This was verified by data logging.

I suspect this was done as a means to reduce torque by reducing boost with a partially closed throttle. Baseline dyno data is shown below:

HP and Torque

Boost and AFR

The test vehicle (09 STI) also had the AEM dryflow panel filter
HP and TQ

Boost and AFR

[2007 STi stock HP vs MPH]

For manual transmissions, for a given gear, there would be a constant relationship between engine speed and road speed and so the results can be displayed as power vs Engine Speed (RPM), a more common format.

[2007 STi stock HP vs RPM]

To complete the picture, torque is then back-calculated from the power results, using the formula: Torque=(Horsepower*5252)/RPM.  Result is the common dyno chart format.

[2007 STi stock HP/Tq vs RPM]

What happens with automatic transmissions? With the torque converter between the engine and the transmission, the relationship between engine speed and road speed is no longer constant.  The dyno doesn’t know (or care) that the torque converter is slipping; it will still determine power as a function or roller/road speed.

[FXT auto HP vs MPH]

So, what to do about engine speed? In addition to the “RPM per MPH” constant multiplier, setups such as Dyno Dynamics provide an option to measure engine RPM with an inductive probe (http://www.fluke.com/Fluke/auen/Accessories/Other-Accessories/RPM80.htm?PID=56651) that detects the pulses of the spark plugs.

The slippage in the torque converter is apparent in the following plot, where the recorded Engine RPM is shown against Road Speed. The dashed line shows where Engine Speed would be if there was no slippage.

[FXT Auto ign RPM vs road speed]

If such engine speed data not available, using the “RPM per MPH” constant to convert road speed to engine speed can lead to large errors in the results. Below are the power data plotted against measured engine speed as well as calculated engine speed.

[FXT Auto HP vs (calculated)RPM]

[FXT Auto HP vs (measured)RPM]

Since the displayed torque values are back-calculated from the Power vs Engine speed data, the RPM error is propagated and leads to large torque values at low RPM.

[FXT Auto HP-Tq vs (calculated)RPM]

[FXT Auto HP-Tq vs (measured)RPM]

100ft-lb at 2500rpm is not as impressive as 330.

Some links that talks about the torque converter:
http://auto.howstuffworks.com/auto-parts/towing/towing-capacity/information/torque-converter.htm
http://en.wikipedia.org/wiki/Torque_converter#Efficiency_and_torque_multiplication
http://www.cdxetextbook.com/trans/auto/torqConvert/torquemulti.html

Great, more power.  Now, what happens if something were to happen to the water/meth system?  If the jet is clogged, or the hose spring a leak? With the lean fuel map and higher ignition advance, lack of water/meth when it is expected would likely lead to engine damage.

Luckily most water/meth systems have a fail-safe function built in to detect these fault conditions.  Once detected, the system would bypass the boost control and force it to run at “wastegate boost”, or the lowest boost that the turbo will run.  Typically, the car would be tuned at wastegate boost on pump gas only, for when the fail-safe system is active.  The high boost zone would be tuned for water/meth.  The graph below is an example of a setup that runs ~8psi of wastegate boost.  The high and low boost areas are well separated so the ECU can be tuned to run in those two different zones effectively.

Boost creep is essentially unintended boost when the wastegate cannot bypass enough exhaust gas to prevent the compressor from building boost.  Shown below is the boost plot for an STI with catless exhaust.  With wastegate duty cycle set to 0%, boost creep is apparent, with boost rising from ~8psi to 14psi.  The severity of the boost creep will change depending on the ambient air temperature and elevation (worse at low elevation and cold air temp).

With boost creep present, dropping to “wastegate boost” after fault detection will not provide the needed protection for the engine since the boost can climb as high as the “high boost, meth ON” setting.  To reduce/prevent boost creep, the exhaust housing on the turbo can be ported to improve flow through the wastegate.  Another method is to add an external wastegate.

Just something to consider if you are thinking of adding water/meth injection.

At 100% pedal, Requested Torque is ~428.
Request Torque table:

In the Boost target table below, at 3200 rpm and Requested Torque of 428, the boost target is 19.9psi.

Since boost target is not directly linked to the pedal position, it can be manipulated by altering the Requested Torque value. The ECU can limit Requested Torque to a value based on which gear it is in. For example, if I set the maximum Requested Torque value to 400 (even though the throttle table calls for 428), it would limit the boost target to those values in the second-to-last column. At 3200 rpm, the target would be lowered from 19.9 to 19psi.

In the example below, boost target was lowered in 5th gear, to reduce stress on the engine at full throttle. Red traces indicate boost target (accounting for elevation and temperature)

The following is an example from a 2008 STI, where 5th gear would run 0.7* less ignition advance than 4th gear, which runs 0.7* less than 3rd gear.  The red trace in each plot is the calculated ignition advance without any gear based compensations applied.

3rd gear

4th gear

5th gear

To ensure that I get the most out of the Cobb software, I went down to Surgeline Tuning in Portland, Oregon, to meet with Tim Bailey. Tim is one of the first tuners to work on the R35 GTR and has seen the development of the software from the beginning. Ryan from Vex Performance also made the trip to take the technical training as well.

The car in the background is a GTR with the AMS Alpha-9 turbo kit. After giving Ryan and myself some hands-on experience tuning his own GTR, Tim showed us the Alpha-9 GTR, running E85 and soon-to-be released Speed Density version of the GTR software. That is one beast of a car, and the tuning process is just starting.

Email tunes are sometimes necessary because there is no nearby shops that can provide the service. However, to maximize your chance of a successful tune, wideband AFR is a crucial piece of data that should be collected.

The following is an example of an Email tune that did not perform to what the parts were capable of. One of the contributing cause is the lack of AFR data.

