Project Matrix: Part 2

Every bit of automotive intuition I have says this can't be done, but we just did it. We turbocharged Toyota's 2ZZ-GE. Five years ago, I didn't think a production car could get away with 11.5:1 compression. One year ago, I didn't think you could turbocharge a production engine with 11.5:1 compression. Even now, I'm still wondering how we got away with it on 91-octane gas.

Despite the technical hurdles, we not only got away with it, but we made it a fun, driveable everyday car that's reliable enough for a 1,000-mile trek to Texas. We also pushed it from 159 hp to 231 hp at the wheels, and cut 1.9 seconds from the quarter-mile time, dropping from a 16.6 at 84.7 mph to a 14.7 at 98.9 mph. Here's how we did it.

Start with a kit
Turbocharging an engine gets a lot easier if somebody else does most of the work. XS Engineering already had a turbo kit for the Celica GT-S, and the Matrix uses the same drivetrain as the Celica, so everything but the intercooler plumbing fit our car.

The XS kit starts with a very short, fabricated manifold. The standard manifold is made of mild steel, which has a limited lifespan in this high temp environment. If you're building a daily driver, a car that will see track use, or even one that will be used only occasionally for many years, add a serious ceramic coating to the inner and outer surface of the manifold to reduce the temperatures encountered by the manifold itself. We've put several thousand miles on ours with no problems, but we don't expect the uncoated manifold to be problem-free for 50,000.

The turbo itself is an IHI VF23. This quick-spooling turbo has a ball-bearing center section for durability and faster response. The VF23 would be a good mild upgrade for a WRX, for example, maxing out at about 350 crank hp. On the Matrix, where it's mounted more than a foot closer to the exhaust ports than on the WRX, it spools instantly, giving the car some much-needed low-rpm torque and dramatically improving around-town driveability.

The VF23 is designed to accommodate water cooling, though, like most aftermarket turbo installations, XS doesn't use this feature. Like the ceramic coating on the manifold, it would be a good idea for long-term durability, and could be easily accomplished by teeing into the heater circuit less than a foot from the turbo.

The downpipe, surprisingly, took a little bit of work to perfect. Exhaust flows from both the turbine and the wastegate have to merge before heading out the exhaust, and with an internal wastegate, that happens right at the turbine outlet flange. Originally, we tried a flat flange. This forced the wastegate exhaust to turn 90 degrees, slam into the turbine exhaust, and turn again to go down the downpipe. Apparently, this caused so much restriction the wastegate couldn't bypass enough exhaust to keep boost at our conservative 7 psi setting. During our first nervous dyno pulls, before the tuning was perfected, and before we were really sure we could get away with this, the boost would jump to 7 psi, hover momentarily, then start a rapid climb. We backed off at 12 psi.

Returning to XS Engineering, Koji Arai ported the wastegate and modified the downpipe flange to allow a more gradual blending of turbine and wastegate exhaust flows. It worked.

We were surprised by the need for this design change because XS had used the flat flange on its turbocharged GT-S with no problems. It tuned that car in the heady days of 92-octane gas, though. With the 91-octane cat pee we're now forced to run, we had to retard the ignition timing a few more degrees, which sends more combustion energy flying out the exhaust. Just enough energy, it seems, to overwhelm the wastegate.

Downstream, a 2.75-inch Car Sound catalytic converter bolts to the downpipe. The stock catalytic converter actually appears to be a high-flow design, but it's integrated into the exhaust secondaries, so using it would require some butchering. We've found Car Sound cats to be consistently good performers, so it was a natural choice. The rest of the exhaust was fabricated from 2.75-inch mandrel-bent tubing by A'pexi using its N1 muffler. A'pexi has a bolt-on exhaust for the Matrix, but its smaller tubing is designed for the naturally aspirated engine.

Intake plumbing on the XS Turbo kit is unconventional. With the turbo stuffed behind the engine, and the engine pressed up against the firewall, there's little room for an intake pipe to feed the turbo. When the company developed this kit on the Celica over a year ago, XS tried building an intake pipe, but the convoluted shape proved too restrictive. In the end, the car made more power with the air filter clamped directly to the compressor inlet. There's no hotter place to draw intake air than right next to the exhaust manifold, but Toyota leaves us little choice. To counter this excess heat and, face it, because it looks good, XS uses an absolutely enormous intercooler. The core is actually twice as big as the opening in the bumper, but we plan to address that with either a different bumper or a hole saw.

