Skip to content
Shop
Explore

INDEPENDENT BMW ENGINE EXPERTISE

Ignition coils. Part 2.

N43 / N53 · NOXEM · DIAGNOSTICS

While working with my ignition coil test stand, I noticed odd behavior in the coil’s performance and decided to look at a real engine. It is hard to accept that, in an age when people plan to go to Mars, BMW could not build a stable, safe ignition output stage. What follows is technical and written for specialists. If that is not for you, skip to “Solution”.

Measurements on the primary side

I connected my oscilloscope to the primary winding of the ignition coil.

In its documents “Ignition strategy” and “BMW engine diagnosis”, BMW states that in homogeneous mixture mode the arc is held at 2–3 kV, with the initial spike 2–3 times above that level.

[[[KEEP1]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_strategy_2.png)

Even in BMW’s own document, the firing line is noticeably higher than the “2–3 times” they claim.

The connection diagram is classical, with nothing unusual:

[[[KEEP2]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_output_stage.png)

The first capture already shows some oddities. Bandwidth: 100 MHz; sample rate: 2 GS/s.

[[[KEEP3]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_diagostics_1.jpg)

The probe is set to 1:10, so one vertical division is 100 V. The measurement was taken at idle in homogeneous mixture mode.

During the arc, the primary voltage is 30–40 V. With the 1:100 transformation ratio, that corresponds to 3–4 kV on the secondary. This roughly matches BMW’s documentation, bearing in mind that the N53 uses spark plugs with a wider gap than a “regular” engine.

The short pulses on the primary, however, reach 380 V, or about 38 kV on the secondary. That is some 10 (!) times the arc voltage. This engine does have situations that demand a higher voltage to sustain the arc, for example homogeneous lean mode, where lambda in all combustion chambers is about 1.5. But a factor of 10 is too much even for the N53. It is also interesting that such a voltage is possible at all, since the arc itself needs much less.

Zooming in shows more detail:

[[[KEEP4]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_diagnostics_2.jpg)

This is the start of the ignition process in full detail, and it looks nothing like what the literature describes. So what is going on?

The output stage: ISL9V5036

The MSD80/81 uses an “intelligent” IGBT ignition switch, the ISL9V5036 from Fairchild. The manufacturer’s datasheet is here:

ISL9V5036S3ST Datasheet

The datasheet shows two interesting points:

  • The over-voltage clamp is connected between the transistor’s collector and gate (not collector and emitter), which under certain circumstances can cause oscillation.
  • The maximum current fall time (Tfall) is 15 µs, but the typical value is about 2.8 µs, roughly 5 times less.

The spread of this parameter is therefore very wide. The behavior of the whole module can differ from one switch sample to another, and also with temperature and other conditions such as primary current and supply voltage.

What happens at turn-off

When the IGBT’s collector current reaches about 6–9 A, the ISL9V5036 receives a low-level (turn-off) command from the MCU. After the turn-off delay, and once the gate reaches the Miller threshold, the transistor starts to close.

The typical current fall time is short, and the collector voltage reaches the 350–380 V threshold within a few µs. That is the transistor’s over-voltage protection level. During this fast voltage rise, the gate current has not discharged the gate capacitance significantly below the Miller threshold, mostly because of the high gate resistance (typically 100 Ω to 1 kΩ). The over-voltage protection therefore reacts almost immediately and forces the transistor to turn on again. Because of the gate capacitance and the limited reaction time of the transistor and clamp circuit, the collector voltage drops almost to 0.

At this moment the gate voltage falls as the gate capacitance discharges through the gate resistor. The transistor turns off, the collector voltage rises again, and an oscillation begins.

The optimistic reading is that the next two spikes are parasitic resonant oscillations and the transistor has finally closed correctly and completely. I am not so optimistic. I think these spikes also occur with the transistor in active mode, because the clamp circuit has not been able to disconnect properly.

After about 10 µs the spark period finally begins. The secondary voltage drops rapidly, but slowly decaying oscillations continue.

At first glance the primary side looks like an unconnected inductance, but it is not. A test on the bench shows that with rapid closing of the transistor, two types of spark appear. One has a higher voltage but is thin and carries less energy. The other is strong but breaks through a smaller air gap. When I slowed down the transistor’s closing (without changing the maximum collector voltage), only the second, stronger spark remained. So this breakdown lasting a few µs is also visible on the secondary side.

