Electronic Fuel Injection

The Otto engine needs a mixture of fuel and air for its operation. It would be the task of the fuel supply—carburetor or injection—to provide the engine with the ideal mixture. Unfortunately, there is no such thing as an ideal mixture.

Perfect combustion, as chemistry calls it, would require air and fuel in proportion of 14.7 parts to 1 (this is the so-called stoechiometric ratio). While this might be satisfactory for the scientists, the real-life conditions of a vehicle call for slightly different characteristics.

We use the ratio of actual mixture to the stoechiometric mixture, called lambda (λ), to describe the composition of the mixture entering the engine: λ=1 denotes the chemically ideal mixture, λ<1 means rich, λ>1 is lean.

The best performance would require a slightly rich mixture, with the lambda around 0.9, while fuel economy would call for a slightly lean one, between 1.1 and 1.3. Some harmful components in the exhaust gas would reduce in quantity between lambda values of 1 to 1.2, others below 0.8 or above 1.4. And if this is not yet enough, a cold engine requires a very rich mixture to keep running. After warming up, the mixture can return to normal, but the temperature of the incoming air still plays a significant role: the cooler the air, the denser it becomes, and this influences the lambda ratio as well.

All these requirements are impossible to satisfy with simpler mechanical devices like carburetors. Electronic fuel injection provides a system that can measure the many circumstances the engine is operating in and decide on the amount of fuel (in other words, the lambda ratio) entering the engine. By carefully adjusting the internal rules of this device, manufacturers can adapt the characteristics of the fuel injection to the actual requirements: a sporty GTi would demand rather different settings than a city car; besides, catalytic converters have their own demands that, as we will later see, upset the applecart quite vehemently.

Earlier fuel injection systems only knew about fuel, the ignition was supplied by traditional methods. Later on, these systems (now called engine management systems) took on the duty of generating the sparks as well. But even with this second incarnation, the fuel injection part remained practically the same, thus the following section applies to both kind of systems.

Fuel injection

The two most important inputs describing the actual operating condition of the engine, thus determining the fuel demand are the engine speed (revolution) and engine load. The engine speed can be measured easily on systems using traditional ignition: the ignition primary circuit generates pulses with their frequency proportional to engine speed (the tachometer uses this same signal to show the rpm to the driver). When the injection system provides the ignition as well, it cannot at the same time rely on it, so an additional sensor is used instead.

The engine load is usually determined by measuring the quantity of air the engine tries to suck in. There are various methods of attaining this: earlier systems used a flap which is deflected by the air flowing through the sensor—the angle of deflection is proportional to the amount of air passing through (air flow sensor, AFS). Later systems used a pressure sensor measuring the pressure inside the inlet manifold (manifold absolute pressure, MAP sensor). Yet another system (although not used on Citroëns) heats a platinum wire and lets the incoming air passing around cool it; by measuring the current needed to keep the wire temperature at a constant value above the temperature of the incoming air, the mass of air can be determined. Some simpler systems do not even measure the amount of air but use a pre-stored table in their computer to approximate it based upon the engine speed and the position of the throttle pedal—not that accurate but certainly much cheaper.

Under ideal conditions, these two inputs would already be enough to control the engine. A large table can be set up, like the one illustrated here (of course, this is only an illustration, the actual values mean nothing here), and for any pair of incoming engine speed and load values the necessary fuel amount can be determined. By keeping the pressure of fuel constant behind the injector valves, the amount of fuel injected depends solely on the time period the injectors are opened for, hence, the table can contain injector opening times:

Engine speed

0% load

5% load

100% load

idle

3

3

3

850 rpm

4

5

5

900 rpm

5

6

7

6,000 rpm

9

8

10

And this is exactly how it is done in modern injection systems: the controlling microcomputer keeps a lookup table like this to determine the base pulse width. Earlier systems were constructed from discrete, analog elements, not like a small computer; a more or less equivalent circuit made of various hybrid resistance arrays and semiconductors were used for the same purpose.

Chip tuning, by the way, is the simple operation of replacing the said table with another one, yielding different characteristics (usually to gain power, allowing for worse fuel economy). As the computer stores this table in a programmable memory—similar in function to the BIOS in desktop computers—, replacing it is possible. The earlier systems with analog circuits cannot be modified that easily.

So, we obtained the base pulse width from the table but as the operating conditions of automotive engines are hardly ideal for any reasonable amount of time, several corrections have to be applied. Our air flow meter measures the volume of the air but we would need to know the mass of the air to calculate the required lambda ratio—remember, colder air is denser, thus the same volume contains more gas, requiring more fuel to provide the same mixture. To accomplish this, the injection system uses an air temperature sensor (ATS)—although on some systems it measures not the air but the fuel-air mixture—and lengthens the injector pulse width according to this input (except for the case of the airflow meter using a heated wire, this one takes the air temperature into account automatically, consequently, there is no need for correction).