Compressor surge is prevented with an A'pexi Twin Chamber blow-off valve. The A'pexi blow-off valve was a rare piece in this project, because we never had to do anything with it. No adjustments, no tuning, no re-thinking. We bolted it on and without touching the pre-load adjustment it worked perfectly. What a relief.

Manage it
Anything is possible with proper engine management, but even the best laid plans can be destroyed with a clumsy tune. XS Engineering's Celica was successfully tuned with what we call the black box method, by adding several single-purpose tuning devices to fool the stock fuel system into delivering the needed fuel. Rough fuel enrichment was accomplished using a rising-rate pressure regulator, and fine-tuning was accomplished with an A'pexi Super AFC. The AFC intercepts and changes the airflow meter signal, allowing you to move to a different part of the stock fuel curve.

Using a rising rate fuel pressure regulator requires ditching Toyota's returnless fuel system, adding a return line, a different fuel rail and a manifold-referenced fuel pressure regulator. That sounded like a lot of work. The black box method, though often effective, also tends to leave idiosyncrasies in the powerband--little bogs, lean spots or areas of poor knock resistance that might be easy to drive around, but only if you know where they are. If, like us, you have to hand the keys over to photographers, Toyota executives and transportation companies, you will want a system with the flexibility to tune out all these problems.

Running the 2ZZ-GE with a stand-alone engine management system isn't exactly trivial, though. In addition to fuel and timing control, the 2ZZ requires a pulsewidth-modulated signal to control intake cam timing (VVT-i) and another simple switched signal to control the switch between low- and high-rpm cams (the L in VVTL-i). To complicate matters, the valve timing needs closed-loop control. In other words, the computer needs the ability to verify that the cam timing is where it belongs and make adjustments to the VVT-i solenoid signal to compensate for changes in oil pressure, temperature or viscosity. A sophisticated system like a Motec, Autronic, the new Electomotive TEC3, or AEM's new Programmable Engine Management might be able to do this, but closed-loop control from a cam timing signal is not something any of these systems were probably designed for, so if it's possible at all, it would take a lot of experimentation.

Short on time and fearful of adventure, we decided to let the stock ECU stay in place and continue to control cam timing and the more mundane features like idle control. The stock ECU also thinks its still controlling fuel and timing, but we disconnected the stock injector leads and removed the stock coils.

In their place, an Electromotive TECII handles fuel and ignition delivery (the TEC3 wasn't available when this project started). The TECII is a speed/density system that uses rpm and manifold pressure to determine fuel requirements, so a manifold air pressure (MAP) sensor had to be added. The TECII also needs a 60-tooth crank position sensor, something the stock engine doesn't have. Since both the TECII and the tuning talent were coming from Vishnu Performance, we also had them make the toothed wheel and the mount for the Hall-effect sensor that gives the TECII its crank position signal.

Because the TECII has integral coils, we also had to have spark plug wires custom made to replace the stock coil-on-plug ignition. Magnecor was able to make wires that fit our unique application. We also switched to a colder, BKR7EVX-11 NGK platinum spark plug.

Adequate instrumentation is critical to tuning, so we added A'pexi fuel pressure, boost, and exhaust gas temperature gauges. The boost gauge helped us recognize our downpipe-induced boost creep, the EGT gauge lets us see that combustion temperatures were safe, and the fuel pressure gauge showed us the stock fuel pump was not to be trusted with boost. We mounted the gauges in a CarbonTrix pod on the dash. The pod, inspired by the light pods on rally cars, turns out to be far more effective than the typical A-pillar mount. With three gauges mounted on the A-pillar, each will be at a different distance from your eyes, requiring you to re-focus to view each gauge. Pillar-mounted gauges can also block your view entering a left turn. With the on-dash pod, the gauges are at approximately the same focal length as the stock instruments.

