WAI vs CAI test results
 

i'll post the details later, but the summary is:

CAI (performed 5 bi-directional runs)
54.3 F avg air intake temp
79.12 km/gal (US) avg consumption

WAI (performed 4 bi-directional runs)
106.7 F avg air intake temp
79.31 km/gal (US) avg consumption

it's always disappointing to put this much work into something and not see any significant results.

in case you're wondering, that's a .2% (point two percent) difference of WAI over CAI. i haven't calculated margin of error yet, but i suspect it's within it, which would make the difference statistically insignificant.

details and methodology to come later on...

That is disappointing. If
 

That is disappointing. If you ever need help with calculating your margin of error, etc just send me the numbers and I'll run them for you really fast.

I honestly believe this is
 

I honestly believe this is the ECU correcting for the warm air. This makes me want to hurry up and throw my Hondata in. One thing at a time though...

Quote:I honestly believe
 

Honda ECUs do not do this, they lean the mixture based on hot air, though I'm not knowing much about obd2 gm stuff.

but the ECU may also be
 

but the ECU may also be controlling the timing advance based on intake air temp. in which case the gains in air temp are erased by less advance.

i can only view "instant" timing advance on the scangauge, though, so i can't really make any comment on what it was doing on average.

btw, for the CAI data:

0.18 = standard error (using the formula: (STDEV(F13:F21))/(SQRT(COUNT(F13:F21))) where F13:F21 are the cells containing the individual data for each run)

correct me if i'm wrong, matt, but standard deviation (a.k.a margin of error) is 2x standard error, or .36% in this case. which confirms my prediction that the .2% difference seen is statistically insignificant.

You're better off just
 

You're better off just posting the results from all four runs if you have it. Post the miles and the MPG. We'll run a t-test or even a two way anova to determine if it was statistically significant. I can post all of the relevant information as well.

On the ECU topic, does obd2
 

On the ECU topic, does obd2 correct timing for it? That would be seriously sucky!

* figures are average of
 

* figures are average of bi-directional runs (cancels effect of wind/grade)

* "test" road: is a nearly level, 6.6km stretch (13.2km round trip)

* speed was set at 89 km/h - cruise control was set once, cancelled with the brake pedal between runs; "resume" feature used for subsequent runs (cruise control consistency verfied by "avg speed" feature on scangauge)

* car was up to steady speed on cruise control when passing a fixed "start" or "stop" marker; scangauge "avg" logging was either reset (at "start") or recorded (at "stop" marker)

CAI (units: km/gal)

78.5
79.05
79.4
79.4
79.25

WAI:

78.9
80.2
78.35
79.8

Re: On the ECU topic, does obd2
 

yes, the level of timing advance is changing constantly, but i don't know what sensors it's using for input for the control.

it's a good point... when you get down to this kind of tinkering, you really need to know the precise details of how the ecu is going to respond.

it makse sense that temperature would factor in as one of the timing controls (but whether it's using coolant or air intake temp, or both, i don't know). it's certainly using throttle position in the formula.

throttle position also changes based on air temp. the hotter the intake air, the wider the throttle has to open to continue producing the same amount of power (thus reducing throttling or pumping losses - one of the several theoretical reasons WAI is more efficient). but the TPS is also a factor the ECU uses to determine a/f mixture and timing advance - a wider throttle would normally change both a/f and timing for the worse, mpg-wise.

it's entirely possible that there's a "sweet spot" intake temp which balances out the pros and cons of all these variables, and i just didn't find it today.

the only way to find it would be spending hours and hours on the road (this took me 3 hrs) testing across range of different IA temps... or simply having the proper info about how the ECU will respond to sensor input.

Group Statistics
 

<p align="center"><strong>Group Statistics</strong></p>
<table width="100%" border="1">
<tr>
<td>GROUP</td>
<td>N</td>
<td>Mean</td>
<td>Std. Deviation</td>
<td>Std. Error Mean</td>
</tr>
<tr>
<td>WAI</td>
<td>4</td>
<td>79.3125</td>
<td>.84101</td>
<td bgcolor="#FFFF00">.42050</td>
</tr>
<tr>
<td>CAI</td>
<td>5</td>
<td>79.1200</td>
<td>.37517</td>
<td bgcolor="#FFFF00">.16778</td>
</tr>
</table>
<p align="center"> <strong>Independent Samples Test</strong></p>
<table width="100%" border="1">
<tr>
<td rowspan="3">&nbsp;</td>
<td colspan="2"><div align="center">Levene's Test for Equality of Variances</div></td>
<td colspan="7"><div align="center">t-test for Equality of Means </div></td>
</tr>
<tr>
<td rowspan="2">F</td>
<td rowspan="2">Sig.</td>
<td rowspan="2">t</td>
<td rowspan="2">df </td>
<td rowspan="2"><div align="center">Sig. (2-tailed)</div></td>
<td rowspan="2"><div align="center">Mean Difference</div></td>
<td rowspan="2"><div align="center">Std. Error Difference</div></td>
<td colspan="2"><div align="center">95% Confidence Interval of the Difference </div></td>
</tr>
<tr>
<td><div align="center">Lower</div></td>
<td><div align="center">Upper</div></td>
</tr>
<tr>
<td>Equal variances assumed</td>
<td>6.372</td>
<td>.040</td>
<td>-.463</td>
<td>7 </td>
<td>.657</td>
<td bgcolor="#FFFF00">-.1925</td>
<td>.41545</td>
<td>-1.17488</td>
<td>.78988</td>
</tr>
<tr>
<td>Equal variances not assumed</td>
<td>&nbsp;</td>
<td>&nbsp;</td>
<td>-.425</td>
<td>3.956</td>
<td bgcolor="#FFFF00">.693</td>
<td>-.1925</td>
<td>.45274</td>
<td>-1.45504</td>
<td>1.07004 </td>
</tr>
</table>
<p><br>
<br>
<br>
Explanation:</p>
<p>In order for something to be statistically significant at the .05 level (that's what psychology uses, medical is often .01) the Mean Difference must be greater than 2 standard deviations of the error. In other words, error is always introduced into an experiment. The two numbers in yellow in the first table are the error. In order for two groups to be statistically different the difference of the means has to be more than TWO SEMs (std. error of the mean). In both cases the mean difference (.1925) was no where near 2 SEMs for either group.</p>
<p>The conclusion is that on MetroMPG's Firefly, there is only a 30.7% chance that the increase of gas mileage was due to the warm air intake and not something else.</p>
<p><br>
<br>
<br>
<br>
</p>