It seems the air ionization time is more than 2–3 µs. If the rise time of the high-voltage pulse is shorter than that, the voltage needed to start the discharge rises significantly.

Consequences

1. For the first few µs the secondary voltage is very high and effectively unlimited. Such a voltage damages the ignition coil’s insulation, can launch a spark discharge outside the spark plug, and can cause damage inside the coil. 2. The output IGBT/MOSFET transistors heat up more. For the first 10 µs they operate in active mode and convert part of the energy stored in the coil into heat. 3. The energy intended for the spark period is reduced. At this important stage of spark formation the secondary potential is not stable and high, and it often drops far below the optimum.

Why this happens

Ignition coil control has long been very simple: a transistor (switch) and a coil, nothing special. The MSV70, MSV80 and MSD80 all use identical output stages. The ignition coils themselves, however, have changed significantly. The MSD80 uses coils with higher inductance, a larger transformation ratio and much more stored energy. With coils of earlier generations (“ordinary” coils) the problems described above were minor and did little harm. With the coils used in the N53, the situation is a catastrophe.

IRFW644B

If these transistors are fitted instead, the situation is even worse. The maximum Uds is limited to 250 V, and these MOSFETs have no over-voltage protection inside the die.

Test with an RC snubber

As a test, I built a simple snubber from a 0.15 µF film capacitor in series with a 10 Ω, 4 W resistor, and connected it in parallel with the coil’s primary winding. This is the oscillogram with the snubber attached:

[[[KEEP5]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_diagnostics_3.jpg)

The vertical scale here is 50 V per division.

The oscillation is gone, and the first spike is about half as high. For the first 8–10 µs there is a clean, even plateau, and after 10 µs the waveform passes into a stable spark period.

With one modified test cylinder I drove for several engine hours in different operating modes. I wanted to be sure there was no misfire and that the snubber resistor could handle the short overloads, which exceed 500 W for several µs in each cycle.

After optimizing the components, I chose a 0.1 µF, 305 VAC foil capacitor and three 22 Ω, 2 W resistors in parallel to increase the peak power capability. With this configuration the first spike rose to 230–240 V in the first few µs, but the thermal load on the snubber resistors decreased.

The dU/dt after the transistor closes also fell, from 200–300 V/µs before the snubber to 50 V/µs after.

The IGBT is also relieved during turn-off. Dynamic losses drop significantly because the transistor’s collector current falls faster than the current through the snubber. The IGBT now works in soft switching, compared with hard switching before.

This is the final result, captured in real time:

[[[KEEP6]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_diagnostics_4.jpg)

The first spike is about 230 V, and almost immediately a break is visible. This is where the spark period starts to form. Once it has formed, the oscillation is imperceptible and dies away quickly. The repeated turn-on of the transistor, the operation in active mode, the power losses and the excessive collector voltage in the first few µs are completely eliminated.

Solution

A simple description of the principle is on Wikipedia: RC snubbers

[[[KEEP7]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_RC_Snubber.png)

Components

  • Resistors: high pulse power capability (typically an overload rating above 300–500 W for a 1 µs pulse), 2–3 W rated dissipation, metal film (not wirewound!!!). An example is the Vishay PR02 series.
  • Capacitors: low dissipation factor (0.1 % at 1 kHz or less), high impulse current capability, polypropylene or polyester dielectric.

The cost of the components is about EUR 5 for a 6-cylinder engine.

In the picture, the 22 Ω, 2 W resistors are at the top and the 0.1 µF, 305 VAC capacitors below:

[[[KEEP8]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_snubbers.jpg)

All components are placed in heat-shrink insulation tubes:

[[[KEEP9]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_snubbers_2.jpg)

Testing in progress:

[[[KEEP10]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_diagnostics_process.jpg)

[[[KEEP11]]](https://bimmerprofs.com/wp-content/uploads/2017/06/BMW_ignition_modified.jpg)

This is what the installation looks like. The snubbers sit in the recess between the ignition coils, and their wires run together with the stock wiring.

Connection diagram

[[[KEEP12]]](https://bimmerprofs.com/wp-content/uploads/2017/06/snubber_schematics.png)

[[[KEEP13]]](https://bimmerprofs.com/wp-content/uploads/2017/06/ignition_coil1_primary.jpg)

Pin 1 and pin 3 are marked in red.

[](https://bimmerprofs.com/management-lambda-probes/)

BEYOND THE FAULT CODE

A clearer picture.
A better next step.

Get in touch
Enlarged technical image
Open original image