In the mid 1980's, I worked at a very prestigious auto racing manufacturer (Drake Engineering), where I ran the high performance street car division. While employed I witnessed the meticulous design and testing methodology that helped them win Indy-500 twenty-six times. Using standard scientific methods were the hallmark of their design and testing methodology. Every design was dyno tested on a 1000hp Henan-Freud dyno. I learned the advantages of real motor dynos, and learned why roller (inertial) dynos did more to guess horsepower than to measure it.
It's my dedication to accuracy and perfectionism that has caused me to test my own car numerous times. When the tests didn't work or gave bogus results, I spent my own money to figure out why. It's this dedication to perfection that has caused me to admit errors in some tests, and ask the forum readers to throw them away. I believe it's worse to mislead people about results that are known faulty, than to admit and explain the error and present the corrected results.
In all previous testing, I relied on the Dynapack Dyno to correct the horsepower to the SAE 1349 horsepower correction standards. These correction factors account for temperature, humidity, altitude, and air density. If all of the dyno instruments are working properly, then the corrected HP measurements are intended to adjust for atmospheric conditions -- thus equalizing the readings taken from one dyno to the next, on days with different weather and atmospheric conditions.
In the previous testing at my normal dyno shop, we placed the air temperature sensor in the top air intake. During a recent discussion with Dynapack USA, I discovered placing the temperature sensor in the air intake leads to erroneous 'corrected' results (even though many dyno testers do the same thing). This causes the previous correction factors to read higher than they should have been, and causes the results to be higher as well.
This problem came to my attention when we captured the dyno results for the AA ECU upgrade at a dyno in Orange California. My car is 400 miles away from my home dyno, and the temperature sensor on this dyno was broken. Therefore I had to learn the math behind the SAE 1349 horsepower correction standards, and write my own computer programs to adjust the uncorrected readings myself.
This extra work caused me to re-investigate our previous methodology, and lead to this discovery that our previous methods were faulty. Therefore to have a real baseline and valid set of results, I was forced to re-open my older dyno tests, collect the hour-by-hour weather temperatures from the national weather service, and use the same computer programs to re-apply the SAE 1349 horsepower correction factors to my older results as well. Therefore my baseline will appear lower than my previously posted results.
To demonstrate how the temperature sensor placement affects the dyno results, look at the following graphs. The first graph shows two dyno runs, each with the temperature sensor in the engine air intake. Sample-1 starts the run with an 8% correction factor (which is already about 5% too high) and increases to 11.5% just sixteen seconds later when the dyno run has completed. Sample-2 starts to run at nearly 12.5%, dips to a little under 12%, then increases back up to 12.5% at the end (also 16 seconds later).
According to Dynapack, when the ambient air temperature sensor is placed in
their preferred position, the typical correction factor will range between 0% -
4%, and remain relatively constant during the entire dyno run. Sample-3 below is
an example of a dyno run using the preferred methodology. The run starts at 3.9%
before fluxuating no more than 15/100th of a percent. Basically, this correction
factor stays constant near 3.8% - 3.9%.
I also made one more methodology change to ensure more reliable results.
Instead of hand-picking the best of three dyno results and presenting it as
'fact' -- I examined all of the dyno runs for each configuration. I threw out
the dyno run that was the furthest away from the other two (whether it was
higher or lower). Then I averaged the remaining two dyno runs. Therefore all of
the results below are an average of two similar dyno runs -- not the "best" dyno
run from each configuration.
Luckily I have all of the dyno files with the raw uncorrected data. This allows
me to re-correct the data using the barometric and humidity readings from the
dyno runs, but relying on the National Weather Service (data.nssl.noaa.gov) for
accurate temperature data. Re-correcting the data isn't easy, because it
requires learning a little more than I wanted to know about the SAE 1349
horsepower correction formulas and how to convert and use the atmospheric data
within its equations. Then I must export the raw data out of the dyno files and
import into Excel; write the Excel/Visual Basic programs to correct the data;
then generate the graphs.
Here's a list of the full methodology:
Collect three dyno runs for each tested configuration
Export the raw dyno data, and import into Excel
Remove duplicate RPM entries
Plot graphs for each dyno run
Remove the dyno run most unlike the other two (the 'outlier')
Average the data from the two remaining dyno runs. This will smooth out the
curves when we create the graphs. It does not enhance the results, in fact
it slightly hurts them (duh, it's an average...how could it ever be
Consult the NOAA for the proper temperature at the specified time of day
near the specified dyno location
Apply the SAE 1349 horsepower correction formulas to generate the corrected
Graph the corrected results