It is not only the external circumstances that require special consideration. While most of the time an engine works under partial load, so it makes sense to spare fuel by basing on a relatively leaner mixture across this range of operation, cold start and warm-up, modest deceleration and fully depressed throttle, idle speed all require different treatment.

The position of the throttle pedal is communicated to the computer by a throttle position switch (TS) or throttle potentiometer (TP). These devices signal both fully open and fully closed (idling) throttle positions. When the pedal is fully depressed, the computer makes the mixture richer to provide good acceleration performance.

Idle speed is more complicated: the throttle is closed, so there has to be a bypass to let the engine receive fuel to run. In simpler systems this bypass is constant (but manually adjustable to set the correct idle speed) in a warm engine, providing a fixed amount of air, although the computer can decide on a varying amount of fuel to be injected. Later systems generally use a controlling device changing the cross section of the bypass, regulating the amount of air coming through (these systems often have no facility to adjust the idle speed, the computer knows the correct revolution and maintains it without any help from mechanical devices). The controlling device can either be an idle speed control valve (ISCV) or an idle control stepper motor (ICSM). The first one can only open or close the idle bypass, so any regulation must be done by rapidly opening and closing it by the computer, the second one can gradually change the bypass, hence fine tuning is easier and smoother.

Just like the choke on carburetors, there is a complete subsystem dealing with cold start and warm up, as the requirements under such circumstances are so different from the normal operation that they cannot be fulfilled by the regular control. The ECU monitors the ignition key switch to learn when the engine is started, then looks for the input from the coolant temperature sensor (CTS) to see whether this is a cold start or a warm one. If the coolant fluid is measured cold, a special warm-up sequence will be started.

The engine needs significantly more fuel, a richer mixture during this period. This extra fuel is used for two purposes: first, part of the fuel injected is condensed on the cold walls of the engine, second, to ensure better lubrication, the engine should run at an elevated revolution during this period.

There are two ways to provide more fuel: through the usual injectors, making the computer inject more gas than normal, or by using an additional cold start injector (CSV)—there is only one such injector even in multipoint systems. This injector is fed through a temperature-timer switch, protruding into the coolant just like the CTS, plus it is heated by its own electric heater. The injector operates as long as the ignition key is in the starting position but its behavior later on is governed by the timer switch. The colder the engine initially is, the longer it stays closed to let the cold start injector do its job. In a warm engine (above 40 °C) it does not close at all.

Without a cold start injector, the computer itself adds about 50% extra fuel initially and drops this surplus to about 25% until the end of a 30-second time period.

From that point, the surplus is dictated by the warming of the engine, communicated by the CTS to the computer. EFI systems without an idle speed control device often use an electromechanical auxiliary air valve (AAV). This valve, which is fully open when the engine is still cold but will close gradually as it warms up, lets an additional amount of air measured by the AFS pass through the system. Because it is measured, it tricks the computer into providing more fuel. The valve is heated by its own heating element as well as the engine, thus it closes shortly.

The injectors are electrovalves. As with any electromagnet, there is a small time delay between the arrival of the control signal and the actual opening of the valve due to the build-up of electromagnetic fields. The length of this delay depends heavily on the voltage the injectors are fed with. The same pulse width would result in shorter opening time, hence less fuel injected if the battery voltage drops below nominal (which is often the case on cold mornings). The injection computer therefore has to sense the battery voltage and to lengthen the injector pulse width if necessary.

The final, total pulse width (also called injector duty cycle) is calculated by summing up all these values received: the base pulse width from the RPM/AFS table lookup, the various correction factors based on the temperature sensors, throttle position and the like, plus finally, the voltage correction.

As the computer has already calculated the exact amount of fuel to be injected, there is only one task left: actually injecting it. There are two possible ways: to inject the fuel into the common part of the inlet, still before the throttle butterfly, or to inject them close to the inlet valves, individually to each cylinder. Depending on the solution chosen, the system will be called monopoint or multipoint. Monopoint fuel injection requires a single common injector; the smaller cost and simpler setup makes it more common on smaller engines (in the case of Citroëns, the 1380 ccm ones). In all cases, the computer actually calculates the half of the fuel amount required as it will be injected in two installments, once for each revolution of the engine.

EFI monopoint

EFI monopoint

EFI multipoint

EFI multipoint

The injectors of the multipoint system can be operated simultaneously or individually. Previous Citroëns on the road today still use simultaneous operation. Individual cylinder injection, however, holds great potential—just to name one, some of the cylinders of a larger engine can be temporarily shut off by cutting off their fuel supply if the car is operating at partial load, saving a considerable amount of fuel—, so we are sure to meet this sort of fuel injection systems in the future.

