Siamese-port injection

"You can't fuel inject a Mini!". We've all heard it many times, but we know it's not true because Rover Minis had fuel injection fitted as standard throughout the nineties.

The SPI Mini (Single Point Injection) used a wet-manifold principle, where fuel essentially forms a reservoir in the intake manifold. The Canems ECU (along with most others) can run an SPI setup just like any other car.

The problems begin when we start looking at the MPI (Multi Point Injection) setups, which is the sensible basis for a performance engine. Along with the standard Mini MPI manifold, any multiple-injector setup will fall into this category. For example, fitting a Jenvey-type throttle body onto a Weber inlet manifold or retro-fitting a pair of motorcycle throttle bodies.

So, what's the problem? If you haven't already done so, have a look at this superb explanation by Marcel Chichak: http://www.starchak.ca/efi/siamese.htm

Put simply, due to the layout and cam timing of the A-series and B-series engines, fuel will always be drawn to the two inner cylinders. Under enough load, the outer cylinders will become very lean and power is lost. Needless to say, engine reliability also becomes a major worry.

Start of development

Throughout the development of our ECU, we've taken nothing for granted. Theoretically, the effects of this 'charge-robbing' certainly make sense. However, we wanted to see some proof from a running engine - preferably recorded with our own ECU and the inbuilt datalogging capability.

To monitor outside and inside cylinders on an engine, our ECUs feature dual oxygen sensor inputs, realtime readouts and datalogging. The ECUs can support either narrow band and / or wide band type sensors.

An MPI Mini injection manifold was retro-fitted to a standard A-series engine. We then configured our injection ECU (at this stage still in development) to run the engine to the best of its ability. Effectively, this would be the starting point for our development work. Oxygen sensors were fitted to one outer branch and the inner branch of a three-branch LCB manifold. Datalogging produced the results below:

Needless to say, the performance of the car was very poor. At low engine speeds and light loads, the fueling wasn't too far out. Put under load however, the engine effectively became a two-cylinder device, with just the two inner cylinders (which were now running too rich) producing meaningful power.

So we now knew that the theory was correct, and we'd have to start looking for ways to solve the problem. There is a lot of speculation and theory involved with fuel injection, especially when other factors become involved like the siamese-port charge robbing. It's important to note that our ECU has been developed in parallel with practical tests, and every change that we made to our ECU was in direct response to a real-life, physically tested problem.

Semi-sequential control

The very last Minis used a camshaft phase sensor, along with a special flywheel-triggering arrangement. This meant that each fuel injector could be fired only during the intake stroke of a cylinder that it supplied. This sequential injection technique ensured equal fuel delivery to all cylinders and an engine with balanced AFRs (Air Fuel Ratios) on all cylinders.

The trouble is, virtually all A-series (and all B-series) engines were never designed for fuel injection - never mind sequential injection. Retro-fitting such a system to an older engine would be extremely complicated and require an engine to be stripped down, to a certain extent. Even then, tuning the ECU would require camshaft timing figures and actually mapping the system would be a specialist task.

In keeping with our 'easy-fit, easy-tune' philosophy, we wanted to avoid these complications. A good management system would be something that everybody could fit, tune and get good results from.

For our ECU, we therefore developed a semi-sequential injection technique which takes advantage of the camshaft timing in the engines with siamesed ports. Effectively, the 'semi' in 'semi-sequential' implies that the fueling is controlled in a fully sequential manner for half of the time. Applying fully sequential injection to the outer two cylinders (1 and 4) means that they always get an equal share of the available fuel. The full details of our semi-sequential injection technique have a patent pending - Patent No. GB0719958.1

The big advantage of our semi-sequential injection technique is that you need just one engine speed sensor. Our ECU uses the 36-1 trigger wheel and crank sensor combination which has proved so common with ignition kits like the build-it-yourself MegaJolt. Needless to say, sourcing and fitting these parts is very easy due to their popularity.

Injection phasing

Further engine testing of the injection system gave some interesting results. With the injection events timed to coincide with inlet valve opening, we found an infinitely-improved idle quality. However, performance was still lacking drastically once the engine was under load. We even went so far as to design a real-time dual oxygen sensor readout so we could monitor the AFRs in each cylinder whilst the car was under load.

