Skip to content
Shop
Explore

INDEPENDENT BMW ENGINE EXPERTISE

N53B30. When adaptations fail

N43 / N53 · NOXEM · DIAGNOSTICS

Select a technical image to view it in detail.

This post covers the N43/N53 engine family: their complicated engine management, the quirks of how it behaves, and how to diagnose it.

The patient is an N53B30. Over a few hundred kilometers the engine became very temperamental. It vibrated at idle, switched roughly between operating modes (for example from low-load Stratified charge to open-throttle Homogeneous mode), and subjectively felt much weaker. All of these symptoms progressed very quickly, and misfires appeared in several modes. In short, the engine was degrading.

What would it cost to fix? EUR 100? EUR 500? EUR 1500? It depends on the diagnostician’s qualification.

First look at the fault memory

The first step was to read the fault memory. It held an error for the control Lambda probe of the second bank. Its status was passive, and the engine was still running in Stratified charge, so everything looked fine. The uneven-running error was also passive, and Stratified charge ran with Lambda up to 3.0–3.2.

The detailed description of the control probe error says that with a rich mixture (as demanded by the MSD80 and confirmed by the wide-band probe), the control probe voltage was only 0.10 V. That is out of range, since the voltage should be above 0.80 V.

In the long-term trim menu we caught a moment when the wide-band probe and the control probe contradicted each other.

The wide-band probe reports Lambda below 1, meaning a rich mixture, yet the control probe shows only 0.17 V.

Had something happened to the control probe of the second bank? I ran the probe test, with these results:

The test shows that the control probe can produce a proper voltage, and the wide-band probe’s range is also normal. The problem is obviously somewhere else.

The history memory held another interesting error: insufficient performance of the second bank’s catalytic converter. Is the catalytic converter damaged too? I’ll hold off on the remaining errors for a moment and come back to them later.

Given my experience diagnosing N-series engines, I did not rush to blame the catalytic converter. The dynamic resistance of the Nernst cells of all probes is within the normal range, and probe heating is as it should be. The learnbits show all segments adapted, including the individual Lambda corrections of the cylinders.

As the previous images show, both the offset and the multiplicative long-term trims are in the normal range. The integrators work, and the engine runs in closed loop.

Cylinders 5 and 6

Strange things kept turning up. In the Rough run menu, the bar for the 5th cylinder deviates strongly from 0 (cylinders are listed in firing order), meaning its efficiency is lowered.

I let the engine idle for 5 minutes, but nothing improved. The engine did not even try to correct the unevenness it had detected itself.

In the individual long-term trim data (multiplicative corrections) for the injectors, the value for injector No. 6 stands out. It is very close to the maximum of 83% and in some conditions even exceeds it.

At the same time, the opening time of the injector for cylinder 5 is relatively short. How is that possible? The measured efficiency of that cylinder is lowered, so why doesn’t the MSD correct it?

Two undocumented nuances

1. A positive deviation from 0 in the Rough run bar actually means insufficient mechanical activity of the cylinder, not increased activity as one might logically expect. 2. The Rough run menu lists cylinders in firing order, not in physical order. INPA evidently shows the cylinders in the same order in which the MSD80 registers them.

The cylinder sequence in the Rough run menu is therefore:

  • 1 = 1
  • 2 = 5
  • 3 = 3
  • 4 = 6
  • 5 = 2
  • 6 = 4

Note for the N43 engine:

  • The firing order is 1/3/4/2.
  • The data are shown for cylinders 1/3/5/6, which correspond to 1/4/3/2.

A detailed description is here.

This leads to the conclusion that the 5th cylinder has a problem, because its efficiency is insufficient, and that the 6th cylinder carries a huge additional trim. Injectors? Valves? I don’t think so. In Stratified charge at idle (Lambda 2.0), for example, the situation is normal: the mechanical efficiency of cylinder 5 and the short-term trims for injector 6 are fine, and the engine runs evenly.

The individual instantaneous short-term injector trims make things even more interesting. The values for all cylinders differ significantly from 0. Cylinder 5 shows an actual efficiency of -16.1%, and cylinder 6 shows +11%. How can this differ so much from the earlier test (see the description of the algorithm below)? And the long-term trims are reported as complete at the same time. This is an absolutely abnormal situation.

The MSD measured the insufficient efficiency of cylinder 5, and the Rough run data confirm it, yet the opening time is still so short.

It turns out that this huge additional short-term trim (as for cylinders 5 and 6) is not actually applied. In other words, INPA (more precisely, the MSD80) shows the desired value as if it were the existing one. We can confirm this by starting a new session: the long opening time for the injector of cylinder 5 is “lost”, because the huge additional short-term trim was never applied.

How the long-term trims are learned

Here is a short description of how the integrating (cumulative) long-term trims for the additional corrections work.

In various operating modes the engine regularly measures the individual “contribution” of each injector (cylinder). It temporarily enriches the mixture in one cylinder, measures how that cylinder changes the total Lambda, and calculates the difference between the real and the ideal situation. The results of these tests are summarized in the first 3 lines of the menu (Adaptation/injectors for Lambda regulation, ../F5/Shift + F6/F3).

After a test ends, these values are added, in “spare” time, to the matching data group of each injector. For example, data measured at short opening times go to the short-opening group.