Tune it
Shiv Pathak, a longtime contributor to this magazine, and for the last year or so the owner of Vishnu Performance, did the tuning. Pathak's tuning experience goes back to well before we met him, and the majority of that experience is with the TECII.

To ensure we didn't blow it up just trying to get it started, tuning started before the car was even turbocharged. The first step in tuning is always the idle, since an engine that won't idle is a pain in the ass to tune. The stock ECU still has control over idle airflow via the idle control solenoid, but we have to provide it with an appropriate amount of fuel and a proper ignition advance curve. The TECII actually allowed us to precisely control idle with closed-loop ignition timing feedback, a technique usually reserved for OEM systems.

Next, it's off to the dyno, where the volumetric efficiency, or VE table is tuned. The TECII bases its fuel delivery on a simple linear model of airflow called the theory of linear thermodynamics. This model is based on the fact that airflow through any given engine is proportional to manifold pressure. In other words, if an engine cosumes 1000cc of air per revolution at 1 bar of absolute pressure (ambient pressure at sea level, that is) it should consume 2000cc at 2 bar (what we would call 14.7 psi of boost).

Of course, this doesn't account for the fact that volumetric efficiency, or the engine's ability to pump air, can be better or worse at different rpm points. The VE table is a map of offsets above or below the simple linear model that help teach the TECII a little lesson in reality.

Pathak tunes the VE table and the ignition map simultaneously, using a mix of intuition, experience, and good old-fashioned guesswork. Working with one area of the powerband at a time, he'll adjust fuel delivery until peak power is found, then do the same with timing, then return to fuel delivery to see if the change in timing affected the need for fuel. This second look at fuel needs is helpful for finding the unique demands of strange and relatively untested engine/ turbo/octane combinations like ours.

After a rough tune this way, a wideband O2 sensor is used to fine tune the VE table and fill in the part-throttle areas of the map. Tuning around the cam switchover turned out to be quite difficult, as intuition and experience clashed with reality. Contrary to all our expectations, when the high-rpm cam is engaged at anything less than full throttle, the VE table had to be leaned out.

With careful tuning, the naturally aspirated driveability was actually improved over the stock car. In stock form, off-idle throttle response is sometimes laggy and, more importantly, there's a pronounced bog after shifting gears. These are probably artifacts of Toyota's emissions-reduction strategies that we didn't replicate. The air's loss is our gain.

Once the turbo was in, the tuning naturally had to be changed. The first step was to add an overboost fuel cut to protect the engine from just the kind of unexpected boost creep we ran into.

Next, to ensure that transient lean spots didn't occur during quick transitions onto boost, Pathak added very aggressive fuel enrichments triggered by sudden changes in either manifold pressure or throttle position. These overly-aggressive enrichments were toned down later when we were sure it was safe. The upper reaches of the VE table and ignition map, where the naturally-aspirated engine simply couldn't reach, could now be tuned. A wideband O2 sensor, the array of fuel pressure, EGT and boost gauges, and six paranoid, detonation-sensing eardrums were all on hand to make it possible.

With all the performance-related parameters dialed in, the last step is cold-start tuning. Pathak tries to tune the raw fuel curve a little on the rich side so any time the system is in closed loop (reading the O2 sensor and correcting fuel mixture on its own), the computer is adjusting lean. This helps cold start because the engine runs open loop when it's cold, which is precisely when it needs a richer mixture anyway. This, combined with some coolant-temp-based fuel enrichments and ignition advance trims ensures smooth running under all conditions.

Driving it
With the tuning finally done, the driving experience is impressive. At idle, the exhaust has an amusing bubbly sound, like Roger Rabbit is cooking popcorn in the trunk, but mash the throttle and the popcorn turns Formula 1. The turbo spools instantly, even at part throttle, giving the normally torqueless 2ZZ-GE a healthy wad of low-end grunt. At 6000 rpm, when the big cam hits, the engine gets pissed. Keep it pissed for long and the ticket will hurt.

Driven hard, the engine is very thirsty. Making power takes fuel, but making it with our ridiculous mix of compression and boost takes a bit more. Dropping the compression to 10:1 or running 95-octane gas would allow far better efficiency and much more power, but even with these limitations it's working spectacularly.