All systems—regardless of the number of injectors—use a similar fuel supply layout. The fuel is drawn from the tank by a continuously operating fuel pump, transported via a filter to the injectors, then back to the tank. There is a pressure regulator in the circuit as well to keep the pressure of the fuel at a constant pressure above that in the inlet manifold (this regulator is a separate unit on multipoint systems while integrated into the injector on monopoint ones). As the pressure difference between the two sides of the injectors are constant, the amount of fuel injected depends solely on the opening time of the injectors. The pressure used in contemporary EFI systems is 3 to 5 bars.

This is practically all there is to it, there are only a couple of safety and economy features in addition. If the engine revolution exceeds a certain limit (between 1,200 and 1,500 usually) and the throttle is closed—this is called deceleration—, the momentum of the car is sufficient to rotate the engine through the wheels. To save fuel, the injection is cut off. As soon as the engine speed drops below the limit or the throttle is opened, the injection is reintroduced—supposedly smoothly and gradually, however, many drivers complain about some jerkiness.

To avoid prolonged operation at revolutions exceeding the specification of the engine, the injection is cut off above a maximum engine speed (6,000-7,000 rpm, depending on the engine). And finally, to avoid the hazard of fire in a crash and the fuel squirting from the injection system with the engine stopped or possibly destroyed, the relay of the injectors is controlled by the ECU, allowing fuel injection only when the ignition (or the signal of the corresponding sensor) is present.

Who will light our fire?

Models with simpler fuel injection have traditional (electronic) ignition systems which are practically equivalent to the solution used on cars with carburetors.

The distributor has two purposes: generating the driving signal for the ignition system and to distribute the high voltage to the four cylinders in turn. This two parts inside the distributor are electrically separate but mechanically coupled—both are driven by the camshaft to keep them in sync with the strokes of the engine.

The ignition signal thus starts from the distributor. A magnetic induction sensor (consisting of a rotating four-sided magnet and a pick-up coil) sends a pulse to the ignition module at each firing point. This pulse will be switched to the ignition coil (an autotransformer; auto here does not mean that it is manufactured for automotive use, autotransformers have their primary and secondary coils connected) by a power transistor inside the module. The current change in the primary coil induces very high voltage spikes in the secondary circuit. These spikes then go back to the HT part of the distributor which in turn sends them to the spark plug of the actual cylinder requiring the spark.

It takes some time for the spark to ignite the fuel-air mixture inside the combustion chamber: this means that the spark has to arrive slightly before the piston reaches its top position (top dead center, TDC), so that it will receive the downward force of the detonation in the right moment. However, as the engine speed increases, so does the speed of the piston or the distance it travels during a given period of time. Therefore, the exact time of the spark has to be advanced as the revolution increases. Traditional systems do this by adding a vacuum line connecting the inlet manifold to the distributor. As the vacuum increases with the engine revolution, its sucking force rotates the inner part of the distributor slightly away from its original position, causing all its timing devices switch earlier, as required by the value of the timing advance.

Clever systems can get away without a distributor: some CXs have such an ignition setup. This systems has two ignition coils, both serving two spark plugs at the same time. These two spark plugs belong to cylinders whose pistons move in unison: one is compressing, the other exhausting. Although both plugs generate sparks at the same time, the one in the exhausting cylinder will be wasted.

Two birds with one stone

We made the ignition seem too simple in the previous section. While it works as described, there are many factors to be considered if we want to build a modern ignition system. For instance, the timing advance depends not only on engine speed but on many other factors as well: engine load, engine temperature and to some extent, the air temperature.

Just like the carburetor was not really good at deciding the amount of fuel required by the engine, the traditional ignition is similarly not perfect in estimating the timing advance and other characteristics of the sparks needed. An electronic system similar to the one used for fuel injection shows clear advantages over any earlier system.

And as they use about the same sensors and rely on each other, what could be more logical than to integrate them into a common system, elegantly called an engine management system?

EMS monopoint

EMS monopoint

EMS multipoint

EMS multipoint

If we compare the schematics of the corresponding EFI and EMS systems, they look almost the same. There are two notable differences: the small arrow on the line connecting the ECU to the distributor has changed its direction and a new sensor, a crank angle sensor (CAS) has appeared. Both changes have to do with the fact that the enhanced system, whose new task is to generate the ignition signals as well, cannot at the same time build on them as inputs. This new sensor—practically a replacement for the induction magnet in the distributor of earlier systems—informs the computer of both engine speed and camshaft position.