Background info on the car:
-2L wrx
-16G turbo
-DW650 injectors
-Injen CAI
-Email tune for Accessport

Reference data before changing maps:
AFR is 9.x:1 (typical value is around 11:1)

Boost: 18, rising to 20psi

Road dyno result: For comparison, torque at 4000rpm (18psi) is roughly what a TD04 would do at 15psi

Fix the AFR…

Run lower boost…

Make more power…

Closed Loop Fueling

First, some background information on how the Subaru ECU controls Air-to-Fuel Ratio at low load (idle, low throttle, cruising) where AFR is kept at ~14.7:1

1. Mass Air Flow (MAF) sensor measures amount of air going into the engine.
2. ECU calculates duration to open the fuel injectors based on injector flow rate, MAF reading, and desired AFR (14.7)
3. Resulting AFR is measured by an Oxygen sensor in the exhaust manifold.
4. If the actual AFR is higher than intended, injector spray time is increased, if AFR is lower than intended, spray time is reduced.
5. Over time, these corrections are stored as a preset of sorts, to help achieve 14.7 AFR quicker.  These correction values are called Long Term Fuel Trims (LTFT)
6. There are four Long Term Fuel Trim that covers different ranges of air flow rates.  The last one is usually for rates  greater than 40 or 50g/s.  This is the important one, as will be shown below.

Open Loop Fueling

When you apply full throttle the ECU will transition to Open Loop Fueling where the injector pulse duration is calculated solely from MAF readings,the Fuel Enrichment table (desired AFR), and the LTFT value that has been stored for the last flow range (i.e. >40 or 50g/s).  The Oxygen sensor data is no longer used (because it is not accurate under these conditions) as a feedback mechanism to correct for errors in the final AFR.  If the value of LTFT is -10% then the injector pulse width calculated from MAF and desired AFR is further reduced by 10%.

The implication of the above is that, if the MAF table is incorrect near, and only near 50g/s, it will develop a value for the LTFT that is not applicable for the rest of the MAF table.

Adding Larger Injectors

So, what happens when you install injectors with higher flow rate?  Without changing anything in the ECU, the same pulse width would  be calculated from the MAF readings but the fuel delivered would be more than before.  In Open Loop operation, this would mean the AFR will be lower than intended.  How do you correct this over-fueling?  The logical method would be to tell the ECU that the injector flow rate has increased so that it can calculate the proper pulse width to use.  An alternate method is to alter the MAF table.  By scaling the values in the MAF table down, the ECU will also end up using lower injector pulse width because it thinks there is less air coming into the engine then the actual amount.

Where is the problem then?

While it is possible to achieve the desired Open Loop AFR by scaling the MAF table, it has to be done across the entire flow range of the MAF table.  Since Open Loop AFR is usually collected under high boost conditions, the low flow area could be overlooked, which was the case for the car that produced the above snapshot, showing high amount of knock and negative fuel trim.

If the MAF table is left with stock values near 2.7V (~50g/s), in closed loop operation, the ECU would have to reduce the pulse width in order to maintain 14.7 AFR, eventually leading to negative values for LTFT.  As it take time to set the LTFT,  it would stay at 0% while testing is being done at full throttle.

So, after the MAF table is manipulated to produce, say, 11:1 AFR under full boost, with LTFT of 0%, the value of LTFT will set over time to a negative value.  The amount would depend on the difference in flow rate between the original and the new injectors.  For example, if the original injectors are 420cc/min and the new ones are 550cc/minute, the flow rate changed by ~30%.  In theory, LTFT would drift to -30% in order to compensate for the injectors.  However, the WRX ECU limits the range of values for LTFT to +/-15%.  In the above screen shot, the value of LTFT at 50+ g/s is -14.99%, the maximum that the ECU allows.

With LTFT at -15%, the calculated pulse width that was necessary to achieve 11:1 AFR in boost is now being reduced by 15%, which means the AFR will now be ~12.6:1, much too lean to support the ignition advance being used and causes detonation.  In response to this, the ECU has to reduce ignition advance to stop the detonation.  The numbers highlighted in the above screen shot indicate the amount of ignition advance that the ECU is removing to combat detonation caused by the lean AFR.

According to Steve Nichols at Dyno Dynamics, Shootout mode does two things:

1) Use pre-configured dyno settings and present these settings in the chart.  The dyno operator cannot access the “eXtra correction” function (referred to in the thread as CF).  In the newer digital dyno software, the ctrl-X shortcut is no longer available.

2) Estimate the engine power (crank power, bhp, or whatever you like to call it).

The different shootout modes that can be used are repeated below:

Shoot_8    Normal eight cylinder passenger cars
Shoot_6    Normal six cylinder passenger cars
Shoot_4    Normal four cylinder passenger cars Normally aspirated twin rotor, rotary passenger cars
Shoot_3R    Triple rotor Turbo/supercharged rotary passenger cars
Shoot_44    Four cylinder 4WD passenger cars eg Subaru WRX
Shoot_81    Very high performance eight cylinder passenger cars requiring higher ramp rates

The modes that I am concerned with are in bold; Shoot_4 and Shoot_44, where the latter is intended for 4WD cars.  Since AWD drive train would have higher loss than a 2WD system, it made sense to me that results generated in Shoot_44 mode would be higher than generated in Shoot_4 mode.

To verify the above, I found a volunteer with a 2008 Subaru 2.5i (MT) and tested it on my 450DS, running 7.3.2.258 software.

You can see below that Shoot_44 mode does indeed generate higher output than Shoot_4 mode.

First, the results in normal mode.

Next, in Shoot_4 mode:

Finally, in Shoot_44 mode: