stratified charge again, Homogeneous lean, and finally Homogeneous (at loads above 60%).
In principle such a solution could make sense, since it would make it easier to tell whether external conditions or an engine problem are to blame. But BMW AG did not get it right.
When this “patch” of Homogeneous lean mixture is created, the required fuel quantity is calculated incorrectly. The efficiency in Homogeneous lean mode is lower than in Stratified charge, so the fuel quantity should be increased to compensate. Instead it is decreased.
As a result, with cruise control on, the engine speeds up and slows down within a second while switching between Stratified charge and Homogeneous lean. It feels quite strange to drive. It is unclear how such a gross mistake could go unnoticed, since it is visible even with the car on a test stand.
To be continued. Planned next steps: load newer SP DATEN and continue programming.
Day 2: more differences in version 8603156
Evaluating version 8603156, I found some additional differences from the initial version:
- At very small load, with engine speed above 1500 RPM, the engine runs at Lambda up to 4.5 (extremely lean).
- After aggressive acceleration the MSD80 stays in Homogeneous mode for about 5 seconds, evidently to be ready for continued aggressive driving without losing a second switching from Stratified charge to Homogeneous.
- The NOx catalytic converter regeneration phases are switched more gently, and the energetic compensation promised in BMW AG after-sales materials is now performed. In the initial version the same fuel quantity as in normal driving was injected during regeneration. Because of the lowered efficiency, power dropped as well, which on an even road was felt as a brief loss of engine torque during regeneration.
Update availability and the .0da files
Back to the main task, updating the MSD80 software. After I loaded the SP DATEN files released in mid-2016, WinKFP did not automatically offer any new version, and ISTA P offered a downgrade to 8603120. It is possible that the downgrade was linked to loss of coding data, since the MSD80 “lost” its wear and sulfur saturation data for the NOx catalytic converter.
The vehicle’s integration level is displayed as appropriate to 2016, yet the MSD80 software contains gross mistakes. How is that possible?
Exploring the *.0da files turned up many interesting things. These are encrypted hex files intended for programming the MSD. Their headers state when the file was compiled, who developed the application, who supervised the programming, and so on.
The files show, for example, that the software of my car (factory release) was developed in 2005 to 2008. The newest version (8603188) was compiled in June 2012, and nothing has been compiled since. That evidently means BMW AG stopped maintaining the MSD80/81 (the N53B30 version for the E-series) at that point.
The file kmm_SG.txt is also very interesting. It describes the attributes of each software release, including engine type and backwards compatibility.
Next step: load the newest possible software release for the N53B30, despite the absence of formal updates.
Day 3: trying releases 8603180 to 8603188
Software releases for the N53B25 and N53B30 were separated, so I naively hoped that every release intended for the N53B30 would cover all its power variants (the versions without index, plus U0 and O0), with the differences held in the coding data.
My engine is the reduced-power N53B30U0 (220 hp / 160 kW).
After programming versions 8603180 to 8603188, error messages immediately appeared for both DISA valves, 2AA8 and 2AA9. So the number of DISA valves is defined by the software release, not by the coding data.
Because of these DISA valve faults, the engine did not enable Stratified charge in normal operation, although the test blocks were allowed at idle.
Differences in release 8603188:
- The Rough run test algorithms are correct in both Homogeneous and Stratified charge modes. The algorithms do differ, but the bar values are stable.
- Switching between operating modes (Homogeneous, Homogeneous lean, Stratified charge) is very exact.
Given all this, I have to change strategy. I will replace the intake manifold and both DISA servomotors with those from an N53B30O0 engine, and then install software version 8603188.
Conclusions
I am quite surprised by the situation. The N53B30U0 was manufactured until the end of 2009. Version 8603156, the last software version intended for this engine, and the whole 86* group were compiled on 3 July 2012, though *156 was clearly finished long before that date. The most surprising point is that the final software was left with such gross mistakes, and that for at least 7 years several hundred thousand vehicles have been affected. Where is the famous German quality?
In addition, these software defects produce false uneven-run error messages, which lead to injectors and ignition coils being replaced without any need. The customer pays to fix defects that do not exist.
Related entries:




