Why Standalone EFI?
One of the major differences between our car and most of the other cars you see in the magazines is our use of a standalone Electronic Fuel Injection system (produced by General Engine Management Systems) rather than the stock ECU.
Standalone EFI is a lot more complicated, but offers substantial rewards for taking the plunge. To understand why, first you have to understand how electronic fuel injection computers work:
The primary purpose of the EFI system is to inject a certain amount of fuel into the engine, and then ignite it. In order to do this, the computer must know how much air is passing through the motor at any given point in time.
The tricky part is that it must measure the amount of airflow by mass. What really counts is the number of oxygen molecules in the combustion chamber at the time of ignition. The only way to determine how many fuel molecules to inject in to react with the available oxygen - given that physically counting them isn't practical - is to compute the mass of air in the engine.
Air is a gas, and as such, the amount of mass in a given volume (the cylinder) is dependant on the temperature and pressure of the gas.
This is the way early OEM and modern racing ECUs function. They measure the pressure in the intake manifold, the pressure in the intake manifold, and the current engine speed (which gives volume), do a little math on these values, and presto! You know the mass of air in the combustion chamber.
These systems are called speed/density systems, because they use engine speed and air density (which is a function of temperature and pressure) to determine air mass.
Speed/density systems are very fast, accurate, and because they measure the characteristics of the intake charge in the intake manifold (which is practically in the combustion chamber) they give very good results.
However they suffer from a serious flaw from an OEM point of view.
The actual temperature and pressure (especially pressure) of the intake charge in the combustion chamber is never exactly the same as it is in the intake manifold. The air charge must pass through the cylinder head and past a valve first, and resonances and other factors can reduce or increase (mostly reduce!) the amount of air that actually makes it into the cylinder. In order to account for these differences, a speed/density ECU needs one further piece of information - a correction factor called volumetric efficiency.
The problem with volumetric efficiency (VE for short) is twofold: firstly, there is no sensor you can use to measure it directly; instead, you must run the engine at each load point, and manually adjust it until you get the air/fuel ratio (as measured by an exhaust gas oxygen sensor, exhaust temperatures, and good old spark plug colour) correct. Secondly, any time you do anything to the engine that changes breathing ability (and thus change the VE), you must recalculate all the VE points.
From an OEM perspective, this means as the engine ages, accumulates carbon deposits, plugs cat converters, and so on and so forth, the VE of the engine is constantly changing, and the state of tune is constantly deteriorating. This is the kiss of death from an emissions standpoint. It also means that if a customer does something to make a drastic change in VE - like fits a performance muffler, for example - that the engine tune will suffer, and the OEM may wind up making a warranty "repair" to adapt the car to its new VE
So most OEM ECUs, starting with the introduction of OBD-1, have used a different method of determining the mass air charge. Instead of the speed/density systems, they use a sensor that measures air mass directly, using a mass airflow sensor. These systems, not surprisingly, are called mass-air systems.
From the point of view of an OEM, a mass-air system is a godsend. As air mass is measured straight off the sensor, they are self-compensating for wear or enthusiasts. As long as the OEM programmer has mapped a given air mass vs RPM "cell" in the computer's lookup tables, you can do what you like to the engine's VE and the computer will do the right thing.
As the sensor is usually mounted at the head of the intake path (rather than in the intake manifold) there tends to be a little bit of lag between when you nail the throttle and the computer compensates for the new air mass readings, but this is a minor nitpick.
The killer tomatoe about the stock mass-air system though, and the reason why we tossed it in favour for a speed/density standalone ECU, has to do with the fact that our engine is turbocharged.
The mass-air system measures air mass directly; there is no intake manifold pressure sensor in this system. Neither is there an intake air temperature sensor. But a turbocharged engine raises the pressure of the intake charge substantially above atmospheric pressure, and air charge temperatures can often be far higher than ambient. Detonation - the explosive and destructive uncontrolled burn of fuel - is a function of both charge temperature and charge pressure.