Eventually, we realised that by moving the injection timing we could bring the power back, but it only lasted in 'chunks' of about 500 or 1000 RPM. Obviously this gave a lumpy (almost turbo-lag type) performance, where the power suddenly came in with a bang. We knew we could find the power at the top-end however, so time was spent on the rolling road.

Rolling-road tests confirmed that the power was as expected at the top-end, so now we needed to concentrate on getting maximum power throughout the entire speed / load range.

Injection advance map

So what was causing these surges in power? The simple answer is time. It takes a certain amount of time for fuel to flow from the tip of a fuel injector to the target cylinder. This time remains more or less constant, irrespective of engine speed or load. Thus, we need to inject the fuel earlier in the engine cycle as RPM increases. Engine load can also affect this injection timing, so effectively we end up needing an injection advance map, rather like the ignition maps which the ECU also features.

With a user-programmable 2D or 3D injection map, our ECU was beginning to solve the siamese-port injection problem. Now that the injection timing could be specified anywhere between 0º and 359º, power delivery was smooth throughout the rev-range. There was a vast improvement over the old carburettor (and distributor) setup. To double check that all was well, we tried running the ECU with a different cylinder head - changing the old CAM4180 to the higher performance 12G940 casting.

With this change, we started to discover more problems. Naturally, the performance engine required more fuel, but this extra fuel mean that the injection timing was being shifted so that it no longer coincided with the open inlet valves on the outside cylinders. Traditionally, sequential ECUs allow you to specify the start-time or end-time of an injection pulse. The trouble is, this makes the fuel map and injection timing map somewhat inter-dependant. If you change the fuel map, you'll need to change the timing map too, otherwise the fuel will be injected at a different angle with respect to top dead centre (TDC).

Independant timing map

We realised that this would cause problems for the after-market fuel injection tuner. At this rate, the ECU couldn't be considered an 'easy-tune' solution. Therefore, we started to modify the injection timing mechanism, so that the specified timing figure would coincide with the centre of the fuel injection pulse. This being the case, it wouldn't matter if the fuel pulse was made shorter or longer (lean or richer) - the injection timing would be unaffected.

Previously, the fuel maps and injection timing map were inextricably linked and this proved to make the mapping process very complicated. With the Canems ECU, once the injection timing map is set up correctly for your engine, you can tune your fuel map independantly.

Injector opening-time correction

Now that the ECU used the centre of an injection pulse on which to base its timing, accurate injector opening-time correction would be needed. This is because an injector pulse is actually made from two different phases. The first phase is the 'injector opening time', during which the valve in the injector is physically beginning to open. During this phase, a negligible amount of fuel flows from the injector. The second phase of the injection pulse is when the fuel is actually injected.

The problem is, the 'centre' of an injection pulse (to which our timing map relates) actually refers to the centre of the second phase of the injection pulse - when fuel is actually flowing. As the first phase of the injection pulse is voltage dependant, the Canems ECU features a voltage correction map for which the opening time can be specified to the nearest 64 microseconds at all voltages.

With an accurately programmed map, you can be sure that the Canems ECU will always position the injection pulse so that its centre of fuel flow corresponds to the figure in your timing map.

Large injector control

Due to the nature of siamese-port engines and their charge robbing problems, they must always be fitted with larger injectors than you would expect. If the injector duty cycle rises above 60% at any time, charge robbing between the cylinders will start to occur. The figure diminishes as camshaft overlap increases, meaning even larger injectors are required.

The Canems ECU has two mechanisms to to help with this. Firstly, extremely accurate fuel metering, with an injector resolution of 64 microseconds. This will help to maintain a smooth idle even with large injectors. Secondly, the ECU supports staged injection control for extremely high powered engines.

Finally....

With the special features and development work outlined above, the Canems ECU provides the simplest, most effective and cheapest solution to the siamese-port injection problem. The datalog below was taken on the same engine as that at the top of this page. The difference is, the Canems ECU is maintaining a balanced AFR between all cylinders

Canems test vehicles have covered thousands of miles in the development of the ECU, all of which have been perfectly reliable. Just see Practical Classics magazine (Jan/Feb '08) for confirmation that the ECU can provide a fit-and-forget solution to the siamese port problem.

Doubtless other companies will copy the work of Canems in the future, but remember that you saw it here first!