Twin Turbo Head Swap

A subtle difference between N/A and TT engines is the design of the heads. TT heads a cast with larger intake and exhaust ports, as well as some smaller differences in the shape of the ports. These differences can be important, however, as they affect the flow of gasses into and out of the cylinders. If the heads are placed side by side, the differences can be noted.

This is a view of the intake ports, taken from directly in front of the opening. The N/A heads are shown on the left, with TT on the right.

Notice the large ridge running down the middle of the N/A port, as well as the wider and more direct path to the valve on the TT. These characteristics are also visible from other angles. These are taken from a 3/4 angle. The ridge is very prominent, and the narrower walls can also be seen.

The difference in the exhaust ports is less prominent. The TT ports are slightly wider, though this is easier to tell by feel than by sight. Again, the N/A port is on the left, with TT on the right. This is also made harder to see by the carbon buildup on the N/A heads, but it's the best I could get.

The combustion chambers appear to be identical between the two heads, despite the port differences. N/A on the left, TT on the right.

The net effect of the port differences in the TT heads is to improve high velocity airflow. The wider intake ports lower viscous drag on the air, and improve the critical mass flow rate through them. To understand why this is true, it is necessary to understand a little bit about viscous flow in a pipe and fluid boundary layers.

There is no such thing as perfectly uniform flow. Fluid particles traveling next to a wall will experience friction with the wall and tend to lose velocity. Particles next to these particles will have friction with them, and also tend to slow down. This produces a curve known as the velocity profile that maps out the velocity of a fluid particle at any distance away from the wall of the pipe. For a smooth straight 2D pipe, it will look something like this. Note that the fluid will eventually return to its freestream velocity as it gets far away enough from the wall.

Mass flow rate is defined as the product of fluid density (ρ), cross-sectional area (A), and velocity (v). For viscous flow in a pipe, total mass flow rate becomes the integral of ρ*A*v over the height away from one of the walls, where v is expressed as a function of height.
Increasing the cross-sectional area does three things. The first is obvious; increasing A increases the mass flow rate. The second is a result of another law of fluid mechanics known as Bernoulli's equation. This is an inviscid law, but it can be applied for viscous flow as well. In simple terms, it states that as the cross-sectional area decreases, the free-stream velocity will increase. This is basically just an application of the definition of mass flow rate in inviscid flow and conservation of mass. In the end, though, the result of increasing the cross-sectional area is to decrease the free-stream velocity and thus the mass flow rate. However, note that both of these effects are linear, and so, they effectively cancel each other.
Which brings us to the last effect of widening the port; the viscous effect. As the port narrows and the velocity increases, the boundary layer also gets larger. At some point, the upper and lower boundary layers will meet, as shown below.

As each boundary layer grows, it tends to decrease the total mass flow rate. If the area continues to shrink, the boundary layers will begin to decrease the freestream velocity. There will come a point where the boundary layers decrease the freestream velocity faster than the contracting pipe can increase it. This is known as choked or critical flow. Continuing to decrease the pipe area past this point will only decrease the total mass flow rate, despite the inviscid increase in freestream velocity.

Few devices will actually operate under critical conditions for very long. Jet and rocket engines do so routinely, but they are an exception. For our purposes, airflow through an engine's heads will probably not be critical. Also, heads must be able to maximize their velocity at widely varying mass flow rates, corresponding to different rpm. This leads to a lot of compromises when designing head ports, usually to optimize them for a particular mass flow rate, while trying to maintain fairly good characteristics when out of this range. Specifically, the TT heads are designed for a higher mass flow rate at high rpm than N/A heads. This makes sense, as turbo engines will be ingesting much more air, especially at high rpm. N/A heads may flow slightly more air at low rpm, but the difference should not be extreme, and when asked to deal with high demand at high rpm, they may be nearing or even exceeding their critical mass flow rate. This is clearly something to be avoided. All in all, I chose to use TT heads on my car.