For a turbo motor, there is a large difference (from a tuning perspective) between off-boost and on-boost tuning, even though it is possible for an on-boost and off-boost load cell to flow the same amount of air mass. Tuning parameters that may be entirely appropriate for the off-boost "version" of that air mass amount may well blow up the motor in the on-boost case
The OEM programmers know this too, but they have the advantage of knowing the characteristics of the car as it rolls off the factory floor. They know, for example, that with the engine in its factory state, that an engine speed of X coupled to a mass flow value of Y means that the intake pressure will be N and the charge temperature will be T - and so they can program accordingly.
But once the engine is modified for extra performance, these factory programming assumptions no longer hold true. It is possible for greater mass flows to occur at lower pressures and lower temperatures than was possible with factory equipment. But as the OEM ECU has no concept of "pressure" or "temperature", it cannot "know" that the engine is operating in a safe state. Instead, these load cells are often only reachable in a factory car by dangerous overboosting conditions. Recognising this, the OEM programmers place values for fuel and timing in these cells designed to save the engine from destruction - typically, they dump in fuel and retard timing.
So as a racing engineer, you make a change that should improve performance, but the stock computer reads this as a dangerous situation, dumps fuel and yanks timing, and undoes all your hard work!
There exists inexpensive aftermarket "piggyback" computers (such as the ApexI Super AFC) that function by intercepting the signal from the mass airflow sensor, and then modifying it according to some programmable functions/positions. (on the Super AFC, they are RPM and Throttle Position) By "lying" to the ECU about how much air mass is entering the engine, the idea is to correct for the modifications done to the car, and get the ECU to "do the right thing"
For the price, these piggyback units can be pretty effective. But like the factory ECU, they don't have any concept of "boost" either. As such, the Super AFC tuner is forced to program for either the off-boost or on-boost conditions (smart ones will do on-boost!) and give up accuracy (and performance) for the other state.
Furthermore, the assumption with the piggyback units is that mass airflow signal changes will scale linearly at the ECU - in other words, a reduction of X percent of airflow signal will result in a similar reduction in fuel injected. This does not have to be the case! It is quite possible that a given signal reduction would move the stock ECU into a load cell where it is programmed to increase fuel instead (or worse, decrease it by a larger percentage than you bargained for). As such, doing a proper tuning job with a piggyback computer turns into a session of reverse-engineering the OEM programming - in effect, poking the stock ECU with a stick and seeing what it does.
As if this wasn't enough, remember that the mass air sensor is the only sensor used to determine not just the amount of fuel injected, but also the ignition timing. As such, by modifying the signal from the mass air sensor, you are not only changing fuel, but timing as well. In the specific case of the Talon ECU, reducing the amount of air reported by the sensor not only leaned out the fuel mixture, but also increased ignition advance. This increases power, but does so at the risk of detonation. It is really very important on a turbo motor to have iron control of these two functions independent of each other, and the piggybacks do not provide this level of control.
With the standalone computer, we do have this level of control, and so can get the most out of our engine. We also get access to other functions, such as anti-lag, launch control, and other goodies. The tradeoff is that we have to do all our ECU programming ourselves - and have to do so every time we make any significant change to the engine. For a dedicated race car like ours, this really isn't much of a drawback.
The power increase (and especially the throttle response increase) from switching from the stock ECU + piggybacks to the standalone computer was significant, and well worth the extra effort.
Comment 2010: In 1999, we were breaking new ground here. The state of the art at the time was piggyback computers like the SuperAFC, and we were considered to be way "out there" by replacing the OEM ECU entirely. As I recall, there were exactly two DSMs with this system - myself, and John Shepherd.
I wrote this article to try and explain why it was I was going to the "radical" extreme of replacing the whole ECU instead of just using the SuperAFC like everybody else.
Time marches on, and with the introduction of the AEM EMS (not coincidently, developed by Jason Seibels who was the key player in bringing GEMS into North America - the AEM uses GEMS software) the concept of standalone EFI controllers is pretty well routine these days. Even my street car has one.
Still, it was pretty cool to have been at the tip of this spear.