Quote:
Originally Posted by swamp2
Is the "problem" the uniformity of the gains across the entire rpm range, the peak gained value or the average value gained?
|
Whichever raises your red flag most I suppose.
For me the uniformity of the gains across the rpm range caught my eye as something unusual enough to quantify.
Quote:
If the variable being condensed to a single value for the distribution plots is simply the average % power gained that is a totally separate issue from the gain vs rpm curves linearities.
|
Perhaps you could use average power gained, but what I looked at was the acceptable range of deviation in percent power gained across the rpm range.
Quote:
Also, somewhat nit picky, but the values in the y axis should be 100 times larger if the title of the chart is "Percentage...".
|
Haha, good catch and absolutely true. I formatted those spreadsheet cells to display in percentage but apparently visual/display formating does not transfer to graphs.
Quote:
In my view the linearity is the clue to a cheat and then the average value of the curve also becomes suspect. Would a very linear curve of gain vs. rpm but at a constant (low) say 4% gain be a red flag? I don't think it would be. It is the combination of high linearity and high average gain that is the "double whammy".
|
I think it is and those normal distribution graphs are representative of this. OE's gain at Gintani's is only ~4.6% over stock which by itself doesn't raise any flags.
The remarkable linearity (as you aptly put it) of the raw gains are:
OE @ Gintani:
4000 9
4500 10
5000 11
5500 12
6000 13
6500 14
7000 15
7500 16
8000 16
ESS:
4000 6
4500 6
5000 7.5
5500 13
6000 12
6500 16
7000 17
7500 15.5
8000 17.5
AA:
4000 13
4500 13
5000 14
5500 20
6000 27
6500 31
7000 23
7500 26
8000 29
GIAC:
4000 7
4500 9
5000 13
5500 11
6000 17
6500 16
7000 21
7500 21
8000 25
Quote:
Really you might want to make 2 sets of plots. 1 that is simply average % gained across all rpms vs. vendor. The other would be some measure of the linearity of each curve. The most simple metric would max-min. You could also use the "RMS" value (square root of the sum of the squared deviations of each rpm point - (Sum((value-mean value)^2))^1/2 ).
|
What I've done is very similar to RMS, but you're right it would be interesting to see the average % gained at each rpm point. I'll try and work on it tonight...