After the long-term trims are deleted, and while they are being rebuilt, the ideal and measured situations differ greatly. The corresponding learnbit is then passive (not colored), because the initial long-term trims are still being created. When the situation is close to perfect (the bars in the first 3 lines are close to 0.0), the learnbit is colored, meaning the long-term trim has been completed successfully.

In our case the learnbits are colored, so the individual additional injector corrections “are” successfully completed.

Here, however, the MSD80 detected huge differences between desired and existing values but, for unclear reasons, did not add them to the corresponding data group. It even demands +16.1% additional correction for the “defective” cylinder 5, yet the opening time remains very short. Clearly something is not working as it should.

These huge Lambda differences between cylinders can explain why false errors are diagnosed for both the catalytic converters and the Lambda probes. In this case the mixture difference between cylinders reaches 30%. No Lambda probe can hold the optimum mixture under such conditions, and no catalytic converter can keep CO/HC low.

The hidden cause: NOx sensor errors

Now to the remaining errors in the history memory. Several hundred kilometers earlier, during maintenance on the exhaust system, the NOx sensor had been disconnected several times. The interruptions never lasted more than 5 minutes, but the fault memory contains several entries, logically time-outs: data exchange with the sensor was interrupted. After the work the errors were deleted and never came back, since there was no reason for them to, as the sensor works correctly.

But experience with the N43/N53 shows that any such error, even a deleted one, leaves a mark on the creation and continuation of the long-term trims.

This case is a very good example of what happens after completely unrelated NOx system errors, even when the problem has been solved and the error deleted successfully:

  • Application of the individual injector corrections stops.
  • As the cylinder imbalance grows, misleading errors for the catalytic converter and for the control Lambda probe are recorded.
  • The injector correction registers “fly away” uncontrollably beyond the permissible limits. They are not actually applied either, just as with non-coded injectors.
  • Nothing indicates that anything is wrong: there are no errors and no learnbit changes.
  • The engine vibrates, shivers, loses power and runs unevenly.

This situation is one step away from bank trim errors, wide-band probe errors, cylinder trim errors and misfire errors.

This is exactly how all the problems of the N43/N53 start when there are errors, even passive or deleted ones on an otherwise perfectly healthy engine, relating to the NOx catalytic converter and/or the NOx sensor.

In such a situation the most common repair strategy is to replace injectors and ignition coils, even though they are in excellent condition. And as mentioned at the beginning, the NOx system here is in perfect order and the engine runs in Stratified charge.

The fix: deleting the long-term trims

After assessing the situation, I decided to delete the corresponding long-term trims of the 2nd group in INPA (Lambda probe long-term trims, individual long-term Lambda trims). As in many other cases, the manufacturer gives no information on which status bits are cleared, and they are not visible. Experience shows, though, that deleting the old long-term trims also resets very specific status bits, which evidently clear the errors related to building these trims, the overfilled registers and so on.

After deleting the long-term trims and driving two short sessions (5 minutes each), the correction amounts for cylinders 5 and 6 normalized. The individually applied short-term trims normalized too, and the appropriate learnbits became colored.

Right after the deletion, the engine starts measuring the injector contributions:

Repeating the test under the same driving conditions gives these results:

The injector data are now measured correctly, and the difference from the ideal situation converges to 0.0. The long-term trim data are therefore being applied and restored correctly.

Scrolling down the menu, we see:

The engine has measured injector flow for both short and long opening. Both results are close to perfect, with the differences converging to 0.0 and never exceeding 0.3%, so the long-term trims work correctly.

The Rough run menu shows correct operation of all cylinders at idle in Homogeneous mode.

These are the correction amounts for cylinders 5 and 6 after the new long-term trims were created:

The engine runs smoothly and evenly again, and its previous performance and power are back.

Conclusion

Neither the Lambda probe nor the catalytic converter was damaged, even though there were errors pointing to both. The injectors were not damaged either, although the engine applied huge individual short-term trims to them. The ignition coils were not damaged, although the misfires suggested an ignition problem. A clean fault memory is not enough for the MSD80: after certain errors, the corresponding long-term trims must be deleted completely before the normal engine algorithm is restored.

An average service center would replace the following parts without justification:

1. The control Lambda probe of the 2nd bank, or even both the control and the wide-band probe, because the self-diagnosis points directly at it and the live data seem to confirm it. 2. The catalytic converters, because self-diagnosis reports weak performance and CO and HC in the exhaust would indeed be elevated. The cause, however, is the abnormal mixture in the cylinders, not damaged converters. 3. Ignition coils, and typically spark plugs as well. A test with ISTA D will mark at least one cylinder, typically several, as “guilty” of misfire. When replacing the coil brings no result, because the coil was actually fine, an injector would be blamed and replaced next.

If the long-term trims were deleted after replacing an injector, as ISTA D requires, the situation would of course improve, because the new long-term trims would build up normally. That would be one more case “solved” only by complicated and expensive repair work.

Related entries

N43 with strange behavior

Flow rate data coding of injectors

Management of Lambda probes

Rough run. Trim errors. Hidden cause

BEYOND THE FAULT CODE

A clearer picture.
A better next step.

Get in touch
Enlarged technical image
Open original image