The flywheel has steel pins set into its periphery. As it rotates, the inductive magnet of the CAS sends pulses to the computer. Two of the pins are missing and this hole passes before the sensor just as the first piston reaches its TDC position. The missing pins cause a variance in the sensor output that can be read by the ECU easily.

The rest is the same: the base pulse width is calculated based on the CAS and AFS/MAP sensors. The correction factors—air temperature, idle or full load, starting, warming up, battery voltage—sum up into an additional pulse width. Besides, the same input signals (AFS, CAS, CTS and TS/TP) are used for another lookup in a table, yielding the correct dwell time and timing advance for the ignition. The dwell period remains practically constant but the duty cycle varies with the chaging engine speed. The ignition signal is amplified and sent to a distributor containing only secondary HT components: it does not create the ignition signal only routes the HT current to each spark plug in firing order.

Some systems also have a knock sensor (KS), sensing the engine vibration associated with pre-ignition (so-called pinking). If this occurs, the ignition timing is retarded to avoid engine damage.

Think green

As we saw, fuel injection and engine management systems are capable of determining the ideal amount of fuel to be injected, depending on the conditions of operation and several other factors in the engine. It is capable of deciding on lean mixture for general, partial load to save fuel, or on rich mixture when performance considerations call for this.

Unfortunately, this is not what such systems are used for today. With the proliferation of catalytic converters, the only concern of our systems is the welfare of the converter.

Ideal combustion would not generate polluting materials in the exhaust gas. Fuel is a mixture of various hydrocarbons (CnHm), which when burned together with the oxygen (O2) of the air, should transform to carbon-dioxide (CO2) and water vapor (H2O). However, combustion is never ideal, besides, fuel contains many additives: the exhaust gas, in addition to the products mentioned, has various byproducts as well, some of them toxic: carbon-monoxide (CO), various unburned hydrocarbons (CnHm), nitrogen-oxides (NOx) and lead (Pb) in various substances coming from the anti-knock additives found in the fuel.

The relative amount of these byproducts depend on the lambda ratio of the air-fuel mixture burned. As shown on the diagram, a value between 1.2 and 1.3 would give a relatively low percentage of toxic byproducts while, as we can recall, being a lean mixture would be in the right direction towards fuel economy.

By using platinum (Pt) or rhodium (Rh) as a catalyst—a catalyst is a substance whose presence is required to enable (or to boost) a chemical transformation while it does not take part in the process itself, remaining intact—the following processes can be carried out:

  • 2 CO + O2 → 2 CO2 (oxidation)
  • 2 C2H6 + 7 O2 → 4 CO2 + 6 H2O (oxidation)
  • 2 NO + 2 CO → N2 + 2 CO2 (reduction)

These precious metals are applied in a very thin layer to the surface of a porous ceramic body with thousands of holes to make the surface contacting the exhaust gases much greater. Actually, a converter does not contain more than 2 or 3 gramms of these metals.

If you compare this diagram with the previous one, you will see that the real gain is the supression of nitrogen-oxides. CO and CmHn will be reduced as well, although to a much lesser extent. Nevertheless, the overall reduction in polluting byproducts is quite high, amounting up to 90 percent. Lead substances are not considered as lead must not reach the converter anyway, it would clog the fine pores of the converter in no time. The fuel used in cars equipped with a catalytic converter has to be completely free of lead.

But there is something of even greater consequence depicted on the diagram: to keep the amount of pollutants down, the lambda has to be kept inside a very small value range, practically at λ=1 all the time. If the lambda drops just a fraction below 1, the CO emission rises sharply, while a small step above 1 skyrockets the NOx emission. The main task of the fuel injection is therefore to ensure that the air-fuel mixture sticks to the stoechiometric ratio all the time. This means higher consumption than the one of a car with fuel injection without a converter to start with.

There are situations where this lambda cannot be observed. A cold engine will simply stall without a much richer mixture, thus the cold start mechanism does not obey the lambda control. The catalytic converter does not work at all below 250 °C, so this is not a significant compromise (its normal operating temperature is 400 to 800 °C, above 800 °C is already harmful; unburned fuel getting into the exhaust and detonating inside the converter could cause overheating, thus ignition and similar problems has to be rectified as soon as possible in catalytic cars).

Dynamic acceleration (full throttle) is also something not observing the welfare of the converter. Reducing pollution might be a noble cause but to be able to end an overtaking is even more important…

The system uses an oxygen sensor (OS, also called lambda sensor) which measures the oxygen content of the exhaust gas. It is located between the engine exhaust and the catalytic converter. Similarly to the converter, it is not functional below 300 °C, hence it has its own heating element to make it reach its operating temperature faster.

The computer uses the input from this sensor to keep the mixture injected always as close to λ=1 as possible. If the sensor is still too cold to give accurate input, the computer can ignore it safely.