Power Modification, Testing, Measuring

This section describes some DIY automotive testing methods.

Similar methods can be used to objectively measure the result of other owner-made modifications.


1. Since 1995 I've been interested in altering my 1991 L98 Corvette for better performance. Changes to Suspension, Brakes, Gears, Tires will be covered elswhere.


2. Various things I tried on the Engine to increase its output include:
Porting the heads - new valves, several varieties of springs & roller rockers
. Intake manifolds - stock, Large Tube, Super Ram versions of the TPI design
. Exhaust manifolds - stock, modified, and LPE Long Tube headers
. Ignition schemes - various MSD-type coils/boxes & some home made tweaks
. Cams - stock, LPE, GM Hot Cam


3. Before engine component testing had gone very far, it became apparent the GM stock ECM programming would not work with all the engine changes being made. To get meaningful MEASURABLE test results, I had to get each test combination running up to a fairly good level of performance. This is mostly a matter of matching fuel delivery & spark curves to the various cams, intakes, etc. Optimizing performance of each combination is a test topic in itself, after each combination is/was roughed-in.

Using Diacom & TunerCat software I became adept at configuring new Spark and Fuel Tables to make the components being tested run well. Adeptness was mostly a case of running several hundred trial & error tests each year, until I got proficient at it. Examples of a few things that need changing for various engine combinations are these:

Stock GM Program

Example of Modified Program

Spark Advance vs MAP vs RPM



Fuel Curves vs MAP vs RPM


Power Enrichment Fuel vs RPM



4. Once the ability to reprogram & optimize the ECM was in hand, it also became apparent I needed a method to objectively measure the result of changing components & ECM parameters. Here is some information on the testing method I eventually settled on, followed by some details of the measuring tools I built to support the testing:

Modifying the stock engine resulted in changes to two general categories of engine behavior - part-throttle and wide-open-throttle (WOT).

Part-throttle characteristics that need adjusting with new engine components include:
. Cold & warm start up
. Idle in and out of gear
. Low and high speed cruise, TCC lock
. General manners in traffic, pump shot (AE)
. Stalling, spark knock, operating temperature, throttle transitions, etc

These are all somewhat subjective measures of engine performance, and can generally be adjusted to personal taste with patience, much ALDL data logging, and lots of tweaking to the ECM program parameters. These tweaks & evaluation methods are not detailed here. Howerver, an excellent summary of the basics of part-throttle tuning is found at: http://www.thirdgen.org/techbb2/showthread.php?s=&threadid=200256


Full-throttle characteristics that need to be adjusted with new engine components include:
. Air-Fuel Ratio (AFR)
. Spark timing
. Power output
. Throttle transients (PE), DFCO, etc.

These parameters can only be measured and adjusted at full WOT. Unless one has constant access to a load dyno, a race track, or salt-flats, these items must be measured via road tests, a little at a time. I decided to do my tuning with fully-loaded vehicle road testing, for convenience, repeatability, and to get real-world results in my driving enviornment. Safety is the paramount concern in all testing.

My preferred test method is to use a steep hill on a private road. I try to keep all the test conditions constant from run to run. The hill incline allows maximum engine load in the shortest distance. Making test runs from a rolling start keeps wheel spin out of the test data. Making all test measurements in one gear gives a common basis for data comparison from run to run. Second gear is my choice for general testing, since it gives the best combination of time in test while keeping terminal velocity within reasonable limits. Once major testing & optimization is complete in one gear, the final design can be tweaked in other gears with little addidional work.


5. Power output of the engine can be determined using Newton's laws, provided you measure and record accurate weight, speed and time for any motion of the vehicle. To remind those who misplaced their high-school physics text:
. Force=mass x acceleration
. Work=force x distance
. Power=work/time
. Kinetic energy=1/2 mass x velocity squared

Thus engine power changes resulting from different components can be inferred directly by comparing time/speed/distance datalogs of each test, with mass held constant. To calculate actual power expended you need to back out the velocity and distance factors in absolute numbers.


6. Since you need to look at the full rpm range of the engine when evaluating hardware or software changes, acceleration testing is a convenient mechanism. It allows all rpms to be considered in a single WOT pass; and changes in rpm, speed and time are easy to measure. The calculation methods to convert the recorded data to real numbers are not presented here, but if you are not sure of how to use the math, Dr. Chris Jacobs' book High Perfornance Ignitions can walk you through the calculations in detail.

Using the road acceleration method to evaluate engine hardware and software changes revealed another measurment problem: Most of the engine parameters of interest (temp, oil press, voltage, rpm, spark advance) can be recorded from the ALDL stream using a scanning/loggong program like Diacom+ or TTS Datamaster. They are quick, compact, and efficient means to log info - except for one shortfall: The ALDL data stream from a P4 ECM produces only 6.7 frames of recorded data per second.


7. Since the P4 ECM is updating its spark and fuel maps approximately 80 times per second, the standard ALDL stream misses ~90% of the data points on a short-duration throttle transient. Also, the ALDL stream only has access to the narrow-band (stock) O2 sensor. This device is not reliable outside steady-state engine conditions as a source of air/fuel ratio (AFR) information, especially for quick acceleration measurements.

Idealy, when measuring a phenomenon with a frequency of 80/sec you need a measuring/logging instrument with at least twice the sample rate - 160 samples/sec would be perfect in this case.


8. Looking around to decide how to do accurate data logging, I discovered the Dataq experimenter's analog-digital converter (ADC) unit from a note on the DIY-EFI board. It is cheap ($30), easy to use, and records two voltage outputs at 120 samples / second - which was close enough for my work. (The next-best solultion was a bigger Dataq board, with 100X the power of the little one, at about 100X the cost - way more than I need.)


9. At about the same time I was looking for the best testing/datalogging method, the DIY Wide Band O2 sensor became available. This is another compact unit that I found from references on the DIY-EFI board. It uses standard hardware to measure AFRs from ~10:1 to 16:1 with reasonable accuracy and transient behavior.


10. A typical test result datalog then looks like this:

The Dataq ADC keeps a very accurate time scale on the X-axis, and RPM and AFR are represented on the Y-axes. From the slope of the acceleration curve you can derive Velocity vs. Time. Knowing the vehicle weight, this can be solved via a simple spreadsheet for the change in kinetic energy over time (ie changes in engine power from test to test). Speed (RPM) and fuel (AFR) are each recorded 120 times/second. This level of sampling gives time slices (0.008 sec each) that are small enough for accurate tuning on transients. The Y-axis is on an 8-bit scale, so increments are each 1/256 (0.004) of the full-scale value. This was sufficient for my needs.

I like to examine the curves in 500 RPM increments. For each increment, I can then vary spark & fuel (in separate tests) until the power curve is at its maximum.


Some details of the measuring tools I built to record data in this format are here:

Copyright 2003 by DrJ. All rights reserved.