The next issue to decide was that of cams. There are a number of factors that effect choice of cam, most of which revolve around high-rpm power or low-rpm torque. A cam lobe is defined by its duration and the position of its peak with respect to some baseline. It is also possible to have an assymetric lobe that rises faster on one side than the other as well as twisted lobes or lobes with varying profile along the axis of the camshaft for use with computer controlled valvetrains, but I won't consider those since they were not choices availible to me.
The camshaft itself is defined not only by its lobes, but also by the separation between its intake and exhaust lobes, and the amount of overlap between them, if any.
In a perfectly ideal engine, the intake valve would open when the piston is at TDC on the intake stroke and close when it hits BDC. Likewise, the exhaust valve would open at BDC and close at TDC. In this case, as shown in the timing digram below, there would be no overlap between the two lobes of the cam.

Under actual conditions, there are other forces at work. Air is a fluid, and as such, it has momentum. It takes a finite amount of time to accelerate from rest change speeds. This can be used to the engine's advantage by clever valve timing. On the pistons intake stroke, once it reaches BDC, it no longer actively draws air in, as it is at the very beginning of its compression stroke. However, the air near the intake valve and in the intake passage of the head is still moving. Its momentum will continue to carry it into the cylinder even though the piston is no longer creating a vacuum. In fact, at this point, the piston is trying to move air out of the cylidner, even as the air's momentum tries to move it in. For a short period of time, when the piston is still near BDC, the momentum is the stronger of these two forces, and by leaving the intake valve open for a few degrees of crank rotation past BDC, the engine can increase the mass of air in the cylinder for the compression stroke.

Fluid momentum can also be taken advantage of during the transition between the exhaust and intake strokes. As the piston moves toward TDC, forcing the combustion products out through the exhaust valve, these gasses gain momentum. Once the piston passes TDC, it is no longer pushing the spend fuel and air out, but the momentum remains. This creates a vacuum in the combustion chamber, in addition to that produced by the downstroke of the piston (which is very slight, at this early point in the stroke). If the exhaust valve were to remain open as the intake valve opens, this vacuum effect can help draw more air into the cyilnder, giving the intake air its all-important momentum. This is known as valve overlap.

The final reason to deviate from the ideal timing model is found in the combustion stroke. During this stroke, it can be helpful to open the exhaust valve slightly before the piston hits BDC. This allows the remaining pressure in the cylinder to begin forcing exhaust products out before the piston begins its upward motion. This reduces the amount of work the piston must do on the gasses to drive them out of the cylinder.

Thus, most actual timing diagrams looks something like this. These two particular diagrams are for the N/A and TT cam profiles on the left and right, respectively. Note that the intake cams are no different between the two motors. The only difference is in the exhaust cam. Specifically, the TT exhaust cam has 4 degrees longer duration and 2 degrees greater overlap with the intake cam.

Naturally Aspirated Twin Turbo
Open Close
Intake 16 deg BTDC 55 deg ABDC
Exhaust 48 deg BBDC 15 deg ATDC
Open Close
16 deg BTDC 55 deg ABDC
50 deg BBDC 17 deg ATDC

All the effects noted above are dependant on fluid momentum. Therefore, it makes perfect sense that they would be most effective when the fluid being considered has plenty of momentum to start with, for example, at high rpm. In general terms, a cam which deviates greatly from the ideal configuration will perform best at high rpm. The same holds true for a cam with lots of overlap between the exhaust and intake valves.

For my application I chose to use the N/A exhaust cam. This was my reasoning.
I am building my car to have excellant torque and responsiveness at any rpm, as well as high peak power. A longer overlap will cost torque at low rpm, because the exhaust valve stays open longer than it needs to. I'm willing to sacrifice that small bit of peak top-end power for the additional torque that a smaller overlap brings.
My car uses relatively high static compression ratio pistons. Because of this, the pressure against the piston will be greater at any point along the stroke as compared to a that of a lower static compression ratio. Because of this, the exhaust valve can wait a little bit longer to open and still have the same amount of exhaust pressure to drive the gasses out. In addition, this extra time is time that that pressure is still doing work on the piston and adding to the net power of the motor.

For these reasons I chose to use N/A exhaust cams with TT heads.
Actually replacing the heads is a bit of a chore. The first step is removing the battery, battery tray, intake piping to both turbos and to the throttle body, and it also helps to detach the rear turbo IC pipe to get some room to move around. The upper timing covers and the upper plenum can also be removed. It's a good idea to plug the lower plenum openings with rags to prevent any loose nuts or washers from falling in. At this point, I chose to drain and remove the radiator for a little more room to work near the front head.

Now that there is some space to work, the next step is to remove the valve covers. Pull all six spark plug wires. There are eleven small bolts on each cover. After removing these, a narrow flathead screwdriver can be wedged under the cover at the point below and levered against the head to break the cover's seal. Underneath will be something like this. Note: the lower plenum and fuel rails have also been removed in these pictures.

Next things to pull are the turbos and exhaust manifolds. This is not technically necessary, but it makes things a lot easier later. Remove the downpipe, disconnect the turbo's oil and water feed and return lines, and remove all seven nuts from each exhaust manifold. This is possible to do with the turbo still attached to the manifold. It just requires some creative wrenching angles and a little patience. The turbo's/manifolds can be worked off the studs on the heads and left sitting free in the engine bay.

Remove the driver's side motor mount next. The A/C compressor bracket can be used as a convenient jack point to relieve the load on the motor mount. Remove the pin through the center of the mount and the three nuts that hold the mount to the engine.

Remove both accessory belts, break the bolt on the crank pulley, and remove it as well. If the engine is raised up a little higher, the motor mount bracket can also be removed. After this, remove the bolts attaching the lower timing cover the engine and pull it out. It will have to be snaked through the crank angle sensor wiring harness. This will reveal the timing belt, timing tensioner and water pump.

The crank pulley bolt can be used to rotate the crankshaft until all four timing marks on the camshaft sprockets line up with the triangular marks on the valve covers (the covers can be layed back down on the heads to do this), along with the mark on the crankshaft. Mark the timing belt clearly at each timing mark with a chalk pen, also indicating the direction of rotation of the belt and which timing mark goes with which sprocket. One of the timing marks is visible here, though it is clearly not lined up with anything.

This is the tensioner and tensioner pulley. To release tension on the belt, loosen the eccentricly placed bolt on the pulley. The belt will snap out of tension and can be removed.

Remove the eight hex 10mm head bolts from each head, and rock the heads to break their seal against the block. Also, at this time, remove the coil pack from the front head, and the four bolts holding the water housing the heads and rock it loose from the heads.

The heads should now be free to lift off, but they are still held together by a piece of sheet metal. If there are an extra pair of hands availible, the two heads can be removed as a unit. If not, some creative contortioning is needed to detach the piece of metal. Regardless, once they are off, things should look much more open in the engine bay.

This is where a problem sprung up, although I was expecting it. My 1993 model engine uses separate cam and crank angle sensors, while the 1991 TT heads I was planning to install used the earlier model combination sensor that mounted on the rear intake camshaft. Consequently, these heads have no place to mount the 93+ sensor, which mounts underneath the rear exhaust sprocket. The 91 heads are on the left, with 93 on the right. The additional mount points are obvious.

The sensor attaches to these mounts like so, so sit underneat the exhaust sprocket. My plan was to make use of the nearby mounts to construct two brackets that would hold the 93 CAS in its proper position, similar to this.

Having the 93 heads to work off of made this little bit of fabrication much easier. After a little bit of time with a dremel and some trial and error, I ended up with this rough product.

They matched the stock mounts almost exactly.

The only problem is that the brackets sit at different heights. This was easily fixed with a few washers to bring the CAS up to its proper height.

These washers ended up being attached to the brackets, along with nuts on the reverse side. I also smoothed the edges down and cut a couple of notches to make them fit better. I then ground all the rust off of them and primered the brackets to keep them from deteriorating over time.

These are the brackets in place on the head. They fit perfectly and held the CAS securely in place.

I also took the opportunity to clean the engine block up a little bit. It had over a decade's worth of road dirt, leaked oil, and who knows what else on it. For example, this is a small taste of what came out of the space between the two cylinder banks. I had already cleaned out the big stuff by this point.